I've got a GTX 570 running in Win7 x64, using the v7 client. I've also now got a VM of Ubuntu 11.10 running with the -smp 8 -bigadv. Since changing from running both the GPU and CPU in the Windows v7 client, to what is described above, my PPD for the GPU has dropped significantly. I was getting around 15k PPD from the GPU, now I'm getting just over 8k. Is this normal? I have 3D aceleration enabled within VMware, could this be hampering my efforts? The reason I changed from running both nativley in the v7 client was the fact I wasn't seeming to get any bigadv WU's I'm still pretty new to folding, but I'm making a push as a colleague has just been diagnosed with Hodgkins Lymphoma stage 3. Poor chap is only 21 I work for an IT company, and hopefully I might be able to get a few VM's running on our local ESX hosts, and possibly at our data centre too. Back on topic! Any help or advice you guys could give would be much appreciated
Hey Mark, From personal experience when you run a virtual machine it uses the maximum amount of resources available without sharing them. This should mean an increase in the smp client but at the expense of the GPU client. You should wait and find out whether the difference is worth it. It may also be worth trying the system without the GPU running, or trying the smp client with smp -7 to enable one spare core for the GPU client. Just experiment.
Hi, thanks for the advice. I'll give it 48hours and see what the points look like. If it's an overall net gain I may leave it. After the testing period I'll try it with one core out. Thanks for the help
Just a thought - but isn't the -bigadv flag not supported when using 8 threads? That could cause some issues ....
Correct. Now you must have at least 16 threads (cores) to fold bigadv. For now, you will still be able to fold with the flag, it just won't get any 'real' big adv work. As for running with GPUs & SMP, the above is correct, you need to work to figure out the real advantage of how to use your setup to it's fullest. For me, I have 2 GPUs and I run -smp 7 in order to use my hardware to it's fullest. If I run -smp 8, everything gets a bit slower due to 'fighting' amongst the GPUs & CPU. With only 1 GPU, you should be fine running -smp 8, but if you were to add another, I'd run -smp 7; as an aside, this only really applies to the Fermi and newer cards.
Thanks for the tips guys. I wasn't sure that the new 16 thread bigadv had gone live yet. I may possibly add a second gtx 570 if one pops up cheap, but otherwise I'm waiting for Kepler. I'll continue the tinkering and try and find the best fit for my system
