very slow WU progress with P4's..

[xnoto]

Hey all,


I've been folding for a while now, I have 5 dedicated machines folding 24/7 and have noticed that my P4's take VERY long to complete 1 WU.

My 3200+ XP and 1600+ chew through WU's with -advmethods and -forceSSE, but my 2.0ghz P4 and 2.4ghz P4 take days to proccess even 1 WU.

For instance, i'm currently working on a 400 size WU, and it's been proccessing for over 15 hours and is only on 91/400 on my 2.4ghz P4.

Is there a certain advmethods flag to use for P4's so they can stack up to my AMD's?

Thanks

 

[CIWS]

The P4 w/ SSE only comes into play if you are processing a protein that is using the Gromacs core. If your machine(s) are running a Tinker based protein/unit it will take much longer to process due to the fact the Tinker core does not utilize the SSE instruction set for it's units in Floating Point calculations. So P4s with Tinker units take longer to complete.

 

[xnoto]

so is it still an advantage to -advmethods -forceSSE on the p4's?

 

[Celerator]

That's my experience. In the past, one was more likely to get Gromacs with -advmethods. SSE is the default with P4's, so -forceSSE will force it on should a WU run into a problem (rare event in my experience). With AMDs, the -forceSSE seems to have a large impact on performance as the AMD doesn't default to SSE. On my faster machines I use both flags. On the slower machines (without SSE) I have set them up to do the "timeless Tinkers" (select Genome in Options of graphical client), and run these pokey computers without any flags.