Message boards :
Number crunching :
@Pre-FERMI nVidia GPU users: Important warning
Message board moderation
Previous · 1 . . . 6 · 7 · 8 · 9 · 10 · 11 · 12 . . . 13 · Next
Author | Message |
---|---|
JanniCash Send message Joined: 17 Nov 03 Posts: 57 Credit: 1,276,920 RAC: 0 |
After seeing this thread I suspended GPU computing on the system, that has a 9800GT with 340.52 drivers and aborted all GPU tasks. Will rolling back the driver to the previous version 260.99 help? Regards, Jan |
Jacob Klein Send message Joined: 15 Apr 11 Posts: 149 Credit: 9,783,406 RAC: 9 |
If you can get a Release 337 driver, that would be best... Until the Release 340 driver is fixed. So, for now, get the 337.88 driver. |
JanniCash Send message Joined: 17 Nov 03 Posts: 57 Credit: 1,276,920 RAC: 0 |
If you can get a Release 337 driver, that would be best... Until the Release 340 driver is fixed. Thanks for the quick response. Installing 377.88 right now. It is a rather old desktop with only a Core2 Duo at 2.14GHz. The GPU is definitely the best component in that box. Would be a shame to disable that permanently. Thanks again, Jan |
Ademers Send message Joined: 16 Nov 09 Posts: 3 Credit: 1,026,395 RAC: 0 |
Hello, since few week i can computer Astropulse but no SETI@home v7 v7.03 (opencl_intel_gpu_sah) at all like WU = 1671312243 can you tell me what wrong ? |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14650 Credit: 200,643,578 RAC: 874 |
Hello, since few week i can computer Astropulse but no SETI@home v7 v7.03 (opencl_intel_gpu_sah) at all like WU = 1671312243 can you tell me what wrong ? No, not in a thread dedicated to nVidia hardware. |
Ademers Send message Joined: 16 Nov 09 Posts: 3 Credit: 1,026,395 RAC: 0 |
Sorry, when post I can't remove the message. |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
Bug Information So, let me know when new driver from that line will be awailable, I'll make build with lifted driver version limitation for testing. |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14650 Credit: 200,643,578 RAC: 874 |
So, let me know when new driver from that line will be awailable, I'll make build with lifted driver version limitation for testing. I was checking this morning for another thread, and nothing yet - not even Beta. Another one for the search teams. |
Jacob Klein Send message Joined: 15 Apr 11 Posts: 149 Credit: 9,783,406 RAC: 9 |
I am checking daily as well :) |
Jacob Klein Send message Joined: 15 Apr 11 Posts: 149 Credit: 9,783,406 RAC: 9 |
Still checking daily... still no updated R340 driver yet. Maybe NVIDIA is actually doing a thorough job of testing to make sure the driver is working correctly. One can only hope. |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14650 Credit: 200,643,578 RAC: 874 |
Yeah. Even though R347 for the newer cards got a refresh last week. |
jason_gee Send message Joined: 24 Nov 06 Posts: 7489 Credit: 91,093,184 RAC: 0 |
The silence has been deafening since Cuda 7 early access testing closed, and it went to RC status (which I've yet to download and install). I wonder if they can or will put Cuda 7 support in that special driver branch (though won't matter to me since I'll continue building older, including 32 bit, versions that run better on older cards) FWIW, current Cuda multibeam still runs best on Cuda 5 here on GTX 980, though that's more to do with system latencies and application architecture, which is due for a major revamp. (forced move to 64 bit application builds is going to take some doing, since there is a performance penalty to work around) Another small part of why the jostling and slow emergence is going on, may be related to google's recent statements/presentations about why they're moving nexus devices to their own renderscript, not enabling OpenCL or Cuda. ( http://stackoverflow.com/questions/14385843/why-did-google-choose-renderscript-instead-of-opencl/14389506#14389506 ) At first that seems full of vested interests etc, though there are very good technical points there. My reason for bringing that up in this pre-Fermi support context, is that combined with the earlier mentioned 32-64 bit scaling transition, and the scaling from very small to very large GPUs (and CPUs too in the case of OpenCL) is indeed much more work than it should be with respect to synchronisation and kernel geometries, so more painful transitions to come yet, while they tread carefully (presumably Google, NV and OpenCL Khronos Group). "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. |
Jacob Klein Send message Joined: 15 Apr 11 Posts: 149 Credit: 9,783,406 RAC: 9 |
Still no updated/fixed R340 driver from NVIDIA, for my Pre-Fermi GPU, to fix the data problems. Remaining on R337, driver version 337.88, until it's fixed. Continuing to try to remain patient :) |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
The R340 drivers will continue to support these products until April 1, 2016 Not too long ... |
Jacob Klein Send message Joined: 15 Apr 11 Posts: 149 Credit: 9,783,406 RAC: 9 |
The R340 drivers will continue to support these products until April 1, 2016 14 months? That's still a long time. :) |
Jacob Klein Send message Joined: 15 Apr 11 Posts: 149 Credit: 9,783,406 RAC: 9 |
Ahem. A new R340 driver, 341.44, was released today, for my pre-Fermi FX3800M GPU. I'll be testing this soon :) |
Wedge009 Send message Joined: 3 Apr 99 Posts: 451 Credit: 431,396,357 RAC: 553 |
The release summary doesn't mention much besides security updates. Would be interesting to know whether there were any changes related to the OpenCL issues mentioned here. Soli Deo Gloria |
Jacob Klein Send message Joined: 15 Apr 11 Posts: 149 Credit: 9,783,406 RAC: 9 |
I'm hoping so (since my NVIDIA contact said they had the issue fixed in December, but it didn't make it in time for the December driver release) ... And I'm going to be testing it within the next couple hours :) |
Jacob Klein Send message Joined: 15 Apr 11 Posts: 149 Credit: 9,783,406 RAC: 9 |
The first test (of the 6 that were failing) ... has PASSED!!! I'll of course test the whole thing, but, progress has officially been made!! |
Jacob Klein Send message Joined: 15 Apr 11 Posts: 149 Credit: 9,783,406 RAC: 9 |
6 for 6 -- All 6 of the "failing SDK examples" now work correctly on R340 driver 341.44. Now commencing with the AstroPulse Beam test, to see if NVIDIA's OpenCL fixes have (hopefully) fixed the problem the SETI app was having! Testing now... [obviously excited] |
©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.