LINPACK: Nehalem vs Shanghai part 2
by Johan De Gelas on December 1, 2008 12:00 AM EST- Posted in
- IT Computing general
The last post generated some very interesting comments and questions, which I wanted to address. Unfortunately, some people misinterpreted the post as a "the best scores Nehalem and Shanghai can get in Linpack" review.
So let me make this very clear: this and the previous blogpost are not meant to be a "buyer's guide". The Nehalem desktop system and AMD "Shanghai" server are completely different machines, targeted at totally different markets. Normally, we should wait for the Xeon 5500 to run these kind of benchmarks, but consider this a preview out of curiosity.
Secondly, we were not trying to get the highest possible LINPACK scores on both architectures. We wanted to use one binary which has good optimizations for both AMD's and Intel CPU's. Fully optimized binaries won't even run on the other CPU.
Our only goal is to get an idea how the Nehalem and Shanghai architectures compare when running a "LINPACK" alike binary which is optimized to run on all machines.
Thirdly, this is not our review of course. This is a blogpost which talks about some of the tests we are doing for the review.
MKL on AMD?
Using the Intel Math Kernel Libraries on an AMD CPU is of course a good way to start some heavy debates. As I pointed out in the last blogpost however, in some cases, the slightly older MKL versions still do a very good job on AMD CPUs when you benchmark with low matrix sizes.
You don't have to take my word for it of course.
Compare the Intel Linpack 9.0 (available mid 2007) with the binary that AMD produced at the end of 2007. AMD made a K10 only version using the ACML version 4.0.0, and compiling Linpack with the PGI 7.0.7 compiler (with following flags: pgcc
-O3 -fast -tp=barcelona-64).
All the benchmarks below are done on one CPU with 4 GB (AMD, Intel Xeon) or 3 GB (Intel Core i7). Speedstep, Powernow! and Turbo mode were disabled.
As predicted, the ACML binary which was compiled with 2007 compiler is slower than the MKL "2007" version also compiled in 2007. The MKL version runs on any CPU that has support for (S)SSE-3, so it continues to be a very interesting one for us to test. As you can clearly see from the Xeon 5472 (3 GHz) score, it is not fully optimized for the latest 45 nm Intel CPUs with SSE-4. It is a good "not too optimized" version which can be used on both Intel and AMD CPUs. You can clearly see this as the 3 GHz Xeon 5472 is behind the AMD Opteron 8384. If this Intel Binary was giving the AMD CPUs a badly optimized code path, this would not be possible.
As we move forward to 2008, we have to create a new binary as both AMD and Intel's fully optimized Linpack versions will not run on the competitor's CPU. Intel released the Linpack benchmark version 10.1, which is not fully optimized for the "Nehalem" architecture, but for 45 nm "Harpertown" family.
AMD has created a new Linpack binary using ACML 4.2 and the PGI 7.2-4 compiler. Below you see how the two CPUs compare.
Bottom line is that these LINPACK benchmarks are moving targets like the SPEC CPU benchmarks, as the compilers and libraries used are just as important as the CPUs.When the Xeon 5500 will materialize, LINPACK performance will probably be higher as the binary is built for the "Penryn/Harpertown" family.
While it is useful for the HPC people to see which CPU + compiler can offer the best performance, it is also interesting to understand what kind of performance you get when you compile binaries that have to run on all current CPUs. It is pretty hard to compare CPU architectures if you are using totally different binaries.
In the next post we'll delve a bit deeper on what is happening with Hyperthreading, Linpack and the new architectures.
Facebook
Twitter
RSS
31 Comments
View All Comments
JohanAnandtech - Wednesday, December 3, 2008 - link
Yes, of course :-).We 'll do everything in the coming months to make sure that the it.anandtech.com community gets high quality debates with a good signal/noise ratio. In a few months the comment box and board should integrate for example. It is still the goal to bring deeply technical benchmarking to the IT world, where good reviews are still scarce IMHO. It will take time, but it.anandtech will get there. :-)
Again, I welcome all constructive and even sharp comments. But preferably well founded with some good reasoning.
superflex - Tuesday, December 2, 2008 - link
Your bias is pathetic. Look at the front page where Intel has been hosting the Intel Resource Center link for over a year. Intel is one of Anand's biggest advertizers and Steve Ballmer enjoys the rim jobs Anand and his douchbag bloogers provide him.Thanks again for confirming my beleifs that AMD (and ATi) will never get a fair shake on this sorry site.
JohanAnandtech - Wednesday, December 3, 2008 - link
As most of our regular readers know, let our trackrecord speak for itself:http://it.anandtech.com/weblog/showpost.aspx?i=443">http://it.anandtech.com/weblog/showpost.aspx?i=443
Comment:
"This is really interesting. It looks like AMD is very competitive in the HPC / Server market. I am glad to hear it. I am currently using a Core 2 Duo system for my desktop machine, and it is very fast. But, I certainly don't want the competition between AMD and Intel to come to an end anytime soon. Any good news for AMD at this point is good news for consumers."
Still, not convinced?
http://it.anandtech.com/weblog/default.aspx">http://it.anandtech.com/weblog/default.aspx
"The very first independent Nested Paging Virtualization tests"
"AMD's K10: a "dead" product or not?"
"AMD back in the quad socket race"
"The revenge of AMD Barcelona's TLB?"
Take the time to research, and judge for yourself.
7upMan - Thursday, December 4, 2008 - link
Thank you, Johan. The last article is a reminder why so many Opteron machines are in the Top 10 of the Top 500 list of supercomputers. And this, folks, you should always remember when arguing Pro or Con AMD.Barcelona is and was first and foremost a server ship (and performing exceptionally well in that role), with Deneb hopefully changing that bias toward gaming. At least I hope so, because I'd love to change my Athlon X2 6400+ to a faster CPU with less power hunger.
Besides, I can imagine why Anand is slightly biased toward Intel (and yes, it really is). After all, it was AMD who wanted to have positive reviews on the first Phenoms by inviting the testers to exotic locations (see Anand's review of Phenom 1). If someone tries to cheat you, you are less inclined to believe every bit of hype he tells you.
Zorblack1 - Tuesday, December 2, 2008 - link
Wow your a moron! Steve Ballmer works for Microsoft not Intel. Microsoft != IntelAnd as for your AMD fanboyness take a hike. You blame the hard working folks at anandtech for AMD's failures.
Mclendo06 - Tuesday, December 2, 2008 - link
Linpack is great and all, but I was wondering if you had any benchmarks for sparse operations that you could run as part of the review, for instance running Pardiso on a 250k equation system (if RAM permits - 3GB will probably limit you to about ~100k-ish depending on matrix sparsity). I may be wrong here, but I think I've heard somewhere that memory is a significant bottleneck for sparse matrix computation, and so it would be interesting to see what sorts of gains Intel has made here with the new memory controller.uf - Tuesday, December 2, 2008 - link
If you pretend being neutral, AnandTech should publish info about its financial interest and its connection with all brands mentioned in the review: Intel, Amd, etc., otherwise you very look like Intel promotion site.JohanAnandtech - Wednesday, December 3, 2008 - link
Understand that Liz and myself are working here in Belgium, Jason and Ross are in Canada. We have no clue whatsoever happens in the financial part of Anandtech, and rightfully so. I have no interest whatsoever to get involved in that.Read the past articles at it.anandtech.com and judge on that whether or work is "neutral" or not.
uf - Wednesday, December 3, 2008 - link
If you are really not biased, and if you are really interested to compare two hardware (CPU,chipset,mem) systems, WHY you didn't use neutral software as someone pointed out. Of cause, we would not reach top speed but it was not our goal.ZootyGray - Tuesday, December 2, 2008 - link
Nobody believes you anymore.Imagine a site that goes to great lengths to present unbiased testing. without the cheap trix you present.
2nd the call for apology - then again, it goes deeper than that - and you know it. Never mind - I am done with you.