Re: Is one wu gpu better then one wu nacI/smp for research?

[bruce]

Though you didn't mention the Android client, I'll include it in my answer.

The value to science depends on the hardware capabilities. Slower {CPUs/GPUs} with fewer threads produce less than faster [CPUs/GPUs] with more threads. As a general statement, GPUs are more powerful than CPUs which are more powerful than the processor in a typiclal phone.

The nacl client is a SMP client of sorts. I'm not sure how it establishes how many CPUs to use. The SMP client can be managed explicitly so you may be able to get more work out of the same hardware. Both of them suffer if you allocate N CPUs if, because of some other long-running process(es) only a smaller number of CPUs are available on a regular basis. GPUs generally don't have the same limitation.

At one time, the GPU client was limited to certain types of analysis whereas the CPU client was more versatile. The vsrsatility of the FAHCores for GPUs has been improved very significantly and I believe they're now equivalent.

When developing a new project, some consideration is also given to maintain a reasonable balance between projects available for client X and the resources being donated by donors running that client. Then, too, the project's due-date may be considered based on the ability of that client to complete the project on schedule.

[Kjetil]

I had hope for yes or no? But tanks.

[Nathan_P]

I was testing some ram this afternoon so I fired up nacl on my daily driver, it used all 24 thread but only to about 70%. As for the GPU/CPU question I have to agree with Bruce, most WU are now equivalent in size and can be processed really quickly on the most powerful hardware. CPU's do have an advantage with really large WU, 1m+ atoms, but they are specialised projects and not widely used

[bruce]

My statement is not 100% true. When the GPU core was limited to implicit solvent, it was certainly more limited, but now that both CPUs and GPUs can perform explicit solvent analyses, that different has vanished.

There probably are some differences in which force-fields can be used. Gromacs (for CPUs) and openMM (for GPUs) are not identical, but from what I understand, the differences are pretty small for the types of analysis that FAH does.

[bruce]

I had hope for yes or no? But tanks.

No.

[Kjetil]

Bruce: So more ppd is better for resarch then more wu, i am running 9 smp/nacI. So is better not to run nacI/smp? But more gpu? =more ppd. sorry for my bad english.

Edit okay

[bruce]

So is better not to run nacI/smp? But more gpu?

Not really.

Run both, since they're both producing good results and for the most part, they use different resources to total PPD goes up, too.

[Kjetil]

I will run gpu/naci or gpu/smp. Now i am running gpu and naci. 11 gpu and 9 nacI now at team 24.

[Rel25917]

Gpu and nacl is a bad idea. The gpu likes to have a core on the processor free and with nacl there is no way to leave one free, it always uses them all. Running them together will slow them both down.

[Nathan_P]

But nacl is fixed points, no QRB so in terms of PPD it makes no major difference

[davidcoton]

GPU and NaCl on the same PC is not optimum. If the v7 client should be installed, it is easy to set up an SMP slot as well, instead of NaCl. It can be set to leave a CPU core per GPU free, and will get QRB points on both slots.

If you still run v6, unless you have a good reason, uopgrade now.

[Kjetil]

Run process lasso?
Running v7 client.
PS: started folding 2004 and upgraded to v7 when it was released
I miss 100-150K ppd, but more wu