PPD guesses and which client?

[Vaulter98c]

I got me all the parts I need for a basic gamer now, hopefully I can get it up an running before I go back on orders!

Looking at 3.4 Ghz 2core/4thread i3 Sandy Bridge water-cooled lol, a GTX 670, 8Gigs of 2133 ram (fast lol), and some HDD's in raid

Question is, would the i3 (2c/4t) at 3.4 be able to finish WU's in Enough time to be worth it using only 3 threads? Would it make the bonus time? Would running all 4 hurt the GPU? I was hoping to go smp but if I have to bust out the old uni client I will lol

And should I go v7 or something else for a 670 GPU? Any special flags?

Sorry if the GPU question is over done but I'm not sure of anyone here running a 6xx series GPU.

 

[jfb9301]

I don't believe that stanford has a client that will fold on the 6XX series yet.

So for now YMMV.

But for what it's worth, I have a 3-4 yr old Cellery Core2 Mobile around 1.8 GHz or so that meets SMP deadlines for1KPPD, where last time I tried Uni, I came up a couple hundred short of that. So 3 core should be no a no brainer. SWMBO's i3 Mobile got around 4KPPD or so -smp 4. I would imagine that the desktop CPU would do much better.

 

[Vaulter98c]

Ok, so like 4k then from smp4, not bad for a base projection. Once there is a 670 client would I need to drop 1 thread for the GPU? haven't folded GPU's and CPU's together since 2010, I feel things have changed a bit since then lol

 

[Nathan_P]

The current cores will fold on 6xx hardware - for v7 you have to add your particular card to the whitelist, however the cores are not optimised and PPD will be roughly the same as a 570. For v6 adding the -forcegpu fermi flag may work but i haven't seen any posts around saying that it has been done

If i find anything i'll post it up

Edit:- found some 680 info, EVGA have managed to get it to work with GPU tracker:- http://www.evga.com/forums/tm.aspx?m=1572933&mpage=1

 

[Vaulter98c]

That's still like 10k atleast tho right? Seems worth it for now. Got any info on this whitelist stuff?

 

[jfb9301]

Took another look a t HFM.net.... 2-3 KPPD depending on WU. I think 4 KPPD will be a good estimate though. Higher clocks on the desktop CPU vs the mobile part. I'm shopping right now for socket 1155 CPUs myself. I want another 2600K, but don't have a lot of $, ebay has them used for 200. 2500Ks are cheaper even. You might look at a cheap K CPU and OC it to within an inch of it's life. I don't know if there is such thing as a K series i3 though.

My 570s are getting almost 20 KPPD...... each...

 

[jfb9301]

Dropping to -smp (N-1) is rather a thing of the past for the most part. Used to be GPU3 used 2.5-5% of a CPU. Most GPU3 units are down to <1.5% cpu usage. Some juicy ones are <0.75%. If you meet deadlines and have a GTX460 or greater, don't drop the CPU thread, the gains far outweigh the losses.

I keep tabs on GPU folding still as I am not a CPU only guy..... My GTX460 and both GTX570s fold... 40KPPD is 40KPPD, and I'm only giving up around 10 KPPD of CPU folding to get that 40K gain.

 

[Nathan_P]

That's still like 10k atleast tho right? Seems worth it for now. Got any info on this whitelist stuff?

The GPU whitelist is supposed to make it easier to support new hardware as it emerges from each gpu maker and their respective partners, the list is updated often and can be updated by the end user to make their board work with f@h v7
GPU whitelist


Edit:- found a whole subforum on the white list:- GPUS.txt info

 

[Vaulter98c]

Eventually I'll get an ivy K series, but from what I'm seeing an i3 at 3.3-3.5 Ghz is still holding it's own in most games compared to the 5's and 7's non k series. Just don't have the dough for a 160$+ CPU, then a heat sink, I've got a water cooler and CPU and my RAM all for nearly that much

It seems at 19x12 the i3 won't keep me from gaming hard. Plus if gives me a nice step up pattern down line

 

[Nathan_P]

Even better, the 670 is already in the whitelist so its should work you download the latest version of the file

 

[Biffa]

I couldn't get my 680 to fold until I:
a) Used the v7 client
b) Added "client-type beta" to the gpu slot
c) Joined the beta program