Message boards :
Number crunching :
@Pre-FERMI nVidia GPU users: Important warning
Message board moderation
Author | Message |
---|---|
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
nVidia decided to drop pre-FERMI GPU support. We already found that 340.52 driver (though simulating work) generates incorrect results with AstroPulse and pre-FERMI hardware. This is thread on nVidia dev forums regarding issue: https://devtalk.nvidia.com/default/topic/774899/cuda-programming-and-performance/340-52-driver-breaks-opencl-support-for-pre-fermi-cc-1-x-cards/ So, if you want to use your pre-FERMI nVidia hardware for AstroPulse crunching stay with pre-340.xx drivers. |
rob smith Send message Joined: 7 Mar 03 Posts: 22202 Credit: 416,307,556 RAC: 380 |
A little bit of clarification: - If your Nvidia Gforce GPU is PRE 4xx then it is NOT a Fermi processor, and you will have the problems outlined by Raistmer unless you stick with the older drivers. (I've not looked to see what of the "Nxxx" or other families of Nvidia GPU are affected - perhaps someone else could fill in the blanks) Bob Smith Member of Seti PIPPS (Pluto is a Planet Protest Society) Somewhere in the (un)known Universe? |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14650 Credit: 200,643,578 RAC: 874 |
The best number to look for is the 'compute capability' of the card. This is listed in the BOINC Event Log, when the card is detected by BOINC at startup. Fermi cards start with cc 2.0, so provided all NVidia cards in your system show cc 2.0 or higher, you should have no problem with the newer drivers. If you have any cc 1.1, 1.2, or 1.3 card in your system, do not upgrade to any 340.xx or later driver unless further advised. |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
Currently issue was demonstrated on 9800GTX+ and 9800GT cards. |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14650 Credit: 200,643,578 RAC: 874 |
Currently issue was demonstrated on 9800GTX+ and 9800GT cards. Those two are Compute Capability 1.1 cards. Borrowing a summary list from Matt Harvey at GPUGrid, other common NV card types are: Compute Capability : 1.3 GTX280, GTX295, Tesla T10 series Compute Capability : 2.0 GTX470,480,570,580 Tesla 2000 series, Quadro non-K Compute Capability : 2.1 Other 400,500 series Compute Capability : 3.0 Other 600,700 series, Tesla K10, Quadro K series Compute Capability : 3.5 GTX780, 780Ti, Titan, Tesla K20 Compute Capability : 5.0 GTX 750, 750Ti It would be helpful if somebody with a card on the cc 1.3 list could test how far back pre-fermi this problem begins. |
BilBg Send message Joined: 27 May 07 Posts: 3720 Credit: 9,385,827 RAC: 0 |
 Add also GeForce GT 240 http://setiweb.ssl.berkeley.edu/beta/show_host_detail.php?hostid=71055 I notified the user, he responded with "Thanks for notifying!" and now the computer shows driver: 337.88 (old tasks have "Driver version: 340.52") http://en.wikipedia.org/wiki/GeForce_200_series GeForce GT 240 (Code name GT215) Compute Capability: 1.2 (GT215, GT216, GT218 GPUs)   - ALF - "Find out what you don't do well ..... then don't do it!" :)  |
BilBg Send message Joined: 27 May 07 Posts: 3720 Credit: 9,385,827 RAC: 0 |
The GPU errored out AP work in seconds. Did you ever been on driver 340.52 ? (if not - your problem is irrelevant to this thread) 'Error tasks for computer 5065145' show only 'Driver version: 337.88' http://setiathome.berkeley.edu/results.php?hostid=5065145&offset=0&show_names=0&state=6&appid= "Entry function 'GPU_fetch_array_kernel_twin_1D_cl' uses too much shared data (0x4034 bytes, 0x4000 max)" You should use Lunatics installer 0.42a - e.g. form link in this post: http://setiathome.berkeley.edu/forum_thread.php?id=75596&postid=1570561#1570561 EDIT: This is the original post: http://setiathome.berkeley.edu/forum_thread.php?id=75391&postid=1560695#1560695 Â - ALF - "Find out what you don't do well ..... then don't do it!" :) Â |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
Here's a 9800 GT producing incorrect results after installing driver 340.32; http://setiathome.berkeley.edu/results.php?hostid=7036438&offset=0&show_names=0&state=3&appid= Bump. Maybe this thread should be a Sticky? |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
Here's a 9800 GT producing incorrect results after installing driver 340.32; Or the new plan class opencl_nvidia_cc1 Eric made over on Beta that resolves this issue could be implemented here. The decision may have been made not to worry about doing it for AP v6 and just add it when AP v7 is rolled out. SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
Currently issue was demonstrated on 9800GTX+ and 9800GT cards. Looks as though the Quadro FX cards are affected also. Seems to be the same problem, not detecting any single pulses, http://setiathome.berkeley.edu/workunit.php?wuid=1601147878 If the task doesn't have any single pulses it will validate. I've even seen cases where the WingPerson found 1 single pulse and the effected card that didn't find the single pulse still validated. |
Wiggo Send message Joined: 24 Jan 00 Posts: 34748 Credit: 261,360,520 RAC: 489 |
Currently issue was demonstrated on 9800GTX+ and 9800GT cards. Seeing as the Quadro FX 1800 is based on the same G94 core as the 9xxx series cards it's to be expected. ;-) Cheers. |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
Currently issue was demonstrated on 9800GTX+ and 9800GT cards. Since you seem to be well versed in that class of cards, care to list other affected cards? I remember someone being 'concerned' about me running a Beta App on TWO tasks on main. Now we have HUNDREDS of tasks being run by cards KNOWN to produce Incorrect results...yet there doesn't seem to be much concern. hmmmm.... |
Wiggo Send message Joined: 24 Jan 00 Posts: 34748 Credit: 261,360,520 RAC: 489 |
Currently issue was demonstrated on 9800GTX+ and 9800GT cards. It'd be easier to say that any cards not based on the GFxxx/GKxxx/GMxxx core code names are affected. This list will help as it lists all Nvidia cards and what their core code names are, http://en.wikipedia.org/wiki/List_of_Nvidia_graphics_processing_units. Cheers. |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
That's probably still too vague for most people. Let's see, how about; Attention, if you have a nVidia card that's 4 years old or older, and have updated to Driver 340.xx, you are now Flooding SETI with Bad AstroPulse Science. This includes just about any nVidia card that's not at least a 400 series or around 4 years old or newer. How about that? |
Wiggo Send message Joined: 24 Jan 00 Posts: 34748 Credit: 261,360,520 RAC: 489 |
That's probably still too vague for most people. Let's see, how about; Actually the thread title is the best description for the hardware effected IMO. ;-) Cheers. |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
That's probably still too vague for most people. Let's see, how about; Except many people don't have a clue about 'FERMI' and the title doesn't mention a thing about Flooding SETI with BAD SCIENCE. To make matters worse, there is a thread about this New Driver, with over 1000 views, without a mention about it causing the older cards to Flood SETI with BAD SCIENCE. Not good. |
Wiggo Send message Joined: 24 Jan 00 Posts: 34748 Credit: 261,360,520 RAC: 489 |
It's really no different to all those AMD/ATi cards out there currently and over the years that are/have been doing bad science due to the wrong drivers being used either is it? Cheers. |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
The bad/wrong ATI drivers cause the App to NOT run or Error out. Much different than completing a task Incorrectly. Why does that have any bearing with the current known software problem producing KNOWN BAD RESULTS? These are Not Errors, they are being Submitted as Successful Results. |
Wiggo Send message Joined: 24 Jan 00 Posts: 34748 Credit: 261,360,520 RAC: 489 |
It really isn't that hard at all to work out what is a pre-Fermi card, but a person has got to love it when the shoe is on the other foot. :-) Cheers. |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
It really isn't that hard at all to work out what is a pre-Fermi card, but a person has got to love it when the shoe is on the other foot. :-) So your position is, since there have been ATI problems in the past, we are to ignore current known nVidia software problems known to produce Bad Science in the present. That's absolutely Brilliant! You just can't make this stuff up... ROFLMAO |
©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.