[QUOTE=Red Raven;382337]Also, mfakto is taking up 25% of my cpu. How can I reduce this?[/QUOTE]

Make sure that you're running a recent version, and that SieveOnGPU is enabled (it is by default). This appears to be a good source:
[url]http://mersenneforum.org/mfakto/mfakto-0.14/[/url]
(GPU sieving was added in 0.13)

[QUOTE=Red Raven;382337]Also, mfakto is taking up 25% of my cpu. How can I reduce this?[/QUOTE]

25%? That seems like the AMD driver bug we've had a while back... not cpu-sieving. What driver version do you have?

[QUOTE=kracker;382343]25%? That seems like the AMD driver bug we've had a while back... not cpu-sieving. What driver version do you have?[/QUOTE]
+1 to this, you should have a 64 bit OS, install the new catalyst/drivers, and use the last mfakto, and sieve on gpu. It should not take CPU resources at all (well, almost. You will notify a small decrease in the performance when all the CPU cores are really busy, the GPU can not run totally by itself). You should reach [U]at least[/U] the [URL="http://www.mersenne.ca/mfaktc.php?sort=ghdpd&noN=1"]numbers shown here[/URL], or better ("at least" because those numbers are a bit conservative) for your card. This as a check that your setup is right.

[QUOTE=LaurV;382345]+1 to this, you should have a 64 bit OS, install the new catalyst/drivers, and use the last mfakto, and sieve on gpu. It should not take CPU resources at all (well, almost. You will notify a small decrease in the performance when all the CPU cores are really busy, the GPU can not run totally by itself). You should reach [U]at least[/U] the [URL="http://www.mersenne.ca/mfaktc.php?sort=ghdpd&noN=1"]numbers shown here[/URL], or better ("at least" because those numbers are a bit conservative) for your card. This as a check that your setup is right.[/QUOTE]

I have Catalyst 14.1. mfakto is version .15pre2.

I downloaded mfakto .14 and it barely uses the CPU. The "Wait" percentage, which seems to have something to do with sieving, is always at zero now, and it's finishing numbers in a few minutes where as .15pre2 took an hour to do one. What's going on? Are the results still correct?

(ignorant of AMD GPUs post deleted by sheepish poster.)

[QUOTE=Red Raven;382360]I downloaded mfakto .14 and it barely uses the CPU. The "Wait" percentage, which seems to have something to do with sieving, is always at zero now, and it's finishing numbers in a few minutes where as .15pre2 took an hour to do one. What's going on? Are the results still correct?[/QUOTE]
You must not have edited mfakto.ini to SieveOnGPU=1. 0.15pre2's default setting (ie. what comes with it in the .zip) is SieveOnGPU=0, while with 0.14 this value is 1.

As TheMawn said, open up mfakto.ini in notepad and Ctrl+F search for "SieveOnGPU", then change the 0 to a 1.

[QUOTE=Jayder;382370]You must not have edited mfakto.ini to SieveOnGPU=1. 0.15pre2's default setting (ie. what comes with it in the .zip) is SieveOnGPU=0, while with 0.14 this value is 1.

As TheMawn said, open up mfakto.ini in notepad and Ctrl+F search for "SieveOnGPU", then change the 0 to a 1.[/QUOTE]

Oh, I thought 1 was the default in both, so I didn't bother. I changed it in .15pre2, so when .14 is done with the batch of 50 TFs I just gave it I'll switch back over.

I'm realizing now that this GPU is crunching TFs much faster than my 2.2Ghz single core laptop, which takes days to crunch TFs. Should it be taking days? At this point I'll probably take it off GIMPS until I reinstall Windows 7.

[QUOTE=Red Raven;382391]Oh, I thought 1 was the default in both, so I didn't bother. I changed it in .15pre2, so when .14 is done with the batch of 50 TFs I just gave it I'll switch back over.

I'm realizing now that this GPU is crunching TFs much faster than my 2.2Ghz single core laptop, which takes days to crunch TFs. Should it be taking days? At this point I'll probably take it off GIMPS until I reinstall Windows 7.[/QUOTE]
Depending on the bit level you're TFing to, it's certainly possible for your laptop CPU to take days to finish a TF job. The CPU is probably better off doing DC (LL double-check) tests or P-1 work, although there's no harm in having it do TF.

Rodrigo

That is completely normal, GPUs are extremely good at TF-ing compared to CPUs. CPUs can better be used for P-1 and LL/DC testing.

Yeah. I don't know why slow (laptops for example) CPUs default to TF.

There once was a debate where the conclusion was that 100.01 km/h is faster than 100.00 km/h but the wear and tear on your laptop that is resulting from crunching TF is completely not worth it.

I would recommend double-checks. If you have some RAM available, you could put one worker to P-1.