Opteron vs Xeon - a good comparison

[Erwin]

Read this for web and database server comparison:

http://www.anandtech.com/IT/showdoc.html?i=1935

Useful.

[The Chief]

Read this for web and database server comparison:

http://www.anandtech.com/IT/showdoc.html?i=1935

Useful.
Thanks for the heads-up :)

[Zachery]

So, just more truth about the AMD systems destroying most Pentium classed systems? I could ahve told you that back in 2001 :p

[Paul M]

I could ahve told you that back in 2001 :pWhich is almost when that was written ;) (well okay, December 17th, 2003, still 2.5 years ago). There are much faster Xeons around now (and presumably Opterons as well ?). Again, a more recent, real life comparison would be nice.

[Zachery]

Still, even if its dated, its true, the fastest Xeons still don't quiet catch up to their AMD counter parts.

I'm going to be switching several of our current Xeon systems to Opterons if I can find a host who supports both SCSI and the Dual Cores. Of course I haven't been looking that hard as I've been busy with projects.

[FlyBoy73]

There is a recent review (late 2005, I think) that put Intel's latest Xeon against Opteron again and the Opteron stomped it with no problem in almost every category.

Here are some other links to Opteron vs. Xeon reviews:

April 22, 2003 - Dual of the Titans
http://www.tomshardware.com/2003/04/22/duel_of_the_titans/index.html

March 2, 2004 - AMD Opteron vs. Intel Xeon: Database Performance Shootout
http://www.anandtech.com/IT/showdoc.html?i=1982

Aug. 13, 2004 - Workstation showdown: Xeon vs. Opteron
http://www.infoworld.com/article/04/08/13/33TCworkstation_1.html

Sept. 13, 2004 - Intel Xeon 3.6 (Nocona) vs. AMD Opteron 250 - Dbase Test
http://www.anandtech.com/IT/showdoc.aspx?i=2205

I'm still looking for the latest shootout I read a few months back where AMD openly (In 100's of newspapers across the USA) challenged Intel to put up or shut up. Intel finally delivered one of their new Xeons and the Opteron that was not even their latest hot thing spanked it.

With the NUMA standard and utilization of hyper-transport Intel is just out of luck and I don't forsee that changing any time soon. AMD is far ahead in chip architecture.

[JohnBee]

There have been some recent reviews of intels latest dual core engineering samples. If memory serves me correctly intel came in at a much lower clockrate than the current crop of Opteron solutions (FX-60) but in turn won the overall benchmarks comparisson between the two by a good margin.

I think they pitted the new intel (2.6ghz) against a 3Ghz dual core Opteron series processor. Although this is not a current technology comparisson it does show that intels upcoming chip solution is strong and should have no problems given AMD a good fight the money.

Nonetheless as things currently stand anyone investing in a XEON workstation or server should have there marbles checked :p

[FlyBoy73]

Do you have a link to one of the reviews?

Intel (and fans) likes to try and point outa couple of specific areas like video rendering in one or two programs and a few other categories where they excel; however, that really isn't relevant to web hosting as far as I know.

Ok, check these links out.. :)

Jan 1, 2006 - Dual core server duel: Xeon vs Opteron benchmark
http://sharikou.blogspot.com/2006/01/dual-core-server-duel-xeon-vs-opteron.html

GamePC Review - Intel “Paxville” Dual Core Xeon (compares to Opteron deeper in the review)
http://www.gamepc.com/labs/view_content.asp?id=paxville&page=1
Performance review between Intel dual core "Paxville" vs. Opterons (ouch!)
http://www.gamepc.com/labs/view_content.asp?id=paxville&page=9

Here is a quote I pulled from the GamePC review that sums it up quite well...

Unfortunately, even a solid platform can’t help Intel’s performance numbers, as their new dual-core chips (while powerful in their own right) simply are bested across the board by AMD’s dual-core Opteron processors. Even worse, the Opterons typically perform much better while running at slower clock speeds and only having half the amount of on-die L2 cache to utilize. AMD’s chips also consume far less power and run quite a bit cooler, giving AMD an edge on nearly all fronts.

Here is another great review re: Opteron vs. Xeon, where an old Opteron steps on a new Xeon.

http://sharikou.blogspot.com/2005/11/benchmark-obsolete-opteron-246-beats.html


Wednesday, November 16, 2005

INTEL has been very shy these days, Paxville was launched over a month, and it's nowhere to be seen, the usual hardware sites such as tomshardware and anandtech have no benchmarks for it. Only GamePC had a benchmark showing dual core Paxville at about 50% of Opteron in Apache benchmarks. Now the rumor is INTEL started sending its Dempsey dual core chips with dual independent bus to reassure its customers, under NDA.

2CPU.com apparrently got a set of Dempsey dual core chips through some NDA crack and was able to do some tests. The candidates are:
Bensley: a 3.46GHZ Dempsey with 1066MHZ dual FSB dual core Xeon on Blackford chipset
Irwindale: 3.8GHZ single core Nocona Xeon with INTEL 7520 chipset
Opteron old dog: the 130nm, 3 year old, Opteron 246 single core at 2GHZ. Currently, the AMD's hot selling top dog is the 2.6GHZ dual core 90nm Opteron 885SE used in Sun's Galaxy line.

I don't think anyone is expecting the $100 single core Opteron 246 to sweep the floor on the $3000 2007 dual core Dempsey/Bensley, nevertheless, the single core Opteron 246 managed to beat INTEL's future dual core chip on some tests.

Sandra 2005 x64 -Memory: Opteron 246, 10400MB/s, INTEL Dempsey 3.4GHZ dual core, 4838 MB/second. Single core Opteron 246 is 200% of dual core Dempsey 3.46GHZ.

Sandra 2005 x64 - Cache/Memory @ 64MB: Opteron 246 single core, 5485MB/s; INTEL Dempsey 3.46GHZ dual core, 4430MB/s. Single core Opteron 246 is 25% faster than dual core Dempsey 3.46GHZ.

Cinebench 2003 x64: Single core Opteron 246 scored 366, dual core Dempsey 3.46GHZ scored 377, a virtual tie with less than 3% difference.

3DS Max 8.0 Ape: Single core Opteron 246 finished in 126 seconds, Dempsey dual core 3.46GHZ finished in 133.8 seconds, Opteron single core 246 is 5% faster than Dempsey dual core 3.46GHZ.

ScienceMark 2.0 FINAL x64 Mol-Dyn: Single core Opteron 246 finished in 26.97 seconds, dual core Dempsey 3.46GHZ finished in 32.97 seconds, Single core Opteron 246 is 20% faster than dual core Dempsey 3.46GHZ.

Power consumption: Dempsey system uses 438 watts, the old single core Opteron uses 262 watts. Remember, newer 90nm Opterons made on Dual Stress Linear Strained Silicon on Insulator technology uses even less power.

Conclusion: INTEL architecture is hopeless, even an obsolete 2GHZ single core Opteron 246 can beat its 2007 future chip on performance, and INTEL's heat production problem is beyond cure.

Update: I found that the Opteron 246 has been discontinued. The current highest AMD CPU is the Opteron 885SE.

[Erwin]

Wow, great links, keep them coming. :)

[FlyBoy73]

All the info that is out there really doesn't leave a lot of room to try and compare them in a head to head heat. I would love to see a shootout between the AMD's latest Opterons against anything Intel can put up against it. I think the results would be highly embarrassing to Intel and that is probably why they won't go head to head officially. If the 246 single core vs. Xeon dual core was bad, this would be shameful.

[kmike]

Intel has some really good processors in their sleeve, set to be released near the end of this year, and based on a completely new architecture (codenamed Conroe). Their performance is approximately 15-20% higher than that of similarly clocked Athlon FX processors, though in business and gaming workloads. You can find some numbers from the leaked engineering sample here:
http://www.xtremesystems.org/forums/showthread.php?t=95021

Please note that the benchmarks were mostly synthetic, we'll have to see how the new architecture fares against Opterons in a typical web serving/db workload, but so far it looks very promising. I however wouldn't wait another 6-8 months for the new processors...

[FlyBoy73]

I need good power last year... ;)

Let me re-emphasize that I am far from "brand loyal".. I will go with the proven winner over time no mater who it is..

Now, to "Conroe" processor, which is at least 6+ months out and is compared to the current AMD FX-60 processor which is not the FX-62+ which will also be out at the same time as "conroe" or maybe before. In that test, the "Conroe" processor wins in the typical enviroments, BUT if you dig into this 'test' you will see that Intel did the setup and test for the AMD system which many are calling highly suspect. Why did they refuse to supply Anandtech or Tom's Hardware (each authorities with great reputations) with a sample to perform an independent test? This is like casting polls. Anyone who wants to see certain results can poll to get the desired results.

Most importantly for those of us running web servers... These processors are built for gamers & graphics guru's.. When was the last time your vBulletin got into gaming loads and processes? That truly (IMO) is like comparing apples and acorns.. Or maybe apples and rocks?

When Intel comes out with a more powerful processor than any higher-end Opteron on the market today - that won't cook eggs for breakfast - I'll look at it as an alternative.. Until then... I'll await the independent reviews..
:)

David

- And to think I used to hate AMD and said I would never use one again..

[eva2000]

Dual opteron 2xx cpus will probably beat conroe for mysql tasks, NUMA support in opteron 2xx server cpus/boards give crazy bandwidth for memory which mysql loves alot!

[FlyBoy73]

The more I have read up on the Intel via AMD "battle" the more I hear about the "Conroe" release and how it trounced the AMD FX60 processor. While wading through some of this I found an interesting article relating to the actual test that took place and how the AMD FX60 system was setup - very badly by Intel... Of course, accidental we're all sure.

Take a read of this if you care about the new Conroe processor and how it beat the FX60 with its arms, legs and ram tied behind its back.
http://voodoopc.blogspot.com/2006/03/if-only-they-had-time-machine.html#links

[kmike]

FYI, Anand has posted an update which covers most of these issues:
http://www.anandtech.com/printarticle.aspx?i=2716
The scores didn't change much, except for the F.E.A.R, but it was Anand's fault - he tested the systems at the different settings. So F.E.A.R. scores are more in line with the other results in the update.

I suspect it's not enough to convince you that Intel didn't play dirty there, since even the abovementioned independent results from the leaked sample didn't do. The point is, the tests gained a lot of publicity for Intel, as was intended, and for a good reason. You can shrug them off, but the fact stands - Intel is already moving, and AMD is not. I really hope it didn't fell on deaf ears at AMD, since the placid days of reaping the fruits of its K8 architecture are already numbered. I find it alarming that there're still almost no information apart from speculations about AMD's new architecture - K9 or whatever it was called.

Putting fanboyism aside, the CPU competition is better for us customers, so I'm looking forward for Conroe (and an accompanied AMD price drop :) )

[dan35]

I used to hate AMD, but now they make a quaility product, and I am a big supporter of them :)

[FlyBoy73]

I think Intel doing that would be very risky, so if it was someone else's "fault", cool... But, unless this changed as well, I think AMD should have been able to deliver their own computer, or at least Anandtech.

I am always happy to see the companies go at it.. The consumer almost always wins when they do.

I also think I know why Intel got as far behind as the did (Itanium) and I am sure they are hungry to try and redeem themselves. Conroe is probably a good start but I also think it is quite premature, not to mention indicative of a worried company, for them to bring out this future processor and put it up against AMD's not-even-the-latest-and-greatest chip. I will be shocked if AMD does not have something ready go... I think they are just keeping their mouths shut and the future will tell us.

All that being said.. Those are gamer chips. Those are not the server monsters that the Opteron has gotten to be. Hopefully they will have an answer for that, but until then, I'm now a proud 2x 270 DC Opteron server owner after many years of Xeon. There is nothing out there that can compete, and certainly at that level, so I will sleep well tonight. lol ;) Maybe next year I will be saying the thing about some new Intel powered box.. :)

David "Fan-boy of the fastest, cheapest coolest & most reliable processors - Whichever they are"

PS. I too used to literally hate AMD. Every computer I had with them was horrible.. But, that was then..

FYI, Anand has posted an update which covers most of these issues:
http://www.anandtech.com/printarticle.aspx?i=2716
The scores didn't change much, except for the F.E.A.R, but it was Anand's fault - he tested the systems at the different settings. So F.E.A.R. scores are more in line with the other results in the update.

I suspect it's not enough to convince you that Intel didn't play dirty there, since even the abovementioned independent results from the leaked sample didn't do. The point is, the tests gained a lot of publicity for Intel, as was intented, and for a good reason. You can shrug them off, but the fact stands - Intel is already moving, and AMD is not. I really hope it didn't fell on deaf ears at AMD, since the placid days of reaping the fruits of its K8 architecture are already numbered. I find it alarming that there're still almost no information apart from speculations about AMD's new architecture - K9 or whatever it was called.

Putting fanboyism aside, the CPU competition is better for us customers, so I'm looking forward for Conroe (and an accompanied AMD price drop :) )

[kmike]

All that being said.. Those are gamer chips. Those are not the server monsters that the Opteron has gotten to be. Hopefully they will have an answer for that
Enter Woodcrest:
http://www.anandtech.com/printarticle.aspx?i=2772
3GHz version simply obliterates the competition. Even if AMD had the similarly clocked offering, it would still be 5-55% slower than the Woodcrest.

It's set to be released in Q3 this year, in July or even June according to the rumors floating around.

[FlyBoy73]

Enter Woodcrest:
http://www.anandtech.com/printarticle.aspx?i=2772
3GHz version simply obliterates the competition. Even if AMD had the similarly clocked offering, it would still be 5-55% slower than the Woodcrest.

It's set to be released in Q3 this year, in July or even June according to the rumors floating around.

I could haggle over estimated numbers (comparing new Intel vs. Old AMD, etc) but I'll just sum up my response for now like this..

If "Woodcrest" or any of Intel's Exec's kids pet named chips are so great... Why has DELL now FINALLY said they are going to offer O-P-T-E-R-O-N servers?? As far into bed as Dell & Intel have been, if there is some super chip on the horizon for Intel, don't you think it would stand to reason that Dell would not offer these after 3 years or retardation? No, you can bet Dell put their hand out and said pony up the best you have and we're going to put it through the paces... And whatever that was wasn't good enough...

Again, all of these tests ass-ume that AMD won't have something better in 6 months or however long until Woodchuck's release date. In the past none of the chip guys were showing their cards so far in advance but Intel is in deep crap so they are laying them down, hoping everyone else won't migrate to AMD/Opteron servers..

Do I care if Intel drops the bomb of all chips? Heck no... If they are so far ahead of AMD the next time I need more servers I will go with them if they have proven themselves.. BUT, my single 2x 270 Opteron box right now is so far beyond the two separate dual xeon servers (one web & 1 database) I was previously using it isn't even funny. There is no comparison and my life has been so so so so much nicer since I went AMD..

So, let's discuss it again when their chips actually materialize for the general public. ;) I've read a bunch of articles in the past touting this or that new super chip from Intel but still nodda.. That doesn't help me now.