The [H] Retrocomputing Thread!

[Red Falcon]

IIRC, there was no SDR SDRAM support for Xeons. You either used RDRAM or DDR SDRAM. Desktop Pentium 4's had SDR SDRAM, but those were quite crippled, performance wise.

Kind of strange this is in the non-Intel/AMD forum. Perhaps we need a retro forum...

It would be very nice if there were a retro or legacy hardware section, and I would gladly move this thread there, but that is really up to Kyle and the mods.
The reason this was put here to begin with is because, as you can see in many of the earlier posts, most of the retro equipment posted was not x86 (Intel/AMD/VIA), and was m68k, PowerPC, SPARC, etc.

Also, I would take SDR SDRAM over RDRAM any day of the week, regardless of the processor.
RDRAM, while technically retro now since it falls beyond the 15-year gap, was terrible back then (ran very hot, was power hungry, and was slower than DDR - not to mention far more expensive after 2002) and is even more ridiculous now - and oh boy, it will not be missed!

 

[dexvx]

In my experience, it's more the front side bus that cripples these systems that were using sdram. And a good agp card seems to make a lot of that slowness go away, although still not as fast as a pentium with a faster fsb.

If we're talking about the same 400 FSB Pentium 4, then SDR SDRAM would lose significantly to RDRAM PC800. By the time 533 FSB Pentium 4's were mainstream, there was no point in using SDR SDRAM because DDR SDRAM was common.

Not sure how some people remembered history, but RDRAM was great. It was definitely NOT good when it debuted on the i820, as the Pentium 3 had no real use for single channel RDRAM that barely added any additional memory bandwidth compared to PC133 SDR SDRAM. The i840 (dual P4/dual channel RDRAM) actually had significant advantages in memory bandwidth applications (like SPEC). It really shone in Pentium 4 early Northwood days.

In the early Northwood days, it was actually cheaper to buy 256MB PC800-40 (which was easily overclockable to PC1066) RDRAM than 256MB PC2700 DDR SDRAM. I actually was in the market for a system at the time, so I remember it quite well. There was a huge trial in which Rambus won millions of $ because it was proved that the major memory OEM's (Samsung, Hynix, Micron, etc) colluded to prevent RDRAM from gaining marketshare. Samsung basically took their RDRAM profits to pad their DDR SDRAM losses. They claimed poor yield, when in fact their normal PC800-40 RDRAM could easily overclock to PC1066 (and PC1066-32 could do PC1200).

 

[Red Falcon]

If we're talking about the same 400 FSB Pentium 4, then SDR SDRAM would lose significantly to RDRAM PC800. By the time 533 FSB Pentium 4's were mainstream, there was no point in using SDR SDRAM because DDR SDRAM was common.

Not sure how some people remembered history, but RDRAM was great. It was definitely NOT good when it debuted on the i820, as the Pentium 3 had no real use for single channel RDRAM that barely added any additional memory bandwidth compared to PC133 SDR SDRAM. The i840 (dual P4/dual channel RDRAM) actually had significant advantages in memory bandwidth applications (like SPEC). It really shone in Pentium 4 early Northwood days.

From the looks of it, a 1GHz Pentium III with SDR SDRAM PC133 utterly destroys a first-generation 1.5GHz Pentium 4 with RDRAM PC800:


That is exactly how I remember history, and granted that software didn't really take advantage of SSE2 in the Pentium 4 at the time, but judging by those benchmarks in the video, the additional memory bandwidth given by RDRAM had next to no effect on any of those applications, and it seemed like the CPU was more of the bottleneck than the memory was (all other components were the same in both systems, so it was tested 1:1), especially when an older Pentium III completely destroys the first-generation Pentium 4; SPEC was also, from what I remember, a synthetic benchmark, when compared to real-world applications like those in the video.
I also remember RDRAM being far more expensive than either SDR SDRAM and DDR SDRAM at that time, but perhaps it was just in the area I lived since I primary shopped locally - I wasn't aware of the online prices of either at that time period, so you could be right about RDRAM being cheaper online in the early 2000s.

By 2005, RDRAM was still very expensive, both online and offline from what I had seen, and DDR (and even DDR2 at that point) were both far more cost effective.
Not to mention that RDRAM was *required* to be paired, needed dummy PCB (CRIMMs) sticks in the unoccupied slots, ran very hot, was extremely power hungry (especially when compared to SDR SDRAM and even DDR SDRAM later), and was generally unpopular with nearly all who used it at the time from what I saw.

If we're talking about the same 400 FSB Pentium 4, then SDR SDRAM would lose significantly to RDRAM PC800.

Well, the benchmarks shown in the video show the exact opposite of what you are stating, and that was a 1GHz Pentium III with PC133 SDR SDRAM beating the 1.5GHz Pentium 4 with faster PC800 RDRAM in those benchmarks, let alone another Pentium 4 with DDR SDRAM, or if it had even been a thing on standard desktops and laptops, a Pentium III with DDR SDRAM (the original XBox had a Pentium III with DDR SDRAM).

SDR SDRAM:
PC66 - 533MB/s
PC100 - 800MB/s
PC133 - 1066MB/s

RDRAM:
PC600 - 1066MB/s
PC700 - 1420MB/s
PC800 - 1600MB/s
PC1066 - 2133MB/s
PC1200 - 2400MB/s

DDR SDRAM:
PC-1600 - 1600MB/s
PC-2100 - 2133MB/s
PC-2700 - 2666MB/s
PC-3200 - 3200MB/s

Both DDR SDRAM dual-channel and RDRAM dual-channel each eventually capped out at 6400MB/s at the end of each of their production lines, but again, RDRAM was expensive, power hungry, ran hot, and was overall far less efficient on all fronts compared to DDR SDRAM by 2004.

Also, I remember Intel starting to phase out RDRAM around 2001, and by 2003 had completely replaced their systems entirely with the far more efficient DDR SDRAM.
All of the gains that RDRAM had over SDR SDRAM and very early DDR SDRAM were basically gone by 2004, making it completely obsolete and far too expensive to continue to support, and I did see this with online and offline prices by 2004 and 2005 across the board.

Granted, the Nintendo 64 and Playstation 2 each used RDRAM, which was appropriate at the time, as well.

 

[dexvx]

From the looks of it, a 1GHz Pentium III with SDR SDRAM PC133 utterly destroys a first-generation 1.5GHz Pentium 4 with RDRAM PC800:

Yes, on cherry picked older games that weren't SSE2 optimized. When Pentium 4 launched, the 1.5 GHz was barely faster than the P3-1GHz. But as time passed, even the Willamette P4 could beat not only the P3/Coppermine, but almost match the P3/Tualatin and Athlon-C clock for clock.

http://www.tomshardware.com/reviews/benchmark-marathon,590-23.html

This is a 2003 CPU Roundup, so around 2 years after the P4-1.5GHz debuted.

That P4-1.5/Willamette/PC800 RDRAM beats:
Quake III Arena -
P4-2.0/Northwood Celeron/DDR
Athlon-C 1400/Thunderbird/DDR
P3-1.4/Tualatin/SDR

UT2003
P4-2.0/Northwood Celeron/DDR
Athlon-C 1200/Thunderbird/DDR
P3-1.26/Tualatin/SDR

Results are quite similar for other software tested, with some anomalies as expected.

By 2005, RDRAM was still very expensive, both online and offline from what I had seen, and DDR (and even DDR2 at that point) were both far more cost effective.
Not to mention that RDRAM was *required* to be paired, needed dummy PCB sticks in the unoccupied slots, ran very hot, was extremely power hungry (especially when compared to SDR SDRAM and even DDR SDRAM later), and was generally unpopular with nearly all who used it at the time from what I saw.

By 2005, RDRAM was dead. It was only active during the late P3 to early P4 days. Around late 1999 to mid 2002.

Common misconception about RDRAM and heat/power consumption. RDRAM is different than SDRAM due to its serial design (hence the need for CRIMM's). Only 1 chip is actually being utilized per RW cycle, so only one chip gets 'hot' (thus the need for heat-spreaders). But overall, the power consumption for PC800-45 is *lower* than PC100 SDRAM. The numbers given was about 900 mW for PC100 SDRAM, but PC800-45 RDRAM is anywhere between 250 mW (idle) to 1100 mW (active).

https://www.anandtech.com/show/545/8

Temperature wise, we're looking about 3C hotter than a PC133 SDRAM stick, so its nothing to huge.

https://www.anandtech.com/show/551/2

That is exactly how I remember history, and granted that software didn't really take advantage of SSE2 in the Pentium 4 at the time, but judging by those benchmarks in the video, the additional memory bandwidth given by RDRAM had next to no effect on any of those applications, and it seemed like the CPU was more of the bottleneck than the memory was (all other components were the same in both systems, so it was tested 1:1), especially when an older Pentium III completely destroys the first-generation Pentium 4; SPEC was also, from what I remember, a synthetic benchmark, when compared to real-world applications like those in the video.
I also remember RDRAM being far more expensive than either SDR SDRAM and DDR SDRAM at that time, but perhaps it was just in the area I lived since I primary shopped locally - I wasn't aware of the online prices of either at that time period, so you could be right about RDRAM being cheaper online in the early 2000s.

Also, I remember Intel starting to phase out RDRAM around 2001, and by 2003 had completely replaced their systems entirely with the far more efficient DDR SDRAM.
All of the gains that RDRAM had over SDR SDRAM and very early DDR SDRAM were basically gone by 2004, making it completely obsolete and far too expensive to continue to support, and I did see this with online and offline prices by 2004 and 2005 across the board.

RDRAM from about late 2000 to early 2002 was about the same price of DDR SDRAM. RDRAM got phased out because the Memory Cartel decided they didn't want RDRAM, so they pushed DDR SDRAM. Again, there was a lawsuit where Rambus won millions of $, as they carefully documented how the Memory Cartel screwed them over. But by that time, it was all moot, because DDR SDRAM just dominated everywhere. Not saying RDRAM was a perfect technology, but the DRAM Cartel actively going for RDRAM's downfall was the critical factor.

One notorious example is how the DRAM Cartel claimed PC1066 shortage was due to bad yields. Yet, go check RDRAM overclocking around the same time period, and literally *every* single stick of Samsung/Kingston PC800-40 was doing PC1066 without breaking a sweat.

P4 SDR vs P4 RDRAM:
https://techreport.com/review/2843/the-pentium-4-gets-sdram-two-new-chipsets/6

It's a massive performance penalty going PC133 SDR. You would literally have more performance if you opted for a cheaper P4 (e.g. 1.5GHz/PC800 RDRAM vs 1.8 GHz/PC133 SDRAM) with RDRAM.

P4 DDR (PC2700) vs P4 RDRAM (PC800/PC1066):
https://techreport.com/review/4186/intel-845pe-and-845ge-chipsets/7

PC1066 is in the definitive lead. Sometimes DDR-2700 trades blows with PC800.

P4 Dual DDR (PC3200) vs P4 RDRAM (PC1066):
https://techreport.com/review/4998/intel-875p-chipset-and-p4-3ghz-processor/8

This is where DDR SDRAM takes the absolute lead. You literally had P4 800FSB (875P) vs P4 533FSB (850E) and dual PC3200 (6.4 GB/s bandwidth) vs dual PC1066 (4.2 GB/s bandwidth). And even then, the 875P setup doesn't win by huge margins.

 

[Red Falcon]

snip

That was actually a lot of great info, and I was really unaware of the power consumption, not to mention the small heat difference - I really thought those were a lot bigger.
Thanks for the links and info, those were a great read and history lesson!

 

[WhoBeDaPlaya]

I just ran across this guide to rework a dead Dallas DS1287/1387. What are your thoughts on it?
http://www.mcamafia.de/mcapage0/dsrework.htm

Don't know how I missed this notification :S
Looks about right to me, though I'm not as familiar with that part as I joined a few years after Dallas Semi had already merged with Maxim.

BTW : We might possibly become NEC / Renensas
https://www.cnbc.com/2018/01/29/ren...cquire-maxim-integrated-in-20b-chip-deal.html

 

[Red Falcon]

Magnavox Computer Monitor 80 (1989)

Well now, this Magnavox Computer Monitor 80 amber monitor has been featured before, but I figured I would share some info about connecting it to a modern day computer via HDMI or DisplayPort.
This Composite-based amber monitor is currently connected to my Intel Xeon x86_64 system from 2017 (cylinder computer with the Elder-sign on it) above.

If the monitor uses composite:
Active HDMI/DisplayPort-to-VGA adapter (powered by USB) -> active VGA-to-Composite adapter (powered by USB) -> Any Composite-based CRT SDTV since the 1950s-present.

If the monitor uses coax or twin-lead RF:
Active HDMI/DisplayPort-to-VGA adapter (powered by USB) -> active VGA-to-Composite adapter (powered by USB) -> Composite-to-RF adapter (can be 75Ω coax or 300Ω twin-lead) -> Any RF-based CRT SDTV since the 1950s-present.

Most of these adapters will take resolutions up to around 1024x768 (4:3 aspect ratio), but some will take higher resolutions if you go for higher-end models.
These adapters can be purchased at Amazon or eBay for relatively cheap, and are all plug-and-play without requiring any software or drivers, outside of the modern system's GPU drivers.

Since the monitors are basically just SDTVs and don't have a true 1:1 pixel resolution that the computer can detect, it may be a good idea to set the resolution on the computer to either 800x600 or 1024x768 first, before connecting the adapters and composite/RF monitor.
I figured I would share a bit of information in relation to the post on our news page today about "Guy Hooks Modern Consoles and PC to 1970s TV", and while this monitor may not be quite that old, this method should allow anyone to connect their modern systems to any SDTV from the 1950s to present!

Stay retro!

 

[Red Falcon]

Game Boy Booster

Now here is quite the interesting find... behold, the Game Boy Booster from Saitek!
Sold in the early 1990s as the "ultimate" upgrade for the original Game Boy DMG-01 brick model, this unique shell has quite the feature set:

- Stereo speakers (utilizes 3.5mm headphone jack)
- Magnifying sens with dual projection lights
- Storage compartment for headphones
- Storage compartment for two additional game cartridges
- Protective cover for link-cable port
- Powered by four C batteries (powers Game Boy, speakers, and projection lights)

This was featured in episode 147 of James Rolfe's Angry Video Game Nerd series episode, "Game Boy Accessories", specifically at the 10 minute & 15 second mark.
The game featured in the screenshots is X, sold only in Japan in 1992, and is the first wireframe 3D game released for a portable system in Japan.

Not too shabby for the "ultimate" portable 8-bit console... stay retro!

 

[GiGaBiTe]

If only the m68k architecture had kept going, maybe the 68080 (non-existent in m68k) could have competed with Pentium II era processors.

The final 680x0 processor in the line, the 68060 had trouble competing with the original Pentium, the P6 architecture was far out of reach.

Motorola had made the same mistake with the 68060 as they did as far back as the 68881/2, by not having a pipelined FPU. While the 68060 was great with integer code and mixed 16/32 bit instruction sets, it was at a severe disadvantage with upcoming software that increasingly relied on floating point math. The FPU in an 060 was really no faster than the FPU in an 040 at the same clock speed. Cyrix made the same mistake with their 6x86, they had a great integer core, but paired it with a relatively unchanged 486 era FPU.

Another issue is the extremely short pipeline (4 stages in the 68060.) It's why Motorola had such difficulty getting higher clocked 68040s out the door and why they topped out at 40 MHz. The 68060 could scale a bit higher to 75 MHz because they die shrunk it, but it had severe problems on higher clocked models and Motorola had to release parts with the MMU and FPU disabled.

The final issue was forwards and backwards compatibility. Without going into rambling detail, no 680x0 CPU was 100% forwards or backwards compatible in both software and hardware (with exception to the 68020 and 68030.) If you wanted to move between various 68k parts, code had to be rewritten or recompiled. This is in contrast to x86 where a well written 8086 application could run on a Pentium, assuming you could get it slow enough

It would be interesting to see if a 68060 would work in a 68040 Quadra 950. The 68060 is not quite pin compatible, but there are interposer adapter boards available from the Amiga community. MacOS is out of the question, but Linux/BSD should support it.

 

[Red Falcon]

snip

Very well said, I don't think I could have done better myself!
From what I've seen, the FPU in the 68060 was a good bit faster than the 68040, but that could also be due to the higher clock speeds of 50-75MHz vs 25-40MHz on the 68040.

Yeah, that compatibility issue between generations was certainly annoying, and could really be seen with the Sega Genesis or Mega Drive when trying to run the 68000 vs the 68010, and especially the Sharp X68000 with the 68000 vs a 68030 (depending on clock speed - 20MHz doesn't have frame-time issues that 25MHz does), 68040, and 68060 with many of the games and software suites.
Yep, I have x86 code myself which will run on modern 2017 x86_64 CPUs that will also run on my 8086 system from 1984, which is definitely a plus to that architecture (as well as a boon as some would say).

It's funny you mention Cyrix with their good integer unit, yet their faulty FPU is what really held them back, as you stated.
You might really enjoy this video on that very subject, and it is quite a fun and informative video on why Cyrix failed against Intel and AMD for those very reasons you speak of:



As for the 68060 working in a Quadra 950, I have never seen any interposer adapter boards for it, and until one is developed, I don't think it would work at this moment in time, but hopefully that will change in the future.
Supposedly there is a way to boot straight into Linux (Debian 4.0 - went EoL years ago) or *possibly* NetBSD directly from a boot diskette on m68k-based Apple Macintosh systems, but the last updates I saw on its development were from 2008, and it is highly depreciated now with modern versions of Debian and NetBSD.

The way I boot into NetBSD on those systems is to boot into Mac OS first, which would require the 68040 (or other compatible m68k CPU), and then use a bootloader to boot into NetBSD from there.
I will give props to the Amiga community, and the same to the X68000 community, as both continue to develop and support their hardware to the fullest extent with modern parts, which is quite exciting to say the least.

Maybe someday the m68k-based Apple Macintosh community will be on par with those communities, but until then, we can keep trying.

 

[GiGaBiTe]

and could really be seen with the Sega Genesis or Mega Drive when trying to run the 68000 vs the 68010

The primary cause of the incompatibility comes from the MOVE from SR instruction being made privileged and requires supervisor mode to execute. It doesn't make much sense to use a 68010 in a Genesis because the functions it adds are useless to the console and is really at most 10% faster at the same clock speed. It's easier to swap in a faster 68000 with a oscillator or overclock the existing 68000 (assuming you don't have a garbage 3rd party version from like Signetics.)

You might really enjoy this video on that very subject, and it is quite a fun and informative video on why Cyrix failed against Intel and AMD for those very reasons you speak of:

More than the slow FPU performance killed off Cyrix. They had a bunch of CPUs that used weird non-standard bus speeds (50, 55, 75 and 83 MHz) which caused all matters of problems. The higher non-standard speeds like 75 and 83 MHz caused stability and reliability problems with motherboards of the time because the various clocks on the board tended to be cascaded off the CPU clock. Most boards expected a 60/66 MHz front side bus and ran main memory at 1:1, AGP at 1:1 (later 2:3 for 100 MHz FSB systems), PCI at 2:1 and ISA PCIClk/3 or 4. When a 75 or 83 MHz Cyrix was used, these went alarmingly out of spec if the system even was able to boot. later boards added more multipliers like 2:3 and 5:6 to try and cope, but this still ran things either too slow or overclocked.

I remember dreading working on Cyrix systems back in the day. My high school computer lab had one Cyrix MII-333 and it was always freezing, BSODing or randomly rebooting. It also ran really hot compared to the other systems which were a mix of Intel and AMD parts.

As for the 68060 working in a Quadra 950, I have never seen any interposer adapter boards for it, and until one is developed, I don't think it would work at this moment in time, but hopefully that will change in the future.

The Quadra 950 has a socketed 68040, the interposer boards that were designed for the Amiga will also work in any mac that has a socketed 68LC040/68040. The problem with it actually working is MacOS doesn't know what a 68060 is. There was some work in the Amiga community to get Mac OS 8.0 running on an Amiga with a 68060, but an unspecified bug was causing it to fail.

 

[KD5ZXG]

I was into z80 far a long time, then switched to 6809.
Mostly missed the boat on anything else 68ish...

Hitachi made an upgraded 6309 with more registers
and math, but I didn't find out till only recently. A little
late now.

https://en.wikipedia.org/wiki/Hitachi_6309

 

[Red Falcon]

I was into z80 far a long time, then switched to 6809.
Mostly missed the boat on anything else 68ish...

Hitachi made an upgraded 6309 with more registers
and math, but I didn't find out till only recently. A little
late now.

https://en.wikipedia.org/wiki/Hitachi_6309

Oh very nice, I've done a little bit with the ZiLOG Z80 CPU, and have liked what I saw on it, but have never used a 6809 or 6309 from Motorola before.
What systems used the 6809 or 6309 did you use them in?

The Z80 I've used is in the Osborne-1 portable computer, featured a bit earlier in this thread, not to mention the Game Boys, heh.
I remember seeing this ad for the Z80 a few years ago as well, which is pretty interesting:

 

[KD5ZXG]

Trash 80 (z80) and the Coco (6809). CP/M drew me back to z80, with the Xerox820-II.
My father was a friend of Jim Ferguson, who designed the Big Board, that became the Xerox.
Mine was cobbled together of spare parts from the Xerox surplus store, nothing recognizable.
But my first computer with any kind of a disk.

Had complete Xerox software though. They were erasing by hand with a magnet, but didn't
wipe the whole floppy. Rarely the same sectors. So you buy a few surplus ring binders with
original disks and manual. Reconstructing the missing pieces was pretty easy. Just copy the
floppy and it would stop whenever it couldn't be read. Just throw in another partially erased
original, till you get lucky and all the sectors are recovered on one target. And finally back to
proper original media....

Then I wrote my own bootloader and boot track with 1:1 interleave. That was a big speedup.
Later, something akin to a cache, that would pack four "normal" 128byte sectors into a 512.
Reclaiming space wasted by inter-sector gaps, like any other kid with nothing better to do.
Maybe should have tried to sell that, but by then CP/M was on the downside of obsolete.

 

[Red Falcon]

Trash 80 (z80) and the Coco (6809). CP/M drew me back to z80, with the Xerox820-II.
My father was a friend of Jim Ferguson, who designed the Big Board, that became the Xerox.
Mine was cobbled together of spare parts from the Xerox surplus store, nothing recognizable.
But my first computer with any kind of a disk.

How exciting!
You and your family have some extensive history with computer technology, and it is really neat to hear that you actually worked with CP/M back in its heyday as well; Gary Kildall would be very proud.

I've never had the privilege of working with the Tandy TRS-80, but I have seen them in videos online and they do look quite interesting to work with... if only I had been born earlier in life.
Were you actually able to work with the datasette (cassette) drive on it to load or store programs?

Had complete Xerox software though. They were erasing by hand with a magnet, but didn't
wipe the whole floppy. Rarely the same sectors. So you buy a few surplus ring binders with
original disks and manual. Reconstructing the missing pieces was pretty easy. Just copy the
floppy and it would stop whenever it couldn't be read. Just throw in another partially erased
original, till you get lucky and all the sectors are recovered on one target. And finally back to
proper original media....

You know, I've never thought about doing that to recover data, and that is an incredibly good idea.
Restoring the data off of diskette after diskette until you get a full copy, really, I seriously never though of that!

Then I wrote my own bootloader and boot track with 1:1 interleave. That was a big speedup.
Later, something akin to a cache, that would pack four "normal" 128byte sectors into a 512.
Reclaiming space wasted by inter-sector gaps, like any other kid with nothing better to do.
Maybe should have tried to sell that, but by then CP/M was on the downside of obsolete.

I've never written a bootloader as well, I'm actually quite envious of everything you've been able to do, as a lot of it I'm still discovering.
I've actually had issues with flash-based media on older buses (SASI and SCSI-1) due to 1:2 interleaving on them, and I was just thinking of this earlier this afternoon; very cool that you would bring this up.

Heck yeah, and those systems are incredibly valuable today, monetarily of course, but more importantly for their historical and technological significance.
If only I had had an older kid or adult who knew their stuff to help me learn about this stuff in the early 1990s onward, I would have had such a better appreciation for older, and newer, systems back then.

Pretty cool that you got to do all of this back when it was mainstream and resources for all of it were readily available, I'm quite jealous!
The computer you built back then almost sounds like this one: http://koo.corpus.cam.ac.uk/chaos/

 

[XoR_]

The final 680x0 processor in the line, the 68060 had trouble competing with the original Pentium, the P6 architecture was far out of reach.

Motorola had made the same mistake with the 68060 as they did as far back as the 68881/2, by not having a pipelined FPU. While the 68060 was great with integer code and mixed 16/32 bit instruction sets, it was at a severe disadvantage with upcoming software that increasingly relied on floating point math. The FPU in an 060 was really no faster than the FPU in an 040 at the same clock speed. Cyrix made the same mistake with their 6x86, they had a great integer core, but paired it with a relatively unchanged 486 era FPU.

Another issue is the extremely short pipeline (4 stages in the 68060.) It's why Motorola had such difficulty getting higher clocked 68040s out the door and why they topped out at 40 MHz. The 68060 could scale a bit higher to 75 MHz because they die shrunk it, but it had severe problems on higher clocked models and Motorola had to release parts with the MMU and FPU disabled.

The final issue was forwards and backwards compatibility. Without going into rambling detail, no 680x0 CPU was 100% forwards or backwards compatible in both software and hardware (with exception to the 68020 and 68030.) If you wanted to move between various 68k parts, code had to be rewritten or recompiled. This is in contrast to x86 where a well written 8086 application could run on a Pentium, assuming you could get it slow enough

68060 seems to have similar construction and performance to AMD K5 - ALU better than Penitium, FPU worse.
If Apple did not switch to PowerPC we would certainly see 68K into >1GHz which would be great because this is great architecture.
68K code density is great, one of the best in fact.
By comparison PowerPC was only better at clock speeds and FPU. By the time Apple switched to PowerPC Motorolla would certainly improve their FPU implementation.

But at the time there was this whole 'RISC confusion'. Many people were hyped about superiority of RISCs... until modern techniques of dynamic code optimization made all of them completely insignificant. Later RISC processors had to start using them in order to compete with X86 and all this RISC thingy did for them is worse code density impacting performance. 68K ISA is in fact best target for all these optimizations than any RISC ISA.

That said 68K died not because of 68060 performance but little to none demand for them after Apple (main user of this arch) decided to switch to PowerPC. 68060 having better FPU would make no difference at all. Even if it was the fastest FPU-wise processors at the time it would still fall into obscurity and only Amiga users would notice.

Compatibility issues were caused by omission of proper microcode circuitry which made emulation of all legacy instructions easy on X86. Modern X86 (basically everything since Pentium Pro but also Post-RISC processors from AMD and Cyrix) emulate all legacy stuff and it shows in benchmarks. Some people speculate that should this emulation was hacked it would even allow to emulate other processors. Intel use this to make changes to processors with BIOS updates. This 'emulation' was never an issue on X86 because sheer processing power made each new processor still faster than older one. For example Pentium 4 were actually terrible at 16bit/DOS stuff and no one really cared.

PowerPC was driven by idea of symmetry and simplification. It only ended in them being slower than cheaper X86... from both Intel and AMD. G5 was worse processor per clock than AMD K6. Of course these PowerPC had much lower transistor count than comparable X86 processors but per transistor utilization was even higher resulting in K6 having much better thermal parameters. In this light PowerPC was total failure. And AMD never had much money to begin with. Certainly if RISC was even partly as good as hype suggested they would have no issues to beat AMD. They did not.

It is nice I can put Sound Blaster PCI 128 and some PCI video card (or use those which do not do up-scaling on VGA - my last Radeon did not, I know NV cards do) to my main modern PC and have pretty much the same DOS experience on my PC that I did back in the time

 

[Red Falcon]

68060 seems to have similar construction and performance to AMD K5 - ALU better than Penitium, FPU worse.
If Apple did not switch to PowerPC we would certainly see 68K into >1GHz which would be great because this is great architecture.
68K code density is great, one of the best in fact.
By comparison PowerPC was only better at clock speeds and FPU. By the time Apple switched to PowerPC Motorolla would certainly improve their FPU implementation.

Well, while there wasn't a huge demand for m68k by the mid-1990s, if I remember correctly, Motorola was having problems with the heat (as mentioned above) at clock speeds higher that 50-75MHz with the 68060.
PowerPC was quite a bit faster than m68k in Integer and Floating Point performance with even the oldest 601, let alone the 603 and great models - clock-for-clock, PowerPC had a higher IPC than even the m68k 68060, and it could be clocked MUCH higher.

Granted that my Quadra 950 has a 68040 33MHz CPU in it, but the PowerPC 601 100MHz CPU is far more than four times faster in it at nearly everything - granted that it isn't a 68060, but still, PowerPC had more going for it than simply a higher clock speed.
Apple gave up on m68k since there wasn't much more they could get out of it, and Motorola wasn't playing ball with them at the time since they themselves were ready to move on from m68k - it was more than just lack of sales, as even Motorola saw the writing on the wall for the technical limitations of m68k.

Also, why would Apple have continued with the 68060 when the PowerPC 601 CPU was faster (again, more than just the clock speed), and the 68060 was the last of the m68k CPUs?
It was a complete dead end for that ISA, and not only did Motorola know it, but Apple knew it as well, and the 68060 would have granted much less performance per clock than PowerPC did, even with the PowerPC CPU emulating legacy m68k-based code - my Quadra 950 equiped with both is living proof of this for a 1:1 apples-to-apples comparison.

But at the time there was this whole 'RISC confusion'. Many people were hyped about superiority of RISCs... until modern techniques of dynamic code optimization made all of them completely insignificant. Later RISC processors had to start using them in order to compete with X86 and all this RISC thingy did for them is worse code density impacting performance. 68K ISA is in fact best target for all these optimizations than any RISC ISA.

That said 68K died not because of 68060 performance but little to none demand for them after Apple (main user of this arch) decided to switch to PowerPC. 68060 having better FPU would make no difference at all. Even if it was the fastest FPU-wise processors at the time it would still fall into obscurity and only Amiga users would notice.

It wasn't "hype", RISC CPUs of the 1990s were legitimately faster than x86 CPUs by quite a wide margin, it's just that x86 CPUs were much more cost-effective in the consumer and lower-end server markets for their respective markets' performance categories.
It wasn't until the Pentium Pro circa 1996 that Intel and x86 even began to lightly compete with RISC CPUs of that era - RISC CPUs were far above x86 for the most part and never bothered to compete directly with them until the mid to late 1990s at the earliest.

I'm really trying to be respectful of your post and your thoughts, it's just that what you are saying here is kind of a distorted history... if you can provide links to anything that backs up what you are saying and it proves me wrong, I will be happy to admit it as well, but please make sure your sources are accurate and credible.
Code density never impacted performance of RISC-based platforms as well - if one was going to do serious computation, they would go with SPARC, PowerPC, POWER, PA-RISC, Alpha, etc. - x86 was laughed at for serious workloads which weren't budget-constrained, especially at that time period in the 1990s with rendering and large databases.

Compatibility issues were caused by omission of proper microcode circuitry which made emulation of all legacy instructions easy on X86. Modern X86 (basically everything since Pentium Pro but also Post-RISC processors from AMD and Cyrix) emulate all legacy stuff and it shows in benchmarks. Some people speculate that should this emulation was hacked it would even allow to emulate other processors. Intel use this to make changes to processors with BIOS updates. This 'emulation' was never an issue on X86 because sheer processing power made each new processor still faster than older one. For example Pentium 4 were actually terrible at 16bit/DOS stuff and no one really cared.

To my knowedge, there is no "emulation" of legacy (i586 and earlier) code in the Pentium Pro (i686 and newer), as the original x86 instructions are still there - I can run 16-bit MS-DOS applications and code on my i7 6700K (in a 32-bit environment) just as easily as I can on my 8088 and NEC V30 (8086) from the early 1980s.
Again, what "emulation" are you talking about, especially emulation that could "allow to emulate other processors"? - Do you mean other CPU ISAs?

x86 can already emulate other processors, and has been able to since the 1990s (ZSNES, MAME, QEMU, etc.) - heck, even my 16-bit NEC V30 can natively run 8-bit Intel 8080 code on it.
To my knowledge, it's actually these legacy instruction sets that are actually holding x86 and x86_64 back more than making them "faster than the older ones"...

Also, the Netburst architecture sucked at everything except email mail merges - was there at that time and lived Netburst fully for the time that it existed, in and out of enterprise - those were the AMD golden days for a reason.
For legacy 16-bit applications, though, I don't think anyone cared since they all took relatively small amounts of processing and computational power to run or complete tasks, and was a non-issue by the late 1990s at the absolute latest, let alone into the 2000s with Netbust and beyond.

PowerPC was driven by idea of symmetry and simplification. It only ended in them being slower than cheaper X86... from both Intel and AMD. G5 was worse processor per clock than AMD K6. Of course these PowerPC had much lower transistor count than comparable X86 processors but per transistor utilization was even higher resulting in K6 having much better thermal parameters. In this light PowerPC was total failure. And AMD never had much money to begin with. Certainly if RISC was even partly as good as hype suggested they would have no issues to beat AMD. They did not.

No, PowerPC was much faster than x86 and early x86_64 in just about everything until the arrival of the Core architecture was released in 2006 - Apple again saw the writing on the wall for PowerPC with the thermal designs, extremely high heat output, power requirements, and limited development - they then quickly moved to x86_64 CPUs in 2006 away from PowerPC, the exact same way they did in the 1990s with m68k and PowerPC.
The IBM 970MP dual-core CPUs in the Macintosh G5 Quad, clock-for-clock curb-stomped x86 and x86_64 CPUs of that time period in just about everything, and really was only held back by OS X in server/web/database applications, but then again OS X also held back x86_64 CPUs in those same areas, which is why Apple XServe products ceased production during 2009, but that's a story for a different day.

I will agree with you that K6 did have better thermal designs, but those CPUs, as good as they were, could not compete with the 970MP CPUs - again, I lived this fully during that time period in and out of enterprise for both productivity and gaming.
PowerPC was not a "total failure", the market for the 970MP CPUs was simply the high-end Apple G5 Quad, which did not sell in droves, and the sales were too low to allow IBM to continue the extremely costly development of that specific iteration of PowerPC, which then eventually segwayed into the PowerPC-based Cell CPUs used in high-end enterprise platforms and the Playstation 3 console.

AMD was also being cornered illegally by Intel at the time, and their purchase of ATI in 2006 set them back even further monetarily - but again, that is a story for another day.
By that point, the whole "RISC vs CISC" design strategy was pretty much at an end and was considered trivial - that battle was far more prevalent in the 1980s and 1990s, but by 2006 x86_64 had become a serious competitor in nearly all markets.

Modern x86_64 CPUs are basically a CISC-wrapped-RISC design anyways, but are still *technically* considered CISC.
However, modern RISC CPUs such as ARMv7 and v8 have more instruction sets in them than CISC designs from yesteryear, so the whole RISC/CISC designation is nearly trivial at this point, though processors are still designated as CISC or RISC (and don't forget about ZISC!).

It is nice I can put Sound Blaster PCI 128 and some PCI video card (or use those which do not do up-scaling on VGA - my last Radeon did not, I know NV cards do) to my main modern PC and have pretty much the same DOS experience on my PC that I did back in the time

I agree with you on that!


Again, after all of that, I want to reiterate, if you can provide links to anything that backs up what you are saying and it proves me wrong, I will be happy to admit it as well, but please make sure your sources are accurate and credible.
Hope to discuss this with you further, and hopefully both of us will be enlightened and learn something new!

 

[NamelessPFG]

Well, I got another addition to the retro fleet after waiting three weeks on international shipping, but here it is:

No accessories beyond the trapdoor RAM/RTC expansion there, and it's certainly no A1200 or A4000, but it's a start! Still cost less than most eBay listings, even when about 40% of what I paid to get this one was just shipping. It helps that the case was all beaten up and I'm not that much of a stickler for cosmetics when I know there's reproduction A500 cases about to be produced.

While I still need a mouse (PS/2 adapters exist, thankfully, and there's even fancy auto-switching ones that let you leave a PS/2 compatible USB mouse like a WMO and a DE-9 joystick/gamepad plugged in simultaneously), an RGB cable (I could try whipping up a DE-15 pin adapter to feed my PEXHDCAP with, since it handles 15.7 KHz RGB without complaint), and Workbench disks (which I do not have the floppy controller hardware to write using my PC or vintage Mac floppy drives), I did manage to whip up a crude ATX adapter using leftover cable bits from a prior MDD G4 ATX adapter rewiring project.

It fires right up. Red power LED, Kickstart 1.3. At least I know it'll power on and output video just fine, though further testing requires some actual software.

I also opened it up, cleaned a distressing amount of dust from the inside (it had clearly never been opened before because the warranty seals were still intact), and there's a revision 5, type 2 motherboard in there. Well, that's gonna make certain things a bit difficult compared to a revision 6 or later, but I'll take what I can get.

 

[Red Falcon]

snip

Oh that is absolutely glorious!
Wow, that's in great condition, and very cool that everything is still working.

Very nice job getting everything going on it, I can't wait to see that bad boy in action.
Nice expansion board as well, that will definitely help with a few larger software titles - how much more memory does that one allow for?

What a find, I'm so glad it is in good hand with someone who can appreciate such a system as well.
Hope to see more from you soon, thanks very much for sharing your find and info!

 

[NamelessPFG]

Oh that is absolutely glorious!
Wow, that's in great condition, and very cool that everything is still working.

Very nice job getting everything going on it, I can't wait to see that bad boy in action.
Nice expansion board as well, that will definitely help with a few larger software titles - how much more memory does that one allow for?

What a find, I'm so glad it is in good hand with someone who can appreciate such a system as well.
Hope to see more from you soon, thanks very much for sharing your find and info!

Thanks! Alas, it's not in as great condition as you'd think; the top half of the case that you see there only has minor chips around a few corners, but the bottom half is significantly cracked around the mono video port on the back, and there's an entire chunk missing on the left side over the sidecar expansion port!

Fortunately, there's a crowdfunding campaign for new A500 cases about to release soon over at https://www.a1200.net/ , which already has replacement A1200 cases and keycaps under their belt from prior campaigns. I'm definitely getting in on that as part of the restoration process I have planned for it.

EDIT: It's now live!


The expansion card seems to be a clone of the A501 512k RAM/RTC trapdoor expansion card, which is the only sort of expansion I know of that goes in the trapdoor slot to begin with. (Others go through the "sidecar" Zorro II male edge connector on the left side.) The downside is that all of these expansions provide Slow RAM - that is, unavailable to the chipset like Fast RAM, but just as slow as Chip RAM (which the CPU can't access at the same time as the chipset). Fast RAM expansions do exist for the A500, but they either go through the aforementioned side slot, or more recently, through a CPU socket passthrough adapter that doubles as an IDE/CF interface.

It's possible to modify the motherboard to allow trapdoor memory to act as Chip RAM for a combined 1 MB (required for many WHDLoad games), but doing so requires me to swap out the Fat Agnus PLCC for a later revision that can address more Chip RAM to begin with.

Naturally, the problem is just finding a compatible Agnus replacement! The usual candidate is the 8372A, but I just can't find them for sale anywhere because Amiga hardware is rapidly going from "expensive" to "unobtanium". The 8375 has a bunch of variants that may or may not be compatible, with different pinouts; some address 1 MB total, others top out at 2 MB. There's also the MegACHIP 500/2000, which offers the full 2 MB and an ECS Agnus to match, but that seems to be in the "unobtanium" category right now.

Note that while a Vampire 500 V2 board provides 128 MB (which is ludicrous for a 68k Amiga, and also what a typical Macintosh II will max out at), that's most likely Fast RAM and won't help one bit with the WHDLoad Chip RAM requirement. Games and other old, non-RTG-compatible software still have to go through the existing motherboard chipset instead of the Vampire's SAGA core, though they're working on full-blown AGA compatibility for their next FPGA core release - this being what I'm banking on to close the gap with the A1200 and A4000. (It's also likely resulted in a spike in demand for the A600 and A500, where most of the Amiga community was rather A1200-centric until recently.)

On the other hand, most games were probably designed around the base A500 with A501 configuration anyway - something that I've already got here. The only drawback is having to swap floppies all the time, since many games tend to take up at least two disks.

 

[XoR_]

Pentium Pro have few fast decoders for limited set of instructions which is something that is very RISC-like. Those instructions are obvious optimization target for compilers. Rest of instructions including legacy modes are decoded by decoder which is running microcode to decode them to format usable by internal units. Simulating other ISAs is just a wild speculation but in theory if fast decoders could be completely disabled it should be possible. Of course no one cracked their protection system and probably only few in the world have knowledge how to program this kind of hardware so possibility of it ever happening are close to zero even if it could be done.

Some PowerPC benchmarks vs PC
https://www.anandtech.com/show/1702/5
http://www.eofw.org/bench/

https://groups.google.com/forum/#!topic/comp.sys.intel/aEpEluPEVKw

I remember reading discussions where people generally complained that eg. in Linux software didn't run any faster on eg. G4 than on comparable x86 from the era (Pentium 3, K7)
This was maybe due to better compiler implementation for x86 and not making any used of AltiVec but this was what they were getting and thus what really mattered.

So I would not say PowerPC were 'much faster'. General consensus always was that AltiVec is faster than any x86 SIMD implementation available at the time but in normal code there was no wow-factor.

On 68060 vs early PowerPC front only Amiga community cared to do any testing and based on numbers I saw 68060 are doing nicely in ALU workloads and very close to early PowerPCs.
I do not have any references but should I stumble on any I will post them.
In this case obviously implementation of system is not optimal because PowerPC lack additional external cache. But 68060 does not have it either and both processors would be faster with it.
Power consumption is something which would be fixed if Motorola continued its development and we would surely see clock doubled versions with larger caches and better per clock performance.
ISA of 68K was very good, great for assembler programmers, a lot of registers, many useful instructions. If PC used 68K then modern CPU techniques would also make use of advantages 68K ISA have. Obviously RISC being designed in ways to make some optimization methods easier made it easier to implement them and get good performance with less effort... but when everything is included and tons of research and development money is thrown at CPU design some other things start to be important eg. how easy it is to write good compiler and good assembler code.

BTW. I just bought beefy PowerMac G5 Dual 2.7GHz 2.5GB RAM 160GB disk and Radeon 9650, all for slightly below 100$ with shipment - for that money I couldn't even get Amiga 600 these days
Always wanted to own one of these beasts but waited for good 2.7GHz offer to pop out.
Will put spare SSD, more RAM and run some benchmarks to see how slow fast it is

 

[Red Falcon]

snip

Wow, you certainly know a lot about Amiga systems, and I've heard of about 1/3 of what you talked about, but the rest was news to me.
You might just be the unofficial Amiga expert on here, especially in this thread!

Oh definitely, those prices... that's exactly how it is any more with the PC-98 and even more so with the X68000 systems.
You're right, 128MB for any m68k-based systems is absolutely crazy, but is so much fun to work with, especially when the only limitation to the system is the CPU/GPU with that much RAM at its disposal.

I've heard that they were going to release that after-market shell for those systems, but it was only a rumor years ago, so it's very good news that it is going to be officially released.
I've often thought about diving into the Amiga area of computing, but the high cost of starting out fresh is going to require saving up, not to mention how expensive all things "retro" are becoming.

Hang onto that system as long as you can - ten years from now, that might be worth as much as an A4000 is now.
That's pretty cool about Chip RAM and Fast RAM on those things, I never new about any of that, and especially with the different revisions.

Thanks for all the great info, and if you make any more developments on it, please share it on here if you can.
You might also checkout buyee.jp for Amiga parts as well - even though it is a Japanese proxy service to Yahoo Auctions (Japan), it can still sometimes provide access to very rare or much needed parts that are not available anywhere else.

Good luck to you, and I can't wait to see more!

 

[Red Falcon]

Pentium Pro have few fast decoders for limited set of instructions which is something that is very RISC-like. Those instructions are obvious optimization target for compilers. Rest of instructions including legacy modes are decoded by decoder which is running microcode to decode them to format usable by internal units. Simulating other ISAs is just a wild speculation but in theory if fast decoders could be completely disabled it should be possible. Of course no one cracked their protection system and probably only few in the world have knowledge how to program this kind of hardware so possibility of it ever happening are close to zero even if it could be done.

Some PowerPC benchmarks vs PC
https://www.anandtech.com/show/1702/5
http://www.eofw.org/bench/
View attachment 63057
https://groups.google.com/forum/#!topic/comp.sys.intel/aEpEluPEVKw

I remember reading discussions where people generally complained that eg. in Linux software didn't run any faster on eg. G4 than on comparable x86 from the era (Pentium 3, K7)
This was maybe due to better compiler implementation for x86 and not making any used of AltiVec but this was what they were getting and thus what really mattered.

So I would not say PowerPC were 'much faster'. General consensus always was that AltiVec is faster than any x86 SIMD implementation available at the time but in normal code there was no wow-factor.

On 68060 vs early PowerPC front only Amiga community cared to do any testing and based on numbers I saw 68060 are doing nicely in ALU workloads and very close to early PowerPCs.
I do not have any references but should I stumble on any I will post them.
In this case obviously implementation of system is not optimal because PowerPC lack additional external cache. But 68060 does not have it either and both processors would be faster with it.
Power consumption is something which would be fixed if Motorola continued its development and we would surely see clock doubled versions with larger caches and better per clock performance.
ISA of 68K was very good, great for assembler programmers, a lot of registers, many useful instructions. If PC used 68K then modern CPU techniques would also make use of advantages 68K ISA have. Obviously RISC being designed in ways to make some optimization methods easier made it easier to implement them and get good performance with less effort... but when everything is included and tons of research and development money is thrown at CPU design some other things start to be important eg. how easy it is to write good compiler and good assembler code.

BTW. I just bought beefy PowerMac G5 Dual 2.7GHz 2.5GB RAM 160GB disk and Radeon 9650, all for slightly below 100$ with shipment - for that money I couldn't even get Amiga 600 these days
Always wanted to own one of these beasts but waited for good 2.7GHz offer to pop out.
Will put spare SSD, more RAM and run some benchmarks to see how slow fast it is
View attachment 63058

Ah, now the legacy modes on i686 x86 CPUs and later models I can understand - that makes much more sense.
Yeah, most CPU architectures can emulate just about anything, but how well they are emulated is up to the architecture itself and the programming involved.

For example, I was able to emulate the Sega Genesis/Mega Drive on a Gamecube with nearly 1:1 emulation - I mean it was basically 99.99% perfect @ 60FPS. (m68k 68000 on PowerPC Gekko)
However, the SNES emulation on the Gamecube was between 60-90% at any given time, and could have simply been poor programming, too low-clock speed for the IPC of the Gekko CPU, or a combination of the two.

The same thing applies to x86 and x86_64 CPUs as well - this is why we really aren't seeing much emulation of PS3 games on the PS4, due to the same reasons of processing power above.
So, those benchmarks you shared are a bit interesting.

The first link shows that clock-for-clock, the IBM 970MP dual-core CPU @ 2.5GHz is clock-for-clock faster than either Netburst or K8 era x86 CPUs, but does indeed lose to the faster clocked variants of those processors due to their high clock speeds.
The second link shows the IBM 970MP is within the lower-tier of Core 2 era processors, which supports the above link as well - also makes sense why Apple went with x86_64 and Core & Core 2 processors in favor of PowerPC in 2006.

The pic of the DOOM 3 benchmarks is pretty neat, but please do remember that DOOM 3 was a single-threaded application, so while the IBM 970MP does lose to the other processors, if an Apple G5 Quad were running an application in full SMP capable of four threads, four cores (two dual-cores) of the IBM 970MP cores would have destroyed any x86 1P dual-core or 2P single-core (and lower-clocked dual-core) systems of that era, especially at the price point the Apple G5 Quad was being sold for.
However, all of that info does show that the IBM 970MP is a bit inferior when it comes to single-threaded applications, which were highly prominent in that era, and it would have mainly beaten x86-based systems at video editing and rendering for the time - which is exactly what I remember from that era as well, circa 2005-2007.

While I was right about the clock-to-clock info, you were definitely right about the x86-based CPUs of that time being faster - thanks for providing evidence to back that up!
As for the last link, I do remember the Motorola PowerPC single-core CPUs in the Apple G3 and G4 systems being similar in performance (still clock-for-clock faster, but clocked low) to Pentium III and early Pentium 4 CPUs of the late 1990s and early 2000s, respectively.

The CPUs in the Apple G4 systems were certainly fast, and actually did hold an edge over Netburst-era x86 CPUs of that time, especially for rendering, but they ran very hot and were quite power hungry, not to mention expensive for the time.
I'm not an Apple fan boy by any stretch, but I do remember the G4 and G5 systems really holding a candle over similarly priced x86 systems of the time for rendering and video editing, and I didn't know too many people who bought such systems for gaming or other applications - those were mainly for our then-powerful Athlon 64 K8 systems of the day!

I agree with you, it would have been very interesting if m68k-based systems could have continued, but Motorola just didn't have enough of the consumer market to really continue this at the time - like I said, even they saw the writing on the wall.
Knowing what I know now about m68k, I do think they could have given Intel a serious run for their money had they continued, but RISC-based systems of the time just outpaced CISC-based systems of the time, especially in clock speed per watt performance, and x86-based systems only really held up due to their vastly lower price tags in the consumer market which allowed them to gain revenue to become more competitive in more of the enterprise markets by the time of the Pentium Pro in the mid-1990s.

Also, that is an amazing price you got that G5 system for - very nice find!
I always liked the higher clock speeds of the single-core IBM 970FX processors in that second G5 model, and especially the fact that they are air-cooled - as opposed to the self-contained liquid cooling blocks on the G5 Quad which all had a high failure-rate (it's what officially killed mine back in 2014).

Since your model of G5 also uses DDR1 RAM (ECC or non-ECC, if I remember right) and 64-bit PCI (PCI-X) expansion cards, they are fairly easy and cheap to upgrade as well, relative to the G5 Quad with PCI-E and DDR2.
Oh, just one thing about the SSD if you do decide to upgrade - those G5 systems all have SATA-I, which most SATA-III SSDs do not have backwards compatibility with, so make sure if you get a SSD that it is either SATA-II or that it will absolutely work with SATA-I speeds.

No SATA-III SSD, even the early ones, would ever work in my G5 Quad, and many other people back in the early 2010s were seeing this same issue as well, so just giving you a heads up in case it doesn't work!
Heck yeah, performance of that system should be similar to an ARM A9 @ ~1.4GHz or ARM A15 @ ~1.0GHz or so, and if you don't like OS X, many variants of Debian Linux run great on them.

Good luck!

 

[XoR_]

Since your model of G5 also uses DDR1 RAM (ECC or non-ECC, if I remember right) and 64-bit PCI (PCI-X) expansion cards, they are fairly easy and cheap to upgrade as well, relative to the G5 Quad with PCI-E and DDR2.
Oh, just one thing about the SSD if you do decide to upgrade - those G5 systems all have SATA-I, which most SATA-III SSDs do not have backwards compatibility with, so make sure if you get a SSD that it is either SATA-II or that it will absolutely work with SATA-I speeds.

No SATA-III SSD, even the early ones, would ever work in my G5 Quad, and many other people back in the early 2010s were seeing this same issue as well, so just giving you a heads up in case it doesn't work!

Thank for the heads up. I do only have SATA3 SSDs. Need to check its SATA2 compatibility mode.

Heck yeah, performance of that system should be similar to an ARM A9 @ ~1.4GHz or ARM A15 @ ~1.0GHz or so, and if you don't like OS X, many variants of Debian Linux run great on them.

Good luck!

I expect quite a bit more more performance than mere 1GHz A15
A57 from what I gather have similar IPC as Core 2. How does C2D compares to G5? Nicely I would say:
Some benchmarks of Core 2 Duo based Mac Pro
http://barefeats.com/quad06.html
https://www.anandtech.com/show/2064/16

For comparison Athlon FX60/X2_5000+ 2.6GHz dual core have 724 in multi core core.

I have Core 2 Duo tablet running at 2GHz and will mainly compare G5x2 2.7GHz to that. I do expect this G5 to be slower because of software optimization being better for X86 than old G5, especially browsers. But I do not expect it to be comparable to Raspberry Pi and instead eat it for breakfast

 

[NamelessPFG]

Wow, you certainly know a lot about Amiga systems, and I've heard of about 1/3 of what you talked about, but the rest was news to me.
You might just be the unofficial Amiga expert on here, especially in this thread!

Oh definitely, those prices... that's exactly how it is any more with the PC-98 and even more so with the X68000 systems.
You're right, 128MB for any m68k-based systems is absolutely crazy, but is so much fun to work with, especially when the only limitation to the system is the CPU/GPU with that much RAM at its disposal.

I've heard that they were going to release that after-market shell for those systems, but it was only a rumor years ago, so it's very good news that it is going to be officially released.
I've often thought about diving into the Amiga area of computing, but the high cost of starting out fresh is going to require saving up, not to mention how expensive all things "retro" are becoming.

Hang onto that system as long as you can - ten years from now, that might be worth as much as an A4000 is now.
That's pretty cool about Chip RAM and Fast RAM on those things, I never new about any of that, and especially with the different revisions.

Thanks for all the great info, and if you make any more developments on it, please share it on here if you can.
You might also checkout buyee.jp for Amiga parts as well - even though it is a Japanese proxy service to Yahoo Auctions (Japan), it can still sometimes provide access to very rare or much needed parts that are not available anywhere else.

Good luck to you, and I can't wait to see more!

To be honest, I just do lots of research on whatever has my interest; I don't really have any actual nostalgia for the Amiga, being a '90s kid in the US where the market was dominated by IBM at home and Apple in the classroom. It wasn't until last decade that I'd heard of its gaming prowess and beating everyone else to the punch to pre-emptive multitasking on a home microcomputer. There's plenty of people in Europe who know more about these things than I do!

The whole Chip RAM/Fast RAM/Slow RAM thing is mainly just a quirk of how the Amiga's custom chipset works. Other systems don't have the memory bus being taken from the CPU like that because the CPU has to do everything on its own anyway (not counting bizarrchitectures like the N64 where the CPU doesn't have DMA at all), which squanders a lot of a Macintosh II's CPU/FPU/clock speed advantage, for instance. It's sorta like the 2D gaming equivalent of running Quake at 640x480 on an AMD K6-2 350 in software mode, and then an Intel Pentium 166 or something else slow paired with one of those newfangled 3dfx Voodoo Graphics cards; the K6-2 doesn't stand a chance, and I doubt a Katmai P3 would've fared much better.

That said, I've discovered a major wrench in the works with this A500, which I sorta foresaw from the seller due to complaints of non-working keys, but not to this extent: the membrane has considerable trace damage, with attempts to remove the plastic above the damaged traces frequently removing the traces themselves by accident as well. I'm gonna need a new membrane - if I can find one! Or, more likely, a USB keyboard controller.

There's also the possibility to go completely crazy and whip up a custom PCB to hold some Cherry MX switches, then get some appropriate keycaps. Heck, I can picture it already: that black Vampire edition A500 repro case getting a blacked-out, red-backlit Cherry MX keyboard to match, like it's some kind of modern gaming battlestation...

 

[Red Falcon]

Thank for the heads up. I do only have SATA3 SSDs. Need to check its SATA2 compatibility mode.

You bet!
Hopefully the SATA-III SSD you use will work with it, but if not, all of the SATA-II based SSDs worked very well in the SATA-I ports of the G5 models.

I expect quite a bit more more performance than mere 1GHz A15
A57 from what I gather have similar IPC as Core 2. How does C2D compares to G5? Nicely I would say:
Some benchmarks of Core 2 Duo based Mac Pro
http://barefeats.com/quad06.html
https://www.anandtech.com/show/2064/16
View attachment 64212
View attachment 64213
For comparison Athlon FX60/X2_5000+ 2.6GHz dual core have 724 in multi core core.

I have Core 2 Duo tablet running at 2GHz and will mainly compare G5x2 2.7GHz to that. I do expect this G5 to be slower because of software optimization being better for X86 than old G5, especially browsers. But I do not expect it to be comparable to Raspberry Pi and instead eat it for breakfast

I think you might be right about the IBM 970MP (dual-core @ 2.5GHz) being faster than that, more like a 2GHz dual-core ARM A15.
The below-quote is from this thread: https://hardforum.com/threads/so-i-have-a-sunfire-t2000-what-now.1877549/#post-1041899444

How can A9 that score 2750 in Geekbench beat G5 with 3700 points?
How can 2.5 DMIPS/MHz times 1700MHz beat 2.9 DMIPS/MHz times 2500MHz?
Better optimizations are one thing but difference in raw power is too great for Odroid U3 to come faster. In some benchmarks made in web browser like Chrome it might be true (due to lack of JIT on PowerPC) but not in normal programs.
I would suspect Odroid XU3/XU4 with Cortex A15 at 2GHz to be be slightly faster than this G5 but not A9 @ 1.7GHz. A9 being faster would not make any sense.

That being said I agree with your point of new computers being much faster using fraction of power. Espeically if we take single core performance which is very important in many situations and which is rather bad on those kind of server processors.

And I definitely agree it would be awesome to see some photos and benchmarks of such PC. I would love to see results of full Phoronix suite benchmarks.

Hmm, your name is XoR_ and that poster's name was XoR... is that really you?!
haha, either way, he/you was correct about the IBM 970MP dual-core CPU being faster, and the IBM 970FX single-core @ 2.7GHz (with two CPUs in the second G5 model you have) would still be pretty fast, but I'm not sure if it would match an A57 - maybe a lower-clocked version of it for sure, though.

Well, I wouldn't mind seeing some benchmarks showing where each of these sits at anyways!
I've always been a huge fan of the PowerPC ISA and CPUs, and wouldn't mind putting ARM down a peg or two with them, just for fun.

Yes, your G5 will eat the Raspberry Pi 3 for breakfast - that A53 @ 1.2GHz doesn't stand a chance; the A53 is only slightly below the older A9 clock-for-clock, but an A15, A57, and those IBM 970 processors would all destroy it at the Pi's 1.2GHz clock speed in both Dhrystone (integer) and Whetstone (floating point) operations.
You need to stop posting about this now.

Seriously.
It is making me want to get a G5 again, and I don't have the space, money, or time for it right now... so stop!

Thanks for those benchmarks, too, that CPU is much faster than I remember it being, especially compared to the x86_64 CPUs of the time - and you are right about the IPC of an A57 being similar to a Core 2 as well.
On a side note, I've been seeing that Apple is planning on replacing x86_64 CPUs in their systems with their own in-house ARM processors starting as early as 2020 - kind of hearkens back to 2006 when they changed from PowerPC to x86/x86_64, and in the mid-1990s when they changed from m68k to PowerPC - here we go again.

You might also consider getting a few of these Macally drive enclosures: https://www.ebay.com/itm/Let-t-of-2...064968?hash=item3d5de181c8:g:TRUAAOSwfqFatsFs
They work really well with the Sonnet Tempo eSata adapters, or the native Firewire 400 1394a ports on those G5 models, and literally look like a mini G5 or Mac Pro case - quite classy.

I used to have four of them daisy-chained on Firewire to my G5 Quad in software RAID0 back in 2012 and was getting 40MB/s (limit of Firewire 400), and then later moved them to eSATA in 2013 and was getting much better speeds.
The G-S350SU model was the best external drive case that I have ever used, and never failed me; the chipset in them natively allows for up to 2TB drives to be used, but they can also be flashed/updated to use 3TB drives as well.

They are old drive enclosures, for sure, but if you want something that is 100% reliable and compatible with those G5 models, they are absolutely your best bet.
Please post more if you have any further updates!

 

[Red Falcon]

To be honest, I just do lots of research on whatever has my interest; I don't really have any actual nostalgia for the Amiga, being a '90s kid in the US where the market was dominated by IBM at home and Apple in the classroom. It wasn't until last decade that I'd heard of its gaming prowess and beating everyone else to the punch to pre-emptive multitasking on a home microcomputer. There's plenty of people in Europe who know more about these things than I do!

The whole Chip RAM/Fast RAM/Slow RAM thing is mainly just a quirk of how the Amiga's custom chipset works. Other systems don't have the memory bus being taken from the CPU like that because the CPU has to do everything on its own anyway (not counting bizarrchitectures like the N64 where the CPU doesn't have DMA at all), which squanders a lot of a Macintosh II's CPU/FPU/clock speed advantage, for instance. It's sorta like the 2D gaming equivalent of running Quake at 640x480 on an AMD K6-2 350 in software mode, and then an Intel Pentium 166 or something else slow paired with one of those newfangled 3dfx Voodoo Graphics cards; the K6-2 doesn't stand a chance, and I doubt a Katmai P3 would've fared much better.

That said, I've discovered a major wrench in the works with this A500, which I sorta foresaw from the seller due to complaints of non-working keys, but not to this extent: the membrane has considerable trace damage, with attempts to remove the plastic above the damaged traces frequently removing the traces themselves by accident as well. I'm gonna need a new membrane - if I can find one! Or, more likely, a USB keyboard controller.

There's also the possibility to go completely crazy and whip up a custom PCB to hold some Cherry MX switches, then get some appropriate keycaps. Heck, I can picture it already: that black Vampire edition A500 repro case getting a blacked-out, red-backlit Cherry MX keyboard to match, like it's some kind of modern gaming battlestation...

Ah, gotcha, that all makes sense.
I'm kind of in the same boat with the Sharp X68000, and only sort of knew about it in the 2000s but didn't officially know about it until late 2013 or early 2014 or so.

That's too bad about the keyboard membrane being damaged like that, but a USB controller might be a good idea to get it fixed.
If my SparcStation's keyboard ever craps out, I was fortunate enough to find a SparcStation keyboard-to-PS/2 keyboard/mouse active adapter, which today is now worth its weight in gold.

I always wanted to get a PS/2 keyboard adapter for the Nintendo Gamecube as well, especially since I had turned mine into a mini PowerPC Linux server (not compatible with the keyboard-controller for Phantasy Star Online, though), and that keyboard controller, if you could even find one, was $$$$.
You know, that last idea you had sounds amazing, and from what you know on these systems, you could totally make it happen.

Thanks for the info on the A500 as well, I never knew much about the Amiga systems outside of what I've seen a few others post about, and this might be quite the adventure for you!
Just like the old saying, "It's not the destination that counts... it's the journey."

 

[XoR_]

I think you might be right about the IBM 970MP (dual-core @ 2.5GHz) being faster than that, more like a 2GHz dual-core ARM A15.
The below-quote is from this thread: https://hardforum.com/threads/so-i-have-a-sunfire-t2000-what-now.1877549/#post-1041899444

Nice looking SATA case. I have two large disks laying around. Might be good stylish investment

Hmm, your name is XoR_ and that poster's name was XoR... is that really you?!

I made account using email account I lost access to so I made new one. Adding '_' to name was surely a very cheap move

You need to stop posting about this now.
Seriously.
It is making me want to get a G5 again, and I don't have the space, money, or time for it right now... so stop!

You can't avoid getting bunch of old Macs forever. Best to buy them now before they get into that recently ridiculously overpriced 'retro' category

 

[NamelessPFG]

You can't avoid getting bunch of old Macs forever. Best to buy them now before they get into that recently ridiculously overpriced 'retro' category

Assuming they already haven't, anyway; everyone saw the Twentieth Anniversary Macintosh (TAM) going up in price miles away due to its nature as a high-priced, limited-edition product (and, frankly, one that would've been a waste of money over a Power Macintosh 9600 if my Power Mac 6500's performance is any indication, since the 6500 and TAM use the exact same logic boards with different jumper settings), but certain other nostalgia-laden models like the SE/30 and especially the Color Classic can command a surprising premium. Must be the compact Mac form factor appeal.

As for more modern models, people are willing to pay more for Power Mac G4 Cubes despite being less capable than something like a Quicksilver or MDD, too. Also, any 17" PowerBook G4 or MacBook Pro goes for more than you expect, just because Apple doesn't make notebooks that large any more.

I was also surprised to learn how much people will pay for an old Macintosh IIcx on eBay, especially considering how much more capable the six-months-later IIci is; it means I'm just leaving money lying around here, though I also can't really bring myself to sell the IIcx until I get down to figuring out what's going on with the logic board that causes it to only work with that 4 MB bank of RAM I have installed (it just won't see the rest, like an address line's cut somewhere, and larger SIMMs = Chime of Death from not seeing any RAM). It doesn't feel right to me, even though the system is currently in a functional state, because there's still parts that could be working better than they are right now.

Oh, and if you do get into really vintage Macs that don't have 1.44MB HD SuperDrives installed, make sure you have some appropriate 400K/800K floppies on hand; you're not going to be able to write those without a beige Mac that has a factory floppy drive, or maybe a KryoFlux. (One area that I definitely want to investigate is the Amiga's capability to read and write Mac floppies if using an actual Mac floppy drive, particularly with ShapeShifter and/or Fusion, but nobody's posted any pinout schematics or software for this that I can find.)

 

[Red Falcon]

^ Really, though, I don't think it is as much about the performance per model as much as it is the form factor or nostalgia relating to a specific unit, or even the components within.
Oh, those RAM settings - I can totally relate when I was maxing out the RAM on my Quadra 950 with 256MB of FPM DRAM SIMMs.

I've never had a chance to work with the KryoFlux before (because they are $$$ and rare), but working with 360KB 5.25" floppies for my Compaq Portable II and Osborne-1 was extra fun, haha.
Actually, I do have a native 24v molex PSU and FDD kit and self-built converter cable for 8" FDDs, and just need to get a working 8" FDD at this point.

Kind of interesting that most disk drive equipment from the 1970s and earlier wasn't 12v, but was actually 24v, mainly for the motors.
Funny you mention the G4 Cube and Quicksilver - those Cubes had hideous overheating and PSU issues if I remember right, though they did look nice, and I did have a Quicksilver myself once upon a time with dual Motorola PowerPC CPUs (don't remember the exact model on them).

The Quicksilver was very reliable and stable with nearly any OS I threw on it.
It would be neat to see you get your Amiga working with Fusion and a 400K/800K Mac FDD - this is the thread were retro dreams and the impossible are both made possible!

But come on, just look how sexy these 1980s Apple systems are:

 

[XoR_]

Nothing beats Mac G3 though

 

[NamelessPFG]

So I went to my first Vintage Computer Festival event here in the Southeast, since it was actually within driving distance. To say that it was a retrocomputing gold mine is an understatement, and I'll just start by listing off the swag I snagged:

-Amiga 2000 rev. 6.2 (8372A Agnus, but standard OCS Denise and Kickstart 1.3) with A2091 SCSI controller (no added Fast RAM, unfortunately), a 4 GB SCSI HDD (that it didn't seem to boot from), a Video Toaster card, keyboard (missing the right Alt keycap and tactile sleeve, but functional), tank mouse, and a loaded floppy that turned out to have Workbench 1.3 on it. You bet I was taking that thing home ASAP!

However, the SCSI HDD wasn't exactly bootable off the bat, and a cable was missing. I haven't exactly had the best of luck getting it to boot from the SCSI HDD without trying to format it elsewhere (which I want to avoid, given what a PITA it is to get Amiga-bootable media going on a PC), but it does boot off the Workbench floppy just fine. Thankfully, so does that Amiga 500 I imported earlier.

-Power Macintosh 9600/200 with a few expansion cards (PATA controller, combo USB/FireWire, ATI Rage 128, SB Live!, Ethernet, other things I can't immediately remember), all the drive covers (you'd be surprised how many of these are missing a few), and some kind of XLR8 G3 upgrade (as I understand, it's sorta like a slotket adapter that G3 and G4 ZIF modules can then be installed in), but only a PATA optical drive.

I knew I had to have this one too when I saw it being unloaded from a van just crammed full of vintage Mac stuff, and as it turns out, the seller had used it as a primary machine for a while, well aware of how expandable it was. The 9600 was my endgame bridge Mac, the one I sought to replace my existing 6500 with, and replace I did.

However, it seems like the pre-installed PATA controller isn't being picked up as a bootable device. I don't have a SCSI HDD currently set up since I also sold the IIcx there, but I do have a spare 1 GB drive I was in the middle of setting up before VCF hit. Didn't think to claim any more SCSI drives while I was there, either.

-17" CRT Apple Studio Monitor, partly for the aesthetic factor, partly because it's still aperture grille, partly because my Sun GDM-5410 is looking primed to join my Sony GDM-FW900 in the FD Trinitron afterlife. Thankfully, this monitor turns on just fine, has no screen scuffs or scratches that I noticed, and doesn't display any apparent color issues. This one's definitely a keeper.

-A fairly stripped-down, parts unit MDD G4 single 1.25 GHz FW400, just in case it had parts I could use on my main MDD. It was pretty cheap, after all.

-Rocketfish 900W 80+ Silver modular PSU, sealed in box. Hardly vintage, but how many PSUs that specced-out do you see being sold for under $20? If anything, I'm running short on ATX PSUs for all my systems!

-Box of 50 "duplicator-grade" blank DSDD 3.5" floppies. Pricing seemed high at first, but it quickly sunk in why: those were actually listed by one of the Amiga enthusiasts with a table there. (A table with a thoroughly-hotrodded Amiga 1200 with an A1200.net transparent case, a Blizzard1260 accelerator with SCSI expansion, Indivision AGA, USB controller, some kind of really compact IDE header SSD... to say that thing alone was worth several hundred dollars is an understatement!)

A bit of negotiation later, and I should be able to prepare myself to get any ol' Mac or Amiga up and running with these, without the kludgery of taping holes on HD disks and hoping it works when formatted as DD.

Now, as for the things I didn't get, but are worth noting because I never expected to see that kind of hardware there:
-Atari 1040ST + SC1224 monitor, both boxed. I was waiting for a price tag to be stuck on 'em, not knowing how VCF consignments went (this was my first), but next thing I knew, they were tagged "SOLD".
-NeXTstation Turbo with monitor, keyboard and mouse. Seriously thought about this one, but ultimately realized I don't know what I'd do with NeXT stuff as someone mainly concerned about retrogaming (and for which I have no idea what id Software were doing on 'em while developing Doom).
-SGI Indy. No accessories. Also passed because the only Indy I would want would have N64 devkit hardware in it.
-Apple DOS compatibility cards, both the official Apple ones and a couple of OrangePC ones with 486DX-33s on 'em.
-Macintosh 128k in its own carrying case, keyboard and mouse included!
-Macintosh SE/30, also with a matching keyboard and mouse. Despite the high list price, someone must've made an offer good enough for it to sell, because it was gone by event's end.
-Macintosh Classic. I tested the Cmd-Opt-X-O startup sequence, and sure enough, it worked. There was also a HDD inside that also worked for booting from.
-IBM portable (as in "luggable") PC with a nice XT-style keyboard.
-Stripped-down Macintosh Quadra 950. If I could find more than a case, logic board and PSU for it, I might have snagged that one too, but I didn't feel like trying to source all the rest.
-Two Power Mac G5s. One of 'em sold.
-Way too many Macintosh IIcx and IIci systems to count, and the occasional full-size Mac II. (No IIx or IIfx, though; I checked!)

Really now, I'm just kinda scratching the surface of what was there. It was a lot to think about buying, but I only have so much space in my car (and at home), not to mention that it's not the sorta environment that's easy to actually test that hardware in before buying. At the very least, my big-ticket items snagged there power on without errors, so that's good.

Pics to come later; I still need to arrange everything that I hauled home, not to mention get the A2000 and PM9600 booting.

Needless to say, if you're in the vicinity of one of these events, I highly recommend going. Lots of stuff to score, lots of cool exhibits run by cool people; heck, one of the guys down here at VCFSE was packing all sorts of imported Japanese computers, ranging from an X68000 XVI with SCSI HD boot, an X68000 Expert alongside it, an FM Towns II, several PC-98s, several MSXs including a turbo R... never thought I'd ever see those running in person.

VCF East is coming next month, and VCF West is scheduled for August. I'd suggest planning in advance, saving up a few hundred beyond any admissions fees in case you wanna score something in the consignment area, and heck, bring in some hardware you're not using anymore. I made a fair bit of money there just by selling mechanical keyboards, even the ones that weren't IBM Model Ms.

 

[Red Falcon]

snip

LOL, you suck!
Well, I can't say that I've ever been to anything even remotely that awesome before, but congrats on getting to go and for the massive haul!

Just, wow!
If I ever get a chance to go, I most certainly will, but I might go broke at the same time, haha.

It sounds like everything you got was really high quality, and +1 to that RocketFish PSU - I still have a 500 watt one in service to this very day, and it was from circa 2009 or so.
Can't believe you saw a transparent A1200, though, those are quite rare, and especially so with all of those upgrades.

I really liked those G4 systems as well, both 1P and 2P models, and never had one fail.
Yeah, that 9600 is very sought-after, and I've often thought about picking one up myself.

Wasn't the 9600 the last unit to support NuBus '90, right before Apple moved fully to PCI?
For those items you didn't get... if one can even find those on eBay, they are $$$ to $$$$ depending on the part - Apple DOS compatibility cards with a 486DX 33MHz in it... you may as well take out a second mortgage.

Can't wait to see the pics, and yes, please share on here if you get a chance.
It is really fun and inspiring to me to see others collections and fixed up pieces on display and/or fully working.

Funny you mention all of those Japanese workstations, as I do have two of the X68000 units, as showcased in this thread, both booting from SCSI/SASI (AztecMonster), and I do indeed have a pre-upgraded PC-98 which I really need to get up and running.
MSX units are extremely sought-after items, though, and are getting tougher to find, including that Turbo R model, dang!

FM Towns units are pretty awful, though they are more rare and very expensive, ironically enough... though I would still really like to have one, but shhh, don't tell.
Nice work on making a bit of cash on those keyboards, too!

 

[NamelessPFG]

Still no pics yet, but I figure I should give an update one week later:

-To my dismay, the Amiga 2000's suffered some battery acid damage near the 68000 CPU socket and related traces. It seems to be screwing with the address lines leading to the RAM, though my current lack of a PLCC extractor isn't helping when it comes down to diagnosing the Fat Agnus chip. It's just not reliable right now like my A500 is, but the A500 is also still awaiting a replacement keyboard membrane, making it useless for anything that requires a keyboard (unless I were to get a DIN-5 female socket and use it to adapt the A2000 keyboard to the A500's header, since the protocol is the same).

I'll be waiting out this week for some PLCC extractor tools, a DiagROM chip to help me figure out the A2000 without hours of multimeter probing, 2 MB of SRAM chips for the A2091 SCSI card (it's useful as Fast RAM, which also allows the SCSI controller to run in DMA mode for faster file transfers on top of adding more system RAM in general), an 8372 Agnus for the A500 so I can turn the trapdoor "Ranger"/Slow RAM into Chip RAM, and of course, that A500 keyboard membrane from the UK that I have no tracking number whatsoever on.

At the very least, I have a Gotek floppy emulator with one of those OLED screens now, all flashed up with FlashFloppy. Good thing I had a CP2102 USB to TTL/UART serial interface lying around; it's the sort you need to use WinDAS on FD Trinitron monitors, and it's coincidentally the same sort needed for flashing a Gotek. Just add pin headers!

-The Power Mac 9600 also took me some work to get going; of the six 64 MB EDO DIMMs, two wouldn't show up at all in the System Profiler, effectively dead regardless of what slot I used, so that brings it down to 256 MB total. I think the G3 400 MHz upgrade card may also be adding a bit of instability along with its speed (can't figure out what extensions I might need for enabling/disabling the backside cache and/or speculative execution, or if there's DIP switch settings for that), and I didn't get the original 604e 200 MHz as a fallback...

Also, if you get one, MAKE SURE YOU HAVE SCSI DRIVES TO BOOT FROM. The original SCSI CD was replaced with a PATA DVD drive, and I would've been thoroughly boned had I not had an external SCSI CD drive to boot an OS 9.1 install disc from! Also, while it had an Acard PATA controller installed in one of the many PCI slots, it wouldn't recognize the PATA drive I pulled from my Power Mac 6500. I mean, when booted from CD, it showed in the System Profiler, but it absolutely refused to mount for some weird reason. Perhaps it wanted me to initialize the drive while on the PATA controller, which just wasn't going to happen as long as I hadn't pulled my files off first.

That said, I ultimately opted not to go with PATA for booting it, because one of the cards I hauled home was a Sonnet Tempo Serial ATA - two internal SATA ports, both quite bootable as long as you have a hard drive set up on 'em. (Don't try it with an optical drive, it won't work for initial Mac OS installation.) I was wondering to myself whether to save it for one of my MDDs instead of the 9600, but after thinking about what a complete pain it would be to route SATA cables from the MDD PCI slots to the drive bays, it went in the 9600. Threw in an 80 GB Hitachi SATA drive (that still has a Molex 4-pin power input for convenience), and it's been relatively smooth sailing.

I say "relatively" because I still sometimes get bomb errors on initial startup (which are random enough that they don't always happen), or it'll randomly freeze for a moment here and there before getting a move on. Also, one of the things I nabbed was a FW400 PATA drive enclosure, which I figured I'd chuck the 6500 HDD into and copy everything from that, except the 9600 just kept seizing up shortly after it got connected. It was pretty frustrating, enough that I actually resorted to pulling the OrangeLink FW/USB combo card out and replacing both it and the Acard PATA controller with a Sonnet Tempo Trio, and somehow, that did the trick. At least until I set up OS 9.2.2 by means of OS 9 Helper, applied the FireWire fix, and then it decided to stop working for some dumb reason. Great.

Even with its quirks, though, this 9600 blows the 6500 out of the water, enough that I feel like anyone who bought a Twentieth Anniversary Macintosh back in the day got ripped off big time when they could've nabbed a 9600 for far less and had a much more powerful and expandable machine as a result, if not nearly as stylish or exclusive. It's still not a patch on either MDD, but you can't write 400K/800K floppies on an MDD, either, let alone use ADB (without one of those elusive iMate adapters) or RS-422 serial if you need them. That's what makes the 9600 such an appealing bridge machine, alongside packing built-in RJ-45 Ethernet (no AAUI adapter necessary!), a crapton of PCI slots, and support for roughly Mac OS 7.6 to 9.1 (and then 9.2.1/9.2.2 with OS 9 Helper, if you so choose).

Speaking of the MDDs, I found out that the parts unit I bought - the 2003 FW400 single 1.25 GHz system reintroduced that year specifically for its OS 9 compatibility - indeed only had a dead PSU holding it back. I hooked up the ATX adapter harness that I personally repinned to that Rocketfish PSU (which I picked up for test bench purposes like this), threw in some DDR-333 and a graphics card, and it fired right up! Guess that puts me in the market for two MDD PSUs (most likely ATX adaptations for reliability and efficiency reasons), a second ATX extension cable to repin, and some hard drive cages since this MDD was stripped. Granted, I don't need two MDDs, either, so the single CPU model here is likely going back up once I'm done getting it up and running.

Oh, and that 17" Apple Studio Monitor looks good on screen with no issues that I've perceived, and it's a hell of a lot more reliable than my aging Sun GDM-5410, albeit only just managing 1600x1200 at 60 Hz rather than 95 Hz. It's also got that graphite G4 aesthetic going for it. Some people may be all too happy to take up LCDs, but I'm still of the opinion that you really can't beat a CRT for retrogaming and retrocomputing if you've got the space.

 

[Red Falcon]

^ Damn, I totally respect your dedication and appreciate for learning about these older systems - keep up the great work, I really enjoy reading everything you are working on or figuring out.
I know *exactly* how you are going about that 9600, and I had nearly the same experience with the Quadra 950 earlier in this thread, especially with the SCSI optical drive.

As for that 80GB HDD, if that model doesn't work very well, you can always use a molex-to-SATA power adapter as well - nearly all modern HDDs only use the 5v and 12v rails (both available on molex) and almost never use the 3.3v rail on the native SATA power plugs, plus SSDs almost always use the 5v rail, with a few that use the 12v rail such as the Intel DC S3500, so those could be a potential option as well.
I've only ever seen a few HDDs actually use the 3.3v rail, and most were from around the time circa mid-2000s when the change from molex to SATA power was becoming a thing; you probably already knew all of that, but just thought I would throw it out there just in case.

Really hope you can get those Amigas up and running 100%, and even though I've never used one, from what I've seen others do with them, they seem like a ton of fun for both games and productivity software.
Good luck with all of this, I really can't wait to hear more about your progress!

 

[NamelessPFG]

So I got a bunch of Amiga-related stuff tod-er, yesterday (past midnight now), with the crown jewel being the top one of these:

I already had the Mitsumi keyboard on the bottom with the A2000, complete with the Video Toaster stickers pre-applied, but saw that Cherry keyboard up top being listed with another Amiga tank mouse (not pictured) and just went "Sure, why not get the best Amiga keyboard out there, and also ensure that both the A500 and A2000 have their own mice - or hook both to one of them for some multiplayer Lemmings action?" I swear, this Amiga thing is not good for my wallet.

Aside from that, there was a lot more that arrived today, mainly to help with the A2000, but also upgrade the A500:
-PLCC extractors (DON'T try to pull a Fat Agnus without one of these!)
-GGLabs F2R16 (This was supposed to be one of the pre-flashed DiagROM ones, but I ordered from the wrong listing and got one with Kickstart 3.1 flashed instead. Hey, at least it was ten bucks cheaper!)
-8372 Fat Agnus (for upgrading the A500 to 1 MB of Chip RAM, instead of leaving the trapdoor memory as Slow RAM)
-2 MB of SRAM chips for the A2091 card (aside from a SCSI controller, it also adds up to 2 MB of Zorro II Fast RAM and enables DMA transfers with Fast RAM installed)

So with all that listed, let me go over how it all went down:
-I thought I noticed something funny with one of the pins in the A2000's Agnus socket, and one safe PLCC extraction later revealed why: one of the addressing pins was almost completely broken off! I don't know how that happened, but I wasn't feeling too happy about it, having to slip the 8372 into the A2000 instead.

Just as a diagnostic measure, I slipped the slightly emasculated 8372A into the A500, and sure enough, it started throwing POST errors too... but I refused to give up on this chip because there was just enough pin sticking out of the package to consider attaching a replacement. I got creative by soldering a paperclip to it (it's not as easy as it sounds due to the tight tolerances between pins), seating that in the A500 again, praying for the best... and it worked! A few board modifications later (the usual "change JP2 and cut the EXRAM trace" affair on the Rev. 5 boards), one "avail" in the Shell later, and sure enough, this A500 has the full 1 MB of Chip RAM!

-The 2 MB of Fast RAM does work... whenever I can get the A2000 to boot properly, which usually entails disabling half of the Chip RAM and RTC with the J500 jumper for a bit of extra stability, and as it turns out, sometimes maintaining pressure down on the 68000 itself after unseating it from its socket for the first time as a diagnostic measure. At this rate, I'm getting an adapter for the CPU slot next to the 68000 socket (the sort that replicates the socket so you can use A500 accelerator boards without the clearance problems from the drive/PSU cage) and dropping the 68000 into that to see if that helps any by circumventing the board socket's corroded pins.

Also, going back to the Power Mac 9600:
-The PATA DVD drive is bootable, but you have to completely unlearn the old "hold C on startup to boot from CD" habit that you have to do on most Macs. I think that's because holding down C makes this particular Old World ROM search on internal SCSI bus 0 specifically.

As for why it's bootable, I suspect it's because the System Profiler sees the Sonnet Tempo Trio in PCI slot A1 as SCSI bus 2 (seriously; the external DB-25 SCSI port is bus 1, and it thinks all PATA and SATA controller cards are SCSI for some reason), and the Sonnet Tempo Serial ATA card is in PCI slot D2, so it's SCSI bus 3, next in the enumeration order.

-I tried having a slave drive on the first channel of the Tempo Trio, with the DVD as master. It keeps locking up the poor 9600 on startup, before it can even show the Happy Mac.

-DO NOT USE OS 9 HELPER TO RUN MAC OS 9.2.1/9.2.2 ON A 9600. No, seriously, that just drove me up the wall with such rampant instability that I had to reinstall fresh with 9.1, which actually does work stably enough. Hell, the file system is showing serious B-tree errors that may require me to initialize the SATA HDD yet again before I do yet another reinstall; I think that's due to some nasty freezes while attempting to copy files from a FireWire drive enclosure.

-As for my main MDD, giving this system a Radeon 9600 in the AGP slot (another one of my VCF scores, probably a G5 pull since it wouldn't work in a G4 without physical modification) gave it a real kick in the pants under OS X. The UI is smoother, BF1942 and SW:BF1 are actually kinda playable (even if the former runs worse than I recall on an Athlon XP 1800+/512 MB DDR-266/Radeon 9600 XT setup), and it makes me wonder if going into 9800 or X800/X850 territory would yield even bigger gains, or if the G4 starts being a huge bottleneck and those games start demanding a G5.

Better yet, booting under OS 9 with the extensions active for my flashed PCI Radeon 9250 doesn't seem to cause any instability problems or reduced performance of the sort that people usually complain about when you have those extensions running on a card too new for OS 9 (because that's the only way to get Core Image support in OS X). It does require dual monitors to make it somewhat convenient, but it's a great way to have an overkill OS 9 and a decent OS X Leopard box in one system.

The 9600 still needs to exist in the fleet, however, because the MDD obviously doesn't have a 400K/800K-compatible floppy drive interface, ADB, or RS-422 serial ports, whereas it's trivial to add USB and Firewire to one of the 9600's six PCI slots, so long as you choose the correct card.

 

[Red Falcon]

Dat keyboard, you need to treat it well... it's cherry.
I know that feeling all too well, those keyboards are literally the keys to the kingdom when it comes to setting up and using these systems.

My X68000's, SPARCstation, Quadra 950, Compaq Deskpro, etc. would all be totally useless without their respective keyboards.
Even keyboard "adapters" don't work very well, if at all - nothing beats the originals for full functionality!

As for your wallet, once you are done, you will be asking yourself, "What wallet???"
That's a good idea for the workaround on the 68000 with that adapter, though, quite resourceful.

Also, very nice find on the 9600 with booting the DVD optical drive, that makes a lot of sense with the older PROM (BIOS) trying to boot from SCSI in its place.
Was your MDD a single or dual CPU model?

It's not a bad idea to upgrade the GPU, but the PowerPC CPU, depending on the clock speed, might actually be the bottleneck, though a 9600 is really just a mid-range card from that era, and a 9800 or X800/X850 would be a BIG upgrade, for sure, and I would be very interested to see the results after the upgrade.
Good luck with your testing, this is all very interesting and your findings are really cool thus far.

Oh, one last thing, not sure if I mentioned this website or not, but you might enjoy it:
http://oldhacker.org/

 

[Red Falcon]

Quoting lostin3d's very inspirational post from this thread (RIP Ted Dabney):

RIP TD. The company you helped co-found had a huge impact on my life.

Sometime in the early 70's I remember visiting a relatives house and they had Pong. A number of years later(around 80-81) I had a paper route and saved something like $300 and bought a used Atari console(the predecessor to the 2600) along with extra controllers, games, and even replacement parts for the joysticks. Sometime after that I learned of a trick involving the switches that often gained access to debug level/dev modes on some games. A year later I saved more money and got my 1st computer. A used Atari 400 with the GTIA chip, cassette drive, and some manuals. It really was the beginning of computers and modding for me. Started getting compute!, and about a half dozen other era magazines. Learned basic, pascal, and some assembly. Later upgraded the keyboard to a mechanical(man connecting those old style ribbons sucked). Upgraded to the next model of cassette drive and then the first of many floppies. I gained an awareness of CP/M and extensively learned about DOS's from those. Upgraded from 16k to 48k with a kit that involved soldering 16k onto a 32k board. That rig was kicking butt till around 83-84.

This all happened before I was even born - what a time it must have been to be alive!
Here is another great video with quite a few stories from the 1960s-1980s:

 

[lostin3d]

Quoting lostin3d's very inspirational post from this thread (RIP Ted Dabney):



This all happened before I was even born - what a time it must have been to be alive!
Here is another great video with quite a few stories from the 1960s-1980s:

Love the thread you started here. Didn't know or remember about it even with the countless hours I've spent on [H].

Lots of cool stories. I too knew people who were in love the Z80 cpu(had friends with Spectrum's and TI99 4a). When I first got my Atari 400 my dad worked with a guy that had TRS80 model II and III along with a lot of cards, drives, etc. Even he called them Trash 80's which was kind of a shock to me. It was a magical time since so many companies were starting or changing. I still feel there's magic today but only when you're able to communicate on the enthusiast level and escape the marketing b.s. Something that is lacking now is that I don't ever look at the other side of the fence and think, how cool would it be to have what they have. These days self built pc's pretty much get you what you want as long as your budget allows, although I'm heavily leaning to getting a PS4pro come black friday. Back then I knew people that had Commodore 64's or Amiga's and thought they were cool. Knew people with Atari 800's and then the ST's. I remember when the Apple IIGS came out and that was awesome but short lived. Knew people with IIe's and those were cool. We briefly lived in Europe and got to see a computer called Dragon that was like a IIe that dominated much of Europe at the time. It was a neat time when you could really appreciate the different 'flavors' out there. These days it mostly blue vs red or green vs red or pc vs console. Kind of boring in relation and honestly sometimes I really have to look hard to see the differences(other than frame rates).

I also remember my first MS-DOS type computer that my mom and dad gave me as a Christmas gift around 1986-87. It was a Tandy 1000ex. Some people scoff at them but it was pretty powerful for its time(my first 8088 http://www.pcmuseum.ca/details.asp?id=100) and one of the first MS-DOS types to break the 4 color barrier(320x240 16 colors or 640-480 4 color- or some such). I never got the math co-processor but did expand the memory in the nineties when I could find some cheap kits. I loved that thing and kept using it until around 90-91. I don't remember how much ram it had but it was enough that I often configured ramdisks for extended gaming sessions. I also find it amusing how much Ms-Dos is still alive in windows.

I briefly stepped out of PC's during the nineties. Around 98 or so I started getting some hand me downs from friends and family. A pentium here, a pentium 2 there, and I did the usual ram upgrades or cables. Maybe found some kind of card to throw in 'em when someone gave a box of misc. junk they didn't need. About that time consoles seemed to get really exciting. NES had been dominating but it wasn't enough to compare to what I'd seen in the 80's home computing era(PC moniker was around but by far usually referred to as home computing by many throughout the 70-80's), Sega had it's things, and then the first PlayStation. That blew me away for what something so small could do. I was one of the suckers who immediately got into the PS2. Paid full price, new. It was dead in under a year. Got another used it died not long after. Finally got a slim and loved it. Total Monty Python skit there. I also remember learning about the Japanese PS2 that were basically a PS computer but never got to see one. Had friends with N64, Saturn, and few other things. My wife still has her Jaguar along with it's box and AVP. Also knew someone with a Falcon. That too was a truly magical time for console gaming. After that I decided it was time to get back into PC's again. I wasn't confident with my building skills at the time but I did like a pre-built at the local BB. It was a VPR Matrix with a pentium 4. Took me about 18 months to pay off. Over the course of 6-8 years I eventually replaced every part of it except the case, literally everything. So proud of that thing. Still have it in a closet it was the first PC I had that could do some 1080p gaming. Then around 2010 I finally got into building my own from scratch. Those are the builds I still use and you see in my profile and signature. Thanks to tips I've gotten from Kyle on overclocking CPU's and GPU's, both rigs hold their own with even some of the best today for gaming. May not be the best, but close enough for me.

Something else I forgot about with the Atari was when I got my first full color printer. It was an Okidata that used a sort of RGB strategy(three colored ribbons) to create full color printing. Slow as anything but man that was cool for home computing in 84 or so. It was also the first printer I didn't have to manually configure bit, parity, and all those other things back then.

I have to say there are some really cool things happening now. I got into 3d around 2012 or so and still play what I can(1080p/120hz and 1440p/144hz). Haven't gotten into VR but it too seems pretty cool. Textures and resolutions are truly getting amazing, if you can afford the hardware to run them. I've recently read about 8k glasses free 3d t.v's that have been favorably demonstrated at some shows in the last year or so. Home digital projectors have been amazing and affordable in the last 10 years. I was happy to give my dad something back a few years ago, his first 1080p projector. A few years before that a Sony Vaio that he still uses today. Quite proud that I've got a rare unusual beast, a Toshiba 3d/120hz laptop. I think every era has it's magic, it's just a matter of seeing and appreciating it.

 

[NamelessPFG]

Well, I'll be! Those A500 cases got funded on the very last day! I'll let you all know when my Vampire SE case comes in.


In other news, here's what's happened since my last post here:

-The Power Mac 9600 continues to be a very finicky machine. It turns out that half the reason for its instability was what I call "insidiously bad RAM", the sort that can somehow pass POST only to give you silent data corruption hell after the fact. Out of frustration, I moved the DIMMs around so it was running non-interleaved, and it wouldn't chime. Not even a Chime of Death, it just turned on and did nothing. Turns out one of the presumed-working DIMMs seems to be bad enough that it fails to let the system POST in non-interleaved mode, leaving me with only 192 MB of confirmed good RAM for this thing. Ouch.

But if that wasn't bad enough, I tried something else: limiting it to 128 MB with known good DIMMs in interleaved mode. It was still unstable. It seems like I have to eat the performance penalty of non-interleaved RAM for reliability's sake.

Also, because I unsoldered the cache-enable resistor jumper, I could change the G3 upgrade's cache to write-back instead of write-through, which improves performance noticeably in benchmarks. Still need to test whether it works out in practice.

-I also dug up this little CompactFlash to IDE adapter I had lying around, with a 4 GB microdrive that I had pulled from an iPod mini that I upgraded to flash storage. Adding this on either IDE channel on the Tempo Trio causes the usual hang on startup. Yes, even the primary one that I normally have the DVD drive connected to. It's quite annoying, to say the least, since I was hoping to use it for an additional OS boot volume, but it really feels like this system wants either SCSI or SATA, not plain ol' IDE/PATA.

-The Amiga 2000's still out of commission, even a month later. I ordered one of those CPU slot adapters meant for fitting a Vampire 500 V2+ accelerator board, which I still intend on doing later, but putting the 68000 into that still makes the system act like it's braindead, as if it's still not seeing the CPU at all. (Yes, I took note of the proper orientations!) And no, I still don't have a proper desoldering gun for removing the CPU socket... it's kind of frustrating, really, knowing this system could easily be repaired, but not having the tools (or the money for said tools) to do so.

-The other MDD G4 I hauled from VCF initially only recognized 768 MB out of 1.25 GB that I had installed, but moving the DIMMs around a bit caused everything to show up fine. Funny how that works.

I also dropped a typical Radeon 9000 into it, since it's cheap and OS 9-compatible. Yes, the ideal way to go for OS X (Leopard especially) would be a later card with Core Image, but I'm restoring this system with OS 9.2.2 as the priority, being one of the fastest systems that can natively boot it, and want to keep costs down.

I just need to recap the PSU (Samsung) and replace a blown fuse. Looks easier to work on than the AcBel one that my dual 1.42 GHz MDD used to run on. Hopefully, that'll get it working again without forcing me to take more drastic measures regarding ATX PSU mods.

-I plopped a new battery into my iBook G4 1.42 GHz. Had to reset the PMU in order to coax it into charging, but sure enough, it actually holds a good charge! I just need to think about what I'd use it for, and whether or not I feel like taking the time to tear down the entire thing just to replace the old HDD with a SSD.

-Back to the 9600 for a bit: I was initially dismayed with its performance in Unreal Tournament being tons worse than the iMac G3 350 slot-loader, despite both running on G3s with a Rage 128 GPU. It seemed like there was more to this than just the iMac having double the bus speed (100 MHz vs. 50 MHz) and SDRAM being interleaved at the DIMM level by default, so I double-checked the renderer and realized I had it running in OpenGL the whole time rather than RAVE! A quick switch later, and it runs about as well as the iMac now, albeit with an awfully banded/16-bit look on a GPU known for having a negligible performance penalty to run in 32-bit color.

Mind you, I wouldn't consider UT'99 performance on either machine acceptable by modern standards; it's well below constant 60 FPS. Best to break out one of the MDDs and boot OS 9 on it (the OS X port doesn't support music for some reason, which I consider unacceptable due to how much I love UMX tracker music in general), which are powerful enough to run it without a hitch (lining up with my impression that first-gen Unreal Engine really wants CPUs in the 1 to 1.5 GHz range, x86 and PowerPC alike), or just do the sensible thing and run it on any ol' modern PC with D3D10/11 and OpenAL.