Slashdot Mirror
EM64T Xeon vs. Athlon 64 under Linux (AMD64)
legrimpeur writes "Anandtech has a nice performance comparison under Linux (AMD64) between the recently introduced 3.6GHz EM64T Xeon processor and an Athlon 64 3500+. It is disappointing to see how the Athlon gets trounced in FPU intensive benchmarks. No memory-bound benchmarks (where the Athlon is supposed to have an edge) are presented, though." Update: 08/09 23:34 GMT by T : As the Inquirer reports, many Anandtech readers take issue with the comparison.
i almost cared just then, for a second
The editors of Slashdot seem to love posting articles whose sole purpose is to evoke flame wars between Intel fans and AMD fans.
For what it's worth, I read the article and the processors seemed pretty well matched except for some "synthetic" benchmarks. I don't know much about the synthetic benchmarks that they used, but I have found that synthetic benchmarks are almost always biased in Intel's favor. Do synthetic benchmark writers optimize for Intel accidentally or is there some kind of conspiracy going on here? You be the judge.
Finally, to try to balance out the article submitter's inflammatory comments about the Athlon being "trounced in FPU intensive benchmarks", here is a nice paragraph from the article summary:
"That's not to say that the Xeon CPU necessarily deserves excessive praise just yet. At time of publication, our Xeon processor retails for $850 and the Athlon 3500+ retails for about $500 less. Also, keep in mind that the AMD processor is clocked 1400MHz slower than the 3.6GHz Xeon. With only a few exceptions, the 3.6GHz Xeon outperformed our Athlon 64 3500+, whether or not the cost and thermal issues between these two processors are justifiable."
Obviously they are not comparing processors which have price parity, so one could spin this either as "look at how slow the Athlon is", or "look at how much money you have to spend to get an Intel chip that is faster than an Athlon", depending upon your bias.
I'm all for the best processor out there. If it is made by Intel, then so be it. This will just give AMD more reason to compete for my dollar.
Wouldn't the larger cahced Opteron, the product actually positioned by AMD to compete with the Xeon series processors, have been a better comparison?
Comment removed based on user account deletion
Athlon64 are for desktop markets, Xeon are for server markets
for this comparison to be fair, Xeon should be compared to Opteron!
So should I save up for an Intel processor or buy 2 AMD machines?
Omnis amans amens
for this, but:
"..No memory-bound benchmarks (where the Athlon is supposed to have an edge) are presented, though."
Why oh why do we continually have "reviews" posted that aren't comprehensive? Hell, i hardly even click on any of the posted reviews anymore...just read the comments later and find out what was missed or just plain wrong in the review.
Where does one go to get the real, straight scoop other than buying both, testing all products involved?
Yeah, i'm a little grouchy this morning...had to get that one out.
What you're looking for is value as well as how much power you need. When your computer is sitting most of the time, hardly doing anything, is dropping $500 on a faster processor really worth it. The human eye pretty much stops distinguishing framerate past 30fps, so, unless your hosting an intensive server or work platform, ensuring a non idle CPU, getting the Intel is just a matter of bragging rights.
I don't understand your post at all. The Athlon line has historically been the better of the two (compared with Intel's P3 and P4 line) CPUs in FPU performance. In fact, the Athlon 64 and the Nocona both support x87, MMX, SSE, and SSE2 instruction sets.
Seriously, Anandtech should just never compare widely available hardware with totally unavailable hardware. And what's with using a 512KB cache, second-rank Athlon64 to compare with Intel's flagship worstation processor? How 'bout the 1MB-equipped Athlon64 FX, or more appropriately an Opteron 150 (in stock at online retailers for $600-$650).
The slowest Socket 939 Athlon versus the fastest Xeon available. PLus the SQL test of the Athlon were in 32bit, not 64 bit (which would have resulted in a win for the athlon).
Some of the other synthetic benchmarks also show slighly suspicious anomalies.
Plus were are the Nocoma 32bit benches? How are we supposed to see how performance improved in 64bit mode without comparison?
A good review would have pitched the 3.6Ghz nacoma vs an Opteron 150, would have tested both in 32 and 64 bit and tried to use some application benchmarks.
Not just picking some old scores out of the datadump to create a "shootout"
HI O WISE PRINCE. WHT TOOK U SO DAM LONG?
For one the Xeon has more L2 cache and for another most of the math benchmarks looked to be integer based. The Xeon gets beat in POVray wich is FPU intensive if im am not much mistaken... I think it is unfair to say the FPU on the Xeon is better...
I would be nice to see more non-synthetic benchmarks.
That's not to say that the Xeon CPU necessarily deserves excessive praise just yet. At time of publication, our Xeon processor retails for $850 and the Athlon 3500+ retails for about $500 less. Also, keep in mind that the AMD processor is clocked 1400MHz slower than the 3.6GHz Xeon.
I think this sums it up (besides the fact Intel kicked their pants). The AMD is running at 2.2 ghz, and retails $500 less. To me this says AMD is working smarter and Intel is working harder. Intel is reaching a (transient) ceiling with their clockspeeds and one day AMD will catch up to it. It will be interesting to see if Intel's multicore plan kicks as much ass as they are presently hoping. It'll also be interesting to see AMDs attempt at the same.
Personally I'm rooting for both. If either company gets screwed, we're all screwed.
Where are the FPU intensive benchmarks that the Athlon is trounced in?
Under normal circumstances a prime finder application does not use the FPU. And I also doubt that the super_pi application uses the FPU. However the powray benchmark (which actually uses the FPU), is one of the benchmarks where the Athlon wins.
So it would seem that it is the Integer benchmarks where the Athlon looses, instead. This also corresponds with how the normal Athlon fares against the normal Pentium.
So you compare a highend server/workstation proc to a highend desktop proc. Sure the server chip will win the majority of the benchmarks.
Where are the 64bit benchmarks? They really didn't do any comparision to 32bit, so you can't say for sure if Intel implementation is good or not. Get the Opteron in there, do the same benchmarks in 32 and 64 modes and see if there is a difference. Also throw say 5 gigs of memory in the machines, that will see how each proc handles addressing above the 4gig limit.
This article should not have been posted here, or on Anandtech for that matter. It has already caused a riot over there, both in the comments section of the article, and the forums. This article was grotesquely sub-par for Anandtech, and should have been removed immediately. Several of us avid AT readers have spotted discrepencies in the charts, stats that are totally bogus in comparison to previous AT articles. Particularly the MySql chart. To put it simply, there is absolutely no way to compare those two chips, as someone in the forums put it, "It's like comparing apples to a slab of meat." The Xeon has double the cache, is double the price, and isa top end server chip, being compared to a midrange desktop chip. The two simply cannot be compared. The article should have included an FX chip and/or an Opteron 150. Discount the article entirely. Hardcore Intel fanboys have spoken out against this article, that should really tell you something.
Please, try not to sound so stupid...
They have a cross licence agreement, so each one has what the other has in production in the term of 6 to 9 months. That is why we see the SSE in AMD processors, and AMD64 instruction in Intel64 processors.
t s/ amd/intel.license.2001.01.01.html
http://contracts.corporate.findlaw.com/agreemen
So I don't see any problem fro AMD in licensing the cp-processor.
I am always disappointed in these reviews because they alway address gaming and multimedia (which I understand are most important to the greatest amount of readers) but rarely address scientific computing. I am most interested in how fast my FORTRAN/C math-intensive code will run (I have seen examples where AMD gets beat soundly in the "FPU" benchmark, but kicks ass in ScienceMark).
AMD has been consistently good at scientific computing, but I haven't seen any performance specs for the 64-bit ones. Has anyone else?
Ed.
Why not RTFA... especially the Conclusions section...
From the article: "That's not to say that the Xeon CPU necessarily deserves excessive praise just yet. At time of publication, our Xeon processor retails for $850 and the Athlon 3500+ retails for about $500 less"
In other news, a Corvette just smooooookkkked a Ford Taurus.
OtakuBooty.com: Smart, funny, sexy nerds.
And the 3500+ and the Xeon are in the same processor class how?
The 3500+ is a mainstream, desktop processor. For a more accurate comparison, the FX series, and the opteron line should have been used.
My guess is that if these same benchmarks had been run on any Athlon vs. the equivalent P4 throughout history, the outcome would've been similar. But the results would also have been as irrelevant yesterday as they are today, since we all know the Xeon isn't 40% faster than the A64 in anything like real-world usage.
This is the real compairson. Overclock the AMD to 3.6GHz and see who wins. As soon as AMD gets tthe 90nm process perfected I think we will see a huge boost in AMDs clockspeed.
Zoid.com
While the Intel chip performed well against the AMD one, hyperthreading appeared to perform badly (i.e. the Intel chip without hyperthreading enabled tended to beat the same chip with it enabled).
It would however be interesting to see a test that somehow say ran two of these benchmarks at the same time to see whether hyperthreading had an effect in such a case. Presumably most of the synthetic benchmarks especially don't really favour hyperthreading.
Finally Linux benchmarks :). They are sooo hard to find and finally they are being used.
The saddest thing is that everyone thinks there's "competition" in the CPU market.
Two companies is not real competition. They cross-license technologies, time their releases, fix their prices.. They work together to gouge as much cash out of us as possible.
Why can I not find a decent CPU to build a terminal out of for less than 50 bucks? How much should a 1.5ghz celeron (or tbred/whatever) be worth? Not anywhere close to what we're paying.
The same thing with ATi/nVidia. Two players means they each get half the market. All the fanboy knobbery (no matter who you're a fanboy for) just builds free hype. So long as whenever anyone anywhere goes to buy a video card, the only names in his head are "ATi" and "nVidia", both companies are happy.
Coke and Pepsi did the same thing to the soft drink market. There were really no "cola wars". They colluded until they dominated and controlled the market. Did you know that vending machine companies will not sell a backlit machine for any non-coke or non-pepsi product?
Just like you all think there's a real option this election day. Yeah, I'm suck of bush, I'm voting for "the other guy". There's no choice, there's no competition.
In a competitive market, Intel or AMD could both be knocked out by a third party. nVidia could go bankrupt tomorrow, like 3DFX did. Bush and Kerry could be golfing together in January, while President Nader is being inaugerated.
This fanboy idiocy creates these situations. It's ridiculous. Quit being such a bunch of stupid douchebags. I don't want to hear whether Intel or AMD is 2% faster on paper today. Tell me about Transmeta, VIA, Cyrix.. Tell me about PPC platforms (that dont cost $3000 extra for a fancy yet unneeded brushed nickel case) Tell me about the companies that may one day offer me an ACTUAL CHOICE and quit licking the balls of your corporate masters.
I don't need no instructions to know how to rock!!!!
Xeon = 3.6 GHz, A64 = PR 3500
Xeon = Server, A64 = desktop
Xeon = L3 cache 1MB, A64 = L3 Cache 512K
Xeon = $??? (probably > 800 when available), A64 = $345 (pricewatch)
Xeon = fastest of Intel's 64-bit chips, A64 = slowest of AMD's 64-bit chips
Anandtech = sold down the river? What the hell?
The reason our eyes don't have a problem with 24 fps film is because movies have lots of motion blur! Video games have no motion blur at all, unless you're playing a PS2, in which case everything is blurry.
To be fair, the Athlon64 processor compared is a 3500+, while Intel's is a 3.6GHz. So AMD chose to rate their processor at that performance level.
To be fair again, Xeons generally outperform Pentium 4s at the same clock speed, due to various things like more cache and hyperthreading (before Intel added it to the Pentium line). The Xeon is normally targetted for servers and high-end workstations.
Finally, at the end of the article, they promise to benchmark the Xeons against the Opterons.
tasks(723) drafts(105) languages(484) examples(29106)
"Math intensive" means floating point intensive, because that is all the math normal people do with their machines. Calculating Pi to a billion digits is not floating point math, it is integer math.
The "math intensive" benchmark in this setup was Povray, and there the Athlon 64 shined. A lot. lame is also a floating point heavy application, and both CPUs are close there.
gzip measure memory performance. Apparently, the dictionary fit completely into the cache of the Xeon. Not a fair test.
I cannot comment on MySQL performance. It should measure integer and memory performance, I would wildly guess.
Bernstein's prime sieve is also integer arithmetic . If you have a prime with 100 million digits, the action is mostly in the CPU caches. Again, no fair test.
The unfairness of the benchmark setup becomes particularly obvious when you look at the chess benchmark. Chess (and other game AI type problems) do a lot of unpredictable jumps. That's the weak side of Pentium 4, and that's why Athlon 64 has historically outperformed Pentium 4s by a WIDE margin. Look at the hardware used by the PC chess tournaments and the chess grandmasters and you see Athlon and Athlon 64 all the time. If Anand now measures that Athlon 64 is outperformed by a Xeon, then the test setup can not have been fair.
I don't know about ubench, never heard of it before.
Password cracking and encryption is 100% integer arithmetic. And it is one of the mainstays of Opterons from the beginning. Anands measurement flies in the face of that.
I call bullshit.
No it's not. There are different engineering trade-offs that were made. P4 traded IPC for clock-speed. AMD traded clock-speed for IPC. All that matters is what performs the best at the retail clock-speed.
A deep unwavering belief is a sure sign you're missing something...
This is still flawed because the AMD model numbers are relative to the class that they compete against. So the 3500+ is supposed to compete against a PIV at 3500 MHz. Just like the Semprom 3100+ competes with a Celeron at 3100 MHz. The Opterons, which compete with the Xeon, don't have MHz ratings they go by pure performance.
Anandtech's claim that the upcoming PIV's are exactly like the Nocona chips are specious if for no other reason than the fact that there would be no reason to differentiate the two.
They should have at least used an FX53 for some semblance of parity.
Dissolve... Resolve... Evolve...
I think that the article demonstrates the effectiveness of cache for some applications. How much would you like to bet that the Xeon was able to run pretty much everything in cache where it won, and the Athlon 64 wasn't?
Very poor comparo.
Jonathan
Strange, I though that overheating is the problem of the past, and manufacturers are making silent coolers. Of course if you have a paranoia you might install too many coolers. By the way if you will look in charts you'll see that amd procesors produce less heat...
As part of a larger project I've recently had to evaluate these two chips technologies. I've been benchmarking the AMD Opteron 246 (2.0 Ghz) against a 3.0Ghz Xeon with 64bit and hyperthreading extensions, using the the same top end memory config, same hard drives, etc.
With the overwhelming majority of our real-world custom application performance numbers, the Opteron system was the better performer by a wide margin.
I'd suggest if anyone is making a real decision about these chips, to test them out yourself under actual-use conditions.
This review is BS. Any running program which would not fit into the Athlon 64's 512KB cache but fit into the Xeon's 1MB cache would have much better performance. Case in point, I downloaded the Windows version of the tspc181 chess program used in the article, and it showed 644KB memory usage in Task Manager. This would explain the much better score of the Xeon, as the Athlon 64 would have to be constantly swapping with main memory while the Xeon ran from the cache. Any test like this will significantly skewer the results. A fairer comparision would be a 1 MB cache Opteron or FX vs the 1MB cache Xeon.
As almost any tech reviewer would have been aware of this, one can only wonder if some money changed hands, as this article seems to be intentionally slanted to make the Xeon look better then the Athlon 64. Also synthetic benchmarks in general tend to be very unreliable, and sometimes worthless, often slanted in design to favor one CPU or another, usually Intels, since they have the most money to throw around.
This ad space for rent.
Right. No kidding this benchmarks sucks. :-)
./configure and make with no optimizations."
Seriously, it makes a great difference what version of GCC they use.
I saw a great boost in benchmarks when I switched from gcc 3.3 to 3.4 on my AMD64.
-O3 -pipe -march=k8 -fomit-frame-pointer -ftracer
That's the way to go!
"We compiled the program using
Anandtech's claim that the upcoming PIV's are exactly like the Nocona chips are specious if for no other reason than the fact that there would be no reason to differentiate the two.
Then you haven't followed the differences between Xeons and Pentium 4s. They are basically the same core with differences in FSB, SMP capability, L2 cache sizes, and sometimes the presence of an L3 cache. Otherwise, the core is the same.
In this case, the Xeon in the article is practically the same as existing Prescotts with the exception of having SMP and EM64T enabled.
I though that these benchmarks looked a little strange when you're using Jack the Ripper as one of your major comparisons. There's a nice thread going on over at Ace's bashing the benchmarks, including a post from the author of the chess benchmark stating:
this test they did was flawed in all respects.
This benchmark put up a server class CPU vs. a desktop class CPU. They should've put the Xeon up against an equivilant Opteron.
-illumina+us "I put on my robe and wizard hat..."
The act of using a 3500+ instead of an Opteron 150 is a minor issue.
/dev/null, and retest both machines.
The major issue is that Anandtech does not know how to compile software.
The Makefile used for TSCP on the A64 is broken, and does not apply -O2 optimization at the right stage.
My A64 3200+ scores 290K n/s when -O2 is properly applied.
On "primegen" most of the time is spent in putchar(), instead of in computation, and they should comment out the putchar() loop instead of directing output to
Also, they should have edited conf-cc and turned on -O2 optimization.
ubench is known to be buggy, and the AMD64 results have been questioned on other sites as being implausibly bad.
They copied their data wrong on the first database test. The A64 3500+ times in at 215 in 64b mode, beating the 3.6 GHz Nocona.
Their encoding benchmarks are equally suspicious.
And gzip was a 32bit executable.
In short, this "review" is HORRENDOUS, and filled with errors. A64 3500+ vs. Opteron 150 is a distraction from the real problem:
These guys don't know how to compile, optimize, and benchmark software.
This is the real compairson. Overclock the AMD to 3.6GHz and see who wins. As soon as AMD gets tthe 90nm process perfected I think we will see a huge boost in AMDs clockspeed.
This always annoys me...
You see, you can't buy an AMD at 3.6 GHz because it wasn't designed to run that fast. The AMD does more work per clock so it CAN'T run at 3.6GHz in 90nm. It is simply not designed to do so. The laws of physics prevent this.
The Intel CPU CAN run at 3.6GHz because it was DESIGNED to run at 3.6GHz AT THE COST of doing LESS work per clock.
If I had a CPU that could execute 2 instructions per clock at 1 GHz and another CPU that could execute 1 instruction per clock at 2 GHz, they would have the exact same performance.
They are different design styles. Sometimes the high frequency, lower-IPC approach is better, sometimes the lower frequency, higher-IPC approach is better. You can see this in the discrepency in performance of 2 CPUs with vastly different design tradeoffs.