Linux Shootout: Opteron 150 vs. Xeon 3.6 Nocona
by Kristopher Kubicki on August 12, 2004 2:35 PM EST- Posted in
- Linux
TSCP
We apologize for the broken TSCP Makefile in the previous review which rendered our initial results inaccurate. Fortunately we posted the file so that others were able to detect the error and not find fault with the processors instead. The large issue many of our readers have brought to our attention are the severe difference in performance between various optimizations. Below you can see how various compile flags affected our benchmark scores.
The first benchmark is run with the optimization flags:
-O2 -funroll-loops -frerun-cse-after-loop
The next benchmark is run with the optimization flags:
-O3 funroll-loops -frerun-cse-after-loop
Finally, we have the architecture optimized flags as well:
(Intel) -O3 - march=nocona -funroll-loops -frerun-cse-after-loop
(AMD) -O3 - march=k8 -funroll-loops -frerun-cse-after-loop
You are reading these charts correctly, the O3 flag actually penalizes the AMD CPU. We also compiled the program with -O2 -march=k8 but we got virtually the same score with or without the march flag.
We were informed others have been capable of much faster nodes per second using GCC 3.4.1 and the flagset:
-O3 -march=athlon-xp -funroll-loops -fomit-frame-pointer -ffast-math -fbranch-probabilities
We did not have time to fully test GCC 3.4.1, although there is a strong likelihood that 3.4 encourages better optimizations (particularly on the x86_64 platforms).
Crafty
For good measure, we have included Crafty into our chess benchmarks section. Crafty was only built using the "make linux-amd64" target. From the Makefile, it seems as though the "AMD64" moniker is slightly inappropriate. The target claims:
# -INLINE_AMD Compiles with the Intel assembly code for FirstOne(), # LastOne() and PopCnt() for the AMD opteron, only tested # with the 64-bit opteron GCC compiler.
The benchmark was generated by running the "bench" command inside the program.
It is clear the difference between both processors is quite severe in this instance. Although it is difficult to pin an exact culprit, there are likely multiple arch optimizations were left untapped, and thus our reasoning for discouraging overusage of optimizations in general.
Facebook
Twitter
RSS
92 Comments
View All Comments
johnsonx - Thursday, August 12, 2004 - link
Crap, now I mixed myself up... none of the charts seem to show the Opteron going from a loss to a win. Indeed, the Opteron is slower than XEON across the board in the Blowfish section.johnsonx - Thursday, August 12, 2004 - link
Kris,Kudos on a good follow-up article, though I felt the original had far more merit than most others did. It must've been difficult to decipher the valid and constructive criticism from all the blather.
That said, I did notice some oddity with the John the Ripper results:
You say "Had we left the default -O2 compilation, Blowfish hashing would have been faster on the Xeon processor than the Opteron. However, as soon as we use -O3, the Opteron outperforms the Xeon processor."
However, the only graph that shows the Opteron going from a loss with -O2 to a win with -O3 is the bottom MD5 graph. Perhaps you meant to say MD5 then in your comments, or are some of the graph numbers wrong?
snorre - Thursday, August 12, 2004 - link
Congratulations with a much better review this time, although your conclusions are almost as bad as the sloppy old review. Based on the results I'd expected something more like this:"Without a doubt, the Opteron 150 trounces over the 3.6GHz Xeon in real world benchmarks."
BTW, AMD's Opteron 150/250 is directly comparable with Intel's Xeon 3.6GHz in 1-way & 2-way systems.
BTW 2, the Performance Test Configuration table on page 1 is unreadable (white text on white background).
TauCeti - Thursday, August 12, 2004 - link
Hi Kristopher,ref#21
Ok, rereading your sentence i agree :)
So before a nice old J. Beam integrates the hours of my day into peaceful, cushioned oblivion, let me assure you that from my point of view you did a very good job today.
You endured a continuous datasphere bashing and gave your best to adress criticism in a constructive way.
Have a nice trip.
Tau
NesuD - Thursday, August 12, 2004 - link
Nice save Kris, Glad to see you square it all up.datacipher - Thursday, August 12, 2004 - link
"When the 3.6F actually shows up at newegg with a price, then i will tell you for sure what it competes against :)"OK, fair enough. Thanks for the replies...
kaoman - Thursday, August 12, 2004 - link
Nice article and better benchmarks.But what I dont understand with the first article's controversy.. if the 3500+ costs ~ $350, and assuming the 3.6F costs ~ $450 by late August (http://www.hardforum.com/showthread.php?t=770923), AND that the performance of the 3.6F EQUALS the Xeon 3.6, what was wrong with the first article's choice of proccessors comparison? The whole apples-to-apples analogy WORKS in the sense that both chips are 64bit x86 processors. If the 3.6F = Xeon 3.6, who cares which one is used to compare to the Athlon64? And that was the basis of the article. A name or price tag doesn't make a processor comparison any more or less proper when the CPUS are identical. Granted I don't know for a fact if they are, but I'm taking Kris's word for it.
DAPUNISHER - Thursday, August 12, 2004 - link
Now that is what I call great damage control :-) Way to turn it around KK, now go enjoy your vacation!I have been among your leading critics after that train wreck you posted, but you have definitely shown the necessary focus and ability to except criticsm be it constructive or otherwise, and rebound quickly, good job.
Lynx516 - Thursday, August 12, 2004 - link
Ah I hadnt noticed you put it there but you have to realy know what you are looking for to be able to see that it is not the standard GCC 3.3.3 compiler. Maybe put a small note at the bottom of the compiler section in bold saying "This is not the standard GCC compiler" or the like as most people will not notice that it isnt standard. but I have to say its nice to see Linux getting such a large main steam coverage.Well done again.
Lynx
AMDScooter - Thursday, August 12, 2004 - link
This should have been the first 64bit article.It is well written and makes sense, a real 180 from that last debacle. Good job! I do however still have issues with the use of synthetic benchmarks. Why bother tossing them in at all when the real world tests in this very review show how utterly useless they are? The AMD chip tracks a mud hole in the a$$ of the Xeon in all but the synthetic tests. This only gives the AMD zealots more ammo for the possibility of some sort of bias toward Intel, and in this case I would tend to agree with them. Keep up the good work :)