Nvidia driver revert?

Message boards : Number crunching : Nvidia driver revert?
Message board moderation

To post messages, you must log in.

AuthorMessage
Profile Miep
Volunteer moderator
Avatar

Send message
Joined: 23 Jul 99
Posts: 2412
Credit: 351,996
RAC: 0
Message 1015804 - Posted: 15 Jul 2010, 10:43:10 UTC

I want to go back to the driver before Nvidia started to show less memory.
Does somebody know for sure on which driver the change was introduced? My logs don't go back far enough.
I've identified 195.62 as the most likely candidate, as with 197.16 it already comes up as 242 MB, but would feel better with independent verification.
Thanks a lot.
Carola
-------
I'm multilingual - I can misunderstand people in several languages!
ID: 1015804 · Report as offensive
Profile Wiggo
Avatar

Send message
Joined: 24 Jan 00
Posts: 34956
Credit: 261,360,520
RAC: 489
Australia
Message 1015807 - Posted: 15 Jul 2010, 10:58:35 UTC - in response to Message 1015804.  

195.62 is what I'm using and it does show all video memory.
ID: 1015807 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14654
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1015808 - Posted: 15 Jul 2010, 11:05:31 UTC - in response to Message 1015804.  

I want to go back to the driver before Nvidia started to show less memory.
Does somebody know for sure on which driver the change was introduced? My logs don't go back far enough.
I've identified 195.62 as the most likely candidate, as with 197.16 it already comes up as 242 MB, but would feel better with independent verification.
Thanks a lot.

I've not done a lot of reverting, but I think you've got the right bracket.

I was about to suggest 190.38 (the first to support CUDA 2.3 DLLs) - you don't want to go back any earlier than that, whatever you do. But your Quadro FX 570M with Vista 32 imposes limitations - they didn't release 190 for that system. 195.62 seems to be the best choice:

http://www.nvidia.com/object/notebook_winvista_win7_195.62_whql.html
ID: 1015808 · Report as offensive
Claggy
Volunteer tester

Send message
Joined: 5 Jul 99
Posts: 4654
Credit: 47,537,079
RAC: 4
United Kingdom
Message 1015838 - Posted: 15 Jul 2010, 13:27:13 UTC - in response to Message 1015804.  
Last modified: 15 Jul 2010, 13:33:07 UTC

197.13 is the first driver that reported less memory, i asked why on Nvidia's developers Cuda forum, but got no answer from any developer, :-(

Cuda-Z and Boinc report less Memory with 197.xx drivers

Claggy
ID: 1015838 · Report as offensive
Profile Miep
Volunteer moderator
Avatar

Send message
Joined: 23 Jul 99
Posts: 2412
Credit: 351,996
RAC: 0
Message 1015845 - Posted: 15 Jul 2010, 13:39:39 UTC

Thank you all. :)

Have gone back to 195.62 - one GPU back into the fold.

Small hiccup - windows panicked and put in the original drivers, which caused the Nvidia setup to hang. Quickly solved, just rerun the setup.

Now to running down cache to put a suitable application into the app_info file.
Carola
-------
I'm multilingual - I can misunderstand people in several languages!
ID: 1015845 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14654
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1015853 - Posted: 15 Jul 2010, 14:22:07 UTC - in response to Message 1015845.  

Thank you all. :)

Have gone back to 195.62 - one GPU back into the fold.

Small hiccup - windows panicked and put in the original drivers, which caused the Nvidia setup to hang. Quickly solved, just rerun the setup.

Now to running down cache to put a suitable application into the app_info file.

Running down cache is purely precautionary - you shouldn't lose a thing, especially if you use the latest Lunatics Unified Installer as a starting point- Jason has been careful to put as much compatibility as possible into the app_info elements: in any event, most of the cache losses were on 64-bit systems, 32-bit is simpler.

Having said that, performing major surgery at the time when WU transfusions are readily available would be a wise precaution. And don't operate powerful editing machinery when tired ;-)
ID: 1015853 · Report as offensive
Crun-chi
Volunteer tester
Avatar

Send message
Joined: 3 Apr 99
Posts: 174
Credit: 3,037,232
RAC: 0
Croatia
Message 1015864 - Posted: 15 Jul 2010, 15:03:39 UTC - in response to Message 1015853.  

190.38 and 191.07 are fastest driver for crunching.
I am cruncher :)
I LOVE SETI BOINC :)
ID: 1015864 · Report as offensive
sarmitage

Send message
Joined: 2 Dec 09
Posts: 56
Credit: 1,123,857
RAC: 0
Canada
Message 1015882 - Posted: 15 Jul 2010, 15:50:00 UTC - in response to Message 1015853.  

Running down cache is purely precautionary - you shouldn't lose a thing, especially if you use the latest Lunatics Unified Installer as a starting point- Jason has been careful to put as much compatibility as possible into the app_info elements: in any event, most of the cache losses were on 64-bit systems, 32-bit is simpler.

Having said that, performing major surgery at the time when WU transfusions are readily available would be a wise precaution. And don't operate powerful editing machinery when tired ;-)


Does the Lunatics installer modify anything outside of the data folder (on my machine, C:\Users\sarmitage\AppData\Local\BOINC\)? If not, couldn't one just stop the client and manager, make a copy of that folder, and then apply the changes? That way, if anything goes wrong, one could just restore the backup and not have lost anything.

-S
ID: 1015882 · Report as offensive
Cosmic_Ocean
Avatar

Send message
Joined: 23 Dec 00
Posts: 3027
Credit: 13,516,867
RAC: 13
United States
Message 1015923 - Posted: 15 Jul 2010, 17:54:03 UTC

One of the safest things you can do without having to run down the cache is set "no new tasks" turn network communication off, then close BOINC, make a backup of the data directory, then do the editing. When you bring BOINC back up, if everything goes haywire, close BOINC, restore the backup, and try again.

Once everything looks fine, allow network and new tasks.
Linux laptop:
record uptime: 1511d 20h 19m (ended due to the power brick giving-up)
ID: 1015923 · Report as offensive
sarmitage

Send message
Joined: 2 Dec 09
Posts: 56
Credit: 1,123,857
RAC: 0
Canada
Message 1015925 - Posted: 15 Jul 2010, 17:58:11 UTC - in response to Message 1015923.  

One of the safest things you can do without having to run down the cache is set "no new tasks" turn network communication off, then close BOINC, make a backup of the data directory, then do the editing. When you bring BOINC back up, if everything goes haywire, close BOINC, restore the backup, and try again.

Once everything looks fine, allow network and new tasks.


OK cool, that was my understanding as well; guess it's about time I try the opp apps for CPU =)
ID: 1015925 · Report as offensive
Profile Zeus Fab3r
Avatar

Send message
Joined: 17 Jan 01
Posts: 649
Credit: 275,335,635
RAC: 597
Serbia
Message 1015934 - Posted: 15 Jul 2010, 18:53:59 UTC - in response to Message 1015925.  

OK cool, that was my understanding as well; guess it's about time I try the opp apps for CPU =)


With such hardware, it would be a shame not to try. :D

Who the hell is General Failure and why is he reading my harddisk?¿
ID: 1015934 · Report as offensive

Message boards : Number crunching : Nvidia driver revert?


 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.