Help with setting up QMD

[tsuehpsyde]

I'm assuming QMD's are better... I've setup all of my folders to do the default thing, but now with the hardbowl around the corner and my new processor getting tweaked, I wanna dominate as strongly as possible. How exactly do you specify QMD's to run on the client for console? I know there's an -advmethod or something similar to specify in the shortcut, but I don't see how that works with a console service? Do I have to edit the registry's startup value? Do they even work on AMD? I feel like such a folding noob.

Full Folding Ahead!

 

[OSUguy98]

all you have to do is use the -advmethods switch and say yes to "big packets" during the config.... there's a link in the sticky at the top that'll give you more detailed info.... right now though ,I think QMDs are an intel-only thing..... unless they've changed that recently....


Keep on Folding!! For the [H]orde!!

 

[Kvothe]

QMD is still all Intel, they fixed that after the glitch when the other servers ran out of WU's. The problem is that QMD was compiled using SSE2 instructions from an Intel compiler and it's not licensed to be modified or optimized for any other platform.

 

[OSUguy98]

Any word on when we might see some QMD action on AMD chips? or would you have to recompile everything just for that? Woudn't it just require the release of a new core?


Keep on Folding!! For the [H]orde!!

 

[Imitation]

Hmm I downloaded the latest console for my P4 lappy and the setup was WAY different. It basically asked me if I wanted to run QMDs. Way easier than having to set the flags manually, I've been pulling QMDs ever since.

 

[Mohonri]

Any word on when we might see some QMD action on AMD chips? or would you have to recompile everything just for that? Woudn't it just require the release of a new core?

As I understand it, it's not Stanford's fault that there's no QMD for AMD. The problem is that the SSE2 optimizations are for intel processors, and the current core runs extremely slowly on AMD (by comparison). Stanford would recompile for AMD, but one of the libraries is supplied by Intel, and optimizing it for AMD would violate their agreement.

So until Intel allow VJ et al to mess with the library, or until someone codes around it (neither is likely to happen any time soon), QMD's won't be available for AMD. They're not worth it at this point anyway.

 

[OSUguy98]

wouldn't they just have to code in a SSE2 equivalent to work on AMDs?
I thought that AMDs could run SSE2 and/or SSE3... in addition to 3DNOW!.... is there THAT big of a difference between the code an AMD would run to use SSE2 and the Intel SSE2? or does it just come down to "optimized code" and the "we're going to use SSE2 alot, so we'll write it so it's slower on AMD machines and people will buy more Intel"?


Keep on Folding!! For the [H]orde!!

 

[Kugar]

I don't know if this is the cause of the AMD-Intel issue with QMDs, but it sure sounds like it might be...

Page 40 of AMD's complaint vs. Intel states the following:.

According to AMD, the Intel compiler builds code that looks at the CPUID at runtime. If it sees "Genuine Intel", then an optimzed code path (SSE, SSE2 I presume) is executed, otherwise a slower code path is run.

So even though AMD supports SSE and SSE2, code compiled by Intel's compiler will not use the optimized instruction set on a non-Intel CPU. It is possible to determine if a CPU supports the extended instructions sets, in fact I've heard Intel invented the mechanism. But rather than use that in their compiler, Intel chose to use the CPUID, allegedly to put competitors at a disadvantage.

 

[Imitation]

wouldn't they just have to code in a SSE2 equivalent to work on AMDs?
I thought that AMDs could run SSE2 and/or SSE3... in addition to 3DNOW!.... is there THAT big of a difference between the code an AMD would run to use SSE2 and the Intel SSE2? or does it just come down to "optimized code" and the "we're going to use SSE2 alot, so we'll write it so it's slower on AMD machines and people will buy more Intel"?


Keep on Folding!! For the [H]orde!!

We need to remember here that AMD's and Intel's cpus are designed completely different. Just because both have SSE2 doesn't mean that the implementation is the same, it can't be because the cpus are so different. So therefore a compiler that is optimized for intel's cpu obviously isn't going to perform as well on an Athlon because they are designed very differently. Does that make any sense?

 

[tsuehpsyde]

Thanks. I'll just run big packets and pray. Almost done finalizing my speeds with the 144 for the Hard|Bowl.

Full Folding Ahead!

 

[BillR]

I don't know if this is the cause of the AMD-Intel issue with QMDs, but it sure sounds like it might be...

Page 40 of AMD's complaint vs. Intel states the following:.

According to AMD, the Intel compiler builds code that looks at the CPUID at runtime. If it sees "Genuine Intel", then an optimzed code path (SSE, SSE2 I presume) is executed, otherwise a slower code path is run.

So even though AMD supports SSE and SSE2, code compiled by Intel's compiler will not use the optimized instruction set on a non-Intel CPU. It is possible to determine if a CPU supports the extended instructions sets, in fact I've heard Intel invented the mechanism. But rather than use that in their compiler, Intel chose to use the CPUID, allegedly to put competitors at a disadvantage.

It is in fact totally an Intel thing. Stanford has lobbied hard to have the code re-compiled but Intel is being their usual stubborn obstinate selves and won’t license the code for use on AMD processors.

Feel like writing an expressive letter? Write one to Intel, be polite of course, but don’t fail to mention they are helping to hold up cancer research. Don’t know if they have any sort of corporate conscience or not, it’s doubtful though.

 

[SpoogeMonkey]

That's what I was going to say Bill. Even when it comes down to helping fellow humans, Intel has to have a thumb up their ass and say F U to the rest of the world. Even if it's something as trivial as allowing a small program run on less than .01% of all cpu's in the world.
I guess thats why I dropped intel chips like a steaming pile of dog doo in the hot summer sun, 6 yrs ago and never looked back.