Impressive!
some questions:
which power supply do you use in this test? (model)
all the tests have been made with the sli activated?
which cpu cooler do you use in this test? (model)
where I can steal a pair of opteron dual core?
thanks!

Proesterchen,

I can't help if you're clinically retared and can't understand it. Are you the fat rich kid..too much drugs in HS?

Scott,

Why do you review these AMD Opterons with crippled 2T command and slow timings (3-3-3)? It's would'nt be so bad if you did'nt use uber Corsiar XMS2 with tight timings (3-2-2) on the Intel setup, but as it is it's severe enough to warrent comment. A64's and Opterons suffer huge performance hits when running at the 2T command rate shown all over the net, about 4% slower. And running those loose timings is another, at least, 5% hit over 2-2-2.. Making a total loss of performance somewhere in the neighborhood of 9%. I'm really surprised AMD did as well as they did crippled like this. But for a more accurate representation of performance you may want to get some real ram when testing AMD's 64 bit chips in the future.

This is even more true when finally getting a real live X2 since using ECC is good for another 3% penalty on top of the 9% due to additional wait state.. I would estimate a real 4400+, with proper ram (2-2-2-7 1T), will be about 10-15% faster than the benchmarks shown for the Opteron 175 here.

I think Anand's conclusions are a tad more poignant:
"Despite AMD's lead in getting dual core server/workstation CPUs out to market, Intel has very little reason to worry from a market penetration standpoint. We've seen that even with a multi-year performance advantage, it is very tough for AMD to steal any significant business away from Intel, and we expect that the same will continue to be the case with the dual core Opteron. It's unfortunate for AMD that all of their hard work will amount to very little compared to what Intel is able to ship, but that has always been reality when it comes to the AMD/Intel competition."

While the tech is fantastic.... I don't think we'll see much adoption outside companies aiming for the very best bang/buck. everyone else will stick with Dell/nobodyevergotfiredforbuyingintel.

The Opteron 175, meanwhile, absolutely runs away from the Opteron 248s. Could this be the new K8 cores' SSE3 support at work?

I don't think SSE3 can account for such a difference on its own virtue. I'd be willing to think that the fact that Opteron 175 is HT-enabled, as has been pointed out by x86-secret, whereas Opteron 248 is not, allow the former to be seen by HT-aware programs as four distinct cores and thus allow it to take advantage of these optimizations, whereas dual O-248, can only be seen as two different cores.

And HT hasn't anything to do with SSE3, has it?

Well anyway, just my two cents....

...SSE3 talk all around. Have none of you seen the Venice Vs Winchester benchmarks at the same frequencies? Venice walks away with Winchester placed very flat under Venice's foot.

The reason the new dual cores chips beat their single core brethren is because the Venice core is faster than Newcastle/Winchester (Or Sledgehammer and Clawhammer.) Although, the name for the Opteron Venice dual cores are Denmark, Italy, and Egypt (Venice 90nm SSE3 1xx, 2xx, 8xx)

And there's Venus, Troy, and Athens (Venice 90nm SSE3 single core 1xx, 2xx, and 8xx)

This is just collective information I've found from roadmaps and such, please correct me if I'm out of date on this stuff...I love to have my stuffs straight.

Athlon64 X2
Dual 2.4GHz, 1Mb cache per core
US$1,001
Dual 2.4GHz, 512Kb cache per core
US$803
Dual 2.2GHz, 1Mb cache per core
US$581
Dual 2.2GHz, 512Kb cache per core
US$537

Pentium-D
Dual 3.2GHz, 1Mb cache per core
US$528
Dual 3.0GHz, 1Mb cache per core
US$314
Dual 2.8GHz, 1Mb cache per core
US$240

I wonder how well the 2.8Ghz overclock?
Maybe a "budget" dual-core setup using these CPUs?

OK, just a wee bit confused here--great article by the way, not confusing at all and a lot of info. Just want some info about purchasing...
From what I understand, Real Soon Now the A-64x2s will come out. These are socket 939 dual core processors?
I soon plan to upgrade to an A-64 system, probably a 3500. Now, when the x2's come out, will I be able to just drop that into a random socket 939 board and get dual-core goodness?

Also, Damage, an idea--have you ever considered, in your tables of benchmarking systems, throwing a price on each? That is, for each benching system, grab a Priceline/Newegg price for all the gear, add together. That way, we not only know that the dual Opteron 265s are 25% faster than the single 175, but also that it costs $2500 as opposed to $1650? (s/fake numbers/real numbers, of course.)

Won't a 2.2GHz 1MiB dual-core Athlon 64 X2 be the 4600+, rather than the 4400+ ? I base that from the 2.4GHz 1MiB X2 being the 4800+.

Oh my, excellent review.

I doubt that most of the gamers who visit this site (except for the Folders) do much that steps into the realm of multiple computation-intensive threads, so I have grave doubts that even the desktop processors (Athlon 64 X2, Pentium D) will make a big difference to the performance most gamers care about. (Witness the gaming benchmarks -- no performance advantage to dual-core over single-core or dual-processor over single-processor; the 152 and 252 win all.)

Fortunately, not everyone who reads the site falls along the gamer/surfer/media-peruser-only stereotype. ;)

This link on page 4 doesn't work: http://techreport.com/reviews/2005q2/opteron-x75/trdemo3.zip

There are multitprocessor MD codes floating around in academia -- there are guys in my research group who have done 16-processor runs on BoB (though I only see one on right now...).

http://bob.nd.edu

so damage when are you going to jump on the TR UT server?

now here's the qustion:

can you pair a 275 with a 248 and have a 3 core system?

*ponders*

Power consumption is pretty high but.....all those things aside. Imagine dualcore in a laptop..?

I was in love with the 170 and 175 as an upgrade for my 146 until I saw the pricing.

Opteron 148 (2.2 GHz): $278
Opteron 248 (2.2GHz): $455
Opteron 175 (2.2GHz): $999

The 4400 X2 at $2.2GHz with 1MB of cache costs $581, and I would imagine performs the same as the 175.

That $419 price difference will more than pay for a 939 mobo.

How long till I can get a quad core laptop to replace the single core one I've got at the moment? :-P

That thing would be a beast, but don't put it on your lap :-D,
-Mole

Re: The power readings... I know you said not to give you a hard time, Damage, but I do think that measuring the power deltas is a little unfair. The Opterons have a memory controller on board, and because of the NUMA archetecture both have fully-powered banks of registered memory. I think that pretty vastly inflates the power deltas for the processors themselves, vs. the duallie Xeons where that isn't an issue.

The real question is how creamy smooth it is. Are we talking Half and Half or Double Devon creme smooth?

Another kick in the pants for Intel !

Nicely done article - what does the device manager report for processers ? 2 or 4 ?

Edit: 2 or 4 on the dual proc board.

Should I wait to purchase an AMD 3500+. I have my eye on the X2 4400+. My question is will this work with games? Will games take advantage of the dual cores?

The total possible bandwidth flowing through the 940 pins of an Opteron 875 is 30.4GB/s—technically, enough to choke a horse.
haha. this isn't the reason I read TR, but it is why I love it so much :D

anyways, could you try to include some software compiling benchies in the future? it's (after f@h) what the vast majority of my cpu's cycles are spent on... (and it should be pretty friendly to dual-everythings... gcc at least has a '-j' flag for the number of jobs to run in parallel)
(according to the benchies at AT, Hammer-derived stuff absolutely trounces anything else, but some independent verification of that would be useful)

Typo or graphic error on page 8. You write:

" where our previous champ had scored 689. That champ, by the way, was a system with dual 3.2GHz Xeons based on the older "Prestonia" core"

The graphs show the Opt 252's clocking that 689 score. So who is it, Xeons or Opts being the older champ?

An excellent and thoughtful writeup! You have really done an outstanding job this time. I have two observations about the ScienceMark single precision Blas SGEMM contrasted with the double precision Blas DGEMM on page 11.

I'm supposing the Opteron's do so well in double precision versus single precision because of the difference between the efficiency of a true 64-bit core implementation versus a (very quick!) 32-bit core. The DGEMM/SGEMM ratio for the Pentium XE 840 is 0.819 (i.e. its slower in double precision). The Opteron 175 DGEMM/SGEMM ratio is 3.36. Wow! Its over 3 times the mflops in double precision. I'd like to see some benchmarks that benefit from or require use of double precision math.

A secondary speedup effect seems to be there due to the dualcore design. Judging from DGEMM results (148 versus 248) versus (175 versus 275); that must be the memory performance difference we saw on page 5. Its really impressive, well done AMD.

i think i just lost a little bowel control.

i wonder if i can convince my boss that i need to get my dual-Xeon workstation upgraded. :)

ps - i doubt he'll go with the "i need a whole new computer" argument. ;)

thought I'd share...

http://www.theinquirer.net/?article=22711

Great review as always... my Envy Meter has gone right up, again. I *must* have four processor cores in my home PC...

Unless AMD's model numbering has confused me (which is possible), isn't there a typo on page 3? Rather than "175 vs 252" I think you mean "175 vs 1 52".

Now I wonder whether Dell, who wants to dominate the server market, is going to tolerate losing market share to everyone else. Supermicro has said they are going AMD, so now Dell is alone, and is going to be stuck with a decidedly inferior product mix.

I know that no server apps were tested, but the results speak for themselves, and those results would indicate that in the highly threaded server usage AMD's new baby is going to kick major butt.

AMD defintely has a winner in the Server/Workstation market. The amount of processing power that a dual-core, dual-chip Opteron has is downright wicked! Intel's quad-chip Xeons and their semi-propertary motherboards are more expensive, have inferior performance and consume far more power.

Fellow ethusiasts don't despair, these monsters are priced and targeted for the Server/Workstation market. They can handle the $1,000+ price tag per chip.

Wait for the upcoming A64-X2s to come out which are still only going to be affordible to the power user, ethusiast segiment for a while. (My estimate is around $500-800)

What's the pricing on these? Did I miss it in the review?