@Pre-FERMI nVidia GPU users: Important warning

Message boards : Number crunching : @Pre-FERMI nVidia GPU users: Important warning
Message board moderation

To post messages, you must log in.

Previous · 1 . . . 4 · 5 · 6 · 7 · 8 · 9 · 10 . . . 13 · Next

AuthorMessage
Profile Wiggo "Socialist"
Avatar

Send message
Joined: 24 Jan 00
Posts: 12599
Credit: 169,203,775
RAC: 86,104
Australia
Message 1603356 - Posted: 21 Nov 2014, 0:21:09 UTC - in response to Message 1603349.  

Just a little more clarification, since 341.05 was released as an R340 driver.

R340 drivers, like 340.xx and 341.xx, aren't *yet* working right for all OpenCL apps... until NVIDIA hopefully fixes things for these R340 drivers. I still have hope! :)

I wouldn't hold your breath waiting for Nvidia to fix their latest drivers for old hardware as that old hardware is now obsolete in their eyes.

Cheers.
ID: 1603356 · Report as offensive
Jacob Klein
Volunteer tester

Send message
Joined: 15 Apr 11
Posts: 50
Credit: 3,939,427
RAC: 8,238
United States
Message 1603358 - Posted: 21 Nov 2014, 0:27:36 UTC - in response to Message 1603356.  

They've indicated that they will fix it, and R340 for pre-Fermi is indeed under bugfix support until April 2016.
ID: 1603358 · Report as offensive
Profile Wiggo "Socialist"
Avatar

Send message
Joined: 24 Jan 00
Posts: 12599
Credit: 169,203,775
RAC: 86,104
Australia
Message 1603361 - Posted: 21 Nov 2014, 0:33:09 UTC - in response to Message 1603358.  
Last modified: 21 Nov 2014, 0:57:46 UTC

They've indicated that they will fix it, and R340 for pre-Fermi is indeed under bugfix support until April 2016.

Yes, I've heard that many times before over many years, which is why I posted what I did. ;-)

Cheers.
ID: 1603361 · Report as offensive
Richard HaselgroveProject Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 11516
Credit: 106,129,012
RAC: 70,009
United Kingdom
Message 1603365 - Posted: 21 Nov 2014, 0:43:36 UTC - in response to Message 1603349.  

Just a little more clarification, since 341.05 was released as an R340 driver.

R340 drivers, like 340.xx and 341.xx, aren't *yet* working right for all OpenCL apps... until NVIDIA hopefully fixes things for these R340 drivers. I still have hope! :)

In the case of the (current) Lunatics app, even a fixed driver won't work, because the application itself has been hard-coded to refuse to run under drivers in the suspect range. Once we have a working driver from NVidia (if?), I'll try to persuade Raistmer to re-release an app which merely notches out the proven bad drivers, and re-allows processing on the far side. If it passes testing, then we can refresh the installer yet again.
ID: 1603365 · Report as offensive
marmot
Avatar

Send message
Joined: 15 May 99
Posts: 71
Credit: 961,485
RAC: 965
United States
Message 1604058 - Posted: 22 Nov 2014, 15:23:33 UTC

After following @Wiggo's advise and installing 266.58 on both the 8600gs and fx 1800m machines I was getting CUDA units for some weeks then I noticed a work drop off and the red message to upgrade to newer drivers to make use of the GPU.
(I'm using BOINC 7.2.42 on one machine and 7.4.26 on the other to keep those variables different.)

My memory (which is aging and often fallible) says that that didn't stop getting packets before but I went ahead and upgraded to 320.78 on both machines and CUDA units started flowing again.

They stopped again last week.

1) Is this because SETI is running out of work or is it another shut off of CUDA units to pre-Fermi cards?

2) Can someone confirm that 266.58 is cut off from work now?
ID: 1604058 · Report as offensive
Claggy
Volunteer tester

Send message
Joined: 5 Jul 99
Posts: 4654
Credit: 47,180,843
RAC: 388
United Kingdom
Message 1604061 - Posted: 22 Nov 2014, 15:33:15 UTC - in response to Message 1604058.  

They stopped again last week.

1) Is this because SETI is running out of work or is it another shut off of CUDA units to pre-Fermi cards?

There's very little work being split, at the moment the Server Status page shows all the MBv7 tapes as done.

Claggy
ID: 1604061 · Report as offensive
Jacob Klein
Volunteer tester

Send message
Joined: 15 Apr 11
Posts: 50
Credit: 3,939,427
RAC: 8,238
United States
Message 1609710 - Posted: 6 Dec 2014, 4:47:05 UTC - in response to Message 1601045.  

I have a minor update. Still waiting on a fix, though.

5 November 2014 9:32 pm -- Kevin Kang
Hi Jacob, thanks for the reporting. We have reproduced this issue and have assigned it to the appropriate developer team for reviewing. Thanks!


12 November 2014 11:28 am -- JacobKlein
Hi Kevin / Team: It has been a week without a status update. Could you kindly please consider answering the questions that I asked in the bug details? Thanks, Jacob Klein


14 November 2014 1:21 am -- Kevin Kang
Hi Jacob, Sorry for update on this issue late. As noted in release notes, the R340 drivers will continue to support the Tesla generation of NVIDIA GPUs until April 1, 2016, and the NVIDIA support team will continue to address driver issues for these products in driver branches up to and including Release 340. However, future driver enhancements and optimizations in driver releases after Release 340 will not support these products. Our developer team is working on this issue actively for the future R340 driver release, we'll keep you posted once it has been fixed. Sorry for any inconvenience! Thanks, Kevin


[NEW]
5 December 2014 8:40 pm -- JacobKlein
[CUDA RegDev Program [DevZone]]
Hi Kevin / Team:
I noticed that new R340 drivers (341.21) were release today for my FX 3800M GPU. I excitedly read the release notes, only to find no mention of this issue. I then spent over an hour thoroughly retesting. To my surprise, these 6 failing test cases are still failing in this latest release. I thought you easily reproduced this issue the same day I reported the bug, 4 weeks ago? Can you please give me an update on the timetable for a solution?
Thanks, Jacob Klein
ID: 1609710 · Report as offensive
Profile Wiggo "Socialist"
Avatar

Send message
Joined: 24 Jan 00
Posts: 12599
Credit: 169,203,775
RAC: 86,104
Australia
Message 1609716 - Posted: 6 Dec 2014, 5:06:10 UTC

I did warn you not to hold your breath on a fix, but is there any good reason why you can't just use a pre-340.xx driver?

Cheers.
ID: 1609716 · Report as offensive
Jacob Klein
Volunteer tester

Send message
Joined: 15 Apr 11
Posts: 50
Credit: 3,939,427
RAC: 8,238
United States
Message 1609721 - Posted: 6 Dec 2014, 5:12:29 UTC - in response to Message 1609716.  
Last modified: 6 Dec 2014, 5:13:55 UTC

I am using 337.88, so I don't damage the science results. I am also testing each and every NVIDIA driver released for my GPU, against their own OpenCL SDK samples, and keeping the pressure on NVIDIA for a fix, since they vow support until April 2016.

I'm not holding my breath. I'm doing my very best to resolve the problem and keep the community informed. I feel like I have very few allies, in this quest.

Lighten up. And have a great weekend!

Thanks,
Jacob
ID: 1609721 · Report as offensive
Profile Raistmer
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 16 Jun 01
Posts: 5804
Credit: 75,992,736
RAC: 50,727
Russia
Message 1609772 - Posted: 6 Dec 2014, 8:10:23 UTC - in response to Message 1609721.  
Last modified: 6 Dec 2014, 8:11:46 UTC

I feel like I have very few allies, in this quest.

Thanks,
Jacob


You have many sympathizers instead, perhaps :)
I already passed that route with own bug reports waiting much longer than 4 weeks. Yours have a possible advantage for mentioning their own examples that fail, that would make situation more intolerable for NV in case they would care. Seems they just don't care. Virtually no updates for OpenCL support given.
Perhaps we should be just glad that OpenCL support not completely revoked in new drivers....
ID: 1609772 · Report as offensive
Jacob Klein
Volunteer tester

Send message
Joined: 15 Apr 11
Posts: 50
Credit: 3,939,427
RAC: 8,238
United States
Message 1609788 - Posted: 6 Dec 2014, 9:40:16 UTC

I don't yet think that they don't care. Rather, I think that it is not yet a priority to them. We will see.
ID: 1609788 · Report as offensive
Profile jason_gee
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 24 Nov 06
Posts: 7471
Credit: 91,023,215
RAC: 10,284
Australia
Message 1609798 - Posted: 6 Dec 2014, 9:58:10 UTC - in response to Message 1609788.  

I don't yet think that they don't care. Rather, I think that it is not yet a priority to them. We will see.


I'd tend to agree. Cuda7 early alpha only just ended, then there'd be DirectX12/Win10, Maxwell midrange products, and 'a bazillion' 64 bit Denver Tegra K1 cores for the Google Nexus 9, absorbing dev resources along with silicon, not to mention the blindside by AMD with Mantle low-latency gaming drivers/apis in there to compete with.

Yeah, spread pretty thin I'd reckon. Not to say they shouldn't address the problem, or have plenty of resources to address it, though I know I'd have trouble justifying dedication of resources for no immediate return pretty hard to take before a shareholder or board meeting ;)
"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.
ID: 1609798 · Report as offensive
Profile RottenMutt
Avatar

Send message
Joined: 15 Mar 01
Posts: 1011
Credit: 230,274,184
RAC: 0
United States
Message 1609966 - Posted: 6 Dec 2014, 19:39:50 UTC

I would thing there is a fixed budget for Tesla Driver support and if they don't use it they lose it, so I think work is being done using some sort of priority system. If it is like any other company, sometimes it is a struggle to find a task and charge number to work to.
ID: 1609966 · Report as offensive
Profile Wiggo "Socialist"
Avatar

Send message
Joined: 24 Jan 00
Posts: 12599
Credit: 169,203,775
RAC: 86,104
Australia
Message 1610255 - Posted: 7 Dec 2014, 6:07:10 UTC
Last modified: 7 Dec 2014, 6:08:03 UTC

I'm sorry if you don't like my replies Jacob (and I don't mean to p*** you off), but even though your GPU was released in 2009 (making it 5yrs old now) it is based on 2008 architecture (making it 6yrs old now and the usual end of support mark) and I've seen these things happen so many times over the last couple of decades I just expect it.

The changes between those cores and Fermi (and newer) cores are like chalk and cheese in comparison and that makes the size of the newer drivers better not supporting those older designs, but also cheaper to produce. ;-)

Cheers.
ID: 1610255 · Report as offensive
Jacob Klein
Volunteer tester

Send message
Joined: 15 Apr 11
Posts: 50
Credit: 3,939,427
RAC: 8,238
United States
Message 1610256 - Posted: 7 Dec 2014, 6:13:01 UTC
Last modified: 7 Dec 2014, 6:15:10 UTC

Don't get me wrong, I know the GPU is ancient, and I expect support to be dropped at some point too. But the facts are that support isn't over until April 2016, and there are some serious OpenCL bugs that they introduced in their R340 line of drivers.

Now is the time for someone to jump up and complain, before support is over. And the complaint needs to be as clear and straightforward as possible, which I have done, knowing full well that their "motivation to fix" must be quite low. I refuse to let up on the pressure. Not holding my breath, but also not going down without the best fight I can put forth. These old GPUs still have some use in them :)

It's the R343 drivers that have dropped support for pre-Fermi GPUs. But the R340 drivers, which will be the "last line before support is entirely dropped", need to be golden... and right now, they're not.

Regards,
Jacob
ID: 1610256 · Report as offensive
Profile jason_gee
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 24 Nov 06
Posts: 7471
Credit: 91,023,215
RAC: 10,284
Australia
Message 1610259 - Posted: 7 Dec 2014, 6:22:42 UTC

FWIW, totally agree with both points of view here. I would only add that for some time there have been chaotic changes going on, no doubts still to be resolved. So in some cases 'right' answers may not yet exist.
"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.
ID: 1610259 · Report as offensive
mrchips

Send message
Joined: 12 Dec 04
Posts: 7
Credit: 8,638,879
RAC: 7,551
United States
Message 1610413 - Posted: 7 Dec 2014, 13:58:13 UTC

maybe someone can help me. I have NVidia GeForce 315 ver. 340.52. I don't get any GPU for seti when I ask for WU. It stopped getting WU's about 1 month ago. My setting should be OK. What am I missing?
ID: 1610413 · Report as offensive
Richard HaselgroveProject Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 11516
Credit: 106,129,012
RAC: 70,009
United Kingdom
Message 1610417 - Posted: 7 Dec 2014, 14:08:35 UTC - in response to Message 1610413.  

maybe someone can help me. I have NVidia GeForce 315 ver. 340.52. I don't get any GPU for seti when I ask for WU. It stopped getting WU's about 1 month ago. My setting should be OK. What am I missing?

You're missing the news which was posted on the front page on 13 November, and also passed on via the 'Notices' tab of your BOINC Manager: SETI@Home: Running out of workunits.

But don't worry, that's now been replaced (on 5 December) by SETI@Home: New Data from AO (Arecibo Observatory). There are workunits being processed now, but the pipeline is still (slowly) re-filling, and they are hard to come by. Your computer should resume processing normally by the middle of next week.

Having said that, your GeForce 315 is one of the older "Pre-Fermi" graphics cards that this thread is warning about. You should revert back to the previous driver while you're waiting.
ID: 1610417 · Report as offensive
Claggy
Volunteer tester

Send message
Joined: 5 Jul 99
Posts: 4654
Credit: 47,180,843
RAC: 388
United Kingdom
Message 1610425 - Posted: 7 Dec 2014, 14:30:34 UTC - in response to Message 1603322.  
Last modified: 7 Dec 2014, 14:32:29 UTC

I'm starting to think that the project has limited Pre-Fermi GPUs to the MBv7 Cuda23 and Cuda22 apps,
and put in a Max driver limit in too, Inspite of the MBv7 CUDA apps NOT being effected by the 340.52 OpenCL Pre-Fermi driver Bug.

Bump.

mrchips has plenty of CPU wu's, and should get MBv7 Cuda Wu's, assuming that there isn't a driver limit put in for the Cuda apps.

Claggy
ID: 1610425 · Report as offensive
Richard HaselgroveProject Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 11516
Credit: 106,129,012
RAC: 70,009
United Kingdom
Message 1610427 - Posted: 7 Dec 2014, 14:37:27 UTC - in response to Message 1610425.  

I'm starting to think that the project has limited Pre-Fermi GPUs to the MBv7 Cuda23 and Cuda22 apps,
and put in a Max driver limit in too, Inspite of the MBv7 CUDA apps NOT being effected by the 340.52 OpenCL Pre-Fermi driver Bug.

Bump.

mrchips has plenty of CPU wu's, and should get MBv7 Cuda Wu's, assuming that there isn't a driver limit put in for the Cuda apps.

Claggy

Well, mine was a more generic reply for the more general readership of this thread, and "the middle of next week" was designed to cover the re-start of the Astropulse database too ;)
ID: 1610427 · Report as offensive
Previous · 1 . . . 4 · 5 · 6 · 7 · 8 · 9 · 10 . . . 13 · Next

Message boards : Number crunching : @Pre-FERMI nVidia GPU users: Important warning


 
©2017 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.