mfaktc: a CUDA program for Mersenne prefactoring - Page 58

TheJudger · 2011-03-13, 14:08

64bit Windows executable!
I've used the same system for building as before: http://www.mersenneforum.org/showpos...&postcount=586

Maximum files size limit here doesn't allow me to include the cudart*.dll.

Oliver

TheJudger · 2011-03-13, 14:10

Last but not least the 32bit Windows executable

firejuggler · 2011-03-13, 14:29

thanks for the binaries. I will be able to test them tonight

drh · 2011-03-13, 15:16

This is my first attempt at trying mfaktc as I finally just got a laptop with a NVS 3100M to get familiar with the program. I'm running driver 257.38, and I get this error message ...

cudaGetLastError() returned 35: CUDA driver version is insufficient for CUDA runtime version

Do I need to upgrade my 257.38 driver?

Thanks,
Doug

James Heinrich · 2011-03-13, 15:58

Quote:

Originally Posted by TheJudger

Maximum files size limit here doesn't allow me to include the cudart*.dll.

Sure it does... it compresses well (only 106kB for 7z)

James Heinrich · 2011-03-13, 16:03

Quote:

Originally Posted by TheJudger

checkpoint files are ignored from foreign versions of mfaktc

Including pre-release variants of the same version, I have discovered, for those that care.

Which, for me, means it'll be another 6d14h before I can upgrade to 0.16-final, once I finish the current bitlevel of my current assignment.

ET_ · 2011-03-13, 16:33

Quote:

Originally Posted by James Heinrich

Including pre-release variants of the same version, I have discovered, for those that care.

Which, for me, means it'll be another 6d14h before I can upgrade to 0.16-final, once I finish the current bitlevel of my current assignment.

I'm still working with v.0.14...

I'll start with v0.16 as soon as my range at 78 bits ends.

Luigi

TheJudger · 2011-03-13, 17:57

Hi Doug,

Quote:

Originally Posted by drh

This is my first attempt at trying mfaktc as I finally just got a laptop with a NVS 3100M to get familiar with the program. I'm running driver 257.38, and I get this error message ...

cudaGetLastError() returned 35: CUDA driver version is insufficient for CUDA runtime version

Do I need to upgrade my 257.38 driver?

Did you receive this error message right after the famous "cudaStreamCreate() failed for stream 0"?

You're right, this indicates that your driver doesn't fit with the CUDA runtime dll. If possible try to update the driver to some 260+ version.

Quote:

Originally Posted by James Heinrich

Sure it does... it compresses well (only 106kB for 7z)

Thank you, James! Actually I didn't try to compress.

Oliver

drh · 2011-03-13, 18:11

Quote:

Originally Posted by TheJudger

Hi Doug,

Did you receive this error message right after the famous "cudaStreamCreate() failed for stream 0"?

You're right, this indicates that your driver doesn't fit with the CUDA runtime dll. If possible try to update the driver to some 260+ version.

I get it right after I try to run mfaktc -st ... mfaktc -tt runs fine, as far as I can tell. I get cudaSetDevice(0) failed. I also get it if I try to run the program with a -tf ...

Thanks,
Doug

firejuggler · 2011-03-13, 18:40

under ubuntu 10.10, with a e8300, gtx460,
for assignment: exp=3321928109 bit_min=75 bit_max=76

0.15
307.36M candidates in 2491ms (123.39M/s) (avg. wait: 250us)

0.16
308.28M | 2.388s | 129.10M/s | 5000 | 34m59s | 34us

so a 5% improvement

under windows i didn't find a significant improvement.

TheJudger · 2011-03-13, 18:50

firejuggler, you're limited by the speed of your CPU (average wait is very small), thats why you don't see ~18%. But your GPU should consume less power that way.
The CPU bottleneck is a common problem for mfaktc on fast GPUs.

Only solution today: start another instance of mfaktc.

Oliver

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

2011-03-13, 14:29	#630
firejuggler Apr 2010 Over the rainbow 2·1,303 Posts	thanks for the binaries. I will be able to test them tonight

2011-03-13, 15:16	#631
drh Jan 2011 Cincinnati, OH 100₁₀ Posts	This is my first attempt at trying mfaktc as I finally just got a laptop with a NVS 3100M to get familiar with the program. I'm running driver 257.38, and I get this error message ... cudaGetLastError() returned 35: CUDA driver version is insufficient for CUDA runtime version Do I need to upgrade my 257.38 driver? Thanks, Doug

2011-03-13, 18:40	#637
firejuggler Apr 2010 Over the rainbow 101000101110₂ Posts	under ubuntu 10.10, with a e8300, gtx460, for assignment: exp=3321928109 bit_min=75 bit_max=76 0.15 307.36M candidates in 2491ms (123.39M/s) (avg. wait: 250us) 0.16 308.28M \| 2.388s \| 129.10M/s \| 5000 \| 34m59s \| 34us so a 5% improvement under windows i didn't find a significant improvement.

2011-03-13, 18:50	#638
TheJudger "Oliver" Mar 2005 Germany 2127₈ Posts	firejuggler, you're limited by the speed of your CPU (average wait is very small), thats why you don't see ~18%. But your GPU should consume less power that way. The CPU bottleneck is a common problem for mfaktc on fast GPUs. Only solution today: start another instance of mfaktc. Oliver