Barcelona Single Thread SPEC CPU2006 scores

[IntelUser2000]

http://realworldtech.com/forum...hreadid=83478&roomid=2

Since the FP can be auto-parallelized, I'll post Intel results for both.

SPECint2006 (base/peak)
AMD 1.9GHz - 9.97/11.3 - Note 2 socket system
Intel 3GHz - 18.9/20.8 - Note 2 socket system
Intel 2GHz - 14.2/15.6 - Note 2 socket system

SPECfp2006 (base/peak)
Without Auto Parallelization:
AMD 1.9GHz - 10.7/11.2 - Note 2 socket system
Intel 2.66 GHz - 14.3/14.5 - Note 2 socket system
Intel 2.00 GHz - 10.9/xxx: - Note 2 socket system

With Auto Parallelization:
Intel 3GHz - 18.4/21.4 - Note 2 socket system
Intel 2GHz - 14.5/16.9 - Note 2 socket system

AMD doesn't have Auto Parallelization so it can't be compared with Intel's results with Auto Parallelization on. But if you compare the scores that have both off you can see that AMD isn't ahead at all, even in FP per clock.

You guys can conclude however you like, but current scores indicate that the performance of Barcelona lies just 5-10% faster than Opteron at single thread. Better scaling?? Some of the CPUs that were quoted as having better scaling were P4, and later with Prescott. What's the similarity with both?? They were both slow...

 

[myocardia]

After having read Gary Keys' article on Barcelona, I can't say that I'm surprised. Disappointed, yes, but not surprised. Too bad, too, because they had the chance to pull another rabbit out of their hat, like they did with the A64's. Hopefully Intel will at least adopt (copy) Barcelona's ability to control the speed of each core individually. That was a very good idea, if you ask me.

 

[harpoon84]

Quite disappointing, based on these results I don't expect Phenom to be competitive with Penryn on the desktop. Lower IPC + lower clockspeed != winner.

 

[undeclared]

I would say this means little or nothing.. Wait till the processor actually comes out.

 

[TuxDave]

Originally posted by: xtwells
I would say this means little or nothing.. Wait till the processor actually comes out.

And then what?

 

[Keysplayr]

Originally posted by: TuxDave

Originally posted by: xtwells
I would say this means little or nothing.. Wait till the processor actually comes out.

And then what?

Seconded! ^

 

[brxndxn]

I would love to see an actual 'server' review coming from a company like Sun Microsystems that actually has the resources to test servers. That review site hardly looks like a serious one to consider.

All I can say is, if Barcelona performance is really that bad why the hell is Barcelona getting all kinds of supercomputer wins right now? Why is Dell refocusing effort on Barcelona servers?

I am still not convinced I have seen any review of Barcelona that attempts to optimize anything for Barcelona performance. I have seen Intel-optimized benchmarks (compiled for Intel processors) tested and compared side-by-side with Barcelona - and that's all there seems to be right now.

 

[heyheybooboo]

Originally posted by: xtwells
I would say this means little or nothing.. Wait till the processor actually comes out.

While I tend to lean in the ""And then what???"" catagory I'm not quite as panty-wadded as some others. I will, however, rephrase your comment to ""Wait till the advanced platforms actually come out""

Originally posted by: brxndxn
I would love to see an actual 'server' review coming from a company like Sun Microsystems that actually has the resources to test servers. That review site hardly looks like a serious one to consider.

All I can say is, if Barcelona performance is really that bad why the hell is Barcelona getting all kinds of supercomputer wins right now? Why is Dell refocusing effort on Barcelona servers?

I am still not convinced I have seen any review of Barcelona that attempts to optimize anything for Barcelona performance. I have seen Intel-optimized benchmarks (compiled for Intel processors) tested and compared side-by-side with Barcelona - and that's all there seems to be right now.

I agree. I don't want to start a flame war but this bench however it was done was clearly **cherry-picked**. I'll concede the power/performance crown to Intel but ~~sigh~~ why do we have to put up with this FUD.

Published SPEC CPU2006 Results from the Standard Performance Evaluation Corporation

 

[Gikaseixas]

here we go again
Can you guys relax and wait till Phenom comes out so we can see for ourselves the real potential of the chip? All these small tests mean nothing to me. I'm sure AMD will not demolish Intel nor will Intel walk over AMD in every test. It's going to be an interesting fight. In the end we will only benefict from that war...

 

[IntelUser2000]

I would love to see an actual 'server' review coming from a company like Sun Microsystems that actually has the resources to test servers. That review site hardly looks like a serious one to consider.

All I can say is, if Barcelona performance is really that bad why the hell is Barcelona getting all kinds of supercomputer wins right now? Why is Dell refocusing effort on Barcelona servers?

I am still not convinced I have seen any review of Barcelona that attempts to optimize anything for Barcelona performance. I have seen Intel-optimized benchmarks (compiled for Intel processors) tested and compared side-by-side with Barcelona - and that's all there seems to be right now.

We already have pretty good idea how Barcelona will perform in server, you see how it has excellent scaling with multiple threads.

However, read the title again. It's named: Barcelona Single Thread SPEC CPU2006 scores. This is very important in PC, aka Desktops. Server guys like SUN are gonna snap up Barcelonas. I don't see that happening in PC.

-All I can say is, if Barcelona performance is really that bad why the hell is Barcelona getting all kinds of supercomputer wins right now?
-Why is Dell refocusing effort on Barcelona servers?

Honestly, those two questions are not really related to this thread. Most consumers use PC. We can see that Barcelona at 2.5GHz is more than competitive with the Harpertown based systems. Servers are massively threaded. Super computers!=Single thread

 

[shabby]

Originally posted by: TuxDave

Originally posted by: xtwells
I would say this means little or nothing.. Wait till the processor actually comes out.

And then what?

They'll release new drivers that will increase performance, thats what...

 

[heyheybooboo]

Originally posted by: IntelUser2000

I would love to see an actual 'server' review coming from a company like Sun Microsystems that actually has the resources to test servers. That review site hardly looks like a serious one to consider.

All I can say is, if Barcelona performance is really that bad why the hell is Barcelona getting all kinds of supercomputer wins right now? Why is Dell refocusing effort on Barcelona servers?

I am still not convinced I have seen any review of Barcelona that attempts to optimize anything for Barcelona performance. I have seen Intel-optimized benchmarks (compiled for Intel processors) tested and compared side-by-side with Barcelona - and that's all there seems to be right now.

We already have pretty good idea how Barcelona will perform in server, you see how it has excellent scaling with multiple threads.

However, read the title again. It's named: Barcelona Single Thread SPEC CPU2006 scores. This is very important in PC, aka Desktops. Server guys like SUN are gonna snap up Barcelonas. I don't see that happening in PC.

-All I can say is, if Barcelona performance is really that bad why the hell is Barcelona getting all kinds of supercomputer wins right now?
-Why is Dell refocusing effort on Barcelona servers?

Honestly, those two questions are not really related to this thread. Most consumers use PC. We can see that Barcelona at 2.5GHz is more than competitive with the Harpertown based systems. Servers are massively threaded. Super computers!=Single thread

Yo, Dude . . .

SPECint is designed to run single-core single-thread. And if you review published SPEC results from vendors on the SPEC website it shows how bizarre are the numbers you posted in this thread.

And btw, Barcelona is a server cpu - Phenom is the desktop cpu.

When you say, **This is very important in PC, aka Desktops. Server guys like SUN are gonna snap up Barcelonas. I don't see that happening in PC.** it's a non sequitur.

And another btw - Applications run single threads. A PC at any given moment is running hundreds of threads.

Have a good evening!

 

[Viditor]

Originally posted by: IntelUser2000

I would love to see an actual 'server' review coming from a company like Sun Microsystems that actually has the resources to test servers. That review site hardly looks like a serious one to consider.

All I can say is, if Barcelona performance is really that bad why the hell is Barcelona getting all kinds of supercomputer wins right now? Why is Dell refocusing effort on Barcelona servers?

I am still not convinced I have seen any review of Barcelona that attempts to optimize anything for Barcelona performance. I have seen Intel-optimized benchmarks (compiled for Intel processors) tested and compared side-by-side with Barcelona - and that's all there seems to be right now.

We already have pretty good idea how Barcelona will perform in server, you see how it has excellent scaling with multiple threads.

However, read the title again. It's named: Barcelona Single Thread SPEC CPU2006 scores. This is very important in PC, aka Desktops. Server guys like SUN are gonna snap up Barcelonas. I don't see that happening in PC.

-All I can say is, if Barcelona performance is really that bad why the hell is Barcelona getting all kinds of supercomputer wins right now?
-Why is Dell refocusing effort on Barcelona servers?

Honestly, those two questions are not really related to this thread. Most consumers use PC. We can see that Barcelona at 2.5GHz is more than competitive with the Harpertown based systems. Servers are massively threaded. Super computers!=Single thread

Firstly, thanks for posting the thread...good quality data there (though still quite preliminary).

The problem I see is that trying to extrapolate Phenom performance from this is close to impossible.

1. There are only a single set of Specint and Specfp numbers, normally this is done on several platforms (and often the results vary widely). For example, I would bet that the AMD Spec system is using a low power chip (HE or highly effecient), and they are also using a PGI compiler. As more systems are submitted, I imagine that these numbers will go quite a bit faster.

2. The differences between Phenom and the HE Barcys is more than just clockspeed...
a.) Entirely different chip revisions
b.) Different platforms
c.) HT3 and buffered/unbuffered Dimms

3. There is a HUGE difference between running a synthetic single threaded bench and running a single threaded app in the real world. For one thing, the desktop environment is always running other threads in the background at the same time (which is the main reason for multi-core in the first place).

4. As for clockspeed, I would be very surprised if we didn't have a 2.8 GHz Phenom by the end of this year. While they've already demonstrated a 3 GHz in June, that was also before they improved the steppings...JMHO

 

[harpoon84]

Viditor, how much improvement do you realistically expect to see between Barcelona and Phenom in desktop applications? I'm thinking there would only be a few percentage points difference, nothing that would enable Phenom to overtake Core2 in IPC, especially compared to Penryn.

How would HT3 make any impact on single threaded performance? Are you seriously expecting to see an appreciable difference from HT3 on the desktop?

Also, from what I have seen there is no appreciable difference in performance between AMD's HE and regular chips, the only difference is obviously TDP. If you can prove otherwise then please do.

 

[Viditor]

Originally posted by: harpoon84
Viditor, how much improvement do you realistically expect to see between Barcelona and Phenom in desktop applications? I'm thinking there would only be a few percentage points difference, nothing that would enable Phenom to overtake Core2 in IPC, especially compared to Penryn.

How would HT3 make any impact on single threaded performance? Are you seriously expecting to see an appreciable difference from HT3 on the desktop?

Also, from what I have seen there is no appreciable difference in performance between AMD's HE and regular chips, the only difference is obviously TDP. If you can prove otherwise then please do.

Jason and Ross did a comparison in July. article here

There was an 8% difference between HE and SE (standard edition) at the same clockspeed, and remember that Phenom will be the extreme edition (so another level up again).
As to HT3 on the desktop environment, I expect it to make a difference in I/O...especially as we move towards CTM based graphics. Things like AMD's upcoming "Triple-Play" should make extensive use of more bandwidth (though I agree that it's been of minimal effect in the past).

Finally, the difference in both the new rev and unbuffered memory will come into play. While I agree that each of these things is only a few percentage points, as you add them up it's a significant amount!
If you add 5% for the new stepping, 8% for the extreme core (clock for clock), and probably 6-9% for the memory and HT3 combined...that's a 19-22% improvement. I don't know about you, but that's fairly significant to me...

At the end of the day though, we are going to have to wait...as I said, making predictions about apples using oranges just doesn't work.

 

[zach0624]

another thing to consider is that penryn's IPC is not going to be much better than conroe's, in most areas 5% or less which is barely statistically significant also you can probably expect a 10% increase in IPC with phenom which from what I can remember whould bring it up to par with conroe and maybe penryn. While viditor's IPC may sound like it is overkill you will see a difference between the new stepping SE chips with unbuffered memory which should be atleast ten percent.

 

[myocardia]

Originally posted by: brxndxn
I am still not convinced I have seen any review of Barcelona that attempts to optimize anything for Barcelona performance. I have seen Intel-optimized benchmarks (compiled for Intel processors) tested and compared side-by-side with Barcelona - and that's all there seems to be right now.

So, you aren't at all interested in seeing a review using current, readily available software, just one optimized for Barcelona?

Originally posted by: Viditor

Firstly, thanks for posting the thread...good quality data there (though still quite preliminary).

Agreed.

The problem I see is that trying to extrapolate Phenom performance from this is close to impossible.

How is that, did AMD completely redesign the architecture, just for Phenom? I heard they didn't, that it's gonna be just like the Opteron/Athlon 64, and Opteron/X2 performance comparisons.

1. There are only a single set of Specint and Specfp numbers, normally this is done on several platforms (and often the results vary widely). For example, I would bet that the AMD Spec system is using a low power chip (HE or highly effecient),

You think AMD used their slower chip, to do these tests? I highly doubt that. Maybe if Intel had been the one who released these numbers, it would be different, but not AMD.

and they are also using a PGI compiler.

Isn't that the same compiler that's always been used, including when the Opterons were defeating the P4-based Xeons at the same tests?

3. There is a HUGE difference between running a synthetic single threaded bench and running a single threaded app in the real world.

Huh? I've owned a computer since 1980, and not once in that time period have I seen a single instance of any processor benchmarking slower than anther, yet outperforming it in "real life". Sometimes the difference won't be as big as the benchmark suggests, like how the C2D utterly destroys an Athlon 64 in 1m SuperPi, but the benchmarks always show which processor is faster at any specific task. I'm even willing to be that that's actually how benchmarks came to be.

For one thing, the desktop environment is always running other threads in the background at the same time (which is the main reason for multi-core in the first place).

I'm sorry to have to say so, but you should do some more research.

4. As for clockspeed, I would be very surprised if we didn't have a 2.8 GHz Phenom by the end of this year.

If you mean a dual-core, you're probably right. If you mean a quad-core, I can just about guarantee you that won't be happening. You realize that there are less than 60 days left in this year, and the fastest quad-core AMD has offered is a 2.0 Ghz part, don't you?

 

[dflynchimp]

It took Intel going from Banias to Dothan to Yonah and finally Conroe to beat AMD's A64 family
AMD's K7 to K8 move had enough momentum to last it all the way until Conroe came out. Barcelona/K10 is only a stepping stone off of which better advances could be made.

 

[Viditor]

Originally posted by: myocardia

Originally posted by: brxndxn
I am still not convinced I have seen any review of Barcelona that attempts to optimize anything for Barcelona performance. I have seen Intel-optimized benchmarks (compiled for Intel processors) tested and compared side-by-side with Barcelona - and that's all there seems to be right now.

So, you aren't at all interested in seeing a review using current, readily available software, just one optimized for Barcelona?

Originally posted by: Viditor

Firstly, thanks for posting the thread...good quality data there (though still quite preliminary).

Agreed.

The problem I see is that trying to extrapolate Phenom performance from this is close to impossible.

How is that, did AMD completely redesign the architecture, just for Phenom? I heard they didn't, that it's gonna be just like the Opteron/Athlon 64, and Opteron/X2 performance comparisons.

Do you mean is the Phenom a different uA than Barcy? No...but as I said, uA is only one part of the performance. And there will be slight differences in the uA between the 2 as well (for instance, allowing for HT3 means they need a slightly different mem controller). Will it be faster or slower? Nobody knows yet...

1. There are only a single set of Specint and Specfp numbers, normally this is done on several platforms (and often the results vary widely). For example, I would bet that the AMD Spec system is using a low power chip (HE or highly effecient),

You think AMD used their slower chip, to do these tests? I highly doubt that. Maybe if Intel had been the one who released these numbers, it would be different, but not AMD.

Yes, they did use the slower chip...but people who should actually care about these tests consider power/performance well before they care about peak performance (in fact, I would bet that peak performance isn't really considered much at all).

and they are also using a PGI compiler.

Isn't that the same compiler that's always been used, including when the Opterons were defeating the P4-based Xeons at the same tests?

Not really...they used the Pathscale and Intel compilers far more often for those comparisons. I can't remember seeing many Portland Group compiled systems at all.

3. There is a HUGE difference between running a synthetic single threaded bench and running a single threaded app in the real world.

Huh? I've owned a computer since 1980, and not once in that time period have I seen a single instance of any processor benchmarking slower than anther, yet outperforming it in "real life". Sometimes the difference won't be as big as the benchmark suggests, like how the C2D utterly destroys an Athlon 64 in 1m SuperPi, but the benchmarks always show which processor is faster at any specific task. I'm even willing to be that that's actually how benchmarks came to be.

A single threaded synthetic benchmark (like these) runs that benchmark and almost NOTHING else...are you saying that you've run benches under those circumstances? No AV, audio, limited graphics, vsmon, nothing in the tray, and your Linux OS limited to only running that bench?
If you have, did you actually consider that a real-world test?

For one thing, the desktop environment is always running other threads in the background at the same time (which is the main reason for multi-core in the first place).

I'm sorry to have to say so, but you should do some more research.

I agree...after 15 years of testing, demonstrating, and designing high-end systems for the broadcast industry, I can safely say that I will ALWAYS need to do more research. Are you saying you don't?
I noticed that you didn't say I was wrong there...did you have a point?

4. As for clockspeed, I would be very surprised if we didn't have a 2.8 GHz Phenom by the end of this year.

If you mean a dual-core, you're probably right. If you mean a quad-core, I can just about guarantee you that won't be happening. You realize that there are less than 60 days left in this year, and the fastest quad-core AMD has offered is a 2.0 Ghz part, don't you?

Hmmm...seems like you might want to read that research request yourself, mate.
Johan also reviewed the 2.5 GHz quad core preliminary part, and they demonstrated a 3GHz quad core in June...

 

[jones377]

1. IBM didn't use a low-power part for the benchmark submission. It has been submitted here for everyone to see. I don't see a BE designation anywhere. Here is another IBM submission for an SE chip, so yes they do specify. Not that it would have mattered anyway.

2. The difference in compilers, specifically between PathScale and PGI. This one surprised me. It is indeed true that most Opteron submissions have been done using the PathScale compiler. But contrary to some belief, the PGI compiler isn't slower, it's in fact faster. The Barcelona submission used the PGI compiler. Here is the data for everyone to see:

Opteron 2222 (3GHz), DDR2 667 CL5 ECC/REG

HP using PGI: 14.5/15.5 (base/peak)
Fujitsu Siemens using PathScale: 13.5/14.9 (base/peak)
IBM using PathScale:13.4/14.8 (base/peak)


3. Comparing K10 with K8. IMO this will give a much better picture of what improvements in singlethreaded performance we will see on the desktop. Unfortunately there are so few submissions for K8 using the PGI compiler so we're forced to use PathScale for the K8 comparison even though it's slower. So here it is:

K10 1.9GHz: 9.97/11.3 (base/peak), here
K8 2GHz: 9.77/10.8 (base/peak), here

So 5% lower clockspeed yeilds 2% higher score in base and 5% higher score in peak. You know I'm starting to have my doubt about this K10 submission myself, I really expected it to be faster than this, compared to K8.

 

[myocardia]

Originally posted by: Viditor
A single threaded synthetic benchmark (like these) runs that benchmark and almost NOTHING else...are you saying that you've run benches under those circumstances? No AV, audio, limited graphics, vsmon, nothing in the tray, and your Linux OS limited to only running that bench?
If you have, did you actually consider that a real-world test?

Are you saying that you believe AMD had the Barcelona doing video compression on all 4 cores in the background, while they had the Xeon running nothing else? Because, if you aren't, that's the worst retort I've ever in my life heard.

You realize that when benchmarking, they run the minimum amount of everything, including processes, on both machines? Were you aware that they do that only so the benchmarks will be the highest on both processors/systems? Oh, or were you saying that Barcelona's are magical, and the more apps you have them run at the same time, the faster they become?

For one thing, the desktop environment is always running other threads in the background at the same time (which is the main reason for multi-core in the first place).

Originally posted by: myocardia
I'm sorry to have to say so, but you should do some more research.

I agree...after 15 years of testing, demonstrating, and designing high-end systems for the broadcast industry, I can safely say that I will ALWAYS need to do more research. Are you saying you don't?
I noticed that you didn't say I was wrong there...did you have a point?

Let me reiterate my point, since you didn't get it the first time: you have absolutely no idea what you're talking about, at least concerning why AMD and Intel are moving towards multicore processors, and you should do more research. Plain enough for you?

4. As for clockspeed, I would be very surprised if we didn't have a 2.8 GHz Phenom by the end of this year.

If you mean a dual-core, you're probably right. If you mean a quad-core, I can just about guarantee you that won't be happening. You realize that there are less than 60 days left in this year, and the fastest quad-core AMD has offered is a 2.0 Ghz part, don't you?

Hmmm...seems like you might want to read that research request yourself, mate.
Johan also reviewed the 2.5 GHz quad core preliminary part,

Do you mean the one that isn't offered for sale by anyone, anywhere, and also failed in the midst of testing?

and they demonstrated a 3GHz quad core in June...

No, they didn't. They demonstrated a highly overclocked quad core, that was obviously hand picked for it's ability to overclock (like Intel's ES chips always are) in June. Or do you believe that all Core 2 Duo's easily overclock to 4 Ghz+, on air alone, like all of the ES's that Intel sent out to review sites were able to do?

 

[Phynaz]

Originally posted by: Viditor

4. As for clockspeed, I would be very surprised if we didn't have a 2.8 GHz Phenom by the end of this year. While they've already demonstrated a 3 GHz in June, that was also before they improved the steppings...JMHO

Not according to This.

 

[Viditor]

Originally posted by: myocardia

Originally posted by: Viditor
A single threaded synthetic benchmark (like these) runs that benchmark and almost NOTHING else...are you saying that you've run benches under those circumstances? No AV, audio, limited graphics, vsmon, nothing in the tray, and your Linux OS limited to only running that bench?
If you have, did you actually consider that a real-world test?

Are you saying that you believe AMD had the Barcelona doing video compression on all 4 cores in the background, while they had the Xeon running nothing else? Because, if you aren't, that's the worst retort I've ever in my life heard.

You realize that when benchmarking, they run the minimum amount of everything, including processes, on both machines? Were you aware that they do that only so the benchmarks will be the highest on both processors/systems? Oh, or were you saying that Barcelona's are magical, and the more apps you have them run at the same time, the faster they become?

I am saying that SPECint and SPECfp are not real-world desktop benchmarks and don't reflect anything to do with the desktop world...what are YOU saying?
Trying to extrapolate Phenom's performance from a single-threaded Spec bench is an effort in futility...

For one thing, the desktop environment is always running other threads in the background at the same time (which is the main reason for multi-core in the first place).

Originally posted by: myocardia
I'm sorry to have to say so, but you should do some more research.

I agree...after 15 years of testing, demonstrating, and designing high-end systems for the broadcast industry, I can safely say that I will ALWAYS need to do more research. Are you saying you don't?
I noticed that you didn't say I was wrong there...did you have a point?

Let me reiterate my point, since you didn't get it the first time: you have absolutely no idea what you're talking about, at least concerning why AMD and Intel are moving towards multicore processors, and you should do more research. Plain enough for you?

That's not really a point though...it's just an insult without anything to back it up.
You should at the very least point out where you think I was wrong...

4. As for clockspeed, I would be very surprised if we didn't have a 2.8 GHz Phenom by the end of this year.

If you mean a dual-core, you're probably right. If you mean a quad-core, I can just about guarantee you that won't be happening. You realize that there are less than 60 days left in this year, and the fastest quad-core AMD has offered is a 2.0 Ghz part, don't you?

Hmmm...seems like you might want to read that research request yourself, mate.
Johan also reviewed the 2.5 GHz quad core preliminary part,

Do you mean the one that isn't offered for sale by anyone, anywhere, and also failed in the midst of testing?

Wow...you mean all of those Penryn's are on sale now??? Where can I get one?
/sarcasm
We are talking about chips that are due out soon...do you also think that Intel will be unable to fulfill their promised Penryn shipping, or is that feeling only reserved for AMD?
Don't you think that is quite fanboyistic (can that be a word?)?

and they demonstrated a 3GHz quad core in June...

No, they didn't. They demonstrated a highly overclocked quad core, that was obviously hand picked for it's ability to overclock (like Intel's ES chips always are) in June. Or do you believe that all Core 2 Duo's easily overclock to 4 Ghz+, on air alone, like all of the ES's that Intel sent out to review sites were able to do?

Sorry, but were you there? Even if it was (and that's quite possible), it was necessarily on an older and less overclockable stepping (BA and B2 weren't turned until late in the following month).
Are you saying that AMD won't top 2 GHz? Till when (in your estimation)?

 

[Viditor]

Originally posted by: Phynaz

Originally posted by: Viditor

4. As for clockspeed, I would be very surprised if we didn't have a 2.8 GHz Phenom by the end of this year. While they've already demonstrated a 3 GHz in June, that was also before they improved the steppings...JMHO

Not according to This.

VR-Zone are the same ones that assured us that we would have 4 GHz Wolfdales and 3.73 GHz Yorkfields last quarter...VR-Zone roadmap...

I don't really see them as a credible source.

 

[IntelUser2000]

There was an 8% difference between HE and SE (standard edition) at the same clockspeed, and remember that Phenom will be the extreme edition (so another level up again).
As to HT3 on the desktop environment, I expect it to make a difference in I/O...especially as we move towards CTM based graphics. Things like AMD's upcoming "Triple-Play" should make extensive use of more bandwidth (though I agree that it's been of minimal effect in the past).

Finally, the difference in both the new rev and unbuffered memory will come into play. While I agree that each of these things is only a few percentage points, as you add them up it's a significant amount!
If you add 5% for the new stepping, 8% for the extreme core (clock for clock), and probably 6-9% for the memory and HT3 combined...that's a 19-22% improvement. I don't know about you, but that's fairly significant to me...

Since you like to do EXTREME CHERRY-PICKING, let me do mine .

http://www.anandtech.com/IT/showdoc.aspx?i=3039&p=11

That page shows that SE version is faster than HE. So??

Also:

http://www.anandtech.com/IT/showdoc.aspx?i=3039&p=9

So, FIXED:

If you add -5% for the new stepping, -8% for the extreme core (clock for clock), and probably -6-9% for the memory and HT3 combined...that's a -19-22% improvement. I don't know about you, but that's fairly significant performance deficit to me...

VR-Zone are the same ones that assured us that we would have 4 GHz Wolfdales and 3.73 GHz Yorkfields last quarter...VR-Zone roadmap...

Either you are too gullible or do not realize the fact that Intel doesn't need 3.5-4.0GHz parts to beat AMD performance wise. It looks like they won't need it for the near future either(phenom launch date included). Schedules change.