Hardware AMD Opteron 2435 CPU Review

[Tim S]

http://www.bit-tech.net/hardware/200...-2434-review/1

Intel might have been the first to launch a six-core CPU, but AMD's new six-core Opterons have far more memory bandwidth and are much more power efficient. But do they have what it takes to dethrone the mighty Xeon W5580?

[Xtrafresh]

I'd REALLY like to see one or two "normal" desktop platforms included in the tests. The way it is now, the numbers only mean anything to me in comparison to eachother, but i'd like to know how much actual difference there is with (for example) a Q9550 DDR2 platform and a Phenom II 955 DDR3 platform.

For one, it would be fun to see the crysis numbers, butit would also give a good sense of exactly how much processing power we are dealing with here in the other tests.

Otherwise i liked the article, it's always fun to see new stuff on the torture table

[SlowMotionSuicide]

Originally Posted by Xtrafresh

I'd REALLY like to see one or two "normal" desktop platforms included in the tests. The way it is now, the numbers only mean anything to me in comparison to eachother, but i'd like to know how much actual difference there is with (for example) a Q9550 DDR2 platform and a Phenom II 955 DDR3 platform.

+1

[Paradigm Shifter]

It would be great if Boston would actually allow individuals to buy from them. Last time I asked, they wouldn't sell me what I wanted (a dual processor Opteron board back in the days of the Opteron 265...) but that was a while ago so if they've started selling to people, rather than just companies... that's good. Otherwise... not so good.

Definitely looks like an early BIOS on some of those results. Others look about what you'd expect given the advantage that i7 has over AMD's offerings.

Sticking some results in for a 'normal' system would be good, too... just for comparisons sake.

[Lizard]

Thanks for the feedback guys, I know we don't have comparable benchmark numbers for a single-CPU system, but its something we can definitely include in the future.

Don't forget though that these sort of workstation/server/HPC platforms have a lot more to offer than brute performance, they also have much better error detection/correction than a standard single-CPU system.

[N19h7m4r3]

Very interesting review.

I'd like to see a another when the new boards are out that support HT3.0.

I'm eagerly awaiting AMD to switch to DDR3 for their HPC boards also.

[naokaji]

With the difference in power consumption it would be interesting to see what happens when AMD would release a high clocked 6 core opteron.

[p3n]

Originally Posted by Xtrafresh

I'd REALLY like to see one or two "normal" desktop platforms included in the tests. The way it is now, the numbers only mean anything to me in comparison to eachother, but i'd like to know how much actual difference there is with (for example) a Q9550 DDR2 platform and a Phenom II 955 DDR3 platform.

For one, it would be fun to see the crysis numbers, butit would also give a good sense of exactly how much processing power we are dealing with here in the other tests.

Otherwise i liked the article, it's always fun to see new stuff on the torture table

AMD loses by miles, shocker!

[Skiddywinks]

Originally Posted by p3n

AMD loses by miles, shocker!

Sure, if all you care about is performance. But you have to take into account the signifcant price difference, significant TDP difference and the power usage.

[tictactoe]

Be interesting to see the results when a Hyper-Transport 3.0 supporting motherboard is used, really ought to hold off a full judgement until then.

[Drexial]

Originally Posted by Skiddywinks

Sure, if all you care about is performance. But you have to take into account the signifcant price difference, significant TDP difference and the power usage.

Originally Posted by tictactoe

Be interesting to see the results when a Hyper-Transport 3.0 supporting motherboard is used, really ought to hold off a full judgement until then.

These are both true statements. But the results are still a bit disappointing. I really think its time for AMD to do what Intel did, ride out on their current line for a while and in a couple years release something that will turn the tides again. So in another couple years Intel can come back and do the same. If AMD rides it out, I think Intel will slow down on its innovations and allow time for AMD to build something up. This is honestly the healthiest competition between companies I have ever seen. If it weren't for Intel's strong arming I feel like the playing field would have been 50/50 when the P4 was out.

[Skiddywinks]

I think AMD would have had the lead back in the P4 days.

[Goty]

I'd reserve judgment on these chips specifically because of results like the LightWave benchmark. Four more active threads at the same clockspeed results in a one second increase in render time? Either the platform or the program has to be bugged.

[Action_Parsnip]

Originally Posted by Drexial

These are both true statements. But the results are still a bit disappointing. I really think its time for AMD to do what Intel did, ride out on their current line for a while and in a couple years release something that will turn the tides again. So in another couple years Intel can come back and do the same. If AMD rides it out, I think Intel will slow down on its innovations and allow time for AMD to build something up. This is honestly the healthiest competition between companies I have ever seen. If it weren't for Intel's strong arming I feel like the playing field would have been 50/50 when the P4 was out.

hmm yes but then again large swathes of this review are not really pushing lots and lots of threads, like you would believe appropriate for a 12 core setup evaluation. I mean theres Crysis and X3, then the custom PC benchmark, which is about as multi-core aware as my best mates cat.

[Action_Parsnip]

Plus like in the anandtech.com review nothing was made of the fact that some programs dont respond well to 'odd'/'unusual' numbers of cores, as in can take advantage of 8 and 16 but not 6 or 12, and an early ropey BIOS looks to play a part in some of the tests here.

[Lizard]

Originally Posted by Action_Parsnip

hmm yes but then again large swathes of this review are not really pushing lots and lots of threads, like you would believe appropriate for a 12 core setup evaluation. I mean theres Crysis and X3, then the custom PC benchmark, which is about as multi-core aware as my best mates cat.

That's why the review includes lots of different types of apps... it allows you to see how well the hardware copes with a variety of loads/stresses.

One of the biggest problems remains though the lack of software that scales about 8 cores - even in the enterprise space. Hence why a pair of Xeon W5580s is so much faster than a pair of Opteron 2435's - the Xeons have far greater performance and memory bandwidth per core.

[Fozzy]

Hmmm....anyone else thinking that ddr2 might be holding this processor back? I mean I see no reason for anyone buying a computer today to go for ddr3 from a price concious standpoint but when you're going all out why wouldn't these processors support ddr3?

I of course skimmed through this article so it may have already been mentioned. Gotta get back to modding!!!~

[looselycoupled]

It appears there is a clear bias against Intel in this review.

Here is the article lead-in on the front page, and quote from the article:

>>> "Intel might have been the first to launch a six-core CPU, but AMD's new six-core Opterons have far more memory bandwidth and are much more power efficient"

>>> "Despite having two more cores than earlier Opterons, the new 2435 has the same low TDP of 75W - nearly half the 130W of Intel's fastest workstation/server CPU, the Xeon W5580.

1) First of all, all of the AMD Operton 2400 series 6-chips have a TDP OF *115 watts*. The value of 75 watts is AMD's "ACP" value -- their so-called average power consumption -- NOT thermal design power.

2) On the same token, only the hot-running, workstation-only Intel Xeon W5580 (which no one will buy) has such a high TDP. The top Nehalem Xeon server chip, the X5570, runs at 2.93Ghz with a 95W TDP. And If you move down to 2.56Ghz, the TDP drops to 80W. If you tested ANY of the mainstream Xeon 5500 series, not the W5580, the power usage would be far lower and the performance/watt far higher.

3) The comment about memory bandwidth in the article lead-inis just plain BS. Your article itself shows the Xeon 5500 series to have far more memory bandwidth than the 6-core Opteron.

[docodine]

I wish that you had tested the Opteron 2427, the price ($449) makes it fairly possible for an average user to buy, at a stretch

I would like one over an i7, anyway. 6>4. :-)

[Xtrafresh]

Originally Posted by Lizard

Thanks for the feedback guys, I know we don't have comparable benchmark numbers for a single-CPU system, but its something we can definitely include in the future.

Don't forget though that these sort of workstation/server/HPC platforms have a lot more to offer than brute performance, they also have much better error detection/correction than a standard single-CPU system.

I'm looking forward to those benches. I'd like to get some perspective performance-wise.

I know that server-grade hardware offers other advantages over desktop-stuff, but this article doesn't deal much about that. Maybe worth an article someday?