Folding Forum

I was browsing this forum and read some stats and it looks like there are a huge increase in GPU-folders (please correct me if I'm wrong). That more people want fold and/or use all the resources they have in their systems, is very good news, but then one question hit me: Are there enough WU's/Projects in the pipeline to feed/use all this GPU-power? Is it possible to give an update to us folders on the future WU-plans (maybe for the next year? - Both for CPU's and GPU's?) - What they do and how big they are?

Secondly: When my GPU-client upload a finished WU, it takes only 5-15 seconds, but my SMP CPU-client use about 15min! Why - Is the CPU-result so much bigger/complex?

http://fah-web.stanford.edu/serverstat.html

171.64.65.20 63015
171.64.65.102 17584
171.64.65.103 16618
171.64.65.106 46589
171.67.108.11 16853

160,659 WUs
Active GPUs: 20,000
That ought to be enough for 3-5 days =P

Mr.Nosmo wrote:I was browsing this forum and read some stats and it looks like there are a huge increase in GPU-folders (please correct me if I'm wrong). That more people want fold and/or use all the resources they have in their systems, is very good news, but then one question hit me: Are there enough WU's/Projects in the pipeline to feed/use all this GPU-power? Is it possible to give an update to us folders on the future WU-plans (maybe for the next year? - Both for CPU's and GPU's?) - What they do and how big they are?

There will be, because the field that this project studies is so incomprehensibly vast that there will be new Work Unit series coming out for many years to come. However you are unlikely to get advance notice of what will be being studied until the day they release that new WU series

Secondly: When my GPU-client upload a finished WU, it takes only 5-15 seconds, but my SMP CPU-client use about 15min! Why - Is the CPU-result so much bigger/complex?

The CPU results are considerably bigger yes. The GPU client at the moment processes units with less than 1500 atoms in them. The SMP client units currently have up to 77,000 atoms in them and I think process a longer time frame so there are more results to return (even though only the final coordinates of the atoms are sent).

I know there are almost endless amount of data to be calculated - The question was more focused at Stanford, asking if they have the resources/enough staff to generate the WU's for us.
I have also been reading this forum long enough to know they don't tell us about new WU's before they are released - I was only trying to get a feeling of what kind of science they are planing and what kind of work-load it will give us folders - Fx. we are going to study protein XX they does YY, so we need to folders to calculate about ZZ-size atoms (ZZ=small, medium, large, or supersize)....

Just to have a general idea of the way we are going to walk......

Mr.Nosmo wrote:Just to have a general idea of the way we are going to walk......

I think you've asked the wrong question.

I would have asked if there are enough servers to hold enough projects to keep all of the clients busy.

Every time that the servers have run out of WUs, it has been because there weren't enough projects ready for a specific client. Specifically, I've seen that for the PS3, for the old PPC Macs, and for one or both of the GPU clients. For the PS3 and GPUs, the problem was eventually solved by adding a new server or two. As the number of clients for any specific client grows, new servers need to be added and that's not something you can do in a few days.

The more flexibility that there is in assigning projects is important, too.

In the case of the PPC Mac, I think it just got overlooked. The number of clients is small and gradually decreasing and apparently only a fraction of the uniprocessor projects can be shared between Windows, Linux, and Mac. It's not easy to figure out if there are plenty of WUs for the Mac, amd if they do run out, it may take a few days to beta test a new project.

I'm sure the goal for the two GPU clients is to share all projects between nvidia and ati, but currently that's not possible. When that happens, they'll probably save one server rather than having to keep surplus WUs two places.