Here's a little teaser....

[Periander6]

Saw this from David Kanter over at RWT. Barc non _rate SPEC benches.

http://realworldtech.com/forums/inde...83478&roomid=2

[Mav451]

Saw this from David Kanter over at RWT. Barc non _rate SPEC benches.

http://realworldtech.com/forums/inde...83478&roomid=2

Good god, those #'s make the Barcelona look exceedingly bad.

[PhilDoc]

Saw this from David Kanter over at RWT. Barc non _rate SPEC benches.

http://realworldtech.com/forums/inde...83478&roomid=2

The link doesn't seem to be working

[JohannesRS]

worked for me.

wish it didn't. it'll make me have nightmares tonight. :P

[Periander6]

The link doesn't seem to be working

Works fine for me. If you want you can try just going to the main RWT site then going to the forums. Not hard to find Kanter's post there about it.

[zpdixon]

Beware guys, in these non-rate SPEC results, Intel actually make use of the Intel compiler auto-parallelization feature when AMD does not (AFAIK), meaning that you are comparing apples to oranges (code automatically multi-threaded by the compiler on Intel vs. mono-threaded code on AMD).

Besides, these non-rate AMD numbers have been available for about a month at [2].

It looks like Intel is working very hard at producing the best SPEC results possible. A few weeks ago the best SPECfp2006 score for a dual Xeon X5365 3.0 GHz was 16.9 [3], and just now they publish a score of 21.4 with the same CPUs [1] (which is the real news brought by this realworldtech.com post). Look at the peak score of the subtest 436.cactusADM for example: it jumped from 20.1 to 95.9 just by using the -parallel compilation flag (check it for yourself in [1] and [3]).

What Intel is doing does not violate the SPEC rules (they explicitely allow auto-parallelization), but it makes any sort of direct comparison of the performance of Opteron vs. Xeon impossible.

[1] http://www.spec.org/cpu2006/results/...903-01960.html
[2] ftp://ftp.software.ibm.com/eserver/b...cpu_091007.pdf
[3] http://www.spec.org/cpu2006/results/...820-01861.html

- Z

[hargen]

I search the non-paralle single-thread test of intel. The search results of non-auto-paralle shock me:
SPECfp2006(base/peak)
AMD 1.9GHz - 10.7/11.2
Intel 1.86GHz - 12.3/12.5 (test data: 2007.6.25)
http://www.spec.org/cpu2006/results/...625-01332.html

Intel 3GHz - 18.7/19.3 (test data: 2007.7.23)
http://www.spec.org/cpu2006/results/...723-01537.html

To surprise, the non-paralle result of SPECint2006 is much faster than the paralle result.
SPECint2006(base/peak)
Intel 3GHz 20.2/22.6 (test data: 2007.7.23)
http://www.spec.org/cpu2006/results/...723-01539.html

Which does one rule the single-thread test?

[zpdixon]

hargen: you link to SPEC results of mono-processor machines (Xeon UP 3040 and C2D E6850). They use faster unbuffered DDR2 RAM, which give them an advantage compared to the 2P results (Opteron 23xx and Xeon 53xx) we have been discussing so far (most of the SPECfp2006 tests are very memory-intensive).

- Z

[hargen]

SPECfp2006 rates tests are very memory-intensive. But SPECfp2006 tests of non-paralle are single-thread floating-point tests,and more depend on the core raw power. Xeon 3040 1.86G uses ecc ddr2 ram, but opteron uses ecc ddr2 ram too. The difference isn't that large - 2% or less?

[hargen]

I have to agree that "AMD will need to increase the frequency substantially to keep competitive with Intel on the desktop", because single-thread or two-thread performance will rule the desktop in future 20 months.

[Motiv]

I have to agree that "AMD will need to increase the frequency substantially to keep competitive with Intel on the desktop", because single-thread or two-thread performance will rule the desktop in future 20 months.

Depends on the application. For the average desktop user, Quad Core is overkill. Browsing the web, spreadsheets and word-processing don't require anything other than duel core.

For a gamer, quad core will rule the roost. All of the latest game coming out in the next 20 months will take advantage of multiple cores.

I've no doubt a 2.6-3.0ghz Phenom will be competitive with equally priced Intels but I very much doubt we will see a AMD take the performance crown for a long time to come.


Speaking of the Phenom. Are any of these chips actual Duel cores or are they all Quads but with two core disabled?

[hargen]

Depends on the application. For the average desktop user, Quad Core is overkill. Browsing the web, spreadsheets and word-processing don't require anything other than duel core.

For a gamer, quad core will rule the roost. All of the latest game coming out in the next 20 months will take advantage of multiple cores.

I've no doubt a 2.6-3.0ghz Phenom will be competitive with equally priced Intels but I very much doubt we will see a AMD take the performance crown for a long time to come.


Speaking of the Phenom. Are any of these chips actual Duel cores or are they all Quads but with two core disabled?

Do not use games to determine the performance of the desktop. game tests are favour of intel: games use too many branch, load/store-dependency-chain and so on. Only 10-20% of the instructoins in games are floating-point instructions. game tests are biased.

[Donnie27]

Here is what David posted:

Oh WOW, I love the Sig LOL!

@hargen
Yes, games should be used since many of us here are Gamers and give our computers a good workout. When they brought up games, they were talking about Desktops, not Servers.

I'm waiting on Quad Core 45nm and that's why I didn't go with the G0 C2Q.

[PhilDoc]

Works fine for me. If you want you can try just going to the main RWT site then going to the forums. Not hard to find Kanter's post there about it.

Thanks, finally got it to work, don't know what the problem was. Yep, pretty sad, still haven't seen anything to turn me off or on either way. Besides, I still have a few months to decide.

What concerns me more is the inability to get the systems to run correctly

[Motiv]

Do not use games to determine the performance of the desktop. game tests are favour of intel: games use too many branch, load/store-dependency-chain and so on. Only 10-20% of the instructions in games are floating-point instructions. game tests are biased.

My post was in response to you saying single threaded will be the way forward to the next 20 months.

This is not correct. Quad core will be the way forward from now on. Most applications other than basic Office apps, are making use of multicore. The more cores, the more it will use.

Dual core will be the norm for low spec machines but expect quad core to be mainstream.

[zpdixon]

To surprise, the non-paralle result of SPECint2006 is much faster than the parallel result.
SPECint2006(base/peak)
Intel 3GHz 20.2/22.6 (test data: 2007.7.23)
http://www.spec.org/cpu2006/results/...723-01539.html

I looked at the SPEC scores in details, and I don't understand where you see that the "non-paralle result of SPECint2006 is much faster than the parallel". Here is a SPECint2006 result with auto-parallelization for a 3 GHz Xeon [1]: 20.9/24.3. This is better than your non-parallel 20.2/22.6.

[1] http://www.spec.org/cpu2006/results/...916-02014.html

[KTE]

I looked at the SPEC scores in details, and I don't understand where you see that the "non-paralle result of SPECint2006 is much faster than the parallel". Here is a SPECint2006 result with auto-parallelization for a 3 GHz Xeon [1]: 20.9/24.3. This is better than your non-parallel 20.2/22.6.

[1] http://www.spec.org/cpu2006/results/...916-02014.html

That's running the -parallel flag for C Benchmarks base optimization as well.

[zpdixon]

That's running the -parallel flag for C Benchmarks base optimization as well.

Yeah, I am not sure but I think that vendors are free to do what they want in the base and peak tests.

- Z

[BrowncoatGR]

I posted this in the news thread as well.

Here's some quick K8 results @2GHz specint base=9.77 rate=10.8
specfp base=10.4 rate=10.9
This is way too close. Either there is something wrong with the K10 IBM system or K10 is going to be a huge disappointment for everybody except the HPC crowd. I would wait for at least a couple more results to get published. Just look at the problems stephen has been having with the various Bioses.
http://www.spec.org/cpu2006/results/...828-01902.html
http://www.spec.org/cpu2006/results/...828-01900.html

[Viper666]

I posted this in the news thread as well.

Um why we looking at single threaded benches on a Native Quad Core,Im confused.The power of the platform is that all the cores can talk to each other very quickly.The only thing I dont understand is the SSE should be twice as fast as K8,even on a single thread,One thing I know for sure is on my Quad FX ,setting memory speed to less that 800,even with lower latency timings slows it down,also setting the bios interleave all(disabling NUMA)kills it.It seemed S7 has that problem.Comparing an Intel with slower memory timing proves nothing because the platforms are totally different.In my tests nothing can use a dual socket AMD system as well as SUSE linux.Windows XP x64 sometimes takes 20 minutes to figure out its not using local memory and then move the thread so it does.Vista wasnt much better.Off node hits kill the AMD 2P .The Desktop Quad AMD will be using HT3 and have alot more memory bandwidth than Barcelona, I also heard that at a certain clock speed it has some kinda power band that kicks in that amazing,since the person who said that was one of the first to provide bench marks .Im gonna guess thats right.Barcelona had to plug into an old socket and run and it does but not as fast as it will in a new socket.I know this Barcelona scales better than lineir that has never before happened in the history of processors,example 1 processor scores 50,2 processors score 110,history says 2 processors are about 30% faster than 1 or 65

[KTE]

Yeah they did. AMD officially stated that you can expect "up to 15% performance gain" going from 2GHz to 2.5GHz. Those were server market releases and based on upcoming SPEC CPU2006 figures though. Nothing of desktop performance was indicated at all. And they're planning to launch a 2.6GHz Opteron 23xx model by end '07 too.

[PhilDoc]

Yeah they did. AMD officially stated that you can expect "up to 15% performance gain" going from 2GHz to 2.5GHz. Those were server market releases and based on upcoming SPEC CPU2006 figures though. Nothing of desktop performance was indicated at all. And they're planning to launch a 2.6GHz Opteron 23xx model by end '07 too.

If you look at the spec web site, AMD has put up some number for the 2360SE and I belive the 2357.

[Motiv]

If you look at the spec web site, AMD has put up some number for the 2360SE and I believe the 2357.

Just to make it easier for people. Thanks for the heads up Phil...

Spec FP Rate (8 cores)

SpecFP Rate (16 Cores)

SpecIntRate (8 cores)

SpecIntRate (16 cores)

[DoubleZero]

If you look at the spec web site, AMD has put up some number for the 2360SE and I belive the 2357.

2360SE it's 18%-22% faster then 2347... not bad, perfect scaling would be 25%.

[BrowncoatGR]

At 8 cores it performs better thana X5355 in rate benches. I'm curious about ST performance though.
@Smartass unless you can prove that Quad xeons communicate faster through the FSB than K10 does internally i say you are making that up.