Folding Forum

So I started folding with my 5870 in Jan. Seeing how my CPU was being unused; I signed up for BOINC (SETI@Home,Rosetta@home(Rosetta@home's been moved to my laptops)) to take advantage of my i7 920.
However; just recently; I read some articles about the i7 920 getting >10k PPD.

Would it be worth it, switching the clients around; to have BOINC using my GPU (Will be switching to Milkyway@Home if I do this) and using the CPU client for F@H?

Thanks;

EDIT: In case you wanted to know; my 5870 gets around about 4800-5200PPD depending on the WU; My 920 is @ 4GHz

CSP,

If you run an i7-920 @ 4GHz with the bigadv workunits you'll get around 25K ppd. You'll need to set up vmware workstation/play and a Linux install (Or download Not Fred's folding image), along with 5GB of memory you can allocate to the VM.

If instead you just fold with the Windows SMP client, it'll be around 15-20K PPD, much lower setup involved, and a lot less memory required.

H.

You don't even need a VM! Just run the v6.29 SMP client in Windows with the -advmethods and -smp switches. The new A3 work units getting killer bonus points.

Yes, you could get even more points running -bigadv in linux, but not worth the hassle to some. One can even conclude that -bigadv will be coming to Windows eventually.

7im,

I'll deal with the hassle, it's not a significant hassle, and for me it's worth the extra ppd. Now, when A3 makes it to Windows, I'll switch it over in a heartbeat ... well a -oneunit beat ....

H.

csp40 wrote:So I started folding with my 5870 in Jan. Seeing how my CPU was being unused; I signed up for BOINC (SETI@Home,Rosetta@home(Rosetta@home's been moved to my laptops)) to take advantage of my i7 920.
However; just recently; I read some articles about the i7 920 getting >10k PPD.

Would it be worth it, switching the clients around; to have BOINC using my GPU (Will be switching to Milkyway@Home if I do this) and using the CPU client for F@H?

Thanks;

EDIT: In case you wanted to know; my 5870 gets around about 4800-5200PPD depending on the WU; My 920 is @ 4GHz

The best option, of course, is to put all of your resources into FAH (at least from my perspective).

Seriously, though, any GPU client will steal a small percentage of cycles from a CPU to keep the data flowing out of/into the GPU and ultimately preparing it for upload. I have no information on how much Milkyway will need, but today's FAH GPU client for ATI needs a lot more CPU resources than today's FAH GPU client for NV. For maximum GPU performance, that client needs to have priority over the SMP (CPU) client so the GPU client will slow the CPU client down some. That MAY turn out to be a performance-killer for the CPU client. Remember that the bonus for the SMP2 client (FahCore_a3) decreases whenever something delays the return of the results.

Haitch wrote:CSP,

If you run an i7-920 @ 4GHz with the bigadv workunits you'll get around 25K ppd. You'll need to set up vmware workstation/play and a Linux install (Or download Not Fred's folding image), along with 5GB of memory you can allocate to the VM.

If instead you just fold with the Windows SMP client, it'll be around 15-20K PPD, much lower setup involved, and a lot less memory required.

H.

7im wrote:You don't even need a VM! Just run the v6.29 SMP client in Windows with the -advmethods and -smp switches. The new A3 work units getting killer bonus points.

Yes, you could get even more points running -bigadv in linux, but not worth the hassle to some. One can even conclude that -bigadv will be coming to Windows eventually.

Well; I do use my 'puter for other purposes as well; I don't think allocating 5GB worth of RAM will allow me to play Crysis whilst folding D:

bruce wrote:
The best option, of course, is to put all of your resources into FAH (at least from my perspective).

Seriously, though, any GPU client will steal a small percentage of cycles from a CPU to keep the data flowing out of/into the GPU and ultimately preparing it for upload. I have no information on how much Milkyway will need, but today's FAH GPU client for ATI needs a lot more CPU resources than today's FAH GPU client for NV. For maximum GPU performance, that client needs to have priority over the SMP (CPU) client so the GPU client will slow the CPU client down some. That MAY turn out to be a performance-killer for the CPU client. Remember that the bonus for the SMP2 client (FahCore_a3) decreases whenever something delays the return of the results.

hmm... I'll see how it pans out; if Milkyway takes up too much CPU then I'll abandon it.

Speaking of which; would it be possible to run both CPU and GPU clients at the same time?

As far as F@H goes, you can run both the CPU Client and the GPU client together. The CPU client's priority should b set to idle while the GPU should be set to slightly higher. As you are using an ATI Card, you may want to use the Environment Virabales so that the GPU Client can be more effective.

PantherX wrote:As far as F@H goes, you can run both the CPU Client and the GPU client together. The CPU client's priority should b set to idle while the GPU should be set to slightly higher. As you are using an ATI Card, you may want to use the Environment Virabales so that the GPU Client can be more effective.

Thanks! And I'm already using Environment Variables >_>