Message boards :
Number crunching :
WUs dying on GTX460
Message board moderation
Author | Message |
---|---|
Send message Joined: 8 Apr 08 Posts: 1 Credit: 5,698,658 RAC: 0 |
Getting this error on Workunit 137186102 (name:de_16_3s_2_1747595_1281330756): <![CDATA[ <message> Incorrect function. (0x1) - exit code 1 (0x1) </message> <stderr_txt> Device index specified on the command line was 0 Looking for a Double Precision capable NVIDIA GPU The device GeForce GTX 460 from the command line cannot be used because a device supporting compute capability 1.3 (Double Precision) is required Found 1 CUDA cards Found a GeForce GTX 460 Device cannot be used, it does not have compute capability 1.3 support No compute capability 1.3 cards have been found, exiting... I thought that GTX460 supported DP |
Send message Joined: 14 Feb 09 Posts: 999 Credit: 74,932,619 RAC: 0 |
It does, but the CUDA app does not support FERMI cards at the moment. |
Send message Joined: 19 Apr 08 Posts: 2 Credit: 3,722,213 RAC: 1 |
Same here on my gtx 460--Cuda wu,s crashes because of a runtime failure. |
Send message Joined: 13 Aug 10 Posts: 15 Credit: 122,278 RAC: 0 |
Wish the GF104 were supported... |
Send message Joined: 26 Apr 10 Posts: 1 Credit: 1,069 RAC: 0 |
i have a GTX460 and have same problem as well.. when will this be fixed ? |
Send message Joined: 13 Aug 10 Posts: 15 Credit: 122,278 RAC: 0 |
Who knows. They're working on moving it all over to OpenCL to be able to support one piece of code for both ATI and Nvidia cards. But now the new Catalyst drivers crash and burn on WUs and the CUDA 2.3 app isn't working for a lot of cards, I'd say something needs to be fix and fixed fast. |
Send message Joined: 26 Jun 08 Posts: 5 Credit: 47,545,674 RAC: 7,018 |
I figured I would just poke this thread. Is there any word yet on making the CUDA app support the fermi line of GPU's? I got a GTX 460 recently and of course it does not work here yet. jna |
Send message Joined: 24 Feb 09 Posts: 620 Credit: 100,587,625 RAC: 0 |
I doubt they will be rushing to do it - the Fermi line of cards had their Double Precision capability cut drastically as a design choice way back, to help compensate on the design nightmare NVidia had trying to get decent yields from production. Therefore even if the app ran for Fermis, the performance would be slow compared to single precision projects. Regards Zy |
Send message Joined: 22 Apr 09 Posts: 38 Credit: 27,377,932 RAC: 0 |
Apart from the fact that a GTX460 is more expensive, compaired to an HD4850 or even higher, although not the HD5700 series (SP) the 5800 are DP. Difference in price is not that big, but they have already a working app for ATI cards, so little reason to invest a lot of time, money and several recourses, IMO! [ADDED] The GTX 460, 470 and 480, all have Compute Capabillity 2.0, don't know for sure if there are also Double Precision, though. Will have a look :) Anyone tried to reach the Lunatics WEB-Site? Look like it's down. Knight Who says Ni |
Send message Joined: 13 Jan 08 Posts: 52 Credit: 59,823,507 RAC: 0 |
From what was said on seti and IIRC the 4XX CUDA cards have the double precision disabled at the factory. I believe it was a choice to only have double precision on the Tesla/Workstation GPU's Blah blah blah you know the rest |
Send message Joined: 20 Apr 10 Posts: 1 Credit: 3,708 RAC: 0 |
GTX460 has double presision. report running GPUGRID atm: CUDA-Z Report |
Send message Joined: 24 Feb 09 Posts: 620 Credit: 100,587,625 RAC: 0 |
Single-precision Float: 637922 Mflop/s That illustrates the problem. The equivalent level card in the ATI lineup - ATI HD 5850 - figures are: (Edit: Ooooppps .... a mere 100% out! Figures corrected, sorry about that ...) Single Precision: 2.08 TeraFLOP Double Precision: 416 GigaFLOP Traditionaly Nvidia cripple the Double Precision to one eight single precision, and ATI cripple to one fifth single precision. Both do that to ensure cheaper consumer end cards do not blow away the much more hansomely priced Scientific Crunching cards (aka Tesla/Firestream). Tesla is half single precision speed (four times the rate for consumer cards). Further up the ATI HD range you go, the more stark the DP performance difference. Software comes into it of course - but given a working ATI app, such a stark performance difference doesnt do a lot to motivate the effort for another codestream. Regards Zy |
Send message Joined: 5 Apr 09 Posts: 71 Credit: 6,120,786 RAC: 0 |
CUDA-Z Report ============= Version: 0.5.95 http://cuda-z.sourceforge.net/ OS Version: Windows AMD64 6.1.7600 Core Information ---------------- Name: GeForce GTX 470 Compute Capability: 2.0 Clock Rate: 810 MHz Multiprocessors: 14 Warp Size: 32 Regs Per Block: 32768 Threads Per Block: 1024 Watchdog Enabled: Yes Threads Dimentions: 1024 x 1024 x 64 Grid Dimentions: 65535 x 65535 x 1 Memory Information ------------------ Total Global: 1248.44 MB Shared Per Block: 48 KB Pitch: 2.09715e+06 KB Total Constant: 64 KB Texture Alignment: 512 GPU Overlap: Yes Performance Information ----------------------- Memory Copy Host Pinned to Device: 5608.38 MB/s Host Pageable to Device: 2997.46 MB/s Device to Host Pinned: 5602.31 MB/s Device to Host Pageable: 3013.9 MB/s Device to Device: 48091.4 MB/s GPU Core Performance Single-precision Float: 1.16563e+06 Mflop/s Double-precision Float: 146494 Mflop/s 32-bit Integer: 585128 Miop/s 24-bit Integer: 584478 Miop/s Generated: Thu Sep 23 23:58:28 2010 Team Alliance francophone, boinc: 7.0.18 GA-P55-UD5, i7 860, Win 7 64 bits, 8g DDR3, GTX 470 |
Send message Joined: 24 Feb 09 Posts: 620 Credit: 100,587,625 RAC: 0 |
Precisely. The equivalent number two ATI card - 5870: Single Precision: 2.7TeraFLOP Double Precision: 540GigaFLOP Regards Zy |
Send message Joined: 23 Sep 10 Posts: 3 Credit: 209,535 RAC: 0 |
So is there an ETA for running Fermis yet? I have a 465 for gaming but would like to use it when im not at my machine for this purpose. So im not about to go get an ATI card or anything. I have a 260 working in my server but id like to up my numbers. Hopefully they will release support soon. |
Send message Joined: 19 Feb 08 Posts: 350 Credit: 141,284,369 RAC: 0 |
From what was said on seti and IIRC the 4XX CUDA cards have the double precision disabled at the factory. I believe it was a choice to only have double precision on the Tesla/Workstation GPU's All compute-capability 1.3 and 2.0 cards have double precision. Find more here: http://www.geeks3d.com/20100606/gpu-computing-nvidia-cuda-compute-capability-comparative-table/ Alexander |
Send message Joined: 24 Feb 09 Posts: 620 Credit: 100,587,625 RAC: 0 |
..... I have a 260 working in my server but id like to up my numbers. Hopefully they will release support soon. If your prime motivation for the 465 is a numbers game, I suggest having a look at GPUGRID (their 260 hassles are over, and they have a good stable Fermi app). Even if the Fermi CUDA app here gets revamped and fielded, and I suspect that will be some time in coming, you will be hit by the reduced processing on Double Precision. GPUGRID is single precision, and therefore more suited to your 465. Regards Zy |
Send message Joined: 23 Sep 10 Posts: 3 Credit: 209,535 RAC: 0 |
Hmm i suppose in that case ill just leave my server to its thing and ill just turn my main pc off at night. I have all 4 cores of my server on the cause as well besides the GTX260. And my prime motivation isnt a numbers game like the people running WCG and F@H. I just wanted to add extra horse power under my name thats all. Seems double precision has been crippled severely on Fermi desktop variant =( |
©2024 Astroinformatics Group