Advanced search

Message boards : Number crunching : New NVIDIA Beta out Windows

Author Message
5pot
Send message
Joined: 8 Mar 12
Posts: 411
Credit: 2,083,882,218
RAC: 0
Level
Phe
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 24351 - Posted: 10 Apr 2012 | 15:43:45 UTC

It appears the new Beta driver is out, and from what I can tell they still have not fixed the Display sleep bug yet, but figured Id still post the news.

http://www.geforce.com/drivers/results/43560

5pot
Send message
Joined: 8 Mar 12
Posts: 411
Credit: 2,083,882,218
RAC: 0
Level
Phe
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 24357 - Posted: 10 Apr 2012 | 19:09:29 UTC - in response to Message 24351.
Last modified: 10 Apr 2012 | 19:15:19 UTC

Maybe I was wrong:

You'll be happy to know that 301.24 fixes the sleeping monitor Bug, althrough i haven't tried it on PrimeGrid yet, only Einstein & Seti so far,

Claggy (Some user on PG)

First post
I did 7 Setiathome offline Benches last night and couldn't get it to fail (But i'm using a different monitor to when i could get to fail with 295.xx drivers),
Before i upgraded i grabbed some BRP4Cuda work and have done some of it this morning no problem,
In a little while i'll downgrade to 295.73 and check i can get offline benches to fail on this monitor.

Second:

I downgraded my i7-2600K/GTX460/HD5770 host to 295.73, ran a setiathome offline bench, proved that the cuda apps do fail with this monitor, then upgraded back up to 301.24,

5pot
Send message
Joined: 8 Mar 12
Posts: 411
Credit: 2,083,882,218
RAC: 0
Level
Phe
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 24362 - Posted: 10 Apr 2012 | 20:47:23 UTC

Sorry to keep posting in my own thread, but using 680 on Einstein app which had problems w/ monitor sleep, it APPEARS that all is well, again though I am using 301.10 and not the 301.24 beta (designed for non 600 series) the 301.10 was only released for 680

ftpd
Send message
Joined: 6 Jun 08
Posts: 152
Credit: 328,250,382
RAC: 0
Level
Asp
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 24376 - Posted: 11 Apr 2012 | 9:57:00 UTC
Last modified: 11 Apr 2012 | 10:20:34 UTC

The sleeping bug is still there with driver 301.24 for xp-pro with gtx480.

I have stopped crunching gpugrid. Sorry!
____________
Ton (ftpd) Netherlands

Profile Retvari Zoltan
Avatar
Send message
Joined: 20 Jan 09
Posts: 2343
Credit: 16,225,565,968
RAC: 1,955,187
Level
Trp
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 24377 - Posted: 11 Apr 2012 | 11:43:20 UTC - in response to Message 24376.
Last modified: 11 Apr 2012 | 11:45:56 UTC

The sleeping bug is still there with driver 301.24 for xp-pro with gtx480.

I have stopped crunching gpugrid. Sorry!

The other projects are also affected by this problem...
I suggest to you (all) to disable the monitor sleep, and manually turn off your monitor until this issue is fixed by nVidia. I do it regardless of what driver I'm using. There's nothing the GPUGrid (or any other BOINC project) staff can do to resolve this issue, it's clearly nVidia's fault.

5pot
Send message
Joined: 8 Mar 12
Posts: 411
Credit: 2,083,882,218
RAC: 0
Level
Phe
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 24378 - Posted: 11 Apr 2012 | 12:32:41 UTC
Last modified: 11 Apr 2012 | 13:12:52 UTC

Very interesting. Reports on Einstein site and my.own exp. are different. Hmm?

EDIT:

From another user at Einstein,

My testing of the 301.24 drivers is successful. The overnight run resulted in several completed Cuda (Binary Radio Pulsar Search (Arecibo) v1.22 (BRP4cuda32nv301)) tasks... and one has already been successfully through the validator. This was on a 9800GT. I have sent him a personal message, as he said he was switching to Foldin@Home. I plan on posting this to NVIDIA forums, and I have postd the failed to Einstein's forums as well.

EDIT AGAIN: Yes, switching projects will do nothing for this, this is an NVIDIA issue, but I'm trying to find out if it just affected you b/c you have XP-pro. Mine was on W7, and I'm working on gathering others graphics cards and OS as well.

EDIT ONE MORE TIME: It appears the individual who was running on the 9800 was also on W7. Currently posting to NVIDIA forums.

ftpd
Send message
Joined: 6 Jun 08
Posts: 152
Credit: 328,250,382
RAC: 0
Level
Asp
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 24379 - Posted: 11 Apr 2012 | 13:31:41 UTC
Last modified: 11 Apr 2012 | 13:32:36 UTC

I always turn off all monitors manually. It won't help! Same problem with gtx295 and gts250 also with other drivers.

I just wait for nvidia for solving this problem. So no more cuda-crunching for the moment.

My RAC went down from 500,000 until now 250,000!
____________
Ton (ftpd) Netherlands

Richard Haselgrove
Send message
Joined: 11 Jul 09
Posts: 1620
Credit: 8,877,067,859
RAC: 19,901,189
Level
Tyr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 24381 - Posted: 11 Apr 2012 | 13:37:39 UTC - in response to Message 24379.

I always turn off all monitors manually. It won't help! Same problem with gtx295 and gts250 also with other drivers.

I just wait for nvidia for solving this problem. So no more cuda-crunching for the moment.

My RAC went down from 500,000 until now 250,000!

If you turn the monitors off manually anyway, why don't you disable the operating system monitor sleep mode? It's the OS which is triggering the NVidia bug - and you don't need that OS feature.

5pot
Send message
Joined: 8 Mar 12
Posts: 411
Credit: 2,083,882,218
RAC: 0
Level
Phe
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 24382 - Posted: 11 Apr 2012 | 13:39:23 UTC
Last modified: 11 Apr 2012 | 14:07:58 UTC

Wait. There's a difference between turning off monitors and switching the setting off. As I've posted this MAY be an xp thing. In XP (haven't used it for awhile), do you have your OS to Never allow your computer to put monitor to sleep. Getting ready to post right now on NVIDIA, and would like to know, since others on W7 are not having this issue. I just want to be sure that I post the correct information. Thanks

EDIT: Can someone take a look at his output file. I don't know anything about that stuff, but that is one crazy looking file. Again, trying to find out if it's just on XP. I see a no hearbeat error, which I thought was from an oversaturated CPU. Also when I copied his Boinc debugger code like excemption into google, it brought back lack of memory. Again, using google and other projects as reference. I thought sleep bug brought back "Couldn't Initialize CUDA driver"

5pot
Send message
Joined: 8 Mar 12
Posts: 411
Credit: 2,083,882,218
RAC: 0
Level
Phe
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 24383 - Posted: 11 Apr 2012 | 14:32:53 UTC
Last modified: 11 Apr 2012 | 14:33:27 UTC

After doing some research, I am fairly convinced that these are not due to the monitor resume issue. If you are running other projects on your xeons, than you will have to free up a core for gpu, or at least a thread. These errors, again from my limited background and research, do NOT seem to be caused by the monitor resume issue. Its either a memory problem, CPU overload, or OC/voltage issue. Again, this is from my LIMITED background on the subject. Seeing as how other cards are failing with 270 drivers, and all are on xeons, im betting youre doing something else on the servers, and the GPUs aren't getting the required resources.

ftpd
Send message
Joined: 6 Jun 08
Posts: 152
Credit: 328,250,382
RAC: 0
Level
Asp
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 24403 - Posted: 12 Apr 2012 | 13:44:34 UTC - in response to Message 24381.

@Richard,

I have already changed the OS-feature in XP-pro that the monitor is NEVER off, also the disk etc.
Is the problem perhaps PhysX? Automatic choice = monitor!

I also installed the swan-sync option, so i use 7 cores instead of 8.

I am back now to 275.33 and the same problem now.
____________
Ton (ftpd) Netherlands

5pot
Send message
Joined: 8 Mar 12
Posts: 411
Credit: 2,083,882,218
RAC: 0
Level
Phe
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 24404 - Posted: 12 Apr 2012 | 14:28:03 UTC

As I said, a monitor resume fail is from,"CUDA failing to load", your's is no hearbeat. This is a CPU issue. It is possible that you would need swan_sync disabled, since yours only runs at 2.5 GHz, you are trying to "force" your CPU to work as much as GPU (time wise), and 2.5 is rather slow.

Have you also told BOINC to use 87.5% if processors? If not, and if you are running other things on the CPUs, SWAN_SYNC is trying to use 100% of one that might not be available (meaning you would be running 9 tasks instead of 8)

ftpd
Send message
Joined: 6 Jun 08
Posts: 152
Credit: 328,250,382
RAC: 0
Level
Asp
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 24406 - Posted: 12 Apr 2012 | 14:45:29 UTC - in response to Message 24404.

@5pot,

Indeed i have told Boic-manager to use 98% of its processors. It uses, confirmed by Boinc-manager 7 instead of 8 processors. I will reduce it now to 87,5%. Thx for the reply!
____________
Ton (ftpd) Netherlands

5pot
Send message
Joined: 8 Mar 12
Posts: 411
Credit: 2,083,882,218
RAC: 0
Level
Phe
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 24407 - Posted: 12 Apr 2012 | 15:10:34 UTC
Last modified: 12 Apr 2012 | 15:17:10 UTC

So just to verify, you should have 8 tasks running (7 CPU + 1 GPU)? And like I said, be careful w/ swan_sync, as your CPU only run at 2.5GHz, you CPU may be bottlenecking your GPU, since it's trying to keep pace. That no heartbeat, means that the GPU is waiting for the CPU (CPU is currently being overused), so it reports no heartbeat.

ftpd
Send message
Joined: 6 Jun 08
Posts: 152
Credit: 328,250,382
RAC: 0
Level
Asp
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 24425 - Posted: 13 Apr 2012 | 10:52:17 UTC - in response to Message 24407.

@5pot,
Correct. I also put off screen security after ... minutes and now it is OK. Energy saving was already off with no success!
I am now using driver 285.58

Good weekend.
____________
Ton (ftpd) Netherlands

Grutte Pier [Wa Oars]~MAB...
Send message
Joined: 8 Jan 12
Posts: 20
Credit: 5,132,859
RAC: 0
Level
Ser
Scientific publications
watwatwatwatwat
Message 24441 - Posted: 15 Apr 2012 | 9:51:54 UTC

Anybody experiencing any improvement regarding speed/credits with 301.24 on XP ?

Post to thread

Message boards : Number crunching : New NVIDIA Beta out Windows

//