New client, new tags question

[drasnor]

Alright, I'm almost done setting up my sister's dual MP2100+ (w00t for RMA upgrades, originals were 2000+) and I'm about to install Folding as a service... only now there's a new console out so I'm having to modify my "service pack".

Anyway, I've noticed that Stanford has added a new (to me) tag to the version 4 client, and I want to hear your thoughts on -forceSSE (new) as well as -forceasm and -advmethods as far as point production on 1) dual Palominos, 2) P3 Coppermines, 3), P4 Northwoods and Williamettes, and 4) dual Opteron 248's.

-drasnor :fold:

[muddocktor]

With AMD rigs, you need to run the -forcesse flag instead of the -forceasm flag to make the system use SSE. The -forceasm flag now only insures that the client starts using the default assembly loops set by Stanford in the client, which is 3DNow! in the case of XP procs, either Pally's or Tbred's or Barton. With Intel, the -forceasm still works just fine to make the client use SSE. I really can't think of any other changes in the flags that will affect you on the v4 clients. One other change they added to the new clients is support for up to 8 instances of the client, double what was supported in previous versions. So if you run across any quad xeon rigs that support hyperthreading, you can fully borg it now. :clap:

[drasnor]

Well, what I mean to ask is are there any benefits to using SSE on AMD processors as opposed to 3DNow!?

-drasnor :fold:

[mmonnin]

Yes about a 30% increase in production. It is quite worth it.

[csimon]

this is your athlon xp folding with 3Dnow!

...and this is your athlon xp folding with extra SSE boost!!!! :clap:

[drasnor]

-forceSSE crashes the dual MP2100+ (Palomino) box. As enthusiastic as I am about FAH, my enthusiasm for not pulling apart the dual MP box outweighs my enthusiasm to give Stanford the S/N's of my processors.

-drasnor :fold:

[csimon]

just remove -forcesse and fold using 3Dnow! which is still a very productive optimization.

[drasnor]

just remove -forcesse and fold using 3Dnow! which is still a very productive optimization.

There was another option? ;)

-drasnor :fold:

[csimon]

nope ...when you have an AMD processor with 3Dnow! capabilities and you don't use the -forcesse flag it automatically defaults to the 3Dnow! optimization ...so basically eliminating the -forcesse flag will give use the 3Dnow! optimization. That simple.

If sse is giving you trouble then don't use the -forcesse flag until AMD fixes it ...or not.

[drasnor]

Sorry, was being facetious in my last post. Hope Stanford and AMD get that problem licked.

-drasnor :fold:

[csimon]

LOL ...you dawg!