mersenneforum.org mfaktc doesn't work for me
 Register FAQ Search Today's Posts Mark Forums Read

 2012-01-25, 13:24 #1 sonjohan     May 2003 Belgium 11616 Posts mfaktc doesn't work for me I want to run mfaktc on my desktop PC. It has a nVidia GeForce 8800 GTS. It's computing capability should be 1.1 if I check nvidia's website. The FAQ for GPU computing mentions compute capability >=1.1. When I run mfaktc, it says the compute capability of my GPU is 1, hence it stops. --> What's going wrong?
2012-01-25, 15:36   #2
Ralf Recker

Oct 2010

191 Posts

Quote:
 Originally Posted by sonjohan I want to run mfaktc on my desktop PC. It has a nVidia GeForce 8800 GTS. It's computing capability should be 1.1 if I check nvidia's website. The FAQ for GPU computing mentions compute capability >=1.1. When I run mfaktc, it says the compute capability of my GPU is 1, hence it stops. --> What's going wrong?
I guess it is a G80 based 8800GTS.

 2012-01-26, 08:37 #3 sonjohan     May 2003 Belgium 2·139 Posts 1. How can I check that? It would be nice to know. I have also another laptop where the graphics card is on the CUDA-list, but it doesn't work either.. 2. How many instances of mfaktc can I run on a GTX 570M? I noticed that 2 instances can run without losing time. 3. Is it possible to lower the screen output from mfaktc? Right now, it sends a screen output every 5 or 10 iterations... Last fiddled with by sonjohan on 2012-01-26 at 08:40 Reason: Extra questions
2012-01-26, 09:53   #4
BigBrother

Feb 2005
The Netherlands

110110102 Posts

Quote:
 Originally Posted by sonjohan 1. How can I check that? It would be nice to know. I have also another laptop where the graphics card is on the CUDA-list, but it doesn't work either..
Try GPU-Z, wonderful little program.

2012-01-26, 14:31   #5

"Kieren"
Jul 2011
In My Own Galaxy!

97·103 Posts

Quote:
 Originally Posted by sonjohan ....... 3. Is it possible to lower the screen output from mfaktc? Right now, it sends a screen output every 5 or 10 iterations...
mfaktc outputs a line for every class completed, if I understand correctly. I don't know of a way to change this. You can, however, keep the output from piling up by setting "PrintMode=1" in mfaktc.ini.

Code:
# possible values for PrintMode:
# 0: print a new line for each finished class
# 1: overwrite the current line (more compact output)
#
# Default: PrintMode=0

 2012-01-26, 15:09 #6 sonjohan     May 2003 Belgium 2·139 Posts Thanks for the information. 1. I'll check it out. 3. I think that's what I'm interested in. Nucleon had shown some screenshots and I noticed only the last class in the output, which made me wonder... 2. To test with trial & error, but if someone knows how far I can go... Again, I understood that Nucleon runs up to 20 instances...
2012-01-26, 15:37   #7
bcp19

Oct 2011

7·97 Posts

Quote:
 Originally Posted by sonjohan Thanks for the information. 1. I'll check it out. 3. I think that's what I'm interested in. Nucleon had shown some screenshots and I noticed only the last class in the output, which made me wonder... 2. To test with trial & error, but if someone knows how far I can go... Again, I understood that Nucleon runs up to 20 instances...
You have to remember that he has more than 1 GPU per machine and is running 4 or 5 machines. A single 570m should have '2' gpu's and you should be able to run 2 instances per gpu. You need to use the -d switch to use the other gpu and probably need to set up affinity switches as well. You also need enough cores to run the instances, so you are probably limited to 4.

 2012-01-26, 18:14 #8 Dubslow Basketry That Evening!     "Bunslow the Bold" Jun 2011 40
2012-01-26, 18:20   #9
sonjohan

May 2003
Belgium

2×139 Posts

Quote:
 Originally Posted by Dubslow A 570M is a dual chip card? I didn't think that was true.
GTX 570M
And when I run GPU-Z (thanks!!) the load with only 1 mfaktc is 50%
With 2 instances it's 99%, so I guess it's correct.

Processing result: no factor for M51xxxxxx from 2^71 to 2^72 [mfaktc 0.17-Win barrett79_mul32]
CPU credit is 9.2449 GHz-days.
Processing result: no factor for M65xxxxxx from 2^71 to 2^72 [mfaktc 0.17-Win barrett79_mul32]
CPU credit is 7.3339 GHz-days.

--> larger factor, less credit???

Last fiddled with by sonjohan on 2012-01-26 at 18:21

2012-01-26, 18:28   #10
chalsall
If I May

"Chris Halsall"
Sep 2002

2×4,643 Posts

Quote:
 Originally Posted by sonjohan --> larger factor, less credit???
Larger candidates take less time to trial factor.

Unlike P-1 factoring and LL work, the work needed to trial factor a candidate to a particular bit level goes down as the exponent goes up.

Therefore (correctly), less credit.

 2012-01-26, 19:02 #11 firejuggler     Apr 2010 Over the rainbow 2·1,217 Posts to expend on what chalsall said, TF exponent are of the form of 2*k*p*+1 (k being any number, P being the exponent you are TF'ing) now take a random prime , lets say 1061, and a trial factoring between 2^30 and 2^31. Every number that can be factorized as 2*1061*k+1 between 2^30 and 2^31 will be tested that's literally thousands of value . If you take another prime, let's say 65xxxxxx, it follow the same principle : it will test every number between 2^30 and 2^31 that can be factorized as 2*65xxxxxx*k+1 There will be a lot less factor possible, therefore less computer power used

 Similar Threads Thread Thread Starter Forum Replies Last Post Antonio PrimeNet 16 2012-11-17 05:15 Flash_ili Software 5 2008-11-22 18:17 petrw1 PrimeNet 8 2008-01-27 01:59 clowns789 Software 7 2003-10-19 21:33 cmokruhl Software 1 2002-10-15 19:04

All times are UTC. The time now is 15:54.

Tue Oct 27 15:54:12 UTC 2020 up 47 days, 13:05, 1 user, load averages: 2.49, 2.81, 2.63