Folding Forum

technic58 wrote:2. There was some discussion about slow GPUs hogging WUs and delaying projects getting done. I haven't messed with overclocking these cards yet, but if I could get these cards to reliably turn in projects, but say I snuck in just under the deadline every time, are you saying that this would be a net negative for F@H? I would think this would be the case if there is a huge shortage of WUs, but if there are enough WUs to go around, I would think even slow (but unexpired) returns would be valuable. Are you saying that people with slower hardware should not contribute even if they can turn in useful work? I would be surprised if this was the position of F@H, but if it is, I guess let me know.

Turning in projects at the last second isn't detrimental to F@H. The total amount of work returned on time per period of time from all donors is more important than the average return time. So, even slower systems that can make the deadlines provide benefit. By the way, don't overclock your GPUs. F@H taxes the hardware enough that you will likely just make them unstable and end up dumping more work units due to errors in calculations.

The only result that matters is the first one completed successfully. If all the fast GPUs are assigned the optimum projects (which is not yet possible until the servers are enhanced) there's still plenty of work that can be completed by slower GPUs. For those of us with slower GPUs, the optimum assignment would be something that I can finish promptly -- where "promptly" is defined based on my peers. How the servers select my peers ultimately leads to the emergence of some kind of "average" that applies to WUs that are being assigned to us.

I would hope that FAH completes the development of an optimum assignment strategy soon, but for the most part, there are a lot of WUs assigned randomly in pools of semi-managed peers.

The total amount of work returned per period of time from all donors would only matter if WUs never had to be reassigned. Actual duplicates are useless to science since only the first one returned produces scientific progress to the trajectory. On the other hand, duplication is necessary because every WU must be completed and some, in fact, do get lost.

Blacklisting the following servers seems to be allowing my GPUs to get more reasonable WUs:

18.188.125.154
140.163.4.200
206.223.170.146

Thanks again for all of the advice!