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."
Test Results can be found here:= 6
http://www.anandtech.com/IT/showdoc.html?i=1982&p
Thank you Mario! But our princess is in another castle!
I remember AMD's K6-3 would blow away the K6-2 at the same clock speed with the major difference being the cache.
Whomever is citing Anandtech as claiming the dual Xeons almost always beat the dual-Opterons needs to read the article again. Both Architectures in a dual configuration tended to perfom about the same with Opteron and Xeon each winning some of the time. The Opteron scales better above dual configurations. However the Opteron is HALF the price of a Xeon! Cost/performance (or else we would all have 12th generation DECAlphas or Power5s by now) is easily handed to Opteron. Nice spin!
Avrice
At two processors Xeon is still ok because the bandwidth of the memory coherency still isn't in serious contention. However, as the systems scale larger support for NUMA is critical to reducing memory latency because it means that memory does not have to flow in from the controllers on other processors.
That is why Opteron is required for good performance with eight to sixteen processors, and you can even see the improvement on the four way tests that Anand ran.
How could this get modded up? The article doesn't use game benchmarks at all. The poster obviously just wanted to get the first meaningful post and didn't even bother looking at the article.
Did you even read the review? Why don't you read before you end up making yourself look bad.
And Anand's Win XP 64 review wasn't bashing it. He was using games to highlight that there are still several things that need to be addressed (Like drivers) before it will be ready for prime time.
Some info here. SSE3 is the big thing.
Belief is the currency of delusion.
Why can't AMD make a Dual AMD64?
They DO. It's called an Opterion!
Guess you are refering to the Athlon64 which is a one way processor (just like the Pentium 4).
BTW the Opterion is made in 3 flavors, the 100 series is a SINGLE way cpu with NO smp support. It's very nearly the same as the Athlon64-FX. The Opterion 200 series is a 2 way (2 cpus), and the Opterion 800 series supports up to 8 cpus. AMD dropped plans for a 400 series, but you can use the 800 series chips to build servers with 1-8 cpu's.
I don't really know, but I think Intel's 64-bit chips will probably use a Tejas-style clip system, not pins. Technology that won't work in a current motherboard. But, once AMD upgrades to socket 939 for the FX-51, it won't work in current boards either. AMD and Intel are both set to release the new sockets at about the same time PCI-Express comes out, so upgrade-happy people will need to buy new motherboards anyway.
If my answers frighten you, stop asking scary questions.
Why can't AMD make a Dual AMD64?
They do - it's called an Opteron 2xx.
"We returned the General to El Salvador, or maybe Guatemala, it's difficult to tell from 10,000 feet"
They call it the "Opteron".
Let's use the Athlon64-FX as a base.
Relative to that, the other chips are:
0. self
1. Plain Athlon64 (not FX) has narrow memory.
2. Opteron for 2-way systems
3. Opteron for 4-way systems
4. Opteron for 8-way systems
So that's 5 chip models.
Full text
Technically since the new Prescott chips have the instructions on them, yes you could buy a 478 Prescott. You would just have to hack the chip to activate the instructions (good luck with that!).
As for a real IA-32e chip with the instructions enabled, Intel has stated that they arent coming out for while, and since Intel is moving the P4 to the new 775 chipset in a few months, I would bet that they would also be released under this new chipset. Heck, they might not even release their IA-32e chips within the lifetime of the 775 chipset.
Personally my Athlon 2600 just fine enough for me.
Intel has said that the initial 64-bit chips will be Xeon server processors -- which don't currently use Socket 478 anyway.
Believe it or not, Intel's compiler generates very good code for the Opteron. Far better than GCC or generic IA32 compilers.
This is the experience I've had with the Intel Fortran compiler (ifort) on an Athlon XP. Codes compiled with ifort are around twice as fast as those compiled with GNU g77 (for Fortran 77), and around 1.5 times faster than those compiled with Lahey lf95 (for Fortran 95).
Tubal-Cain smokes the white owl.
the very definition of l3 cache, is level 3 cache, hence the name, l3. do you not remember when l2 cache used to be on the mobo?
When I tell an object to delete this, am I killing it or telling it to kill me?
Anand seems to conclude something a bit different then the submitter:
The comparison we've made here is a very important one; it identifies Intel's strengths and their weaknesses with Xeon, and it crowns Opteron a clear multiprocessor winner. An area that we didn't touch on is cost, which is where AMD truly shines. The Opteron 848 processors we tested are around 1/2 the price of Intel's 2MB L3 Xeon MPs and we have not seen retail data on how expensive the 4MB parts will be.
In a 4-way configuration AMD's Opteron cannot be beat, and thus it is our choice for the basis for our new Forums database server.
L1,L2,L3 caches relate to caches either on the CPU or very close to it. Only when a bit of code isn't found in the L1,L2[,L3] cache, does the processer go all the way out to the system memory to fetch the next set of instructions.
Now, a database generally runs a fixed set of instructions during a query, though those instructions may not fit in the smaller L1 and L2 caches (the L2 on the dual Xeon we just bought was 512k, but 1mb is also available). Having the larger L3 cache can help keep those instructions for doing queries closer to the CPU, which can significantly speed things up.
As for database design, good database will use available index(es) when possible during a query since they shortcut looking for data. This is why having proper indices is important to maximize database performance.
The database will use as much memory as you allow it to cache data. Only when the data isn't resident in memory does it go out to the HDDs to find it, regardless of whether the query was able to make use of index(es).
Both databases and CPU caches make use of MRU/LRU (Most/Least Recently Used) to determine what to keep and what to throw away. The MRU stuff stays, and the LRU stuff is flushed out when something new is stored in the cache/database memory.
A complex SELECT statement that returns millions of rows, but makes use of 2mb of instructions would significantly improve performance with the addition of a 3mb L3 cache since the CPU wouldn't have to go to system memory to find the next set of instructions, per row. Save a few ms here and there fetching instructions, and it adds up over time.
It's interesting to notice that in these tests, the Opterons were clocked quite a bit slower and had a lot less cache than the Intel CPUs, yet performed comparably in 2-way and better in 4-way than the Intel chips.
The Opteron clocked at 2.2ghz with 1MB of cache was very close in 2-way performance with the Xeon 3.0 and 3.2 ghz each with 4 and 2 mb of cache respectively. The 1.8ghz Xeon compared well with the Xeon 2.8ghz with 2MB of cache. The Opterons were typically within 3% or so of their Intel counterparts in 2-way benchmarks and closer to 10% ahead in 4-way.
If nothing else, this says a lot about the efficiency of the Opteron's design. Less silicon, and more importantly for AMD, less expensive silicon, manages to achieve very close results.
L3 is just the next larger cache after L2. Where it is in the physical circuitry is an implementation detail.
As a general rule: Get rid of the consultant
(B) + (D) + (B) + (D) = (K) + (&)
$332x2x55=$36,520
Those are 2.8ghz Xeons, btw, so more expensive. They're also the price on their own, bought singularly. $37k for 55 developers CPUs is pretty good, and could undoubtedly be bought down to nearer $20k when bought bulk.
Seriously, price arguments about Intel don't work that well any more. They used to cost shedloads, now it's only their top-of-line processors (such as the extreme edition, and new models) that carry the weight (after all, if you need the fastest processor available, you can afford to pay for it :-P).
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.
Actually, one can start to get a statistically significant result at around as little as 30 data points depending on a few factors.
For giggles, let's take a large population (the MS Windows install base: ~300M) and if I wanted to be 95 percent (SixSigma!) sure that I calculated within a confidence interval of plus/minus 5 points, how many must I sample? Thousands? Hundreds of thousands?
Nope, a scant 384! A generally accepted confidence interval is +/-4 at 95% and you see this used in several polls. To get to that level, I'd need to sample only 600 of the 300,000,000 population.
See for yourself.
Hmm... moderated yourself nicely.
IBM supports native x86_64 distros (SLES8/AMD64 now, and looks like RHEL3/AMD64). So IBM supports you running 64-bit today.
Windows is working towards 64-bit for Opteron, so yes, MS is recompiling their crap for Opteron.
Simply because there is a logical reason for images being flash makes it no less annoying as hell.
Opteron is a fascinating platform, and very cool, *especially* with respect to 64 bit computing.
XML is like violence. If it doesn't solve the problem, use more.
unless Intel implements their own support for HyperTransport, uses AMD's 8000 chipset or comes up with their own switch fabric architecture, they're not going to beat AMD for 4,8,16 and 24 way systems. Pumping up Front side bus isn't going to fix the problem. Why do you think IBM, NEC and HP all use their own cell chipsets for their large 64 way servers, like HP SuperDome? It's because FSB blows chunks under those conditions.
But the Xeon you show isn't the one in the review! The review was for a just released mega-super-expensive 4MB cache version of the 3.4GHZ Xeon.
XJS*C4JDBQADN1.NSBN3*2IDNEN*GTUBE-STANDARD-ANTI-U
In fact, in some cases, the quad Opteron outperformed the quad Xeon.
Wasn't it in fact in _all_ cases?
"although dual Xeon configurations almost always beat dual Opterons."
/. ?
Perhaps the submitter's screen reader doesn't work well with flash, but in the 2-way benchmarks, Opteron was on top twice, and Xeon was on top 3 times. All the 2-way benchmarks were fairly close (within 5%), and the Xeons never beat the Opterons by a margin greater than 1.7%. I don't quite know where 40% wins translate into "almost always" loses. In other words, the story submitter is a moron, or simply didn't look at the article.
"Results were a little varied as 4-way Opteron systems seemed to fare the best,"
Seemed? Let's see, out of five 4-way benchmarks, Opteron won... all of them - performing about 10% better than Xeons each time.
Since when did we start letting Tom Pabst submit articles to
Note to editors: When the submission is non-sequitur, either reject it or edit it.
-- "Government is the great fiction through which everybody endeavors to live at the expense of everybody else."
"When Intel made the PIV, they either refused to license AMD to build compatible CPUs, or AMD declined to manufacture them."
Actually, AMD came out with an entirely new "socket" with the original K7, the Athlon. It actually debuted in a slot form factor, but the resulty remains the same - AMD has been independent of Intel sockets since '99. AMD and Intel have numerous cross-licensing agreements, which is why Intel is offering an AMD64-compatible bunch of CPUs fairly soon. AMD could have used socket 423 or socket 478, but why would you want to redesign your chip to work with a socket that'll make it slower? It also would have killed upgrade routes for AMD's existing customers had they gone to Socket 423 when the P4 came out.
Most people seem to forget that the Athlon family was originally competing against, and beating, the P3 long before the P4 arrived. Thus, AMD came up with a chip that has competed against two major products from Intel for about 5 years now. That's impressive no matter which way you slice it.
-- "Government is the great fiction through which everybody endeavors to live at the expense of everybody else."
Dual systems usually use Xeon DP, which are a LOT cheaper than Xeon MP (not to mention faster due to the faster FSB, despite having less cache) (the same is also true for the Athlons, the 248 is much cheaper than the 848, in this case the chips though are identical safe the HT links).
That "Xeons cost twice as much" statement is correct for the Xeon MP 3Ghz/4MB cache vs. Opteron 848, which was the primary focus of this article.
The current opteron memory bandwidth per memory controller is 6400MB/s. So it is twice that in a 2 way system, and four times that in a 4 way system. At least if properly configured. The memory bandwidth in an Itanium2 system with it's shared FSB is 6400MB/s total. Opteron should do much better in 2-way and 4-way benchmarks once OS's begin to optimize for a NUMA architecture. But even with out that the performance is quite good. The only thing I have seen that the Itanium2 has is a larger physical address space support. Which is great if you are building a 512 CPU box and irrelevant everywhere else.
Try the printable version of the article...
"Currently Intel's latest 3.0+ GHz offerings are spanking Athlon 64s in benchmarks with 32 bit applications."
...
What a bunch of crap! That's almost as big a lie as Intel makes of AMD cpu's. I didn't even bother reading the rest when you are obviously delusional.
BTW A64 chips can be had for the same prices as their A32 counterparts in the same speed rating, Motherbaords are almost equally expensive whether A32 or A64, & outperform them by up to 30%.
Ok well now I have read the rest of your comments & I have to reply to those as well:
"they're going to go to Intel because it has more upgrade possibilities"
What possibilities? They force upgrades way more often than AMD, & are known for being the best money sink for performance users.
"is cheaper than the Athlon 64 for the same level of computing power"
Uh not really, A64 costs the same or less than Intel for comparable performance as long as you don't follow the rule of 'Mhz/Ghz equals performance'. Take a look at pricewatch or Newegg and see for yourself...
"currently performs better"
Hogwash.
"So this is more of a plea for AMD to extend the Athlon "32" line a bit further. Please AMD, don't prematurely kill off 32-bit Athlon chip development!"
Uh they are building them at least through the end of this year, & probably next to some degree. The thing is their is no real future for A32, performance has been decreasing performance-wise in comparison & wasn't keeping up. AMD realized their best bet was to focus on A64 with it's integrated memory controller & higher IPC than A32 has. A64 is a real contender where as A32 wasn't keeping up & they knew it.
we are all invisible unless we choose otherwise
I think time will show (barring any major gaffes) that AMD made the right move. First and foremost, the Athlon XP appeared to be reaching the end of its run. Cranking up the speed of the processor didn't really seem to yield comparable performance gains. In the early days the performance ratings on the XP line were a little iffy, but they were close enough that most people didn't really care. With the XP 3000, and 3200 though those ratings were dubious at best. The speed ratings ratings for the Athlon 64, however, are pretty accurate, even understated in some areas.
I recently upgraded my principle system and at the time, I was determined that the P4 2.8HT CPU was going to be the chip the machine was built around. It clearly trounced the Athlon XP 2800 in all tests for about $100 more. I was about to order when the AMD Atlhon 64 3000+ was released. It outperformed the Intel chip in most areas, was 64 bit, backwards compatible, and only $20 more. In my opinion, and Anandtech and Tom's Hardware agree, the AMD Athlon 64 3000+ is the best bang for the buck CPU out there.
Pricewatch's Lowest Prices are...
AMD Athlon 64 3000+ - $205
AMD Athlon 64 3200+ - $251
AMD Athlon 64 3400+ - $401
Intel Pentium 4 3.0 (800) - $209
Intel Pentium 4 3.2 (800) - $273
Intel Pentium 4 3.4 (800) - $420
AMD Athlon XP 3000+ - $158
AMD Athlon XP 3200+ - $194
Looking at these benchmarks here and on the following page here. You can see, with the exception of pure media encoding, the Athlon 64's perform better. The margin is slim in some areas, and quite large in others. All this and the chips are a little cheaper.
In my opinion, the XP line was dead. It had gone as far as it was going to go. I think AMD knew that the 3000 and 3200 were more like the 2850 and 2900 and they weren't going to get any more mileage out of the design. I definitely think AMD made the right call putting all their eggs in the Opteron/Athlon 64 basket.
"The words of the prophets are written on the Slashdot walls."
> Now its obvious you're trolling. [...] The long and short is that all PC sounds cards equally suck until you get to professional grade gear.
You must be joking. Most of the integrated sound I've had the joy to listen to produced noticeable background-noise. The most obvious one was the Eden-M board. Most mb-producers don't give much about seperating the analogue part from the digital, so accompanied with an intergrated graphics card, you can practically hear how a window is restored. It is usually not the quality of the on-board sound, which sells it, but purely the capability of producing some sound.
"Between strong and weak, between rich and poor [...], it is freedom which oppresses and the law which sets free"
The 4MB Xeons MP are MUCH more expensive than the normal dual Xeons.
One 3Ghz XEON MP cost about 3500$
HI O WISE PRINCE. WHT TOOK U SO DAM LONG?
L3 cache is a generic term. You can think of the "L" as meaning "Level". L3 is 3 "steps" away from the CPU registers/core. Think of it this way (although it isn't necessarily implemented this way as there are optimizations that can be done to make this faster):
The CPU needs a value. If it isn't in a register, look in L1 cache first. If not in L1 cache, look in L2 cache. If not in L2 cache, look in L3 cache (etc through as many L* cache as you want). If not in L* cache, go ahead and fetch it from main memory.
Whether or not L3 cache runs at the core clock frequency or whether it is on die or not is completely dependent upon the implementation (which CPU are you talking about?).
For the P4 Xeons, the L3 is on the die and runs at core frequency (just has a few clocks of latency to access it).
If you're interested in the architectual aspects, you really need to watch the presentation linked to by Ace's. It's an hour long, but it's well worth it if you're into that sort of thing. This is the 'father' of AMD64 talking about his baby, and then taking questions from reasonably intelligent people at the end.
-- "Government is the great fiction through which everybody endeavors to live at the expense of everybody else."
Not only that, but when they were comparing prices, they said that the 2Mb cache Xeons were twice the price of the Opterons. Howevere in the article they tested the *4Mb cache Xeons*, which you can expect to be even more expensive.
In terms of bang for your buck Opterons rock.
meh
No Xeon MP as fast as 3.0 GHz or with the 4MB cache is available via pricewatch.
The fastest Xeon MP processor available they list is 2.8 GHz with 2MB cache, for $3788
A single Opteron 848 processor costs $1469
Oh, lookie, the AMD processor is less than half the price of a Xeon product inferior to the one that article benchmarked it against. Retract your lies.
Actually, AMD came out with an entirely new "socket" with the original K7, the Athlon. It actually debuted in a slot form factor, but the resulty remains the same - AMD has been independent of Intel sockets since '99. AMD and Intel have numerous cross-licensing agreements, which is why Intel is offering an AMD64-compatible bunch of CPUs fairly soon. AMD could have used socket 423 or socket 478, but why would you want to redesign your chip to work with a socket that'll make it slower? It also would have killed upgrade routes for AMD's existing customers had they gone to Socket 423 when the P4 came out.
I'm not disputing that AMD came out with their own 'socket' with the Athlon. I was wrong, in fact, about the generation of CPU that forced AMD's hand.
AMD came out with their own packaging starting with the Slot-A Athlon. This is because Intel refused to grant them licenses to manufacture Slot-1, and then Socket-370 compatible CPUs. As a result of this, AMD was forced to go its own way and develop independent CPU/Chipset x86 platforms. It was either that, or continue making Socket-7 compatible CPUs and spiral down the obsolete technology commode.
I'm pretty sure that, had they been granted a license, AMD would have manufactured Slot-1 and then Socket 370 compatible CPUs and saved themselves the costs of developing their own chipsets, partnering with other chipset manufaturers to develop Athlon chipsets, convincing motherboard manufacturers to build on their platforms, etc. As it was, once they had gone down the Athlon path, there was no going back.
"Anyone that has ever gotten an idea based on any of my work and done something better with it-good for you."--J.Carmack