Conroe debunked

[robberbaron]

Netburst wasn't necessarily bad, it's just that the P6 should have been further explored. The Tualitin was a hint at what the P6 was capable of. The Pentium M confirmed it.

 

[Eva_Unit_0]

Netburst wasn't necessarily bad, it's just that the P6 should have been further explored. The Tualitin was a hint at what the P6 was capable of. The Pentium M confirmed it.

P6 certainly was a great architecture (arguably the best ever), no doubt about that...but it had some major drawbacks that make it perform very poorly compared to netburst in some tasks. Namely, P6 has quite weak FPU performance (okay, so does netburst but netburst sees huge performance increases from SSE). P6 (in the form of Pentium M's) competes incredibly well with netburst in tasks like gaming, but in other tasks like video and audio encoding it gets its ass handed to it.

I still think that's why P6 never officially made a return back to the desktop...the FPU performance in things like encoding is simply too weak.

 

[GilmourD]

Oh please...I hate it when everyone bashes netburst becuase here, at the end of its life, it has a hard time competing. Did everyone really forget that it's been around since november of 2000? Or that all the way up until 2004 when prescott and the Athlon 64 came around it was undisputed king of the hill? Or that it ran all the way from 1.3ghz to 3.8ghz, all on the same architecture, and can hit over 5ghz with good cooling?

For the majority of netburst's life (with northwood) it was the best cpu out there...yes, the XP's were great little chips (and I ran mostly XP's, not northwoods for cost reasons) but P4's were faster in pretty much everything. Go dig up some old XP vs. P4 articles and take a look for yourself.

Is netburst in need of replacement? Yes. Was it a bad architecture? No, of course not!

I laughed at Netburst from day one. I was a system builder at that time and suggested that people stick with either the last couple PIIIs that came out or move to the Athlon T-birds. One thing to note... At that period in time, I had a long-in-the-tooth PII and my brother had an OC'd Celeron.

 

[Poncho]

I laughed at Netburst from day one. I was a system builder at that time and suggested that people stick with either the last couple PIIIs that came out or move to the Athlon T-birds. One thing to note... At that period in time, I had a long-in-the-tooth PII and my brother had an OC'd Celeron.

Are you kidding me? The 845 and 865/875 platforms were hands down the best performers at the time. AMD couldn't hold a candle to Intel at that time.... hell, AMD's performance crown didn't have that big of a lead during the prescott days. The reality of it is that AMD caught up to intel when intel when to prescott and then inched ahead.

 

[coldpower27]

thanks man, and hell yeah i'm gonna get an AM2 setup. if anyone is an AMD fan here, it's me
if it makes you feel any better though, i'm also looking into an intel based rig to do comparisons between intel and amd as far as memory overclocking and tweaking goes.. conroe isn't gonna be out in time though, i'm probably gonna order some stuff within the next week, just gotta get some people to buy off my spare ddr and an extra dfi nf4 i've got sitting around

very interesting that anand already knows that the conroe is very data hungry. i don't see any review showing the performance difference with different fsb speeds and memory speeds, but oh well. i'll take their word for it now, and push for benches proving it when it's possible down the road

also, a note about memory efficiency on K8.. it's VERY cpu mhz dependant. running my ram at 250mhz with a 2ghz cpu gives 5709mb/s for a meager 71.4% efficiency.
bumping the cpu speed up to 3ghz while keeping the ram at the same timings and mhz brings the bandwidth up to 7415mb/s.. an impressive 92.7% bandwidth efficiency.
my guess is that anandtech has their hands on a slower clocked AM2, and ddr2's effeciency will really pull through when we get cpu's oc'ed to 3ghz and beyond

Ah then it will be awhile before we see the DDR2 bandwidth advnaatge really pull through if we need 3.0GHZ and beyond to see great differences...

Though it will be interesting to see what evel of bandwidth we can extract out of the Athlon FX 62.

 

[(cf)Eclipse]

Oh please...I hate it when everyone bashes netburst becuase here, at the end of its life, it has a hard time competing. Did everyone really forget that it's been around since november of 2000?

i know netburst used to be king, i'm very aware. it remained king until K7 almost caught up, but was a lot cheaper, and then K8 overtook it (but the price benefit kinda went away)
it's more the idea behind netburst that caused me to hate it. sacrificing so much efficiency to get high mhz never stuck me as an intelligent idea

Ah then it will be awhile before we see the DDR2 bandwidth advnaatge really pull through if we need 3.0GHZ and beyond to see great differences...

Though it will be interesting to see what evel of bandwidth we can extract out of the Athlon FX 62.

yeah, that remains to be seen, but you get the gist of where i stand on that issue

 

[Donnie27]

i'd extend that to netburst and you have me

Hugs my very trusty P4 2.6C. I'll never extend that to netburst and think Intel moved away from Northwood way the hell too soon! I never agreed with everything, hell maybe about half of the things Intel did. I'd put my i875 and 2.6C against any of my old P3 systems and BX or i815. 3.52GHz on Air is its best overclock.

One of my buds was using it downstairs in my Den and thought I still had it overclocked, hehehe! I unoverclocked and it using a CNPS 7000CUAl on low and it very quiet. The Liteon 52X burner drowns out the whole computer when it spins up and it is not that loud itself.

 

[Donnie27]

Yeah, rehire those that were laid off might be good. Give jobs back to those that were kicked in the kiester!

http://www.wsws.org/news/1998/may1998/sil-m9.shtml

Their next layoffs were worse!

 

[(cf)Eclipse]

Hugs my very trusty P4 2.6C. I'll never extend that to netburst and think Intel moved away from Northwood way the hell too soon!

a 90nm and 65nm netburst would have been interesting

 

[cyks]

I came back to this thread believing that there may be some real conroe news updates, wtf is my problem? Can this get any more OT? Is it a crime to ask for a real conroe thread, and have these retired one put to rest to make way for em?

 

[GilmourD]

I came back to this thread believing that there may be some real conroe news updates, wtf is my problem? Can this get any more OT? Is it a crime to ask for a real conroe thread, and have these retired one put to rest to make way for em?

I think that this thread has digressed due to the point that any real news on Conroe will arrive upon release of actual product.

 

[DejaWiz]

I came back to this thread believing that there may be some real conroe news updates, wtf is my problem? Can this get any more OT? Is it a crime to ask for a real conroe thread, and have these retired one put to rest to make way for em?

Start one!

 

[Donnie27]

i know netburst used to be king, i'm very aware. it remained king until K7 almost caught up, but was a lot cheaper, and then K8 overtook it (but the price benefit kinda went away)
it's more the idea behind netburst that caused me to hate it. sacrificing so much efficiency to get high mhz never stuck me as an intelligent idea

yeah, that remains to be seen, but you get the gist of where i stand on that issue

Dewd, K7 was long gone and didn't really last that long August 99 and K8 launched about April or May 2000. Netburst launched over a year later than K7 in Nov 2000. Remember, K8 was just the first 18um with Full speed L2 TBirds. K7 was just the Classic Athlons with 333MHz L2 that was on chip but not on die. The faster the chip ran, the more the cache hurt it.

We take SIMDs for granted now. When the SIMD dependant Netburst launched, folks didn't even bother running the Intel Launcher 1.12 software that was made up of On-the-shelf apps and Games. Ever see anyone benchmark/test "Incoming Forces"?

 

[Donnie27]

a 90nm and 65nm netburst would have been interesting

http://www.tgdaily.com/2006/04/19/intel_to_ship_core_inq3_2006/

Intel to launch Woodcrest, Conroe and Merom in Q3
Wolfgang Gruener
April 19, 2006 22:52

Santa Clara (CA) - Faced with increasing competitive pressure from AMD, Intel apparently has accelerated its launch schedule of its new Core microarchitecture. While Woodcrest, which will carry the Xeon 5100-series designation, was always scheduled for a Q3 introduction, chief executive Paul Otellini on Wednesday confirmed that the company will also be announcing the desktop processor "Conroe" (Core Duo E4000 and E6000 series) as well as the mobile CPU "Merom" (Core Duo T5000 and T7000 series) in the third quarter. All three variants will be shipping "in volume."

E4000s are the very Cheap Conroes with 2MB instead of 4MB L2 and 800MHz FSB instead of 1066MHz FSB. Slowest speed listed was 1.6GHz for $150. IMHO, 666 isn't the only reason AMD moved theirs up two weeks.

 

[Mav451]

http://www.tgdaily.com/2006/04/19/intel_to_ship_core_inq3_2006/

E4000s are the very Cheap Conroes with 2MB instead of 4MB L2 and 800MHz FSB instead of 1066MHz FSB. Slowest speed listed was 1.6GHz for $150. IMHO, 666 isn't the only reason AMD moved theirs up two weeks.

So would the E4000's be the "Durons" of the Conroe line? E.g. nerfed cache and bus, but wholly the same (hopefully) OC-ability? Though for $150, it's not quite Duron pricing yet =D.

 

[$BangforThe$]

Thanks man . Thats the first I have heard of that . Even tho It was exspected by a few here. I wonder If Intel is willing to move the time frame ahead further.

 

[coldpower27]

So would the E4000's be the "Durons" of the Conroe line? E.g. nerfed cache and bus, but wholly the same (hopefully) OC-ability? Though for $150, it's not quite Duron pricing yet =D.

Probably the Pentium D 805 or Pentium D 925 of the Core Architecture line...

A ture Celeron edition by the name of Millville is in the works I beleive though it's Single Core.

 

[(cf)Eclipse]

Dewd, K7 was long gone and didn't really last that long August 99 and K8 launched about April or May 2000. Netburst launched over a year later than K7 in Nov 2000. Remember, K8 was just the first 18um with Full speed L2 TBirds. K7 was just the Classic Athlons with 333MHz L2 that was on chip but not on die. The faster the chip ran, the more the cache hurt it.

no, you're off by a number. you're thinking K6 and K7.

K8 was launched with the first a64's and opterons.

 

[Mav451]

Yeah, I remember K8 = 2003, solely b/c the launch of the PowerPC G5 was during the same summer.

 

[xcetera]

It's likely the Celeron or value priced processor will be the last of the Presslers. If Intel can launch Conroe earlier then they must be cranking up saleable processors as we speak. Unless they pull a paper launch, it's going to be a great for us consumers.

 

[Mav451]

Probably the Pentium D 805 or Pentium D 925 of the Core Architecture line...

A ture Celeron edition by the name of Millville is in the works I beleive though it's Single Core.

Der...Celeron != Duron. When I say Duron, I mean a highly capable chip that actually competes with its higher priced cousins.

http://www.xbitlabs.com/articles/cpu/display/celeron-d_9.html

The Celeron-D, even with it's improvements takes more clocks to equal the same performance (seems like 3.8Ghz Celeron D similar to 3.0ish Prescott)

 

[GilmourD]

no, you're off by a number. you're thinking K6 and K7.

K8 was launched with the first a64's and opterons.

K75 would be the code for the later Athlons with onboard L2.

I love when people are wrong with specificity.

 

[(cf)Eclipse]

K75 would be the code for the later Athlons with onboard L2.

i learn something new every day

while we're talking about amd's core names though, i'm sad that K9 won't be making it... though it'd probably be a dog of a chip

 

[$BangforThe$]

I came back to this thread believing that there may be some real conroe news updates, wtf is my problem? Can this get any more OT? Is it a crime to ask for a real conroe thread, and have these retired one put to rest to make way for em?

Good threads don't just die they smolder this one's still burning

 

[coldpower27]

Yeah K8 was anything Opteron or Athlon 64 technology...
K7 went form the original 0.25 micron Pluto SlotA Athlon to the final 0.13 micron
Barton Socket A Athlon on Desktop.

With the first K8 technology derivative intorduced with the Opteron Sledgehammer 0.13 micron SOI core.. talking about desktop it was Clawhammer 0.13 micron SOI in September 2003.

P7/NetBurst goes from Willamette on 0.18 micron to 65nm Presler core and hopefully ends here for the desktop at least.. we also have the Foster core on 0.18 micron to the final Tulsa 65nm core... and hopefully it ends here as well. 4 Process nodes is pretty damn long for an architecture.

If past architecture is to be used as a starting point K8 technology should end on the 65nm node with the Dual Core Brisbane and we should start K8L technology sometime in the H2 2007 with possible Quad Core derivatives. Though who knows AMD may makes some K8 tech based Quad Cores?

P6 would be a tough one to categorize, it debuted for servers back on the 0.6 micron node, and ended it's life in servers with the 0.18 micron Cascades Pentium III Xeon core, but it didn't get down onto the desktop till the Pentium II
Klamath core on the 0.35 micron node and ended it's life on the 0.13 micron node as Pentium 3 S Tulatin... which is quite good for a desktop processor 4 process node generations..

PM is 3 nodes from the original Banias core on 0.13 micron to the final Yonah core on 65nm node... which is pretty average but makes sense since Core is replacing it.

So going by Intel, expect Core Architecture as we know it now to last to at least to the 32nm node... and if it scales well by then still to the 22nm node as well.

 

[PGHammer]

Wow, talk about opening up a can of worms.

Without "Asians", you don't have nVidia. You don't have Yahoo! You don't have ASUS, Abit, DFI, Leadtek, et al

You can pretty much kiss the PC industry dead without Asians.

Your spelling leaves much to be desired ...are you even American? You sound like a hard-core, right-wing 'America!' kind of guy, yet you don't spell demonstrate or property correctly.

RAMBUS pretty much did *themselves* in with their attempted price-gouging. Not only did the Asian memory manufacturers (especially Samsung and Infineon) resist in droves, Micron (based in the far-off province of *Nampa, Idaho*) also wanted nothing to do with RDRAM and the ridiculous licensing fees that Rambus wanted for it, Intel or no Intel.
(At this point in history, Micron still made PCs in addition to being the largest US-based DRAM manufacturer extant, a title it still holds.) Intel itself had a backstop (the infamous i820SD, a variation of the previously RDRAM-only i820). Because of the resistance from memory-manufacturers (not just over licensing fees, but over Rambus' strong-arm tactics) and consumers (over the extremely high prices being charged for RDRAM compared to still-viable SDRAM), RDRAM was an even larger flop than the Intel i820SD's MTH problem. There were mutterings that the next two chipsets on Intel's roadmap (Allendale and Canterwood) wouldn't even support RDRAM. (Said rumors were proven dead-on with the release of the 865PE (Allendale) and 875P (Canterwood), which supported DDR SDRAM, not RDRAM.) Worse (for Rambus and RDRAM) DDR SDRAM pretty much punctured the balloon of bandwidth advantage that RDRAM claimed by using a pair of idea that Rambus couldn't patent (because Intel owned the patents): dual-channel memory architecture and Performance Advancement Technology (both apply to the communication between chipset and memory, and, as it turned out, both were critical to explaining away the RDRAM peformance advantage). Lastly, DDR SDRAM was *affordable* (which could never be said about RDRAM).

 

[$BangforThe$]

Intel is saying in articles they exspect terehertz trans. @ 45nm and 3D gates @ 32nm. So the improvements made at thoses nodes should be considerable.

 

[(cf)Eclipse]

Intel is saying in articles they exspect terehertz trans. @ 45nm

one transistor is nothing though.. i'd be surprised if we saw more than a ~25-35% gain in clock speed from 65nm to 45nm

 

[coldpower27]

Der...Celeron != Duron. When I say Duron, I mean a highly capable chip that actually competes with its higher priced cousins.

http://www.xbitlabs.com/articles/cpu/display/celeron-d_9.html

The Celeron-D, even with it's improvements takes more clocks to equal the same performance (seems like 3.8Ghz Celeron D similar to 3.0ish Prescott)

ok, thanks for not being exactly specific on that, AMD was just plain stupid in that area to create a chip that requires only a mild overclock to reach the performance of it's full fledged brother hence why Intel goes to such great lengths to castrate the Celerons. It doens't want people buying Celerons overclocking em and getting Pentium level performance

I have read the Xbit article on Celeron performance...

Though Core E4200 does look tempting from the point that you could get the highest Allendale performance from it assuming it is a Dual Core in the first place.. if you get it to 1066FSB. Though it's castrated enough to ensure you would not match a Conroe unless you can get it to clcok 50% higher then it already is and then some more to compensate for the additional LV2 cache.

 

[coldpower27]

RAMBUS pretty much did *themselves* in with their attempted price-gouging. Not only did the Asian memory manufacturers (especially Samsung and Infineon) resist in droves, Micron (based in the far-off province of *Nampa, Idaho*) also wanted nothing to do with RDRAM and the ridiculous licensing fees that Rambus wanted for it, Intel or no Intel.
(At this point in history, Micron still made PCs in addition to being the largest US-based DRAM manufacturer extant, a title it still holds.) Intel itself had a backstop (the infamous i820SD, a variation of the previously RDRAM-only i820). Because of the resistance from memory-manufacturers (not just over licensing fees, but over Rambus' strong-arm tactics) and consumers (over the extremely high prices being charged for RDRAM compared to still-viable SDRAM), RDRAM was an even larger flop than the Intel i820SD's MTH problem. There were mutterings that the next two chipsets on Intel's roadmap (Allendale and Canterwood) wouldn't even support RDRAM. (Said rumors were proven dead-on with the release of the 865PE (Allendale) and 875P (Canterwood), which supported DDR SDRAM, not RDRAM.) Worse (for Rambus and RDRAM) DDR SDRAM pretty much punctured the balloon of bandwidth advantage that RDRAM claimed by using a pair of idea that Rambus couldn't patent (because Intel owned the patents): dual-channel memory architecture and Performance Advancement Technology (both apply to the communication between chipset and memory, and, as it turned out, both were critical to explaining away the RDRAM peformance advantage). Lastly, DDR SDRAM was *affordable* (which could never be said about RDRAM).

Just to nit pick i865 Series Chipset are Springdale not Allendale, Allendale is the codename for the 2Mb cache Dual Core of Core Architecture basically a Conroe with 1/2 LV2 Cache.

 

[$BangforThe$]

one transistor is nothing though.. i'd be surprised if we saw more than a ~25-35% gain in clock speed from 65nm to 45nm

I think that terahertz trans with High K gates is going to be more than a ghz gain . Effiency gain . Also performance gains based, On gate speed.

 

[DejaWiz]

ok, thanks for not being exactly specific on that, AMD was just plain stupid in that area to create a chip that requires only a mild overclock to reach the performance of it's full fledged brother hence why Intel goes to such great lengths to castrate the Celerons. It doens't want people buying Celerons overclocking em and getting Pentium level performance

Granted, however....

Users that oc their cpu's only make up a very small fraction of overall processors in use. And those that oc the value cpu's is a small percentage of that. Crippling a cpu so that it can never perform like it's big brothers even when oc'd or simply slowing it down while offering less cache are both good tactics to make a value line, imo. The average Dell or HP users aren't going to (and can't) oc their cpu's.

 

[(cf)Eclipse]

ok, thanks for not being exactly specific on that, AMD was just plain stupid in that area to create a chip that requires only a mild overclock to reach the performance of it's full fledged brother hence why Intel goes to such great lengths to castrate the Celerons. It doens't want people buying Celerons overclocking em and getting Pentium level performance

really? think about what you said again. if it's so easy to make up for the performance difference it implies the performance difference is small. i'm sure we've all seen benches of semprons vs celerons...

 

[coldpower27]

Granted, however....

Users that oc their cpu's only make up a very small fraction of overall processors in use. Crippling a cpu so that it can never perform like it big brothers even when oc'd or simply slowing it down while offering less cache are both good tactics to make a value line, imo. The average Dell or HP users aren't going to (and can't) oc their cpu's.

Yes, I agree, with the use of intel motherboards that are the mainstream or value series there aren't any overclocking options in thr BIOS's to begin with if they strictly use the intel motherboards.

 

[PGHammer]

I laughed at Netburst from day one. I was a system builder at that time and suggested that people stick with either the last couple PIIIs that came out or move to the Athlon T-birds. One thing to note... At that period in time, I had a long-in-the-tooth PII and my brother had an OC'd Celeron.

I steered folks AWAY from Classic Athlons due to chipset issues (especially with Creative sound cards). I was still personally driving a two-year-old P-II 350 (because of a weird pricing quirk, it cost *less* than the P-II 333 I was originally considering, so I jumped on it; good thing I had already picked up a 440BX-based board for it as future-proofing). I bought my first Northwood (P4 2.4 which is currently sitting on a shelf) primarily *because* of the one strength it had over the laptop-targeted M: encoding performance (primarily audio encoding, not video). The system it replaced (a VERY odd duck for me, because I broke several of my own rules acquiring it: it was a *workstation* motherboard, it had features I wouldn't use, and it had a non-Intel chipset; the BIQ was the ABIT VP6, which replaced a TYAN Trinity 400 that barely lasted a year) I held onto the VP6 for three years (then passed it to my Mom, who used it for another year; still capable, it's on the shelf, preparing for a third life as an e-mail/IM client for visiting family members).

 

[DejaWiz]

Yes, I agree, with the use of intel motherboards that are the mainstream or value series there aren't any overclocking options in thr BIOS's to begin with if they strictly use the intel motherboards.

Doesn't have to be strictly intel motherboards. For example, HP uses custom BIOS images for all their systems which cripple the BIOS options severly. This is both for intel and AMD based systems. And the motherboards come from a variety of manufacturers, but mostly Foxconn, Asus, and Gigabyte.

 

[coldpower27]

really? think about what you said again. if it's so easy to make up for the performance difference it implies the performance difference is small. i'm sure we've all seen benches of semprons vs celerons...

There wasn't much of a difference between the Durons and the Athlon's that I ackowledge at the same clockspeed mind you.

Semprons which one? the K8 or K7 based ones? The Celeron D's were quite even agains the K7 based Semprons but the k8 based Semprons have since pulled ahead.

Again the difference bwtween the Athlon 64 and Sempron (K8) isn't very large again... though this time at least they castrated it to the aging Socket 754 platform...

a Sempron 3100+ is only barely slower at best compared to the Athlon 64 2800+.there just isn't a very large difference at all.

Same thing with Sempron 3400+ vs Athlon 64 3200+ S939 mild difference due to Dual Channel and additonal cache. if you can clock your Smepron to 2.5GHZ you have easily entered Athlon 64 territory...

Yes the Sempron K8 tends to be faster due to less crippling methods, but Intel Celeron can be used on i945 Chipset with possibility of upgrading to Dual Core, AMD's Semprons cannot, and no I don't really count the S939 Semprons only available to certain OEMs.

 

[Donnie27]

no, you're off by a number. you're thinking K6 and K7.

K8 was launched with the first a64's and opterons.

I should have said K7 then K75, not I didn't mean K6.

http://en.wikipedia.org/wiki/Athlon

 

[coldpower27]

Doesn't have to be strictly intel motherboards. For example, HP uses custom BIOS images for all their systems which cripple the BIOS options severly. This is both for intel and AMD based systems. And the motherboards come from a variety of manufacturers, but mostly Foxconn, Asus, and Gigabyte.

So yes additonal safety measures are placed to provide another layer of protection against overclocking in additon to Intel crippiling it's processor to the degree it has? Your point???

 

[Donnie27]

Oh please...I hate it when everyone bashes netburst becuase here, at the end of its life, it has a hard time competing. Did everyone really forget that it's been around since november of 2000? Or that all the way up until 2004 when prescott and the Athlon 64 came around it was undisputed king of the hill? Or that it ran all the way from 1.3ghz to 3.8ghz, all on the same architecture, and can hit over 5ghz with good cooling?

For the majority of netburst's life (with northwood) it was the best cpu out there...yes, the XP's were great little chips (and I ran mostly XP's, not northwoods for cost reasons) but P4's were faster in pretty much everything. Go dig up some old XP vs. P4 articles and take a look for yourself.

Is netburst in need of replacement? Yes. Was it a bad architecture? No, of course not!

Here, here, WFS!