Piddling Around With Big Compute Power

NBtarheel_33 · 2013-04-25, 06:52

So, I've been granted some trial time on a cluster. I'm still learning just what is available, but I know that I have the login system at my disposal to start (it has 8 x Xeon E5-2670 2.6GHz CPUs and 16GB RAM). I then apparently have the ability to call on other nodes in the cluster (to what extent, I am still trying to work out), each of which has 2 of the aforementioned Xeon CPUs, 2 NVIDIA Tesla K20 or K10 (choice seems to be mine, depending on what is available) GPUs, and 32GB of RAM.

Anyway, I have been playing around with CUDALucas (wow!), as well as doing some Extreme GIMPS stuff, like P-1 with ungodly amounts of RAM. I thought I'd post interesting results here. Like this one, for instance...I never knew E could be 12...

UID: ANONYMOUS, M100002677 completed P-1, B1=1105000, B2=33702500, E=12, We8: xxxxxxxx, AID: xxx

This took under 24 hours using 4 threads and 30GB of RAM. I have the option of asking for 256GB of RAM...wonder what might happen then?

Let me know if there's something interesting you'd like to try out...

bearnol · 2013-04-25, 06:59

Superfac13 on big mersenneplustwos (3M+), please
J

Uncwilly · 2013-04-25, 12:29

Quote:

Originally Posted by NBtarheel_33

UID: ANONYMOUS, M100002677 completed P-1, B1=1105000, B2=33702500, E=12, We8: xxxxxxxx, AID: xxx

This took under 24 hours using 4 threads and 30GB of RAM. I have the option of asking for 256GB of RAM...wonder what might happen then?

Let me know if there's something interesting you'd like to try out...

If you are doing things like that, could you try out some of the exponents in the 332M range?

Someone will also likely suggest looking at some numbers down on the lower end of the spectrum.

NBtarheel_33 · 2013-04-25, 14:53

Quote:

Originally Posted by Uncwilly

If you are doing things like that, could you try out some of the exponents in the 332M range?

What's your favorite number?

Quote:

Originally Posted by Uncwilly

Someone will also likely suggest looking at some numbers down on the lower end of the spectrum.

I'm trying to kill off a few 800-1000 day old low-50Ms that I have in the queue...

Puzzle-Peter · 2013-04-25, 15:20

Quote:

Originally Posted by NBtarheel_33

Let me know if there's something interesting you'd like to try out...

Does it behave like one big machine, i.e. can software be run parallellized over several nodes? If so, try to prove 2^73845+14717 from the Five Or Bust project using PRIMO on 32 cores. It will still take many weeks though.

kracker · 2013-04-25, 17:57

Quote:

Originally Posted by Uncwilly

If you are doing things like that, could you try out some of the exponents in the 332M range?

Someone will also likely suggest looking at some numbers down on the lower end of the spectrum.

Ask him to do TF on the gpu, not cpu

CPU TF efficiency compared to GPU is terrible.

henryzz · 2013-04-25, 21:54

Quote:

Originally Posted by kracker

Ask him to do TF on the gpu, not cpu

CPU TF efficiency compared to GPU is terrible.

I think he was talking about P-1 actually.

frmky · 2013-04-26, 01:01

CUDAPm1 on the K20 would be helpful.

Aramis Wyler · 2013-04-26, 02:23

You don't run unstable code on big iron, silly! It's like pouring the good wine into plastic cups!

Uncwilly · 2013-04-26, 03:57

Quote:

Originally Posted by henryzz

I think he was talking about P-1 actually.

Yeah that is what I was thinking. That is what he was talking about, big RAM looks interesting.

Quote:

Originally Posted by Aramis Wyler

You don't run unstable code on big iron, silly! It's like pouring the good wine into plastic cups!

Quote:

What's your favorite number?

332212009 is available and hasn't had any P-1.

frmky · 2013-04-26, 06:15

Quote:

Originally Posted by Aramis Wyler

You don't run unstable code on big iron, silly! It's like pouring the good wine into plastic cups!

Alpha isn't unstable. It's just some work is still in progress. In fact, I have it running right now on a K20 here.

2013-04-25, 06:52	#1
NBtarheel_33 "Nathan" Jul 2008 Maryland, USA 2133₈ Posts	Piddling Around With Big Compute Power So, I've been granted some trial time on a cluster. I'm still learning just what is available, but I know that I have the login system at my disposal to start (it has 8 x Xeon E5-2670 2.6GHz CPUs and 16GB RAM). I then apparently have the ability to call on other nodes in the cluster (to what extent, I am still trying to work out), each of which has 2 of the aforementioned Xeon CPUs, 2 NVIDIA Tesla K20 or K10 (choice seems to be mine, depending on what is available) GPUs, and 32GB of RAM. Anyway, I have been playing around with CUDALucas (wow!), as well as doing some Extreme GIMPS stuff, like P-1 with ungodly amounts of RAM. I thought I'd post interesting results here. Like this one, for instance...I never knew E could be 12... UID: ANONYMOUS, M100002677 completed P-1, B1=1105000, B2=33702500, E=12, We8: xxxxxxxx, AID: xxx This took under 24 hours using 4 threads and 30GB of RAM. I have the option of asking for 256GB of RAM...wonder what might happen then? Let me know if there's something interesting you'd like to try out...

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
New GPU Compute System	airsquirrels	GPU Computing	90	2017-12-08 00:13
Analog hardware to compute FFT's...	WraithX	Hardware	1	2012-11-28 13:29
Doubled compute power for a day?	Christenson	PrimeNet	19	2011-10-26 08:29
New Compute Box	Christenson	Hardware	0	2011-01-15 04:44
My throughput does not compute...	petrw1	Hardware	9	2007-08-13 14:38

2013-04-25, 06:59	#2
bearnol Sep 2005 127 Posts	Superfac13 on big mersenneplustwos (3M+), please J

2013-04-26, 01:01	#8
frmky Jul 2003 So Cal 4104₈ Posts	CUDAPm1 on the K20 would be helpful.

2013-04-26, 02:23	#9
Aramis Wyler "Bill Staffen" Jan 2013 Pittsburgh, PA, USA 2³×53 Posts	You don't run unstable code on big iron, silly! It's like pouring the good wine into plastic cups!