mfakto: an OpenCL program for Mersenne prefactoring - Page 96

tapion64 · 2014-05-09, 07:05

Testing the latest version with R9 295X2. I tried varying a bunch of different settings, and the best I get is around 575 GHz-days/day. Not sure if I need to look at tweaking some AMD settings or if that's just the limit it can achieve. Definitely not hitting that predicted 1229 over in mersenne.ca though.

axn · 2014-05-09, 07:22

Quote:

Originally Posted by tapion64

Testing the latest version with R9 295X2. I tried varying a bunch of different settings, and the best I get is around 575 GHz-days/day. Not sure if I need to look at tweaking some AMD settings or if that's just the limit it can achieve. Definitely not hitting that predicted 1229 over in mersenne.ca though.

Have you tried running two instances?

Stef42 · 2014-05-09, 07:26

Quote:

Originally Posted by tapion64

Testing the latest version with R9 295X2. I tried varying a bunch of different settings, and the best I get is around 575 GHz-days/day. Not sure if I need to look at tweaking some AMD settings or if that's just the limit it can achieve. Definitely not hitting that predicted 1229 over in mersenne.ca though.

Did you use the -d switches? For example: the first GPU on the 295X should be started with a .bat file > mfakto.x64 -d 1, the second: > mfakto.x64 -d 2

tapion64 · 2014-05-09, 12:33

Ah, that was it. I didn't realize it would treat them as cross-fired devices instead of a single card. Making it separate instances works fine. Is there a way to have it draw work from one worktodo.txt for both, or do they just have to be in different directories?

kladner · 2014-05-09, 12:43

Quote:

Originally Posted by tapion64

Is there a way to have it draw work from one worktodo.txt for both, or do they just have to be in different directories?

MISFIT will take care of the two worktodo.txt files for you, and feed them from a common staging file. AFAIK, mfaktX of all flavors requires separate directories and working files for multiple cards.

kracker · 2014-05-09, 14:54

Quote:

Originally Posted by tapion64

Testing the latest version with R9 295X2. I tried varying a bunch of different settings, and the best I get is around 575 GHz-days/day. Not sure if I need to look at tweaking some AMD settings or if that's just the limit it can achieve. Definitely not hitting that predicted 1229 over in mersenne.ca though.

Nice, first look at the Hawaii core!
Well, with a bit of fiddling you should be able to get 600 GHz per core.

tapion64 · 2014-05-09, 20:30

I guess I need to look into setting up MISFIT. The second core of the card is getting ~600, so I think just the cost of running graphics for my new monitor lowers the first one by about 50. Not that bad though, ~1175 combined.

kladner · 2014-05-10, 02:35

Quote:

Originally Posted by tapion64

I guess I need to look into setting up MISFIT. The second core of the card is getting ~600, so I think just the cost of running graphics for my new monitor lowers the first one by about 50. Not that bad though, ~1175 combined.

I've used MISFIT for a long time. You can have it do as much, or as little as you like. You can make it a "set it and forget it" controller, which can handle much more than two instances, fetching, distributing, and submitting the results of processing. You can also use it as a convenient monitoring utility, and a flexible manual control interface.

kracker · 2014-05-24, 00:03

I ran the oldest version of mfakto I could just for fun

0.10: 14m36s
0.14: 7m59s

tului · 2014-06-24, 04:55

Quote:

Originally Posted by kracker

I ran the oldest version of mfakto I could just for fun

0.10: 14m36s
0.14: 7m59s

Any newer or beta mfakto we could test on our GCN cards? I'd love to squeeze some more performace out of them. They're not 79xx renames, they're the real deal new chips.

LaurV · 2014-06-24, 05:34

If you want to squeeze more performance from mfakto, try to factor lower than 73 bits only. Contrary to mfaktc, where there is no big drop in performance for higher bitlevels (or, say, no big gain in performance for lower bit levels), for mfakto, especial for higher GCN cards, the "shorter" kernels are much faster. For example, I get from my HD7970 GHz edition, something like: 450GHzD/D when factoring 6xM to 74, but I get 500GHzD/D when factoring to 73 only, and so on. Decreasing the bitlevel increase the "gain" (but helps GIMPS less) and also decreasing the exponent increase the "gain", but only a little. For example, the same card I described above, gives 630-650GHzD/D when factoring 4xM exponents to 69 bits. Right now, Chris made them unavailable from GPU72, to channel the workers toward Cat4 exponents, but there are still 35 thousands of them (44-47M) at 68 bits, you can take them to 69 directly from PrimeNet, or ask Chris to make them available. For this range of expos and bitlevel, the performance of the card (kernel) is about 50% higher.

2014-06-24, 05:34	#1056
LaurV Romulan Interpreter Jun 2011 Thailand 25B6₁₆ Posts	If you want to squeeze more performance from mfakto, try to factor lower than 73 bits only. Contrary to mfaktc, where there is no big drop in performance for higher bitlevels (or, say, no big gain in performance for lower bit levels), for mfakto, especial for higher GCN cards, the "shorter" kernels are much faster. For example, I get from my HD7970 GHz edition, something like: 450GHzD/D when factoring 6xM to 74, but I get 500GHzD/D when factoring to 73 only, and so on. Decreasing the bitlevel increase the "gain" (but helps GIMPS less) and also decreasing the exponent increase the "gain", but only a little. For example, the same card I described above, gives 630-650GHzD/D when factoring 4xM exponents to 69 bits. Right now, Chris made them unavailable from GPU72, to channel the workers toward Cat4 exponents, but there are still 35 thousands of them (44-47M) at 68 bits, you can take them to 69 directly from PrimeNet, or ask Chris to make them available. For this range of expos and bitlevel, the performance of the card (kernel) is about 50% higher. Last fiddled with by LaurV on 2014-06-24 at 05:36 Reason: /s/bit/little/ it was confusing

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
gpuOwL: an OpenCL program for Mersenne primality testing	preda	GpuOwl	2718	2021-07-06 18:30
mfaktc: a CUDA program for Mersenne prefactoring	TheJudger	GPU Computing	3497	2021-06-05 12:27
LL with OpenCL	msft	GPU Computing	433	2019-06-23 21:11
OpenCL for FPGAs	TObject	GPU Computing	2	2013-10-12 21:09
Program to TF Mersenne numbers with more than 1 sextillion digits?	Stargate38	Factoring	24	2011-11-03 00:34

2014-05-09, 07:05	#1046
tapion64 Apr 2014 125₈ Posts	Testing the latest version with R9 295X2. I tried varying a bunch of different settings, and the best I get is around 575 GHz-days/day. Not sure if I need to look at tweaking some AMD settings or if that's just the limit it can achieve. Definitely not hitting that predicted 1229 over in mersenne.ca though.

2014-05-09, 12:33	#1049
tapion64 Apr 2014 125₈ Posts	Ah, that was it. I didn't realize it would treat them as cross-fired devices instead of a single card. Making it separate instances works fine. Is there a way to have it draw work from one worktodo.txt for both, or do they just have to be in different directories?

2014-05-09, 20:30	#1052
tapion64 Apr 2014 5·17 Posts	I guess I need to look into setting up MISFIT. The second core of the card is getting ~600, so I think just the cost of running graphics for my new monitor lowers the first one by about 50. Not that bad though, ~1175 combined.

2014-05-24, 00:03	#1054
kracker "Mr. Meeseeks" Jan 2012 California, USA 2³×271 Posts	I ran the oldest version of mfakto I could just for fun 0.10: 14m36s 0.14: 7m59s