Running on /--Prime95--/ Lucas-Lehmer on Nvidia Tesla S1070

optimus · 2013-08-30, 21:05

Hello,

I would like to find out if I can run Prime95 on Nvidia Tesla S1070 system. It has 4 GPUs with 240 cores each, for a total of 960 cores.

http://www.nvidia.com/docs/IO/43395/...V_LR_Final.pdf

If there is someone who has experience using Prime95 with this or similar system, I would appreciate any information you could share regarding system configuration and running the software.

kracker · 2013-08-30, 21:36

Yes, LL testing with CUDALucas.

You will however, have to give it work manually.

optimus · 2013-08-30, 22:25

That's great. Will I be able to use all 960 cores to run 960 LL tests simultaneously? And also, do you think the average completion is going to be around 70 days, like it is on my i7 iMac 2.8GHz? Thanks again.

kladner · 2013-08-31, 00:49

In general, each GPU takes on one exponent at a time. In your situation you should be able to run 4 CUDALucas instances at a time. I am not certain of the performance of your GPUs, but for the sake of comparison:

A GTX 460 OC'd at 830 MHz core will do a 3xM double check LL in about 48 hours. The chip has 336 CUDA cores, CC 2.1.

A GTX 570 at about the same speed can do the same assignment in roughly 24 hours. The chip has 480 cores, CC 2.0.

frmky · 2013-08-31, 03:01

The S1070 has roughly the same throughput as a GTX 570. So expect a 3xM double check LL to take about 4 days or so, and you can run 4 of them in parallel.

2013-08-30, 21:05	#1
optimus "Derek" Aug 2013 Los Angeles 2 Posts	Running on /--Prime95--/ Lucas-Lehmer on Nvidia Tesla S1070 Hello, I would like to find out if I can run Prime95 on Nvidia Tesla S1070 system. It has 4 GPUs with 240 cores each, for a total of 960 cores. http://www.nvidia.com/docs/IO/43395/...V_LR_Final.pdf If there is someone who has experience using Prime95 with this or similar system, I would appreciate any information you could share regarding system configuration and running the software.

2013-08-31, 00:49	#4
kladner "Kieren" Jul 2011 In My Own Galaxy! 2·3·1,693 Posts	In general, each GPU takes on one exponent at a time. In your situation you should be able to run 4 CUDALucas instances at a time. I am not certain of the performance of your GPUs, but for the sake of comparison: A GTX 460 OC'd at 830 MHz core will do a 3xM double check LL in about 48 hours. The chip has 336 CUDA cores, CC 2.1. A GTX 570 at about the same speed can do the same assignment in roughly 24 hours. The chip has 480 cores, CC 2.0.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Lucas-Lehmer test	Mathsgirl	Information & Answers	23	2014-12-10 16:25
lucas-lehmer theorem	Robot2357	Math	6	2013-06-15 03:10
lucas lehmer outstretch	science_man_88	Miscellaneous Math	7	2010-07-14 12:35
Lucas-Lehmer Test	storm5510	Math	22	2009-09-24 22:32
Lucas-Lehmer	Dougal	Information & Answers	9	2009-02-06 10:25

2013-08-30, 21:36	#2
kracker "Mr. Meeseeks" Jan 2012 California, USA 2⁷·17 Posts	Yes, LL testing with CUDALucas. You will however, have to give it work manually.

2013-08-30, 22:25	#3
optimus "Derek" Aug 2013 Los Angeles 2₈ Posts	That's great. Will I be able to use all 960 cores to run 960 LL tests simultaneously? And also, do you think the average completion is going to be around 70 days, like it is on my i7 iMac 2.8GHz? Thanks again.

2013-08-31, 03:01	#5
frmky Jul 2003 So Cal 19×137 Posts	The S1070 has roughly the same throughput as a GTX 570. So expect a 3xM double check LL to take about 4 days or so, and you can run 4 of them in parallel.