Howdy

[spikey_richie]

Hello there AMD users, haven't posted on here for a while.

I took a break from crunching when I moved back home from the USA, but I bought an nVidia GTS 450 last week so I thought I'd put it to the test running some Einstein@home.

Have I missed much?

[Dirk Broer]

Hi Rich,

I wouldn't use my GPUs for Einstein if I were you, as they take a complete CPU core for support. Effectively this means that when you have a dual core CPU plus a GPU you can run just one normal project besides Einstein, when you have decided to let Einstein use your GPU. You can of course run non-cpu intensive projects, susch as WUProp of FreeHAL (but the latter not just now), but Einstein makes a rather costly use of your CPU cycles. It is better to run Einstein CPU-only and run another project GPU-only, such as MilkyWay, DNETC, PrimeGrid or Collatz. This will leave another core free for another project. Einstein does not run all that much faster when it uses both CPU and GPU either. All in all a shining example of bad CUDA programming, IMHO.

[spikey_richie]

Hi Dirk,

Thanks for that. Once I've added Milkyway back into my BOINC account, how do I configure it to exclusively use GPU?

[Dirk Broer]

At http://milkyway.cs.rpi.edu/milkyway/
Go to 'Your Account',
Under MilkyWay@Home preferences, set "Use CPU" to 'No'.
Choose the appropriate Video card family, so set either
"Use NVIDIA GPU if present"
to 'Yes' or
"Use ATI GPU if present"

[spikey_richie]

Ah ha, a sneaky new section

Thanks for that Dirk.

Einstein is now happily sat on my CPU(s) and MilkyWay is on my GPU