mfaktc: a CUDA program for Mersenne prefactoring - Page 134

Dubslow · 2011-12-26, 20:01

I've noticed mine doing similar things. I don't have a graph, but the avg. rate seems to drop ~10% for around 30-60 seconds, before going back to full throttle for usually at least 90+, often 120-180+ seconds, before it drops 10% again.

Note: BigBrother, your CPU wait times are sky high. They should be around 2-5% -- try increasing SievePrimes. (Your card does seem to be among the slower ones, but you should still be able to get a decent throughput increase with an increase in SievePrimes.)

When these fluctuations happen to me, around 80% of them CPU wait stays the same, in the low 2.xx% range. On the other 20% though it spikes to 10% or 20% -- but like I said, not all the time. Rather weird.

@kladner: 'start' is not its own executable, which is (AFAIK) what shortcuts can recognize. cmd.exe is (obviously) its own executable.

BigBrother · 2011-12-26, 20:45

Quote:

Originally Posted by Dubslow

Note: BigBrother, your CPU wait times are sky high. They should be around 2-5% -- try increasing SievePrimes. (Your card does seem to be among the slower ones, but you should still be able to get a decent throughput increase with an increase in SievePrimes.)

I know, I set SievePrimesAdjust=0 and SievePrimes=5000 for this test.

Dubslow · 2011-12-26, 20:56

Oh. My apologies.

bcp19 · 2011-12-26, 23:30

Quote:

Originally Posted by BigBrother

Something odd is happening when I run mfakct 0.18 on my GT 555M. After a short while, the GPU usage drops from 100% to +- 70-80%, and a little later, it goes back again to 100% for a short time, before it drops again, ad infinitum. Mfakct 0.17 always runs at 100% GPU usage on this card.

I was having a similiar problem on my 560Ti, one of the 2 instances of mfaktc I had running would 'pause' and the gpu would drop from 100 to 60 or so %. I ended up opening a ticket with nVidia support, and they had me install their new beta driver and the problem went away. I thought I was having heat related issues since it was running the GPU at around 88C, but they said it can run up to 99C without problems.

BigBrother · 2011-12-27, 07:56

Quote:

Originally Posted by bcp19

I was having a similiar problem on my 560Ti, one of the 2 instances of mfaktc I had running would 'pause' and the gpu would drop from 100 to 60 or so %. I ended up opening a ticket with nVidia support, and they had me install their new beta driver and the problem went away. I thought I was having heat related issues since it was running the GPU at around 88C, but they said it can run up to 99C without problems.

I'm pretty sure this problem is not heat related, at least not on my card. I'll try installing some new drivers tonight, what version are you using now?

bcp19 · 2011-12-27, 14:24

Quote:

Originally Posted by BigBrother

I'm pretty sure this problem is not heat related, at least not on my card. I'll try installing some new drivers tonight, what version are you using now?

GeForce 290.53 beta for desktop GPUs

BigBrother · 2011-12-28, 22:00

Quote:

Originally Posted by bcp19

GeForce 290.53 beta for desktop GPUs

I installed 290.53 beta for laptop, still the same problem with mfaktc 0.18, I'll stick with mfaktc 0.17 for now.

TheJudger · 2011-12-28, 22:35

Quote:

Originally Posted by BigBrother

Something odd is happening when I run mfakct 0.18 on my GT 555M. After a short while, the GPU usage drops from 100% to +- 70-80%, and a little later, it goes back again to 100% for a short time, before it drops again, ad infinitum. Mfakct 0.17 always runs at 100% GPU usage on this card.

Do you have AllowSleep enabled?

Oliver

Dubslow · 2011-12-28, 23:57

http://gpuz.techpowerup.com/11/12/28/csv.png

I do not have AllowSleep enabled. I was doing literally the exact same thing for all the time shown in the shot, but for whatever reason load dropped down to 80% from 92%, and avg. rate dropped from ~185 to 170's and 160's, and for one class even down to 130's before going back up to 180. CPU wait remained more or less constant. (The portions with reduced load are the ones that are flat, as opposed to bumpy. I do wish the graph was taller.)

kladner · 2011-12-29, 05:28

Quote:

Originally Posted by Dubslow

http://gpuz.techpowerup.com/11/12/28/csv.png

I do not have AllowSleep enabled. I was doing literally the exact same thing for all the time shown in the shot, but for whatever reason load dropped down to 80% from 92%, and avg. rate dropped from ~185 to 170's and 160's, and for one class even down to 130's before going back up to 180. CPU wait remained more or less constant. (The portions with reduced load are the ones that are flat, as opposed to bumpy. I do wish the graph was taller.)

I do have AllowSleep=1. Most of the time the GPU runs at 95-98%. I do see variations in throughput, but lots of times it has to do with other things going on in the system which is taking over more CPU time and starving mfaktc. For example, anything involving Adobe apps almost always eats a chunk of performance. This shows up as big dips in GPU utilization. Norton and its like can certainly do the same. Are you sure what else is going on in your machine?

BigBrother · 2011-12-29, 11:35

Quote:

Originally Posted by TheJudger

Do you have AllowSleep enabled?

Oliver

I did. I disabled it and now it seems to run without performance dips. I'll try to check it out further tonight.

2011-12-26, 20:01	#1464
Dubslow Basketry That Evening! "Bunslow the Bold" Jun 2011 40<A<43 -89<O<-88 3×29×83 Posts	I've noticed mine doing similar things. I don't have a graph, but the avg. rate seems to drop ~10% for around 30-60 seconds, before going back to full throttle for usually at least 90+, often 120-180+ seconds, before it drops 10% again. Note: BigBrother, your CPU wait times are sky high. They should be around 2-5% -- try increasing SievePrimes. (Your card does seem to be among the slower ones, but you should still be able to get a decent throughput increase with an increase in SievePrimes.) When these fluctuations happen to me, around 80% of them CPU wait stays the same, in the low 2.xx% range. On the other 20% though it spikes to 10% or 20% -- but like I said, not all the time. Rather weird. @kladner: 'start' is not its own executable, which is (AFAIK) what shortcuts can recognize. cmd.exe is (obviously) its own executable. Last fiddled with by Dubslow on 2011-12-26 at 20:02

2011-12-28, 23:57	#1472
Dubslow Basketry That Evening! "Bunslow the Bold" Jun 2011 40<A<43 -89<O<-88 3×29×83 Posts	http://gpuz.techpowerup.com/11/12/28/csv.png I do not have AllowSleep enabled. I was doing literally the exact same thing for all the time shown in the shot, but for whatever reason load dropped down to 80% from 92%, and avg. rate dropped from ~185 to 170's and 160's, and for one class even down to 130's before going back up to 180. CPU wait remained more or less constant. (The portions with reduced load are the ones that are flat, as opposed to bumpy. I do wish the graph was taller.) Last fiddled with by Dubslow on 2011-12-28 at 23:59

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
mfakto: an OpenCL program for Mersenne prefactoring	Bdot	GPU Computing	1676	2021-06-30 21:23
The P-1 factoring CUDA program	firejuggler	GPU Computing	753	2020-12-12 18:07
gr-mfaktc: a CUDA program for generalized repunits prefactoring	MrRepunit	GPU Computing	32	2020-11-11 19:56
mfaktc 0.21 - CUDA runtime wrong	keisentraut	Software	2	2020-08-18 07:03
World's second-dumbest CUDA program	fivemack	Programming	112	2015-02-12 22:51

2011-12-26, 20:56	#1466
Dubslow Basketry That Evening! "Bunslow the Bold" Jun 2011 40<A<43 -89<O<-88 3×29×83 Posts	Oh. My apologies.