F@H - My Experiences Using both AMD and Intel

[randomizer]

It depends if you want ultimate performance or ultimate bang for buck. Gulftown could wipe out Magny Cours for folding as far as raw performance goes.

 

[jennyh]

I'm not sure about that. How can 6 cores with SMT hope to beat 12 real cores? The figures on show here prove that more, slower cores are performing better than fewer, faster cores.

 

[randomizer]

Oh, I was thinking of Istanbul for some reason 😀

 

[MU_Engineer]

I'm not sure about that. How can 6 cores with SMT hope to beat 12 real cores? The figures on show here prove that more, slower cores are performing better than fewer, faster cores.

If the program is optimized well enough for Intel's CPUs and the Intel CPU is highly clocked, it's very possible. FAH is compiled with the Intel compiler, according to Randomizer, and that sounds very probable to me. That would give Intel's CPU a good optimization over AMD's CPUs by itself. The FAH data set I linked shows that a pair of 2.5 GHz Opteron 2380s (8 cores) scores about the same in FAH as an i7 (4 cores w/SMT) clocked somewhere around 3.3-3.4 GHz. I don't see that changing in the future unless Pande decides to offer a considerably different set of optimization for the next round of FAH clients that don't favor Intel CPUs or AMD suddenly increases the performance of its CPUs greatly. I don't think Pande is going to change his code, but AMD might outperform Intel with the Bulldozer-based CPUs. We'll just have to see.

What won't happen is an i7-based system beating a dual 12-core Magny-Cours system. Magny-Cours would have four times the cores of an i7, and as we saw, an i7 clocked to within an inch of its life can't beat even four Barcelonas running at about half the clock speed. There's no way even Gulftown will beat a 24-core Magny-Cours. I doubt that unless Gulftown is an excellent overclocker and can hit over 5 GHz, it probably won't beat two 8-core Magny-Cours units either.

Oh, I was thinking of Istanbul for some reason 😀

Depends on what Istanbuls. An overclocked Gulftown ought to outrun dual Istanbuls (particularly lower-clocked Istanbuls) but you're not going to outrun four Istanbuls with a Gulftown.

 

[randomizer]

I was referring only to a single Istanbul, not a MP system.

As for the Intel compiler, it is mentioned on the F@H site however I can't find where now.

EDIT: Found it: http://folding.stanford.edu/English/FAQ-PROTOMOL (towards the bottom)

Specifically talking about this particular core, but it's a safe bet they use it on other scientific cores as well.

 

[yannifb]

The Bulldozer based opterons are going to be insane imo. Even the desktop version will be, I mean 8 cores and 16 'Modules' will be amazing, considering the os recognizes each as a core. And yes, when the roadmap says 8 cores, that means 16 modules. Google bulldozer modules in google and theres an article clarifying this.
Anyway back to what I was saying- the 12 core BD Opterons will be recognized as 24 cores by the OS, thats a lot of parallel processing power.

 

[archibael]

What won't happen is an i7-based system beating a dual 12-core Magny-Cours system. Magny-Cours would have four times the cores of an i7, and as we saw, an i7 clocked to within an inch of its life can't beat even four Barcelonas running at about half the clock speed. There's no way even Gulftown will beat a 24-core Magny-Cours. I doubt that unless Gulftown is an excellent overclocker and can hit over 5 GHz, it probably won't beat two 8-core Magny-Cours units either.

I'm not sure we're going to try to position any i7 against Magny-Cours, though I'm not with Marketing and am not privy to those kinds of plans. Isn't its target market 4P+ enterprise servers?

 

[MU_Engineer]

The Bulldozer based opterons are going to be insane imo. Even the desktop version will be, I mean 8 cores and 16 'Modules' will be amazing, considering the os recognizes each as a core. And yes, when the roadmap says 8 cores, that means 16 modules. Google bulldozer modules in google and theres an article clarifying this.
Anyway back to what I was saying- the 12 core BD Opterons will be recognized as 24 cores by the OS, thats a lot of parallel processing power.

You have this backwards. The desktop units will be up to four modules, which you'll see as eight cores/threads. The biggest server chip currently planned on the Bulldozer architecture is the 8-module Interlagos, which has 16 cores/threads per chip.

 

[yannifb]

You have this backwards. The desktop units will be up to four modules, which you'll see as eight cores/threads. The biggest server chip currently planned on the Bulldozer architecture is the 8-module Interlagos, which has 16 cores/threads per chip.

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=3683

When they say 8 Bulldozer cores, they mean 8 bulldozer modules. The roadmap says 4-8 cores, which means 16 to the os, max. That articles clarifies this, from an AMD spokesperson.

 

[jimmysmitty]

I'm not sure about that. How can 6 cores with SMT hope to beat 12 real cores? The figures on show here prove that more, slower cores are performing better than fewer, faster cores.

Nehalem-EX.... 8 real cores 16 threads and so far it looks like it should be fine for it. Also called Beckton.

http://arstechnica.com/hardware/news/2009/05/intel-touts-8-core-xeon-monster.ars

Not to mention it has 4 QPI links instead of one that we are used to on DT CPUs It also looks to allow for 8 socket servers. Might be interesting.

As for CrySlayer, seriously a lot of people do F@H for the benefit of man kind. Yes there are some who grow their epeen with it but man you need to educate yourself on this kind of stuff before commenting.

I for one don't do F@H because I don't need my electric bill to be more than its already $100/ month.

 

[Chad Boga]

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=3683

When they say 8 Bulldozer cores, they mean 8 bulldozer modules. The roadmap says 4-8 cores, which means 16 to the os, max. That articles clarifies this, from an AMD spokesperson.

Man,
Your reading comprehension sucks, although I will admit that article was not as clear as it should be.

Find a different source to help you out.

MU_Engineer was right.

 

[Cryslayer80]

Nehalem-EX.... 8 real cores 16 threads and so far it looks like it should be fine for it. Also called Beckton.

http://arstechnica.com/hardware/news/2009/05/intel-touts-8-core-xeon-monster.ars

Not to mention it has 4 QPI links instead of one that we are used to on DT CPUs It also looks to allow for 8 socket servers. Might be interesting.

As for CrySlayer, seriously a lot of people do F@H for the benefit of man kind. Yes there are some who grow their epeen with it but man you need to educate yourself on this kind of stuff before commenting.

I for one don't do F@H because I don't need my electric bill to be more than its already $100/ month.

And you need to understand I did never say F@H will not accomplish anything, I just doubted and questioned it. That is disputable, but it is an undeniable fact that most people who use it do not give two cents about curing diseases... That is all.

PS: Educate myself about someone who goes on a forum every day praising his folding score with other noobs thinking: "he's da lord!"? I don't think it takes an elementary school to get what is going on here...

 

[Chad Boga]

@Chad Boga: Dude, it's okay, ease off a bit... He didn't kill your friend, he just misread something... But I guess when it comes to misreading that will apparently give AMD an advantage harsh words must be used, huh?

He tends to misread a lot of things, this ain't the first time and even when things get explained to him, he still wants to stick to his original incorrect interpretation.

 

[Cryslayer80]

Sorry then, my bad. Seeing that avatar plus that statement... You know :lol:

 

[Chad Boga]

Sorry then, my bad. Seeing that avatar plus that statement... You know :lol:

My avatar use to be of AMD coz that was what was powering my computer, but when I changed to Intel I decided to change my avatar accordingly.

Having said that, I only started off using an image of which ever company was powering my computer because I couldn't think of anything better to use, but I have found a non-computing related picture I will change my avatar to, in the near future. 😀

 

[Cryslayer80]

Yeah, I'm Intellophobical a bit 😀

:lol:

 

[Chad Boga]

Nice. :lol: :lol: :lol:

 

[verndewd]

I donty fold because of the DOD income that FAH gets. Between QMC and FAH there are far too many open doors to militarizing that type of information and since the DOD does fund FAH Id rather fold for the LHC for now.

 

[yannifb]

He tends to misread a lot of things, this ain't the first time and even when things get explained to him, he still wants to stick to his original incorrect interpretation.

Misread a lot of things? Tell me one other time that you corrected me....

And now that was one little mistake... calm down.

 

[ElMoIsEviL]

Actually, if you look at the data, four Barcelonas beats even ridiculously-overclocked i7s. Jack57000's quad Opteron 8358 setup gets 26 min 42 sec frame times on p2681s. The highest-performing i7 system on there was ParrLeyne's i7 975XE running at 4.37 GHz, which gave 27 min 44 sec frame times on p2681s. The highest-clocked i7 on there was road-runner's i7 920 @ 4.50 GHz, which got 28 min 18 sec times. And last but not least, the highest-performing system on thee was Slash's glorious 8-socket Opteron 8356 unit, which knocks out a p2681 frame in a mere 14 min 51 sec, which is six and a half minutes less per frame than the next guy, who was running Xeon X5560s. The best part about his unit is that he built the whole thing for $2500, while the X5560s cost that just for the two CPUs, let alone the board and RAM.

So yes, core for core an i7 will outperform an Opteron- particularly when the i7 is heavily overclocked and the Opteron is a 2007-vintage 65 nm unit. But a lot of slower cores trumps a small number of fast cores in highly parallel code like F@H, which is why the "Opterons are crap, AMD can't compete against the Xeons" crowd is wrong. It will be interesting to see how well a dual Magny-Cours with twice the core count of a Westmere Xeon at approximately the same price tag will perform in F@H. I predict that Intel won't even be competitive unless those Westmeres get massively overclocked on that EVGA dual LGA1366 motherboard while the Magny-Cours run at stock on a Tyan or Supermicro board. And even then, you're just comparing how well a non-overclocked system compared to an overclocked system. That's not even a fair fight as the Magny-Cours units could be overclocked too. An NDA-breaking guy got an early multiplier-unlocked but non-VID-adjustable Magny-Cours EX up to 3.2 GHz.

You're comparing 8x Quad Core Opterons (32 Cores) against 2x Xeon Nehalem Quad Cores and claiming victory for the Opteron?

What you aren't taking into consideration is the amount of Cores there. If 2x Xeons can keep up with 8x Opterons there is a problem there

Especially with 4P and 8P Intel Setups being released in droves this quarter (Beckton 8 Core CPU (16 threads) in 4P and 8P configs).

Beckton

Beckton is a Nehalem-based processor with eight cores and uses buffering inside the chipset to support up to 16 standard DDR3 DIMMS per CPU socket without requiring the use of FB-DIMMS.[20] It has four or more QuickPath interfaces, so it can be used in at least quad-socket configurations. Designed by the Digital Enterprise Group (DEG) Santa Clara and Hudson Design Teams, Beckton will be manufactured on the P1266 (45 nm) technology. It is expected to be launched in Q1 2010.[21] Beckton is also referred to as Nehalem-EX (EXpandable server market).

 

[ElMoIsEviL]

I ran SETI for a while too, then I figured out that aliens probably don't exist, amusingly enough when I stumbled upon the Fermi Paradox.

OT - Didn't Elmo link something about intel trying to manipulate the folding results recently?

Certain Folding@Home work units make use of the Intel Compiler in order to boost performance (because at the end of the day the goal is scientific advancement not a benchmark).