51 GPUs for F@H

Very impressive I just hope they don't alter the GPU client in Points or output that makes that farm unattractive. Still that is pretty [H]ard folding there... now if only you were folding for the [H]ard team! :p

 
they should unless they dont want anyone to run the console client anymore, quite frankly im running about 800 installs of the console client, doing 500WU's per day, and doing as much as 16 8800's? ya thats bull crap, quite frankly i think im done with this project
 
Hang in there Flecom advancements are coming for the SMP client as well. The linux version is already out hitting some 5k PPD putting it on par with the GPU. So there is light at the end of the tunnel.

Although if you project out the SMP client really should be getting 5k Per CORE.

The console client on the other hand just doesn't seem to get the love it should, which is sad!



 
FLECOM said:
they should unless they dont want anyone to run the console client anymore, quite frankly im running about 800 installs of the console client, doing 500WU's per day, and doing as much as 16 8800's? ya thats bull crap, quite frankly i think im done with this project
Click to expand...
I have to agree the points structure in this DC project is WAY out of balance with the resources invested. Stanford should take a hard look at WCG and see how they might revise the way work is evaluated. The way things are now, anyone with a couple of hundred bucks can outproduce a farm of a dozen recent servers that are running CPU clients. IMO, that's unfair and totally disproportionate even if the science accomplished is more in the end.

Sunin said:
Hang in there Flecom advancements are coming for the SMP client as well. The linux version is already out hitting some 5k PPD putting it on par with the GPU. So there is light at the end of the tunnel.

Although if you project out the SMP client really should be getting 5k Per CORE.
Click to expand...
Indeed, but where have you heard about improvements to the SMP client as far as WU evaluation is concerned? I'm running a few Linux SMP clients and even though I don't have them monitored yet, I haven't seen a noticeable improvement in production. Perhaps one of the reasons is the fact I'm receiving a lot of P2665 WUs. Thought I'd be rid of them with the Linux clients, guess I was wrong. :confused:

 
Flecom, don't dispar.

I think we will get a revision on the points. Thats the main reason I have not sold my Q6600s. I could get 2 8800GS for each of them and produce more than 2 times the points. I think Stanford has not learned it's lesson and wants to get a bunch of GPUs going, then when they do they will drop the points. (Making many mad.)

I fully expect this (yet I'm still getting GPUs when I find a deal. They hold their value ok)

There have been a few that say keeping a diverse farm is a good idea. From what I've seen (GPU1 the perfect example) it is a darn god idea.
 
You must log in or register to reply here.
Back
Top