Bigger and better GPU sieving drive: Discussion - Page 6

mdettweiler · 2010-10-09, 11:02

Quote:

Originally Posted by vaughan

Did you download and install and run ntune? You can sometimes (see below) get it to control you GPU fan speeds so you can lower the temps when crunching 100%

Alternatively you can download and install and run Rivatuner. It triggers unsigned driver messages from Win 7 but will also control the fans and therefore the temps.

Google is your friend to find these apps.

On several of my machines, all Win 7 or XP 64-bit so far, ntune triggers a BSOD and forces a reboot. Win 7 says the system recovered from a serious error (yeah lousy programming by nvidia). On my 32 bit Windows machines ntune works as it should.

Both Rivatuner and ntune bring the GPU temps down at least 10C so are very useful utilities.

I didn't actually have the chance...by the time you first suggested it the computer was already toast.

From some research I did last night, it seems that the motherboard wasn't really supposed to be used with a Q6600, according to the manufacturer's website. (That'll teach me to trust Newegg's "Supports Core 2 Quad" label without further checking in the future...) The C2Qs that it did support all had lower voltage than the Q6600 (which seems to be an outlier in this regard). It would appear that it can take the extra stress OK as long as you don't stick something big and power-hungry like a GPU on it.

This also makes me a little worried about my own quad, which is nearly identical to Gary's 4 Q6600s (I having picked out the parts for all of them around the same time). Hopefully I'll be fine as long as I don't ever stick a GPU in it...

Meanwhile, though, thanks for the tip on ntune and Rivatuner--I'll give them a try when we get the machine working again, since even if the GPU temp wasn't the source of this particular problem it would still be good to have it running as cool as it can (so as not to unnecessarily trigger thermal throttling in the CPU). Quick question: do either ntune or Rivatuner work in Linux?

vaughan · 2010-10-09, 13:21

I'm not sure Max. I asked my son (Linux guru) and he cannot think of anything similar for Linux off the top of his head.

henryzz · 2010-10-09, 16:42

Quote:

Originally Posted by mdettweiler

The C2Qs that it did support all had lower voltage than the Q6600 (which seems to be an outlier in this regard).

The Q6600 was one of the first quads and was 65nm unlike most which were 45nm. This kept the price down so most people bought them.

Ken_g6 · 2010-10-14, 19:39

Since it hasn't been mentioned yet, PPSieve-CUDA V0.2.2 is out. It should be twice as fast on non-Fermi GPUs as 0.2.1a! (And maybe 50% faster on Fermi.)

It should also handle multiple different GPUs better; but I can't test that at all. Frmky, would you care to give it a go?

Edit: P.S. A new version of the OpenCL client will be coming too, eventually.

frmky · 2010-10-14, 19:43

Quote:

Originally Posted by Ken_g6

It should also handle multiple different GPUs better; but I can't test that at all. Frmky, would you care to give it a go?

The only multi-GPU system I have is the one with 4xC1060, and the old client worked there as well. I'm using 0.2.2-alpha on the current range, and it works well.

gd_barnes · 2010-10-16, 22:07

Max,

Am I remembering correctly that PST/PrimeGrid was going to do something special at P=50T? Perhaps merge it with the Proth side or something like that? I just want to make sure before we do too much P>50T.

Gary

mdettweiler · 2010-10-16, 23:00

Quote:

Originally Posted by gd_barnes

Max,

Am I remembering correctly that PST/PrimeGrid was going to do something special at P=50T? Perhaps merge it with the Proth side or something like that? I just want to make sure before we do too much P>50T.

Gary

Last I heard they're planning to move their half into BOINC at 50T (er, 60T since we're doing 50T-60T). That shouldn't affect our half.

The Proth side is currently at 1405T (more precisely: part of it is at 1405T and suspended, the rest is catching up to that). When the Riesel side reaches that level, the two will be merged and we'll continue from there with tpsieve. (The latest versions of tpsieve support sieving non-twin searches on both sides; it's mostly similar to ppsieve under the hood, so it was actually easier to build this functionality into tpsieve instead of ppsieve.)

mdettweiler · 2010-10-16, 23:16

Quote:

Originally Posted by mdettweiler

Last I heard they're planning to move their half into BOINC at 50T (er, 60T since we're doing 50T-60T). That shouldn't affect our half.

The Proth side is currently at 1405T (more precisely: part of it is at 1405T and suspended, the rest is catching up to that). When the Riesel side reaches that level, the two will be merged and we'll continue from there with tpsieve. (The latest versions of tpsieve support sieving non-twin searches on both sides; it's mostly similar to ppsieve under the hood, so it was actually easier to build this functionality into tpsieve instead of ppsieve.)

Further clarification per a note I just read from John over at the PST forum: they're not quite going to be moving their half into BOINC at 60T. What they're going to do is run it through BOINC under their existing "PPS Sieve" subproject (since the application is the same); the Proth side is fast approaching 1405T, so they'll move their half of the Riesel side into BOINC as soon as that occurs. I'm not sure at what point that will occur but it will likely be above 75T on the Riesel side.

mdettweiler · 2010-10-16, 23:32

Quote:

Originally Posted by mdettweiler

Further clarification per a note I just read from John over at the PST forum: they're not quite going to be moving their half into BOINC at 60T. What they're going to do is run it through BOINC under their existing "PPS Sieve" subproject (since the application is the same); the Proth side is fast approaching 1405T, so they'll move their half of the Riesel side into BOINC as soon as that occurs. I'm not sure at what point that will occur but it will likely be above 75T on the Riesel side.

Oops, hate to contradict myself twice in one night but according to an email Lennart just sent me, PrimeGrid is processing quickly enough on their PPS Sieve subproject that they'll be up to 1405T within 3-4 days. So it is likely after all that PST's half of this sieve will enter BOINC at around 60T.

Lennart · 2010-10-16, 23:41

Quote:

Originally Posted by mdettweiler

Oops, hate to contradict myself twice in one night but according to an email Lennart just sent me, PrimeGrid is processing quickly enough on their PPS Sieve subproject that they'll be up to 1405T within 3-4 days. So it is likely after all that PST's half of this sieve will enter BOINC at around 60T.

3-4 days before all work is out

We need to clean it up and that may take 4-7 days so we have time to make it up to 70T-80T deep.

Lennart

Ken_g6 · 2010-10-29, 22:00

FYI, for those waiting for an ATI/OpenCL app that will work on the new range...you'll have to keep waiting awhile.

The good news is that yesterday I solved almost all of the problems, I think. I haven't tried building a TPSieve version, but that shouldn't be awfully hard.

Except that once I finished off the main debugging, I noticed that an nVIDIA card had arrived at my door. I dropped everything (shouldn't have done that), tried to install it, and scratched my motherboard so badly that I had to order a new one.

Soo...maybe I'll have something late next week.

2010-10-14, 19:39	#59
Ken_g6 Jan 2005 Caught in a sieve 18B₁₆ Posts	Since it hasn't been mentioned yet, PPSieve-CUDA V0.2.2 is out. It should be twice as fast on non-Fermi GPUs as 0.2.1a! (And maybe 50% faster on Fermi.) It should also handle multiple different GPUs better; but I can't test that at all. Frmky, would you care to give it a go? Edit: P.S. A new version of the OpenCL client will be coming too, eventually. Last fiddled with by Ken_g6 on 2010-10-14 at 19:41

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
New PRPnet drive discussion	mdettweiler	Conjectures 'R Us	89	2011-08-10 09:01
Sieving drive Riesel base 6 n=1M-2M	gd_barnes	Conjectures 'R Us	40	2011-01-22 08:10
Bigger and better GPU sieving drive: k<10000 n<2M	mdettweiler	No Prime Left Behind	61	2010-10-29 18:48
GPU sieving drive for k<=1001 n=1M-2M	mdettweiler	No Prime Left Behind	11	2010-10-04 22:45
Sieving drive for k=301-400 n=1M-2M	MyDogBuster	No Prime Left Behind	42	2010-03-21 01:14

2010-10-09, 13:21	#57
vaughan Jan 2005 Sydney, Australia 5·67 Posts	I'm not sure Max. I asked my son (Linux guru) and he cannot think of anything similar for Linux off the top of his head.

2010-10-16, 22:07	#61
gd_barnes May 2007 Kansas; USA 3³×5×7×11 Posts	Max, Am I remembering correctly that PST/PrimeGrid was going to do something special at P=50T? Perhaps merge it with the Proth side or something like that? I just want to make sure before we do too much P>50T. Gary

2010-10-29, 22:00	#66
Ken_g6 Jan 2005 Caught in a sieve 5×79 Posts	FYI, for those waiting for an ATI/OpenCL app that will work on the new range...you'll have to keep waiting awhile. The good news is that yesterday I solved almost all of the problems, I think. I haven't tried building a TPSieve version, but that shouldn't be awfully hard. Except that once I finished off the main debugging, I noticed that an nVIDIA card had arrived at my door. I dropped everything (shouldn't have done that), tried to install it, and scratched my motherboard so badly that I had to order a new one. Soo...maybe I'll have something late next week.