An early look at Phenom dual / tri core performance

[elbert]

Easy?! Yeah, just slap two 100W 280mm^2 dice in a MCM and call it a day... no, sorry, it doesn't work like that in the real world.

65nm runs FAR too hot, and the die is FAR too big, to support octo core in any form. MCM is also more difficult to implement on an IMC based architecture, and do you honestly think AMD will trumpet MCM just 6 months after hyping up the advantages of a 'true native' quad core design over MCM?!

It just ain't gonna happen.

AMD should focus their R&D on getting 65nm QUADS right, as well as 45nm design&research. NOT octocore in 6 - 9 months, that is just an insane and doomed strategy, and no offence Elbert, but I'm glad you're not at the helm of AMD right now. Hector isn't the greatest but even he wouldn't try something as ludicrous and risky as this.

PS. Here is a funny and well known image with good ol' Henri Richard, but it also shows how big the K10 die is.

http://www.geckoandfly.com/wp-content/uploads/2007/05/amd_barcelona_quad_core.jpg

Now, you tell me how the feck AMD is going to fit two of those on the PCB?! LOL snap out of it man, back to reality! 😗

Wow AMD is going to have no revisions, size change, or 45nm by summer. Can I quote you on that? So Kentsfield was an insane and doomed strategy? Please thank before you type as it more or less Intel's strategy for C2D and kentsfield. I guess you only like it when AMD is only trying to play catch up. Thats a poor mans game and AMD jumping ahead to a full quad doing what Intel said was impossible was doomed but its here. This may depend on the move to 45nm and its a shame I have to point this out to you.

AMD would be insane to not try and, get a leg up on Intel to, return to high profits.

P.S. Did you note the size of the 2 PCB's aren't the same. PCB can be made to size and as long as it will fit inside the holder for the heatsink which is a big area. Nearly the lenth of this finger from the lockdown arm to the Heatsink lockdown. This is a price issue at best.

 

[epsilon84]

Revisions don't bring down the die size.

Just tell me, when do you think AMD will have 45nm ready? Are you honestly expecting a 9 month gap between nodes? Barcelona was launched only in September, and you expect a 45nm octo core by mid 08? Yes, that is a doomed strategy alright. Stick to quad, and get it right!

For comparison, A64 X2 was launched during August 2005 at 90nm. It took until December 2006 for 65nm to debut, a gap of 16 months. You are expecting AMD to cut this down to around 9 months, AND have a perfectly working octo core at that?! Again, sorry mate, doomed strategy.

PS. Such a chip, should it exist, would have had to been taped out LONG AGO if it was on the roadmap. Nehalem, which is due in Q4 08, was demonstrated fully working at IDF during September, or roughly 12 months out from release. AMD has yet to even provide a WORKING demonstration of a 45nm chip QUAD, let alone a 45nm Octo core.

You're off in dreamland buddy, but keep living it up, if it keeps your hopes of AMD regaining the crown alive.

 

[yomamafor1]

I quoted "I imagine that if the L3 bug IS affecting general perf (which I think it is) we will see at least 20% and maybe 30%." There is nothing random about basing this on the excellent David Kanter K10 analysis over at RealWorld tech.

And what data are you using to back your claim?

Anand says it's around 15% faster than K8. Some things like HL2 show almost 30%. The fact that a 3DMark run crashed says MAYBE the L3 TLB errata is limiting performance.

How is BSOD affect performance? How is the fact that Phenom is unable to run a specific benchmark, affected its general performance?

Anand has a new review up with Penryn and Barcelona. There are a "variety of workloads that show ~40% over even Penryn." And that's with B2 NOT B3.

Variety of workloads? Ahh... You mean Sandra Memory Bandwidth benchmark? I'm not even sure if Barcelona can win in latency test. 40% even Penryn? Have you lost your mind?

In Anand's words,

This, together with our own benchmark data, makes us believe that the Xeon 54xx based on the Penryn architecture will beat the best AMD chips on every aspect of raw processing performance: integer, legacy x87 FP, and SIMD (SSE).

Also FYI,
http://www.anandtech.com/IT/showdoc.aspx?i=3162&p=12

Please show me a single benchmark where Barcelona 2.5Ghz is able to gain 40% over Clovertown?

 

[elbert]

Revisions don't bring down the die size.

Just tell me, when do you think AMD will have 45nm ready? Are you honestly expecting a 9 month gap between nodes? Barcelona was launched only in September, and you expect a 45nm octo core by mid 08? Yes, that is a doomed strategy alright. Stick to quad, and get it right!

PS. Such a chip would have had to been taped out LONG AGO if it was on the roadmap. Nehalem, which is due in Q4 08, was demonstrated fully working at IDF during September, or roughly 12 months out from release. AMD has yet to even provide a WORKING demonstration of a 45nm chip QUAD, let alone a 45nm Octo core.

You're off in dreamland buddy, but keep living it up, if it keeps your hopes of AMD regaining the crown alive.

Think the last date around was may for AMD's 45nm fab. How did Intel tap out Kentsfield? LOL I think you have your wires crossed as Nehalem is a full quad and has nothing to do with how Intel did the Kentsfield. I think your just not that up on tech as you bring up something totally outside the kentsfield being 2 C2D issue.

P.S. So the G0 revision couldn't be smaller and OC the same as the B3? LOL While revisions dont bring down the die size it would allow for the smaller size with about the same heat. Why must I explain this to you? Its all tied together. You should know the lower the heat the smaller the size can be or higher the clock.

 

[epsilon84]

Think the last date around was may for AMD's 45nm fab. How did Intel tap out Kentsfield? LOL I think you have your wires crossed as Nehalem is a full quad and has nothing to do with how Intel did the Kentsfield. I think your just not that up on tech as you bring up something totally outside the kentsfield being 2 C2D issue.

No, I think you just hoping for miracles from AMD that simply aren't going to happen. Just like your hoping tri core will get 'optimizations' that would bring it almost up to par with a quad. :lol:

Again, let me state, as of yet, AMD has shown NO working sample of 45nm. No tape out, nothing.

To make a working MCM Octocore, you need a working QUAD core first. Has AMD shown a working 45nm chip? No, they haven't. Even a task manager PR event would be better than nothing. This indicates 45nm is still some way off, but you keep on living the dream of a octo core Phenom by mid 08! LOL

Again, hey, if it keeps your hopes of an AMD comeback alive, more power to ya. I just think you'll be bitterly disappointed come summer.

How about we make a friendly bet?

I bet AMD will not release 45nm until Q4 2008 (if not later), and it will be quad core only.

You make your prediction, and we'll see how right (or wrong) we are this time next year. 😉

 

[yomamafor1]

Think the last date around was may for AMD's 45nm fab. How did Intel tap out Kentsfield? LOL I think you have your wires crossed as Nehalem is a full quad and has nothing to do with how Intel did the Kentsfield. I think your just not that up on tech as you bring up something totally outside the kentsfield being 2 C2D issue.

Here is the problem to AMD's 45nm process.

Unless IBM gives AMD the necessary technology, AMD is a sitting duck. So far IBM has not demonstrated 45nm functional logic circuits. During Q3 quarter earning conference call, Dirk (or was it Hector) said they're already testing their 45nm process, and will ramp 45nm node in H108.

Now, from those who has true experience in fabrication plant, many agreed that 45nm production in H208 is a wishful thinking. Fab 30 was shut down not long ago, and is going through some major overhaul, presuming 45nm tool sets. It will take at least a year for them to move in the equipment, validate it, test production, increase yield, before they can actually use it for production. Assuming they have the financial ability for overhaul now, it will be 2009 before AMD can actually start mass 45nm production.

Knowing AMD, they will probably produce a handful of 45nm ES chips, and sell it on the market to maintain their credibility.

Another obstacle lies in IBM's 45nm process. For some very very odd reason, IBM LOVES SOI. But SOI is killing AMD on 65nm, and the problem will only exacerbate on 45nm. Without HK/MG, implementing 45nm with SOI will be extremely difficult.

I'm willing to bet, that it will be 2009 before we see any AMD 45nm chip in real volume.

P.S. PM sent.

 

[elbert]

No, I think you just hoping for miracles from AMD that simply aren't going to happen. Just like your hoping tri core will get 'optimizations' that would bring it almost up to par with a quad. :lol:

Again, let me state, as of yet, AMD has shown NO working sample of 45nm. No tape out, nothing.

To make a working MCM Octocore, you need a working QUAD core first. Has AMD shown a working 45nm chip? No, they haven't. Even a task manager PR event would be better than nothing.

Again, keep dreaming, if it keeps your hopes of an AMD comeback alive.

LOL I'm hoping optimizing would bring the tri core up to the quad? I only suggested performance against the C2D. AMD has had a working quad has they launched back in september. Nothing stop them from doing R&D with the 65nm.

Again I will, keep dreaming, as 4X4 in January will make give some major twists and with out question if its a spider 4X4.

 

[epsilon84]

LOL I'm hoping optimizing would bring the tri core up to the quad? I only suggested performance against the C2D. AMD has had a working quad has they launched back in september. Nothing stop them from doing R&D with the 65nm.

Again I will, keep dreaming, as 4X4 in January will make give some major twists and with out question if its a spider 4X4.

Tri core is only 20% behind quad core (better efficiency per core), and you expect further 'optimizations'. Yes, that is wishful thinking. It doesn't matter if you suggested performance against C2D, gaining against C2D is still gaining against Phenom X4. 😉

I edited one of my earlier posts, to show the timeframe between dual core 90nm and 65nm for K8. It was 16 months. The timeframe between SINGLE core 90nm and DUAL core 65nm was 20 months. You are suggesting AMD pulling off the same feat (doubling in cores + node shrink) in a matter of 9 months.

If AMD keeps the same pace as their previous shrinks, it won't be Q1 2009 before they launch 45nm. However, indications are they want to speed up the ramp of 45nm, so I predicted a Q4 2008 launch (probably in limited quantities), and QUAD CORE ONLY (at least initially). This is assuming there are no delays with AMD, and we all know how AMD is with delays, but I'm willing to give them the benefit of the doubt with the company on the line.

Finally, good luck with your Spider platform purchase, I hope it doesn't disappoint. I strongly suggest waiting until B3 stepping Phenoms are released though, and V2 Spider boards with the SB700 instead of the dated SB600, but hey, if you wanna jump right in during January, no one is stopping ya, I was never a fan of adopting brand new chipsets, I tend to give it a few months to work the kinks out with never revisions.

PS. 4x4 Spider? That's the first I've heard of it, you got any more details? Sources, links, etc?

 

[elbert]

Here is the problem to AMD's 45nm process.

Unless IBM gives AMD the necessary technology, AMD is a sitting duck. So far IBM has not demonstrated 45nm functional logic circuits. During Q3 quarter earning conference call, Dirk (or was it Hector) said they're already testing their 45nm process, and will ramp 45nm node in H108.

Now, from those who has true experience in fabrication plant, many agreed that 45nm production in H208 is a wishful thinking. Fab 30 was shut down not long ago, and is going through some major overhaul, presuming 45nm tool sets. It will take at least a year for them to move in the equipment, validate it, test production, increase yield, before they can actually use it for production. Assuming they have the financial ability for overhaul now, it will be 2009 before AMD can actually start mass 45nm production.

Knowing AMD, they will probably produce a handful of 45nm ES chips, and sell it on the market to maintain their credibility.

Another obstacle lies in IBM's 45nm process. For some very very odd reason, IBM LOVES SOI. But SOI is killing AMD on 65nm, and the problem will only exacerbate on 45nm. Without HK/MG, implementing 45nm with SOI will be extremely difficult.

I'm willing to bet, that it will be 2009 before we see any AMD 45nm chip in real volume.

P.S. PM sent.

Without any volume until 2009 would have AMD at octal while Nahlem is just getting volume. This is side stepping the agreement AMD has with IBM to produce 45nm AMD CPU's. This means no overhaul is needed to see 45nm from AMD by summer. Granted not major volume but not a problem for high cost CPU's like an octal.
http://www-01.ibm.com/chips/techlib/techlib.nsf/techdocs/C7973408D5AE2181002572E90079A79E
http://arstechnica.com/news.ars/post/20070130-8732.html
http://www.extremetech.com/article2/0,1697,2071433,00.asp

 

[BaronMatrix]

Really Baron? No one, and I mean NO ONE that I know of make that claim, only YOU. Even those who has extensive knowledge and experience in the field never claimed such number, only YOU

I guess we're all biased rabid Intel fanboy. 😗

That's highly likely as Intel has 80% of the market.

 

[epsilon84]

Without any volume until 2009 would have AMD at octal while Nahlem is just getting volume. This is side stepping the agreement AMD has with IBM to produce 45nm AMD CPU's. This means no overhaul is needed to see 45nm from AMD by summer. Granted not major volume but not a problem for high cost CPU's like an octal.
http://www-01.ibm.com/chips/techlib/techlib.nsf/techdocs/C7973408D5AE2181002572E90079A79E
http://arstechnica.com/news.ars/post/20070130-8732.html
http://www.extremetech.com/article2/0,1697,2071433,00.asp

http://techreport.com/discussions.x/13232

Intel's Nehalem to feature eight cores, 16 threads

The design is now complete, and Otellini claimed it's on track for delivery in the second half of 2008. In fact, he displayed a wafer of Nehalem chips and reported that each chip will be comprised of approximately 731 million transistors.

In its "largest configuration," Nehalem will pack eight CPU cores onto a single die. Each of those cores will present the system with two logical processors and be able to execute two threads via simultaneous multithreading (SMT)—a la HyperThreading. So a single Nehalem chip will be able to execute 16 threads at once.

I don't see AMD coming back into the game until Bulldozer. Phenom is dead and buried, at 45nm it may end up competitive with Core 2 if they can get the clockspeeds up, but I don't see a 45nm 8 core MCM Phenom competitive with a native 8 core Nehalem. Unless Intel pulls off another Netburst, of course.

Again, sorry to crush your dreams of an AMD comeback, it just won't happen, in the near future anyway. 😉

 

[yomamafor1]

That's highly likely as Intel has 80% of the market.

Huh? I fail to see the connection between that and YOU making outrageous claim with no data.

Hold on, let me find my Sharidouchlish translator. :lol: :lol:

 

[elbert]

Tri core is only 20% behind quad core (better efficiency per core), and you expect further 'optimizations'. Yes, that is wishful thinking. It doesn't matter if you suggested performance against C2D, gaining against C2D is still gaining against Phenom X4. 😉

I edited one of my earlier posts, to show the timeframe between dual core 90nm and 65nm for K8. It was 16 months. The timeframe between SINGLE core 90nm and DUAL core 65nm was 20 months. You are suggesting AMD pulling off the same feat (doubling in cores + node shrink) in a matter of 9 months.

If AMD keeps the same pace as their previous shrinks, it won't be Q1 2009 before they launch 45nm. However, indications are they want to speed up the ramp of 45nm, so I predicted a Q4 2008 launch (probably in limited quantities), and QUAD CORE ONLY. This is assuming there are no delays with AMD, and we all know how AMD is with delays, but I'm willing to give them the benefit of the doubt with the company on the line.

Finally, good luck with your Spider platform purchase, I hope it doesn't disappoint. I strongly suggest waiting until B3 stepping Phenoms are released though, and V2 Spider boards with the SB700 instead of the dated SB600, but hey, if you wanna jump right in during January, no one is stopping ya, I was never a fan of adopting brand new chipsets, I tend to give it a few months to work the kinks out with never revisions.

PS. 4x4 Spider? That's the first I've heard of it, you got any more details? Sources, links, etc?

Yes as even the quads need further optimization. Please stop incorrectly twisting my words and make up a lame excuss just admit you was wrong.

I think I am suggesting both a (doubling in cores + node shrink) but not both has to be AMD. I guess the FX not being released until January pretty much forces me to wait till b3. If you check the FX phenom 1207+ I think you will find some information on atleast the 4X4 and I don't think The 790FX is just for AM2+ as AMD will need a new chipset for 1207+.

 

[epsilon84]

Yes as even the quads need further optimization. Please stop incorrectly twisting my words and make up a lame excuss just admit you was wrong.

I think I am suggesting both a (doubling in cores + node shrink) but not both has to be AMD. I guess the FX not being released until January pretty much forces me to wait till b3. If you check the FX phenom 1207+ I think you will find some information on atleast the 4X4 and I don't think The 790FX is just for AM2+ as AMD will need a new chipset for 1207+.

Oh, so now you're saying both tri core AND quad core need further optimizing! :lol:

I am not twisting your words, I am just frankly finding it amusing how you think programmers will specifically 'optimize' for tri core when most massively multithreaded software (and I'm NOT talking 3DMark, I'm talking real applications) just spawns the appropriate amount of threads based on the core count. This is clear to see when you look at the scaling between 2 -> 3 -> 4 cores in examples such as rendering or encoding.

In regards to Spider, I honestly haven't heard anything of a dual socket version, but I do recall a scraped project codenamed 'FASN8' which would have been what you are describing.

 

[elbert]

http://techreport.com/discussions.x/13232

Intel's Nehalem to feature eight cores, 16 threads

The design is now complete, and Otellini claimed it's on track for delivery in the second half of 2008. In fact, he displayed a wafer of Nehalem chips and reported that each chip will be comprised of approximately 731 million transistors.

In its "largest configuration," Nehalem will pack eight CPU cores onto a single die. Each of those cores will present the system with two logical processors and be able to execute two threads via simultaneous multithreading (SMT)—a la HyperThreading. So a single Nehalem chip will be able to execute 16 threads at once.

I don't see AMD coming back into the game until Bulldozer. Phenom is dead and buried, at 45nm it may end up competitive with Core 2 if they can get the clockspeeds up, but I don't see a 45nm 8 core MCM Phenom competitive with a native 8 core Nehalem. Unless Intel pulls off another Netburst, of course.

Again, sorry to crush your dreams of an AMD comeback, it just won't happen, in the near future anyway. 😉

In its "largest configuration," Nehalem will pack eight CPU cores onto a single die. Each of those cores will present the system with two logical processors and be able to execute two threads via simultaneous multithreading (SMT)—a la HyperThreading. So a single Nehalem chip will be able to execute 16 threads at once.

We only know when Nehalem will launch not its largest configuration. We will see it as a quad only unless your wrong and a octal can be done at 45nm. See you don't like others twisting your words.

 

[epsilon84]

In its "largest configuration," Nehalem will pack eight CPU cores onto a single die. Each of those cores will present the system with two logical processors and be able to execute two threads via simultaneous multithreading (SMT)—a la HyperThreading. So a single Nehalem chip will be able to execute 16 threads at once.

We only know when Nehalem will launch not its largest configuration. We will see it as a quad only unless your wrong and a octal can be done at 45nm. See you don't like others twisting your words.

That is exactly what is says. Native octo core Nehalem at 45nm. It's probably not what you want to hear, but sometimes the truth hurts. I don't see how an MCM based Phenom X8 will come anywhere close to a native Nehalem octocore.

It's ironic isn't it, AMD was touting the advantages of their native approach so much, and yet they need MCM for octo core, whereas Intel is going native.

http://www.anandtech.com/cpuchipsets/intel/showdoc.aspx?i=3101&p=2

Intel announced that in its largest configuration, Nehalem (2H 2008, 45nm) will feature 8 cores on a single die, each core supporting 2 threads per core (welcome back Hyper Threading) for a total of 16 threads per physical chip.

Are you suggesting (or hoping) Nehalem WON'T launch as octo core, but as quad core only? What the heck are they meant to do with the 8 core version then, sit around and wait for AMD to release theirs first? :lol:

I don't think Intel is that stupid, especially when there is already a native octo core design complete and ready for production, to just sit there and let AMD one up them and be first to octo core.

Intel was first to quad core, and they sure as hell would want to be first to octo core as well, if nothing but for the PR.

 

[elbert]

Oh, so now you're saying both tri core AND quad core need further optimizing! :lol:

I am not twisting your words, I am just frankly finding it amusing how you think programmers will specifically 'optimize' for tri core when most massively multithreaded software (and I'm NOT talking 3DMark, I'm talking real applications) just spawns the appropriate amount of threads based on the core count. This is clear to see when you look at the scaling between 2 -> 3 -> 4 cores in examples such as rendering or encoding.

In regards to Spider, I honestly haven't heard anything of a dual socket version, but I do recall a scraped project codenamed 'FASN8' which would have been what you are describing.

I would find it funny to if I wasn't a professional programmer for 9 years turned college teacher for the past 8 years. Yes rendering or encoding used on 2s and 4s will but the rest will not. WinRAR, PovRay, and X265 being those on these benchmarks. I got to admit I miss jumpingjack but you come close. Thats a good thing as he was smart.

 

[epsilon84]

Clearly multithreaded programming was before your time then. 😉

Not all programs can fully load all cores at 100%, and not all programs scale at 100% with core count. No amount of 'optimizing' will get around that hard fact.

Oh, and if you miss jumpingjack, come to the 'new' forum, a lot of ex THG posters are there. 😉

 

[elbert]

That is exactly what is says. Native octo core Nehalem at 45nm. It's probably not what you want to hear, but sometimes the truth hurts. I don't see how an MCM based Phenom X8 will come anywhere close to a native Nehalem octocore.

It's ironic isn't it, AMD was touting the advantages of their native approach so much, and yet they need MCM for octo core, whereas Intel is going native.

http://www.anandtech.com/cpuchipsets/intel/showdoc.aspx?i=3101&p=2

Intel announced that in its largest configuration, Nehalem (2H 2008, 45nm) will feature 8 cores on a single die, each core supporting 2 threads per core (welcome back Hyper Threading) for a total of 16 threads per physical chip.

Are you suggesting (or hoping) Nehalem WON'T launch as octo core, but as quad core only? What the heck are they meant to do with the 8 core version then, sit around and wait for AMD to release theirs first? :lol:

I don't think Intel is that stupid, especially when there is already a native octo core design complete and ready for production, to just sit there and let AMD one up them and be first to octo core.

Intel was first to quad core, and they sure as hell would want to be first to octo core as well, if nothing but for the PR.

Now I guess its all depends on which comes to market first and with IBM producing 45nm Phenom I would guess AMD. Its kind of like saying 4 months ago I dont see how a 65nm Kentsfield can compete with a full quad 65nm phenom. Just to many unknowns. Its kind of good you got by the AMD cant do that to Intel will beat it argument. Some progress. 😉

 

[TechnologyCoordinator]

Wow AMD is going to have no size change or 45nm by summer. Can I quote you on that?

Yes. You can quote me or anyone on that.

Intel won't try to put 4 duals together. Intel will move to a native quad and then use two quads to make an Octo.

 

[epsilon84]

Now I guess its all depends on which comes to market first and with IBM producing 45nm Phenom I would guess AMD. Its kind of like saying 4 months ago I dont see how a 65nm Kentsfield can compete with a full quad 65nm phenom. Just to many unknowns. Its kind of good you got by the AMD cant do that to Intel will beat it argument. Some progress. 😉

Are you still sticking by your 'Octo core by summer 08' prediction then? :lol:

There aren't that many unknowns actually. We already know the performance of Phenom. A shrink to 45nm will not bring miraculous gains in performance. I don't see it matching Yorkfield, let alone Nehalem.

Of course, you may have different views on that, and only time will tell who is right. 😉

 

[elbert]

Clearly multithreaded programming was before your time then. 😉

Not all programs can fully load all cores at 100%, and not all programs scale at 100% with core count. No amount of 'optimizing' will get around that hard fact.

Oh, and if you miss jumpingjack, come to the 'new' forum, a lot of ex THG posters are there. 😉

Multithreading was the in thing with learning C, C++, smalltalk 80, pascal, and PL1 while in college. Visual programming I self taught visual programming while working. Not any programs can fully load all cores at 100% a realistic number would be 60 to 80%. Very few show anything like that on the benchmarks by the way.

 

[epsilon84]

Yes. You can quote me or anyone on that.

Intel won't try to put 4 duals together. Intel will move to a native quad and then use two quads to make an Octo.

Actually, Nehalem will be native for both 4 and 8 core configs.

And oh, I missed the comment about quoting. Sure, you can quote me if you want... 😉

My question to Elbert is - can we quote YOU in saying AMD will have an 8 core CPU by mid 2008? :lol:

 

[coret]

And what data are you using to back your claim?

Anand says it's around 15% faster than K8. Some things like HL2 show almost 30%. The fact that a 3DMark run crashed says MAYBE the L3 TLB errata is limiting performance.

How is BSOD affect performance? How is the fact that Phenom is unable to run a specific benchmark, affected its general performance?[/quote]

I know I have no real standing anywhere on these forums, but yomama... you're being a bit harsh on BM with this post, or part of post anyway. BM is making a fairly legitimate suggestion here.

True - what AMD has told us is that under certain circumstances (which were replicated during the 3DMark testing) a problem within the L3 cache can force an error to occur in the chip, causing a proper old school crash/freeze which we haven't really seen much of since MS figured out how to make explorer.exe commit suicide to preserve overall system stability. To make this occur, the error would probably have to quite a major one.

If this is the only error the L3 is producing then I can understand and agree with your comment. However, AMD only mentioned this one serious error due to hardcore instability issues - what if the L3 cache is also provoking other errors, which themselves are less serious (ie not infringing on system stability) but still bad enough that the CPU coughs/chokes for a nanosecond while it corrects itself? This would produce a, albeit slight, performance hit. Removing the cause of the issue in the L3 cache would then negate this problem, and the CPU would work smoothly instead.

Now, I know this has no fact behind it ... but to me it seems to be quite fair reasoning. It also makes sense that if it's fixed, and any minor errors produced are cleaned up, the performance of K10 vs. K8 could rise a couple to a few percent. Not enough a gain to challenge Intel, but then phenom could do with every ounce of performance it can get, right?

 

[TechnologyCoordinator]

Intel Pumpers! The end is here now!

Phenom is the 100% super-chip destroyer of evil Nazi Intellers! At its super-high clocking of GREATER than the 2Ghz it is so fast that evil Intel "Benchmarks" Cannot be used in the measuring of its fastitude!

Fact 1: Phenom is already the most popular CPU in all of the desktop history!! AMD has the sales of at least 10 BILLION in this month! This is obvious since Phenom comes in many different cores while Stupid Evil Intel only has the 2 cores or the "glue" in the 4 cores!!
3 Cores is better than the 4 cores unless AMD says that the 4 cores is better!

FAct 3: All Intel CPUs are the VAPORWARE. Oh, Intel says it makes some lie called "core 2" but here is truth: Those chips are only old AMD chips that the INTel STOLE from the AMD and then lied and said it made them. INTEL HAS NEVER MADE A SINGLE CHIP EVER!! IT ONLY STEALS FROM THE AMD!!!
AMD will always be the 10x generation leader because Intel can only steal the old AMD chips!

Penryn is the most delayed CPU in all of the history of the computing!! This is because the dumb Intellers could not steal the 10 year old AMD chips!!
Nehalem is a lie!!! Just compare: Phenom is the most widely used and tested chip in the history of all the world! Randy Allen has powerpoints proving Phenom is the bestest of AMD!! But Nehalem is just a stupid demo! INTEL VAPORWARE!!

Fact B: Intel is ALREADY BK!! Do not believe some lying pro-Intel pumper who says Intel will survive until 2009. Do not believe some fanboy who is saying that the 2Q 2008 BK will come true!! These are all Intel lies! It uses its illegal money to buy off the pumpers, but HECTOR knows that it is already the BK!!!

Fact VIII: All Intellers are the evil Nazis as the Sharikous proved! Intel people believe in evil ideas like "thinking" and "facts". All good PURE AMD peoples know that this is evil and dumb. Only the Hector is needed for thinking! Only things that the AMD approves are allowable!
Do not believe evil Intellers who talk about "choice" or "competition" These words are the stupid and bad. Since Intel is EVIL any choice that is not AMD IS THE EVIL! Only AMD must be the processor! Only AMD may be legal in the usage!
Choice is the Evil! AMD MUST RULE AND ALL OTHERS MUST DIE! Heil Hector!

(The Burninator 10/7/07)