Folding Forum

bruce wrote:

erich56 wrote:I consider to buy a PC with a RTX 3070 inside. Will I be able to crunch FAH without problems?

Based on the proteins being studied today, the 3070 may be more than you need. You'd probably be better off with a system that can handle two GPUs that are somewhat smaller. There's no way to predict what proteins will be studied in the future.

well, I bought them already

erich56 wrote:

bruce wrote:

erich56 wrote:I consider to buy a PC with a RTX 3070 inside. Will I be able to crunch FAH without problems?

Based on the proteins being studied today, the 3070 may be more than you need. You'd probably be better off with a system that can handle two GPUs that are somewhat smaller. There's no way to predict what proteins will be studied in the future.

well, I bought them already

However, I have older PCs with "smaller" GPUs inside, so I can always use them for FAH crunching, which I have done so far anyway

is there a provision to run two work tasks concurrently on one GPU? I am aware of the Configure->Expert tab, but am not sure where the is documentation

jchang6 wrote:is there a provision to run two work tasks concurrently on one GPU? I am aware of the Configure->Expert tab, but am not sure where the is documentation

As far as I'm aware, you cant run 2 WUs at the same time on the same GPU. Somebody, correct me if I'm wrong.

It is not officially supported. There has been talk about doing something like that, or at least match GPUs to projects in a more fine-grained way, but there's no timeline for when that might happen.

If you have a professional GPU that supports multiple users on the same GPU, you might be able to manually split the GPU into compute units that Folding@Home sees as separate GPUs, but most gaming focused GPUs don't support that in the driver.

gunnarre wrote:It is not officially supported. There has been talk about doing something like that, or at least match GPUs to projects in a more fine-grained way, but there's no timeline for when that might happen.

If you have a professional GPU that supports multiple users on the same GPU, you might be able to manually split the GPU into compute units that Folding@Home sees as separate GPUs, but most gaming focused GPUs don't support that in the driver.

If a WU can reliably saturate [80-100%] a Gpu, what would the point be in running 2 WUs at the same time? If it can't I can see the rational in running more than 1 WU at any time.

erich56 wrote:@gunnarre - thanks for the hint.
If I understand right, I could choose 17220 or 17231. Both are on top of the list with 707.654 atoms. However, how can I know to which disease these two numbers are assigned to?
The setting in the web control window only let's you choose between diseases, but not project number - at least from what I can see.

bruce wrote:looking up those two projects shows that they are owned by Neha Vithani and they are both part of the CORONAVIRUS PROJECT

So I set "Covid-19" as my target disease, but got project 13446 was downloaded for crunching, neither 17220 nor 17231.
Sometimes life does not seem to be easy

13446 simulates the main viral protease of Covid-19 and has about 80 000 atoms in it. It's part of a Covid-19 Moonshot sprint. The Moonshot Sprint projects tend to favor smaller GPUs disproportionally: https://foldingathome.org/2020/09/28/fo ... a-support/

Once the sprint finishes you're more likely to get one of the receptor binding projects which have more than 200 000 atoms. The sprint is now at about 70.4% according to the front page of https://foldingathome.org. You will occasionally get non-Moonshot WUs intermingled though, depending on what the work servers have ready.

@gunnarre: thanks for the thorough information
well, as you say, 13446 with it's about 80.000 atoms is more on the lower side what concerns the number of atoms, but it could be even worse, if I look at 13447 with it's 4.082 atoms only

So eventually I will get the other WUs with more atoms, thus bringing more work to my two RTX3070

The good thing about not utilizing a 3070 fully, is that they will run higher boost frequencies, and at the same time, you'll be able to cut power to the GPU by a lot more.