AMD Zambezi/FX/Bulldozer chess benchmarks

[rbarreira]

What desktop Intel chips are faster than the i7-2600k except the insanely expensive i7 980x which costs 3 times as much as the top Phenom X6?

You already answered your own question (i980, i990, some X and W series 6-core Xeons, and of course the very high end >6 core Xeons).

As for the price difference: blame AMD. As long as they aren't competitive at the upper end, Intel gets to set the price almost where they want. This is very bad news for us. And Bulldozer doesn't seem to improve the situation.

I don't really see the point of involving $2000 - $5000 server CPUs in this debate which started from a benchmark of relatively affordable desktop CPUs.

AMD sells a $190 6-core CPU which is faster in NPS than Intel's $320 4-core CPU. Yes, the two additional cores will hurt effective speedup a little and nullify the NPS advantage, but when you're saving $130 on the CPU that seems quite acceptable...

I don't care who you blame for the price difference, the fact is that Intel's top-performing CPUs are simply not competitive in terms of bang per buck. The i7 980x is a great example as it costs more than 3 times as much as a Phenom X6 but doesn't even get twice the performance (and that's with the same number of cores, so no speedup argument here).

[Gian-Carlo Pascutto]

I don't really see the point of involving $2000 - $5000 server CPUs in this debate which started from a benchmark of relatively affordable desktop CPUs.

I presume you mistyped a 0 too much, because several of the CPUs I listed are 500, not 5000 USD.

AMD sells a $190 6-core CPU which is faster in NPS than Intel's $320 4-core CPU. Yes, the two additional cores will hurt effective speedup a little and nullify the NPS advantage, but when you're saving $130 on the CPU that seems quite acceptable...

The Phenom II X6 is great. No disagreement here.

AMD is now selling an 8-core CPU which is drawing more power, is more expensive, and in many circumstances, slower. I don't think that one is so great.

I don't care who you blame for the price difference, the fact is that Intel's top-performing CPUs are simply not competitive in terms of bang per buck. The i7 980x is a great example as it costs more than 3 times as much as a Phenom X6 but doesn't even get twice the performance (and that's with the same number of cores, so no speedup argument here).

I don't disagree with what you say here. I simply pointed out Intel can do this (give crappy price/performance to customers) because AMD is not able to compete in raw performance at the top end. It's clear the current Bulldozer does not change that, and that's a bad thing.

[rbarreira]

I presume you mistyped a 0 too much, because several of the CPUs I listed are 500, not 5000 USD.

It was not a typo. As far as I know the server CPUs (i.e. Xeons / Opterons) which cost $500 are not really better than their slightly cheaper desktop counterparts, due to a price premium on servers, supporting multi-CPU configurations, error-correcting memory and other server features. So in that $500 price range the best you can get is probably the i7 980 which we already discussed.

The Phenom II X6 is great. No disagreement here.

AMD is now selling an 8-core CPU which is drawing more power, is more expensive, and in many circumstances, slower. I don't think that one is so great.

I agree it's not looking so great, although it's not a disaster. Especially if it turns out that the final hardware steppings (or those coming out soon), coupled with compilers tailored for this novel architecture can improve the performance. At least us enthusiasts who don't mind recompiling programs to get more performance would be well served...

I don't disagree with what you say here. I simply pointed out Intel can do this (give crappy price/performance to customers) because AMD is not able to compete in raw performance at the top end. It's clear the current Bulldozer does not change that, and that's a bad thing.

It might be by design, AMD may have realized that it doesn't have the capital right now to build the top performing machines, so decided to keep the status quo of having great value for the price instead. It would be great to see AMD having top performance again, but it doesn't seem likely as the company is not doing great financially, and Intel has a lead in manufacturing process.

$DEITY help us if AMD goes bankrupt (I don't think Intel even wants that as it would put the anti-monopoly regulators on their ass even more than they already are)

[bob]

According to that page, the Bulldozer FX-8150 does outperform the top Phenom X6 at Fritzmark (11861 vs 11630), despite not being the top Bulldozer chip.

The FX-8170, being clocked at 3.9 GHz instead of 3.6 GHz, should theoretically get around 12,900 Fritzmark.

I've heard rumors that a future stepping of the Bulldozer chip would improve integer performance, but that was a while ago and I haven't kept up with the news about it...

Not sure about the benchmark. 200K faster? the overhead for additional cores in the parallel search will make that a net loss in performance...

Never want to see about same nps with more cores, a net loss can't be avoided.

[Albert Silver]

http://www.hardware.fr/articles/842-19/ ... fritz.html

Summary: even at 8 cores, it does not outperform the Phenom II X6 (let alone Intel chips) in NPS. Real speedup will be lower, and the individual cores are slower as well. It also uses more power on load.

This is not the chip to get.

I'm very disappointed frankly. Not that I'm affected as I own a Sandy Bridge 2500K, and have been deliriously happy with its overclockability, and sheer bang for the buck, but I was very optimistic that for chess particularly, the Bulldozer would beat it, and would be a prime choice.

[georgerifkin]

http://www.hardware.fr/articles/842-19/ ... fritz.html

Summary: even at 8 cores, it does not outperform the Phenom II X6 (let alone Intel chips) in NPS. Real speedup will be lower, and the individual cores are slower as well. It also uses more power on load.

This is not the chip to get.

disappointing results
I was looking forward to these new bulldozer processors.
do you know when 16 cores processors will be out?
thanks

[Gian-Carlo Pascutto]

I agree it's not looking so great, although it's not a disaster. Especially if it turns out that the final hardware steppings (or those coming out soon), coupled with compilers tailored for this novel architecture can improve the performance. At least us enthusiasts who don't mind recompiling programs to get more performance would be well served...

I doubt you will get any interesting gain through recompiling (for integer workloads. AVX/XOP is another matter). There are apparently some OS-level tweaks that can give a few percents here and there in multiprocessor loads, but don't expect magic.

There is good news though: the die size of Zambezi 8-core is less than that of Phenom X6. This means AMD can price them even more aggressively.

[FlavusSnow]

Just to note, the intel chips may be 4 core, but those benchmarks are definitely using HT... so 8 threads of intel vs. 6 threads of AMD on the Thuban chips. And of course 8 vs 8 when Bulldozer is what you speak of.

[Terry McCracken]

Just to note, the intel chips may be 4 core, but those benchmarks are definitely using HT... so 8 threads of intel vs. 6 threads of AMD on the Thuban chips. And of course 8 vs 8 when Bulldozer is what you speak of.

Hyperthreading doesn't help chess. Turn it off and run the tests. I think most do unless they're trying to inflate scores with NPS.

[Gian-Carlo Pascutto]

Only the Intel 2600K has hyperthreading, the 2500K does not. From what I can see, I suspect the Fritzmark used them for the 2600K, but Houdini didn't for any CPU.