Socket 3 Pentium?

[Elledan]

I was just playing around with some old parts, trying to construct a router system, when I suddenly came across this weird Pentium I CPU. While trying to search the Intel site, it does not list this CPU's ID (SX948) with other Pentium CPUs, but does find a PDF file, listing a number of CPUs, including this one, most of them with an SXnnn ID as well. It specifies it as a 60 MHz core, 60 MHz FSB part.

The CPU does not fit into a socket 7, but appears to be socket 3 (whilst comparing it with a 486).

So, are there socket 3 Pentiums (judging from the sheer weight of it, it's an old part (the marks on it say 1992)), will they work with any socket 3 mainboard, and will a 486DX2-66 kick this Pentium's lower regions?

Thanks

 

[Dan_D]

That part you have is a Pentium Overdrive processor for 486's. It will work in a socket 3 board and it would beat out a 486 at just about anything if not everything. Unlike today Intel used to actually improve the performance of the processors architecture per clock cycle. It takes a 486 DX-4 100 to equal the power of the Pentium 60 or 66. But in a 486 board it likely won't perform that great due to the bus design of the 486 boards vs. earlier Pentiums. For example most 486's didn't have PCI. They were all ISA and VLB bus. So the 486 board wasn't as advanced, so that hurts it. As does the memory architecture. Fast Page Mode DRAM instead of EDO DRAM. A small increase to be sure, but an important one.

Just like the Pentium vs. the 486, the Pentium Pro was better per clock cycle than the regular Pentium. It wasn't until the Pentium 4 days that Intel started to suck at improvement of CPU cores. And actually they still can do it (Pentium M proves it) but they went with the highly scalable "netburst" core for marketting reasons. They helped perpetuate the "MHz myth".

Another thing to point out is alot of Pentium 60's and 66's had the FPU bug which caused them to misscalculate during multiplication. To put it bluntly that pile of shit can't add. Although it's FPU is alot faster than that of a 486.

That's why the Pentium 75 got a new socket and a new FPU. Intel wanted to distance themselves from the most embarrassing incident they ever had. A processor recall that affected thousands of processors and systems.

Shortly thereafter Intel started it's Intel Inside campaign to try and promote their processors as reliable and better performing than the Cyrix/AMD solutions of the day.

 

[Stugots]

a 486/66 or 486/100 would be a better choice than a pentium 60. i remember something about the pentium 60 having serious bugs in it. besides the 486/100 is probobly faster anyway

 

[Dan_D]

a 486/66 or 486/100 would be a better choice than a pentium 60. i remember something about the pentium 60 having serious bugs in it. besides the 486/100 is probobly faster anyway

They are about the same. That assumed though that you were using a PCI 486 board with a more advanced memory controller and had an Intel 486/100 as the AMD's and Cyrix chips of the day sucked hard.

It is true that the P60/P66 had a bad FPU bug. Anything over a certain amount of digits where math was concerned caused the chip to produce the wrong numbers when it was made to calculate math equations.

There is in Windows 95 on up a Pentium FPU workaround that it loads automatically that does infact fix this issue through software so the FPU does work.

 

[Stugots]

i had a 486/100 with 80mb of ram, that i used for a router running coyote linux for a long time, that thing ran great. never crashed, i only shut it down when i moved, i had it on a little ups used for hubs and switches and it worked fine on it. passive cooling on the cpufan, booted from floppy then worked from ramdisk, removed the fan from the psu. completely silent once it was up and running.

 

[phaelinx]

If you run programs that require "Pentium" instruction sets, the 486dx4 will simply not run it. I used to have a 486dx4 100mhz and a pentium 60mhz, I didnt notice a huge difference other than the fact that the pentium 60 would run programs the 486 could not because of that instruction set that the 486 didnt have.

 

[Dan_D]

True the added instruction sets were a plus. However a 486DX was only a match for the Pentium 60 if it was equipped with a PCI motherboard. The bulk of them were ISA and VLB.

So ISA video cards and VLB video cards couldn't match the performance of PCI. So overall even if the CPU was about the same the Pentium 60 was superior all around.

 

[phaelinx]

yeah both my 486dx and my p60 had pci slots..i ran a diamond v2200 pci card at the time along with the matrox m3d 3d accelerator addin card, basically allowed 3d accel games to run on a 2d card..and both games ran quake2 smoothly using the PowerVR options

 

[Elledan]

http://www.intel.com/design/pentium/SPECUPDT/24332601.pdf

The CPU I've got is the D1 stepping, which apparently does have the FP-bug.

Unfortunately I haven't got any decent socket 3 mainboards to try this chip with, only ISA-only '486'-mainboards.

If I were to get a socket 3 mainboard with PCI slots, would it be safe to use the CPU with it for a router or something similar? I mean, a router wouldn't possibly use x87-instructions, now would it?

Thanks for the info so far. It's good to know that not everyone here is only interested in the 'latest, greatest' CPUs

 

[phaelinx]

i wouldnt use any sort of windows varient on it tho..i'd install linux and use linux as the router

 

[The_Mage18]

First off, MEGA props to Sir-Fragalot. He (Or she, I want to be polically correct here) is the first person I've seen post here who knows a great deal of old school tech like myself.

In answer to your question Elladan it would make a great router. Keep in mind the comercial routers for home use use a chip that has far less power or flexibility. Any router you setup with Linux will outperform a shelf bought router hands down. Also, if you do get a board that has PCI you can drop a PCI WIFI card and make it an access point as well. Your only real limitations are going to be RAM amount and drive space. You'll be limited in RAM so a GUI would not be wise, stick with comand line interfaces. In case of drive size, a PCI hard drive controller card will take care of that.

When I went to Michigan Tech back in 1997, one of the offices built a pentium 75 with 5 NICs to use as a router for internal proxy and web use and it ran perfectly fine.

 

[Elledan]

First off, MEGA props to Sir-Fragalot. He (Or she, I want to be polically correct here) is the first person I've seen post here who knows a great deal of old school tech like myself.

Hear, hear.

On a sidenote, I always thought that the FP bug was limited to early Pentium Pro chips.

In answer to your question Elladan it would make a great router. Keep in mind the comercial routers for home use use a chip that has far less power or flexibility. Any router you setup with Linux will outperform a shelf bought router hands down.

Heh, that's the very reason why I'm constructing this router

I bought a Sitecom router a month or so ago, and while it doesn't crash randomly like the P166 Compaq (FreeSCO) it replaced, its inability to support a decent number of connections makes it a joke for anything more than just browsing the 'net.

Going to use Coyote this time, though.

When I went to Michigan Tech back in 1997, one of the offices built a pentium 75 with 5 NICs to use as a router for internal proxy and web use and it ran perfectly fine.

I've used a 486DX2-66 w/ 16 MB EDO RAM and a Conner 450 MB HDD as a FreeSCO (and Smoothwall) router for some time. Configuring the ISA NICs is a pain compared to PCI, though.

 

[phaelinx]

Haha..conner...I had a conner hd..It was a 40 meg scsi hd that sounded like someone was taking a chissel and scratching the platters because it was loud as shit. I had linux on it for years. I ran it as a gateway/router..

 

[Dan_D]

As far as I know the Pentium Pro didn't have the FPU bug as it was created well after the Pentium 60/66. However the Pentium Pro did have bugs with 16 bit code which made Windows 95 virtually unusable on those machines.

On a side note I am a "he" actually.

I also agree that a Pentium class machine of any sort would make a badass router. They are powerfull by router standards. I know a place in Dallas where you can get PCI 486 boards.

With the other comment concerning running Windows on it. It doesn't make sense to run Windows on a router. Windows 95 will work with the Pentium 60/66 as long as you add the Windows FPU bug patch to it. I don't know if it came with later versions of the OS such as Windows 95B OSR2.0/2.1.

Windows 98 should have the fix built in.

 

[ColinR]

Trudging back through the mists of time, I recall that the FP bug was a specific kind of long division (computers don't like division anyway) possibly involving a negative number. The majority of users word processing, playing games, and balancing the company accounts on spreadsheets would never encounter it. However, if you were the kind of user who did encounter it, the nature of it meant that you'd encounter it a lot.

The concept of Overdriving sockets (essentially using today's technology in yesterday's sockets) continued beyond the 486's Socket 3. A company called Evergreen had a load of products which used the concept, although true performance and specifications were hidden under a glaze of pseudo-technical marketing speak.

On a side note (seeing as it seems to be a running theme ), the AMD 486DX4s didn't suck at all, and I believe the Cyrix had write back caching (as opposed to write through). Either way, it was much more preferable to upgrade a chip to an AMD/Cyrix DX4 than a whole system. Unfortunately, my rig of the time (486DX33) wouldn't support an Overdrive Pentium.

 

[Elledan]

As far as I know the Pentium Pro didn't have the FPU bug as it was created well after the Pentium 60/66. However the Pentium Pro did have bugs with 16 bit code which made Windows 95 virtually unusable on those machines.

Weird. Either I've been told lies, or my memory is quite unreliable. I think it's the former, though.

I also agree that a Pentium class machine of any sort would make a badass router. They are powerfull by router standards. I know a place in Dallas where you can get PCI 486 boards.

Unfortunately, Dallas is an ocean too far for me

I've decided to revive an old P166 system and use it for a router for now. But who knows, one day I might come across a decent socket 3 board and use that one instead.

With the other comment concerning running Windows on it. It doesn't make sense to run Windows on a router. Windows 95 will work with the Pentium 60/66 as long as you add the Windows FPU bug patch to it. I don't know if it came with later versions of the OS such as Windows 95B OSR2.0/2.1.

Windows 98 should have the fix built in.

I've actually found that Win98 runs faster on old Pentium systems than Win95. Or at least it was on the P166 with 24 MB RAM I once used as a backup system for two months

 

[Lazn_Work]

I rma'ed a p90 that had the bug.. I think 120mhz pentiums were the first to not have the FPU bug.

The PPRO did not have that bug. (slowest PPRO was 150MHZ)

http://www.computerhope.com/help/cpu.htm

"One of the most famous and most known bugs is the Pentium FPU flaw / bug discovered by a a mathematician in October 1994. This bug involved the Pentium incorrectly performing floating-point calculations with certain number combinations, with errors anywhere from the third digit on up. This issue does not occur on 120MHz and above Pentium computers however is known to occur on Intel Pentiums 100MHz and below."

and as you can see from that page, his processor is probably a socket 4 pentium.. unless it is a socket 3 overdrive, but those are like 80mhz not 60 if I remember right.

edit: I was almost right: http://www.fact-index.com/p/pe/pentium_overdrive.html they are either 63mhz or 83mhz depending on model.

==>Lazn

 

[Dan_D]

I knew the Pentium Pro didn't have the bug. The P6 core is alot different than the P5 core.

I didn't think that Pentium's below 120MHz all had it though. I thought it was confined to the original Pentium 60/66. Especially since I've worked with many systems from 75MHz to 120MHz that didn't seem to have the bug. Win95 had a built in method for checking this. Processors that had the bug showed the yellow ! mark in device manager on the math co processor.

I have never seen it on anything but Pentium 60's and 66's.

 

[Elledan]

Weird, I don't seem to be able to find my CPU among the OverDrive CPUs on Intel's site, e.g. here: http://support.intel.com/support/processors/overdrive/prt.htm

Because the FSB for this CPU is specified as 60 MHz, that means that it's not even meant for 486 mainboards, right?


By the way, thanks for the information on the FP bug. It's nice to finally know the truth

 

[ColinR]

If it is an (original) Pentium Overdrive, it will have the pins arranged in square patterns on the underside (like a 486). On modern chips they are arranged in triangular arrangement.

Additionally, there is an extra row of pins arround the edge that aren't used by 486es but are used by Pentium Overdrives.


Like the DX2 and DX4 chips, the Overdrive gets it's MHz bump by increasing it's internal clock speed, leaving the motherboard's speed at it's existing speed (usually 25 or 33, but early DX50s were mounted on 50MHz boards). Iirc, 33MHz got the boost to 83MHz (x2.5).

The FSB is unlikely to be rated at 60MHz; as you say - no 486 mobos were rated at 60MHz. In latter times, they stopped producing the 50MHz boards, and relied on DX2s and SX2s to supply the 50MHz parts. If you have it in black and white, I'd say it was a typo by someone who's only been in the industry 5 years!

Tbh, I'd plug it all in and see if it works (set the board speed to 20MHz, ramp it until it doesn't work or you hit 40MHz - most boards won't go much further).

 

[Dan_D]

The 50MHz boards for the 486 DX 50 won't have PCI slots. Alot of later boards dropped support for that chip. It had heat issues mainly.

 

[Lazn_Work]

I knew the Pentium Pro didn't have the bug. The P6 core is alot different than the P5 core.

I didn't think that Pentium's below 120MHz all had it though. I thought it was confined to the original Pentium 60/66. Especially since I've worked with many systems from 75MHz to 120MHz that didn't seem to have the bug. Win95 had a built in method for checking this. Processors that had the bug showed the yellow ! mark in device manager on the math co processor.

I have never seen it on anything but Pentium 60's and 66's.

Most 75mhz Pentiums were built later, and did not have the problems, same with many of the 100mhz Pentiums. But I know my 90mhz one did have the problem. It had to do with what stepping the CPU was. (it was the first time I had heard of steppings within a rated MHZ back in 94)

==>Lazn

 

[Elledan]

Like the DX2 and DX4 chips, the Overdrive gets it's MHz bump by increasing it's internal clock speed, leaving the motherboard's speed at it's existing speed (usually 25 or 33, but early DX50s were mounted on 50MHz boards). Iirc, 33MHz got the boost to 83MHz (x2.5).

So why would this CPU be rated at 60 MHz?

The FSB is unlikely to be rated at 60MHz; as you say - no 486 mobos were rated at 60MHz. In latter times, they stopped producing the 50MHz boards, and relied on DX2s and SX2s to supply the 50MHz parts. If you have it in black and white, I'd say it was a typo by someone who's only been in the industry 5 years!

Take a look at the PDF I linked to earlier. It clearly specifies it as 60 MHz core, 60 MHz FSB part.

I think I'll just have to try this chip in a 486 mainboard and see what happens

 

[ColinR]

Alot of later boards dropped support for that chip. It had heat issues mainly.

I think it was due to the motherboards not being able to keep all their components synchronised.

That .pdf has no mention of Overdrives (okay, it has one, but only in passing, and in reference to Interrupt generation). There are no references to Overdrive specs.

We are going back some time here... I think it could be that the core is running at that speed and the core is the same. Outside the core is the FSB, running at 60MHz, then (perhaps) there's the multiplier tech, THEN the socket, running at 30, 33 or 25 or whatever.

Whatever, it's an old chip, probably won't be running Doom 3 any time soon, but could make a perfectly serviceable router. It may not even work! Just plug it in, if it works, run with it.

 

[Dan_D]

Some 486 boards did have a overdrive sockets. Just the way he described it in the first post I thought overdrive. However it could be socket 4 in all reality which is similar to 486 socket 3 in most respects.

We are talking about something that none of us has likely seen in 5+ years.

 

[Elledan]

Heh, I finally put this CPU next to a 486DX2-66 (socket 3), and it's way, way larger

The 486 has three rows of pins, while the Pentium has 4 rows, it also doesn't physically fit in a socket 3 ZIF-socket, so I think it's safe to say that this is a socket 4 or 5 (or 6?) CPU.

BTW, I didn't know until very recently that there ever existed something between socket 3 and socket 7

 

[Lazn_Work]

Heh, I finally put this CPU next to a 486DX2-66 (socket 3), and it's way, way larger

The 486 has three rows of pins, while the Pentium has 4 rows, it also doesn't physically fit in a socket 3 ZIF-socket, so I think it's safe to say that this is a socket 4 or 5 (or 6?) CPU.

BTW, I didn't know until very recently that there ever existed something between socket 3 and socket 7

Look at the link I provided in my earlier post. It showes all the sockets that Pentiums have used. (scroll down)

Like I said in that post, it is probably a Socket 4 pentium, as that is what the 60mhz Pentiums were (with the exception of the 63mhz overdrive)

==>Lazn

 

[Elledan]

Look at the link I provided in my earlier post. It showes all the sockets that Pentiums have used. (scroll down)

Like I said in that post, it is probably a Socket 4 pentium, as that is what the 60mhz Pentiums were (with the exception of the 63mhz overdrive)

==>Lazn

Thanks. My fault for not taking a look at that page sooner.

Anyway, this CPU is definitely socket 4 (all 273 pins are there; I just counted them ). Problem is, I don't have, or have seen a socket 4 mainboard so far. I'll just have to try and find one, I guess.

 

[Lazn_Work]

Thanks. My fault for not taking a look at that page sooner.

Anyway, this CPU is definitely socket 4 (all 273 pins are there; I just counted them ). Problem is, I don't have, or have seen a socket 4 mainboard so far. I'll just have to try and find one, I guess.

I truely doubt it is worth the effort, I would think it would make a better keychain or paperweight.

If you have a local swapmeet that has old computer junk you can probably buy a P2 with motherboard for ~$5 that would be 2 -3 times faster than that.

==>Lazn

 

[Elledan]

I truely doubt it is worth the effort, I would think it would make a better keychain or paperweight.

If you have a local swapmeet that has old computer junk you can probably buy a P2 with motherboard for ~$5 that would be 2 -3 times faster than that.

==>Lazn

I know what you're saying, and I do already have a pile of P2 systems to prove it, but if there's one thing I don't like it's functional hardware sitting around, unused

And when I say 'find', I mean pretty much stumbling over a socket 4 mainboard and being offered it for (nearly) free. There's no way I would pay more than 2 Euro for such old hardware.

 

[ColinR]

My old 486 motherboard had a row of pin sockets free around the outside of those occupied by the 486 itself. It was also blue, and the retention arm was all plastic, and had arms going down either side of the socket like a horseshoe with rightangles. Socket 3? Socket 2? So long ago...3.3V or 5V?

[nostalgia]
I remember my brother killed our 486DX33 by putting it in the wrong way round (I have no idea how he did it either) and switched it on. The damn thing SCREAMED and went snap (the motherboard was absolutely fine). His next trick was to stick a network card in our Pentium 100 (I was out at work). He didn't bother to set the jumpers on it and it conflicted with the CD-ROM. Mechwarrior 2's intro movie was stuttering all over the shop when I got home and started playing (he'd scarpered, cos he thought he'd broken the machine). I spent hours cleaning the CD, checking drivers, making boot disks, before I opened the machine and noticed this extra card. 30 seconds later everything was fine... Ahhh, old hardware
[/nostalgia]

 

[The_Mage18]

[nostalgia]
I remember my brother killed our 486DX33 by putting it in the wrong way round (I have no idea how he did it either) and switched it on. The damn thing SCREAMED and went snap (the motherboard was absolutely fine). His next trick was to stick a network card in our Pentium 100 (I was out at work). He didn't bother to set the jumpers on it and it conflicted with the CD-ROM. Mechwarrior 2's intro movie was stuttering all over the shop when I got home and started playing (he'd scarpered, cos he thought he'd broken the machine). I spent hours cleaning the CD, checking drivers, making boot disks, before I opened the machine and noticed this extra card. 30 seconds later everything was fine... Ahhh, old hardware
[/nostalgia]

Set nostalgia = True

Imagine having your modem set to the same IRQ as the COM port your mouse is using. Windows 3.1 is NOT fun to use via keyboard. Those of you old schools out there know the mistake I made, you made it too at some point admit it.

Admittedly, plug and play has made hardware installation extremely simple but there's still something to be said about placing hardware on it's own dedicated channel.
Then again with modern motherboards and all the devices now integrated into them, have you ever looked at the ESCD table for IRQs? I didn't think it was possible for the Enter key on my keyboard to share the same IRQ as my video card.