CPU Speed and WU's

I posted this at the folding community board.


mrm1957 wrote:
I brought this up on our team site when the question was asked about how to get work units in faster.
If Stanford can incorporate a way to determine the CPU's speed during the initial setup of client. If it were possible that the CPU speed be broke into two pairs say 300 to 1200 Mhz do work units from 0.1 to 2.0 and
from 1200 and up to do the work units from 2.1 to 6.0. This would make the larger work units get done by the faster machines and the slower ones to get the smaller ones done. This would get results in faster and
increase the speed of the project.
I know in my case with using two machines to fold it seems that I always get the larger WU's on my 800 Mhz machine and the smaller ones get on to my 1600 Mhz machine. Some of the larger WU's on the small machine can take up to 30 hours to complete so it would make sense that if I were doing smaller WU's I could probally get in four units instead of one.


I agree and this is in the works. Once we get the new servers up (and thus have a broader array of WUs to ship out), we'll start assigning WUs from the AS based on CPU speed, with the slower CPUs getting the smaller WUs, etc.

Vijay

cool i hope this happens soon, i like the 603-606, 1.5 pointers rule!

Yeah that is what it's all about to get the production up. It's a killer getting those 5 & 6 pointers on a small CPU. It took me 33
hours to get a 6 pointer done on the Duron 800. If I remeber right
those servers should be there this week and probally be up in two weeks.

mrm1957 said:

Yeah that is what it's all about to get the production up. It's a killer getting those 5 & 6 pointers on a small CPU. It took me 33
hours to get a 6 pointer done on the Duron 800. If I remeber right
those servers should be there this week and probally be up in two weeks.

Click to expand...

i have done the math maybe 2 weeks ago and it took the same amount of time (basically) to do 1 6 ponter and 3 2 pointers, so i dunno if it will really make a difference on our end...but will most likely let them finish those high pointers faster since the slower cpu's wont be wasting time with them...get my point? (im sure it is obvious, and that most people already knew this)

That's exactly it getting more WU's done. Info from 3 proteins is worth more than 1.

mrm1957 said:

That's exactly it getting more WU's done. Info from 3 proteins is worth more than 1.

Click to expand...

you sure that is true? maybe since the bigger ones are much more complex they could be more valuable than 10 (random #) small ones