Slashdot Mirror
Xeon vs. Opteron Performance Benchmarks
QuickSand writes "Anand got his hands on some of Intel and AMD's enterprise processors including 4MB L3 Xeons, and put them to the test. Results were a little varied as 4-way Opteron systems seemed to fare the best, although dual Xeon configurations almost always beat dual Opterons. The exact benchmarks are here."
dual xeons have owned the market for a long time...it will be difficult (although not impossible) for AMD to topple this.
many people did not upgrade to Intel's Itanium, but rather were upgrading to their high end dualie xeon systems -- they run very reliably, and very fast. a few instances where we've put in dual 2.x ghz xeons for web/mail servers...and only a slashdotting could bring them down...(well, an exaggeration...but you get the point).
We're like rats, in some experiment! -- George Costanza
Hmmm, you should read the article before commenting.
The last two articles on Xeons used their forum database as the workload for the benchmark. In the current article he even managed to use an unnamed enterprise order management system.
Then, if you have the games and the 64 bits systems at hand, why not do a quick benchmark?
Their review of windows64 highlighted some obvious problems, probably with drivers/PCI, that may be relevant for professional use (think of CAD).
Believe it or not, Intel's compiler generates very good code for the Opteron. Far better than GCC or generic IA32 compilers.
So in any evaluation, the compiler and binaries that are used is an important question.
There was no mention of this in the article.
The jist of the whole thing is that Intel's achitecture has a huge bottleneck in its FSB. All the processors share the same FSB and quickly max it out if there are more then 2 processors. So anyone building or buying systems with more then 2 processors will get much better performance out of an AMD opteron system then an Intel.
We were all warned a long time ago that MS products sucked, remember the Magic 8 Ball said, "Outlook not so good"
I attend a university that is currently building a beowulf cluster, and when it came down to making a decision, the deciding factor was price/performance ratio. While it may make sense for enterprises to go with the Xeon, the Opteron is a clear winner, in my mind, when money is an object. Of course, if you have the money to burn, the Xeon may seem to be the more obvious choice.
I've owned 5 AMD processors from the K5 to an Athlon64 and all are still in perfect working order. But these sorts of anecdotes aren't very helpful in determining average chip reliability.
Ita erat quando hic adveni.
Alright I have had about 3 AMD processors die on me. I have owned about 4 Intel processors all the way back from original Pentium. Not one has ever had a problem.
Now... given this kind of statistics, as sad as it may sound I'd say I am willing to pay anything for an Intel just to avoid the headaches.
That is an interesting use of the word statistics. In order to determine if your next processor is likely to break, you should look at thousands or hundereds of thousands of Intel procs and AMD procs. Your 7 processor study is inherently flawed.
Thank you Mario! But our princess is in another castle!
Wow, I have a working AMD 386/40. Yet I have a score of dead Intel 286/386/486's. I just evened out your "statistics". Not to mention the 5th gen and above x86 class processors I have.
Slashdot is proof that Sturgeon's Law applies to mankind.
Firstly, Anandtech uses flash for its images so that people w/o the plugin can't see the data. This forces you to install it, so that you can see their OTHER Flash pieces... ads.
Secondly, you are not going to get MS to recompile an MS-SQL for Opteron. You're not going to get IBM to support a Linux installation, after you've rolled your own ueber-NUMA-patch-level-42 kernel.
The test was clear - out of the box, plug in servers, load OS, load app, run benchmark.
And the outcome was clear, the Opteron architecture is vastly superior, both performance and price-wise.
The MHz myth is over, at least in Slashdot and Anandtech circles.
I want to delete my account but Slashdot doesn't allow it.
I disagree. I'm running a Windows XP workstation with dual 2.4 GHz Xeons, and I'm not at all disappointed... neither are the 50 or 60 other developers surrounding me which are running on the same boxes.
What exactly would be our grounds for dissapointment?
That your company spent $3750x2x55= $412500 on processors alone (assuming you have the 1mb MP model Xeons), when you could have the same performance for a quarter of that price.
When anger rises, think of the consequences.
Confucius (551 BC - 479 BC)
need to buy new motherboards
This makes me want to throw up. The last motherboard purchase I made, it was a chore finding one with the _least_ amount of features. Need an AMR riser slot? Fuck no, I'd rather have the PCI slot back. Need integrated sound? No, integrated sound makes my already bad speakers sound worse. It must've been tough figuring out how to make a decade's worth of improvements in technology amount to nothing. I have an ISA soundblaster from 10 years ago that sounds better than the onboard sound on my last motherboard. Need integrated video? I won't begrudge you this. Some people build clusters with their motherboards, and a video card is needed to boot, but if I have a choice I won't buy a mobo with integrated video.
In short, I want a motherboard with slots for RAM, an AGP slot, a socket/slot/hole for a CPU, PS/2 hookups, serial and USB connectors, and the rest of the board filled up with PCI (or PCI express) slots. That's the ticket.
You will find that most high-end Xeon systems are also NUMA systems. IBM, Unisys, HP all construct their really big xeon boxes as NUMA-clusters of 4-processor SMPs. They create a distributed memory machine at the chip-set level. This is actually what the opteron does, except that the chip-set (well, the memory controller part of it) is built into the processor.
I think the above poster had the correct idea about NUMA, but worded it in a misleading way. A NUMA design (either of opterons, or of Xeon-quads) will have to do some memory access through the memory controllers on other nodes. This increases the latency of memory access, and can clog up the inter-processor links if lots of memory loads/stores go to remote memory. Thus NUMA-aware operating systems and system libraries are necessary to maximize the amount of memory access that is local, and minimize the usage of the inter-processor links.
While the opteron design is elegant, and fast, it is not the only smart way to do things. It offers great aggregate memory bandwidth, but can slow things down in the worst case. Most large NUMA systems are created by linking 4-way SMP nodes. (Examples: Sunfire, HP alphaservers, Cray X1, NEC SX-6, Unisys 7000, IBM xseries 4xx xeon, IBM xseries 4xx itanium,...) Apart from opteron systems, the only systems I can think of that do NUMA per processor are the cray T3E, SGI origin, and intel paragon, all of which are Massively parallel supercomuters.
It is safe to say, however, that a shared bus system does not scale well beyond a few processors. This is best demonstrated by the 36 processor SGI-challengeXL, which was significantly bottle-necked at the memory bus.
food for thought.
How about cost? The Xeons cost twice as much as the Opterons, and the Opterons give equivalent or better performance! Although you are correct that the performance difference may not be staggering (and between top of the line chips, who would expect it to be?), the price/performance ratio certainly is.
When you lose something irreplaceable, you don't mourn for the thing you lost, you mourn for yourself. - Harpo Marx
The purpose of the test is not to test the memory, but to test the processors. Thus, they used the same memory in testing each processor configuration.
= 9. All on dual processor configurations. There is definitely no Intel bias in that test.
One of the purposes of the test was to show how the memory bandwidth bottleneck of the Xeons limits their effectiveness in 4-way configurations, which the Opterons do not have that problem. Doing this comparison with different memories would make things more complicated.
Additionally, you'll notice that Anand's final words recommend the Opteron for being at least equivalent and much cheaper than Xeon. This was also the selection process for their new forum servers, so you can bet that they aren't getting any kickback from Intel, or those would be Xeons.
If you still have doubts about the validity of Anandtech's testing, check out the benchmarks from their AMD vs. Intel web server test in December: http://www.anandtech.com/IT/showdoc.html?i=1935&p
Really, I think some people ought to think before they flame like this. The benchmarks are showing the Opterons to be equivalent or faster in 2-way configurations and definitely faster in 4-way configurations, so what is there to complain about? The fact that Anandtech has consistently recommended AMD's processors just makes it doubly silly.
" Can somebody tell me if the IA-32e processors will be in the socket 478 format to work with existing boards, or will they require a whole new socket and chipset (rather than a bios update)"
They're disabled in all socket 478 chips. The new Pentium 4e chips (Prescott core) supposedly have the extensions, but they remain disabled. Technically, it may be possible to gain access to those instructions through some sort of BIOS hack, but you would also need to use an Operating system that can detect, support, and make use of those new instructions. Also, you risk using unfinished or untested parts of the CPU if you do manage to gain access and use the extensions. There would be no benefit other than simple tinkering.
"If they really are just "extensions" then I don't see why anything special would need to be on the motherboard correct?"
You still need a CPU that supports the instructions, and which has them enabled. Technically, if Intel released Prescotts in S478 form with IA-32e enabled, it should work fine with an existing motherboard which would otherwise support the Prescott chip you're using. The probablility of Intel taking the time and effort to do this less than a quarter away from a whole new socket is virtually nil.
"The cpu should switch into 64bit mode whenever the OS tells it to right?"
That's not entirely accurate. Technically, what happens under AMD64 (the basis for IA-32e), is that specific instructions can be sent to the CPU to have it run code in what's called "Compatibility mode", which essentially allows it to behave as though it were a 32-bit CPU. The difference is that you're not 'switching' to 64-bit mode. You're either in 64-bit mode with the option for compatibility mode when needed (meaning you need a 64-bit capable OS), or you're in 32-bit and you're stuck in 32-bit.
If you're looking for 64-bitness, you may simply want to get an Athlon64. If you're waiting for 64-bitness on the Intel side of things, you'll be waiting until some time towards the end of this year. Good luck.
-- "Government is the great fiction through which everybody endeavors to live at the expense of everybody else."
4 Xeons (@Intel's announced pricing): $14768 ($3692 ea)
Did the quad Xeon system outperform the quad Opteron by a factor of 2.5:1? No. In fact, in some cases, the quad Opteron outperformed the quad Xeon. The Xeon had advantages of hyperthreading, 4x as much cache, and a clock speed 800mhz higher than the Opteron, ans still got beat.
Clock speed may sell in the consumer market ("Me want bigger!"), but in the server market, Opterons getting better performance for half the price are going to win more and more converts.
- Greg
Start a happiness pandemic
*raises hand* I'm a corporate IT type, and I read his benchmarks. Along with about three others on a regular basis. Because sometimes, "real work" tends to scale pretty similarly to game performance--especially when that real work is a lot of 3D graphic operations.
"America has done some terrible things. But I know that Americans don't cheer when innocents die." -Dave Barry
They WILL standardize on a socket, it's just that the socket will be Socket 939 and not the current one.
It's pretty much the same story with SlotA/SocketA. They had an initial design that was quickly replaced. The second socket then stuck it out for the duration.
Intel did pretty much the same thing with their P4, initially releasing it on socket 423 and then quickly moving to socket 478 which has lasted for several years now (though it too will soon be replaced).
Markets change, technology changes, and sometimes sockets need to change with them. Remeber that the specification for Socket 754 and Socket 940 for current Athlon64 chips was set in stone about 3 years ago (before the first beta chips tapped out), and a lot has changed since then. AMD has gone to great lengths to minimize socket changes, but there's only so much that they can do.
Definitely not BS, though whether or not it's useful depends heavily on your application.
The idea behind hyperthreading is that the P4's long pipeline will often stall with only a single thread going through. With hyperthreading you run two threads at once, so when one thread stalls you just start up the other thread and go with that one for a while. In a way it's almost like a poor-mans dual-processor system, giving you two logical processors on a single chip.
Now, obviously there are a few things to consider here. First off, if ALL of your processing is being done in a single thread then you aren't going to see any benefit to hyperthreading, and in fact the extra overhead might even make things a bit slower (usually only 1-2% slower).
Games almost always do all their major processing in a single thread. Even if they have extra threads hanging around, you almost always spend 99%+ of your time in a single thread. For this reason, games see virtually no benefit to hyperthreading (they don't see much/any benefit from dual-processor setups either).
On the other end of the spectrum, some applications see up to a 25% performance boost when hyperthreading is enabled. The tests I've seen show the biggest improvement have been things like Photoshop and rendering applications. Some server applications should benefit as well.
The other boost that hyperthreading gives you, like with a real dual-processor setup, is that it makes multitasking a bit "snappier". This is by no means a night-and-day difference here, but it is there.