2CPU.com » News » August 2004 » Linux and EM64T; Intel's 64-bit Suggestion

I'm backkkkkk... from vacation that is. Anyways, Tom let me know about an article over at Anandtech that's talking about Linux and Intel's EM64T.

Although the Athlon 64 3500+ and the Xeon 3.6GHz EM64T processors were not necessarily designed to compete against each other, we found that comparing the two CPUs was more appropriate than anticipated, particularly in the light of Intel's newest move to bring EM64T to the Pentium 4 line. Once we obtain a sample of the Pentium 4 3.6F, we expect our benchmarks to produce very similar results to the 3.6 Xeon tested for this review.

We'll have to wait awhile for their next article comparing dual 3.6GHz Nocona's with Dual Opterons. Take a look.

02/23/2004 10:32 PM: Linux and Hyper-Threading @ 2CPU.com! by Jim
As promised, I've thrown together a second hyper-threading article, this time detailing performance in Linux. I rounded up over 12GHz worth of processing power for this article (including a 3.2GHz Pre...

07/15/2000 03:51 PM: Linux and Gaming by Hooz
I was doing my AM surfing today and stumbled across this link over at VoodooExtreme. It's sort of a "State of Linux Gaming" article, but it does include some interesting points. Here's a c...

04/26/2000 12:02 PM: Linux and the Police force? by J0rdan
In my early morning wandering, I found a really nice article over at Linuxnews. The St. George Police department (SGPD), switched their dispatch service to a Linux system. They stated uptime, and a sm...

duraid Registered User Posts: 378 Joined: 2002-03-31

#31051 Posted on: 08/09/2004 07:26 PM Does anyone know if programs built with Intel's new 8.1 C/C++ compiler will run, unmodified, on AMD's x86-64 parts?

i_wolf labhair dom as gaelige Posts: 2034 Joined: 2002-11-19

#31052 Posted on: 08/09/2004 07:32 PM One problem I noticed with their benchmarks is that they were using generic builds and generic flags when they were building . I would have liked them to build specifically againts each architecture with optimal flags and then compare. From what we see in these benchmarks it would appear that Intel is back in the performance game with Nocono. P.S. welcome back from vacation Jim. Hung like a donkey. Go like a horse!

El_Angelo Registered User Posts: 79 Joined: 2004-04-15

#31053 Posted on: 08/09/2004 07:52 PM the whole review seems a little fishy... best that you start reading the comments posted by the users before you actually believe this review... (maybe you can post some of them on the frontpage, jim_ ) most severe one : mysql bench for the 3500+ should be 215/223 in stead of 289 (and +- 250 for the xeons)... seems like the 3500+ is tested in 32bit mode while the xeon was tested in 64bit mode

i_wolf labhair dom as gaelige Posts: 2034 Joined: 2002-11-19

#31054 Posted on: 08/09/2004 07:55 PM Does anyone know if programs built with Intel's new 8.1 C/C++ compiler will run, unmodified, on AMD's x86-64 parts? I know that in the past you could get a binary that was compiled with ICC to run on AMD's parts if you removed the check for the cpu string ID with a hex editor. I am not sure if its still possible to remove the cpu string id check from the resulting binaries created with ICC 8.1. However , does 8.1 C/C++ support EMT64/iAMD64/x86-64 presently? Probably the best AMD64 compilers presently are from Pathscale. Also i see here Sun, AMD and Pathscale are collaborating together on these compilers. http://www.pathscale.com/pr_080204.html Hung like a donkey. Go like a horse!

rmn oh my, it's huge! Posts: 5894 Joined: 2002-01-26

#31055 Posted on: 08/09/2004 08:55 PM Why the Athlon 64, and not an Opteron or Athlon FX...? :confused: Also, lots of synthetic benchmarks, but not many real applications. If you want to test real floating-point, why not run 3DS MAX or POV Ray or some science stuff? P.S. - Just noticed, they did run POV Ray (on page 2), and the Athlon was 30% faster. RMN ~~~

i_wolf labhair dom as gaelige Posts: 2034 Joined: 2002-11-19

#31056 Posted on: 08/09/2004 11:32 PM one thing to also keep in mind is that these were uniprocessor tests. There is no way a SMP Xeon system can scale as well as a similar Opteron rig, primarily because the Xeons are still using a shared FSB. i.e. these results are in no way indicative of the type of SMP performance that can be expected. Hung like a donkey. Go like a horse!

XWRed1 Registered User Posts: 185 Joined: 2001-08-27

#31057 Posted on: 08/09/2004 11:58 PM One problem I noticed with their benchmarks is that they were using generic builds and generic flags when they were building . I would have liked them to build specifically againts each architecture with optimal flags and then compare. You can't do this with amd64 binaries and gcc right now. Up until recently there was only one implementation of the amd64 arch in hardware, so gcc doesn't have any processor-specific optimization flags to choose from since they are (well, were) all the same.

Hydra Performance Junkie Posts: 174 Joined: 2001-03-09

#31058 Posted on: 08/10/2004 01:20 AM Testing Xeon 3.6GHz and comparing it with Athlon 64 3500+ is almost the same as testing Opteron 150 and comparing it with P4 2.4GHz, not very fair is it? They obviously should have compared the Xeon 3.6GHz with Opteron 150, that almost goes without saying. This review smells like money under the table, and it really surprised me that Anandtech were capable of such terrible journalism. Hopefully some proper 64-bit benchmarks will appear soon, as this one is totally useless! :mad: See here for more bad comments, this review utterly sucks: http://www.anandtech.com/talkarticle.aspx?i=2158 Overclocking is a blast! Do I smell something burning?

i_wolf labhair dom as gaelige Posts: 2034 Joined: 2002-11-19

#31059 Posted on: 08/10/2004 04:45 AM the inquirer is also running a story about it here.... http://www.theinquirer.net/?article=17754 Hung like a donkey. Go like a horse!

Hydra Performance Junkie Posts: 174 Joined: 2001-03-09

#31060 Posted on: 08/10/2004 06:30 AM More comments over at Slashdot: http://linux.slashdot.org/linux/04/08/09/136230.shtml?tid=137&tid=142&tid=118&tid=163&tid=1&tid=218 And over at Ace's Hardware: [url=http://www.aceshardware.com/forum?read=115093783]Anand's test is a joke[/url] This review is a joke, a really bad one too Overclocking is a blast! Do I smell something burning?

LRSeriesIII Aspiring Rocket Scientist Posts: 1120 Joined: 2002-08-29

#31061 Posted on: 08/10/2004 07:58 AM Honestly, while the inclusion of the Athlon64 3500+ for comparison was...questionable, to say the least, the choice of benchmarks and the explanation of what the benchmarks stressed (and why they were chosen) seemed extremely lacking. Also, while the big thing about this review was supposed to be the fact that it was using AMD's and Intel's 64 bit x86 processors, it failed to give any 32 bit comparison. Very sloppy work in my opinion. ->Computers ->Folding for team 3074

stmok23 Registered User Posts: 797 Joined: 2002-02-02

#31062 Posted on: 08/10/2004 11:15 AM The article should be pulled and corrected. That simple. If they leave the article as is, they get more bad press from those who know what's wrong with what they've done. (some people already say they're on the level with tom's hardware.) Sempron (Socket 754): 2x Abit NF8-V (nForce3 250Gb) and ASRock K8SLI-eSATA2 (ULi M1697) Dual CPU love: Supermicro P6DBE (i440BX), PIIIDRE (i840), 2x PIIIDR3 (i840), 4x ASUS P3C-D (i820), and ACorp 6A815EPD1 (i815EP) OSs?: Linux, Solaris and BSDs.

duraid Registered User Posts: 378 Joined: 2002-03-31

#31063 Posted on: 08/10/2004 03:57 PM Originally posted by i_wolf I know that in the past you could get a binary that was compiled with ICC to run on AMD's parts if you removed the check for the cpu string ID with a hex editor. that's right. However , does 8.1 C/C++ support EMT64/iAMD64/x86-64 presently? Yes, it does. Specifically, it claims to support EMT64 and says nothing else. So, one wonders what would happen to, say, Opteron 64-bit SPECint scores if you could actually put Intel's compilers to work.... Probably the best AMD64 compilers presently are from Pathscale. Also i see here Sun, AMD and Pathscale are collaborating together on these compilers. That's really unfortunate, because the Pathscale compilers are pretty crappy - they're actually something of a rush-job X86 backend added to an open source Itanium compiler from a few years back. The question remains - have Intel declared war on AMD64 by refusing to support it with their compilers? Or are they playing nice, and by simply being compatible, offering the best Opteron compiler out there as a free download? (I actually hope it's the latter, not only because it means we get to use a good compiler (FINALLY!) on our Opteron hardware, but because it means the lame pathscale people get to f!@k off and die.)

i_wolf labhair dom as gaelige Posts: 2034 Joined: 2002-11-19

#31064 Posted on: 08/10/2004 04:48 PM Yes, it does. Specifically, it claims to support EMT64 and says nothing else. So, one wonders what would happen to, say, Opteron 64-bit SPECint scores if you could actually put Intel's compilers to work.... Yup it would be very interesting to see. In the past AMD got big gains from Intel's compilers. I had no idea that Pathscale were using the front end from an Itanium compiler. The only thing is, by using an existing frontend ( lexical analysis, syntax analysis, semantic analysis, global optimization ) , they aren't really doing anything that other compiler designers do not do already i.e. hook a different back end for an ISA onto a front end. I assume that the front end they are using (to generate Internal Representation language for the back end) is tried and trusted and probably wasn't just used with the Itanium? What would be more interesting for me to know is what backend they are using for generating x86-64. I assume they wrote it from scratch?? Hung like a donkey. Go like a horse!

stmok23 Registered User Posts: 797 Joined: 2002-02-02

#31065 Posted on: 08/10/2004 05:01 PM From Anand Lal Shimpi Weblog : Topic #2: Kris' article. I didn't read the article before it went up (my fault), and I saw the instant backlash. I've already talked to Kris and he's planning a followup with the fastest Opteron available, but to his credit here's what he was trying to do: show an Athlon 64 vs. Pentium 4 F (64-bit Prescott) comparison. He did a good job of that in the article, but I think what was lacking was exactly pointing out that the Xeon test bed is basically a preview of the new Pentium 4 F (desktop P4 with 64-bit extensions). There were a few errors in the benchmark graphs which he has since corrected and he's explained the theory behind the comparison as well, which should hopefully get the point across a bit better. The problem with benchmarking under Linux is that there aren't that many good benchmarks, so you're left with a handful of decent tests and a lot of synthetic benchmarks. I went over some ideas with Kris about other more real world tests to run under Linux for part 2 of the article, we'll see how many can get in there. To those who are questioning him, I trust Kris enough to leave him in charge of the site while I'm gone. His loyalty's are in the same place as mine, with the readers, and he will do any and everything to do the best job possible. Kris did a good job of making the Pentium 4 3.6F vs. Athlon 64 3500+ comparison, but I think a lot of his intent got lost in translation so to speak (another movie I need to see). In any case, the Xeon vs. Opteron comparison will be coming thanks to your demands - Kris will be working all night making sure it happens asap. Ask and ye shall receive, that's how things have always worked here and that's how they'll continue to work. The Author has a 2nd chance, if he screws up the next one...Well, Anandtech will get alot hits, that's for sure! Sempron (Socket 754): 2x Abit NF8-V (nForce3 250Gb) and ASRock K8SLI-eSATA2 (ULi M1697) Dual CPU love: Supermicro P6DBE (i440BX), PIIIDRE (i840), 2x PIIIDR3 (i840), 4x ASUS P3C-D (i820), and ACorp 6A815EPD1 (i815EP) OSs?: Linux, Solaris and BSDs.

LRSeriesIII Aspiring Rocket Scientist Posts: 1120 Joined: 2002-08-29

#31066 Posted on: 08/10/2004 06:44 PM Anandtech wanted a review comparing the Pentium 4 3.6F against the Athlon64 3500+, but did not have a Pentium 4 3.6F so they substituted a Xeon 3.6F. The whole motivation for writing the article in the first place was, according to my understanding, Intel's release of 64 bit x86 processors. From that point of view, I do not see a big problem with using the Xeon as a stand in for the Pentium 4, given that this article was focused on processor performance and that the Xeon and Pentium 4 are, basically, the same processor. My suspicion is that much of Anandtech's readership simply did not understand the level of similarity between the Xeon and the Pentium 4 and that the article simply failed to educate those readers effectively. The real problem here, as far as I can tell, is in fact the choice of AMD processor to compare with the Intel. Intel's 3.6F parts, whether Pentium 4s or Xeons, are top of the line, flagship processors. The Athlon64 3500+, on the other hand, while a respectable processor, is more of a fast mainstream processor. The far more appropriate choice to represent AMD's offerings would have been an Athlon64 FX-53 (or even a 55) which Anandtech has already reviewed. Even an Athlon64 3800+ would have been a better choice than the 3500+. This article was Intel's fastest pitted against AMD's third stringers. ->Computers ->Folding for team 3074

rmn oh my, it's huge! Posts: 5894 Joined: 2002-01-26

#31067 Posted on: 08/12/2004 01:02 AM They have made a few "corrections" to the article, namely to the SQL tests, where the Athlon 64 now beats the Xeon (as well as in POV Ray, and most of the other real benchmarks are a tie). The conclusion has also been changed. It used to say: Without a doubt, the 3.6GHz Xeon trounces over the Athlon 64 in math-intensive benchmarks. Intel came ahead in every severe benchmark that we could throw at it, particularly during John the Ripper. Even though John uses several different optimizations to generate hashes, in every case, the Athlon chip found itself at least 40% behind. Much of this is likely attributed to the additional math tweaking in the Prescott family core. Now it says: Without a doubt, the 3.6GHz Xeon trounces over the Athlon 64 3500+ in math-intensive synthetic benchmarks. Again, not that it is really a comparison between the two chips yet anyway, but perhaps something of a marker of things to come. However, real world benchmarks, with the exception of John the Ripper is where AMD came ahead instead. Even though John uses several different optimizations to generate hashes, in every case, the Athlon chip found itself at least 40% behind. Much of this is likely attributed to the additional math tweaking in the Prescott family core, and the lack of optimizations at compile time. Cough, cough... :rolleyes: RMN ~~~

Hydra Performance Junkie Posts: 174 Joined: 2001-03-09

#31068 Posted on: 08/13/2004 01:48 AM Anandtech gives it another go here: Linux Shootout: Opteron 150 vs. Xeon 3.6 Nocona Guess what chip comes out on top this time around? Overclocking is a blast! Do I smell something burning?

rmn oh my, it's huge! Posts: 5894 Joined: 2002-01-26

#31069 Posted on: 08/13/2004 02:22 AM On top by over 80% in some cases! :eek: RMN ~~~

stmok23 Registered User Posts: 797 Joined: 2002-02-02

#31070 Posted on: 08/13/2004 11:08 AM Now that's a better article. Much improvement shown. (I'm glad they listen and made this improved article!) Sempron (Socket 754): 2x Abit NF8-V (nForce3 250Gb) and ASRock K8SLI-eSATA2 (ULi M1697) Dual CPU love: Supermicro P6DBE (i440BX), PIIIDRE (i840), 2x PIIIDR3 (i840), 4x ASUS P3C-D (i820), and ACorp 6A815EPD1 (i815EP) OSs?: Linux, Solaris and BSDs.

HFU Registered User Posts: 2000 Joined: 2002-06-15

#31071 Posted on: 08/13/2004 01:08 PM Opteron is much ahead of D-0 stepping Nocona in UP setup, I can't imagine when running it in dual or 4 way configurations! AnandTech could at least use Iwill DH800 or Asus NCCH-DL for a fair DDR400 memory setup rather than the much slower DDR2-400 on Supermicro board. Comparing DDR400 CL2 to DDR2-400 CL3 is similar to comparing DDR400 to DDR266. So much about not biased article. HFU

stmok23 Registered User Posts: 797 Joined: 2002-02-02

#31072 Posted on: 08/13/2004 02:54 PM Have you noticed, after reading, that the tone of the article is, well, not "ballsy" enough to say one chip is better than another for a majority of the benchmarks? It sounds like the author is fearing that he might anger either AMD or Intel fans. Its worded in a way, so that it won't provoke another 100+ flames at him. (Sounds like he's treading lightly). Sempron (Socket 754): 2x Abit NF8-V (nForce3 250Gb) and ASRock K8SLI-eSATA2 (ULi M1697) Dual CPU love: Supermicro P6DBE (i440BX), PIIIDRE (i840), 2x PIIIDR3 (i840), 4x ASUS P3C-D (i820), and ACorp 6A815EPD1 (i815EP) OSs?: Linux, Solaris and BSDs.

rmn oh my, it's huge! Posts: 5894 Joined: 2002-01-26

#31073 Posted on: 08/13/2004 10:55 PM Originally posted by HFU AnandTech could at least use Iwill DH800 or Asus NCCH-DL for a fair DDR400 memory setup rather than the much slower DDR2-400 Isn't Intel trying to push DDR2? RMN ~~~

i_wolf labhair dom as gaelige Posts: 2034 Joined: 2002-11-19

#31074 Posted on: 08/14/2004 12:33 AM I wonder what difference GCC3.4 will make to these benchmarks are GCC has received quite a bit of work regarding Opteron optimizations. Hmmmm..... can't wait for the dual core dothan versus Opteron bench's next year : p Hung like a donkey. Go like a horse!

HFU Registered User Posts: 2000 Joined: 2002-06-15

#31075 Posted on: 08/14/2004 08:31 AM Originally posted by rmn Isn't Intel trying to push DDR2? RMN ~~~ Intel is trying to push DDR2 standard along with PCIe, though the current stance on DDR2-400 performance is not on par with much matured DDR400 specs. We won't see major memory performance advantage until DDR2-667-800 speed. Not to mention the E7525 chipset used on Supermicro could be one of the bad stepping chipset that is released after much bashed i915/i925X desktop counter parts. For sure, Intel rushed out D-0 Nocona with quick aim at AMD64 that has been optimized for more than a year now. Well, swapping the DDR400 wouldn't totally revert the benchmark result, at least we know for sure how much slower Nocona performed at current stage. I'm inclining to the idea i_wolf suggested, let's compare dual-core Dothan vs dual-core Opteron, forget about this miserable Nocona crap. If Intel doesn't improve Nocona upon E-0 launch this October, then savvy customers know what to do next. HFU

2CPU.com » News » August 2004 » Linux and EM64T; Intel's 64-bit Suggestion