Message boards :
Number crunching :
Computation Errors in home 1.20 (opencl_amd_ati)
Message board moderation
Author | Message |
---|---|
Send message Joined: 29 Mar 10 Posts: 3 Credit: 34,560,651 RAC: 0 |
Darned annoying. It seems to have started with a long overdue driver update. amd hd6950 gpu. Keep getting notices about the catalist driver and about not having a double precision gpu as well. This was working fine for the last few years, it's only been a problem for 3 weeks or less. I'm NOT replacing this card Just to crunch number for this project. So if this can't be fixed, I'll just stop. I already dumped seti for a similar reason - they didn't want to support the older ndiva card I kept installed to just do their work. |
Send message Joined: 8 May 09 Posts: 3322 Credit: 520,664,656 RAC: 33,166 |
Darned annoying. You may have upgraded too far, YES the 6950's CAN do the MilkyWay gpu units just fine, but the driver could be the key. You are all the way upto 1.4.1848 and that could be the problem. The most current NON beta driver is 13.9 and can be gotten here: http://sites.amd.com/us/game/downloads/Pages/radeon_win7-64.aspx Although you MAY have to use an uninstall utility if you run into problems, which can be gotten here: http://sites.amd.com/us/game/downloads/Pages/catalyst-uninstall-utility.aspx I brought all of my AMD gpu's up to the 13,10 Beta level and some just had nothing but problems. I even got error messages telling me I was using 'an unsupported driver version'. Other pc's worked just fine, but there are too many projects that they WILL work at so I just moved on, and in the case of my own 6950 to Collatz and it worked great! I have now moved it on again this time to Rainbow, aka DistRTgen. I was unwilling to downgrade my own version of the Catalyst so just moved on to other projects. |
Send message Joined: 29 Mar 10 Posts: 3 Credit: 34,560,651 RAC: 0 |
......... I was unwilling to downgrade my own version of the Catalyst so just moved on to other projects. Thanks for the info. I'll try the old driver, but I shouldn't Have too. I donate the cycles to this project, it's one of the few that interest me, but it sure would save electricity to shut the system down after hours every day. If they payed me to crunch this project, that would be different, but They need to fit MY system not the other way around. That's why SETI lost me, I left my older geforce GPU in when I change to the AMD. That way SETI could have their 'cuda' stuff running without interference between projects. I admit that card wasn't quick but at least they got computing time for free. |
Send message Joined: 18 Jul 09 Posts: 300 Credit: 303,562,776 RAC: 0 |
... but They need to fit MY system not the other way around. That's where you are mistaken. You volunteered to participate in the project the way the science team has it set up. There is no obligation for them to cater to anyone, we are all free to leave at any time. |
Send message Joined: 29 Mar 10 Posts: 3 Credit: 34,560,651 RAC: 0 |
... but They need to fit MY system not the other way around. Well since your the expert... your advice just cost them my cooperation. GOOD RIDDANCE. |
Send message Joined: 26 Nov 10 Posts: 2 Credit: 11,226,655 RAC: 0 |
This is a odd thought and I had to respond with a disagreement on your logic. We are what make these projects work. If the project operators alienate the users then how will the Distributed Computing Model work? If these teams don't cater to the community then they need to rethink what they are doing and need to build their own super computers and use their own power to crunch these numbers. The only way this works is by cooperation not through the alienation of users and equipment who are trying to assist in a project they in essence get nothing for other than a higher power bill, wear and tear on equipment and a warm feeling of helping these projects. |
Send message Joined: 8 May 09 Posts: 3322 Credit: 520,664,656 RAC: 33,166 |
This is a odd thought and I had to respond with a disagreement on your logic. We are what make these projects work. If the project operators alienate the users then how will the Distributed Computing Model work? If these teams don't cater to the community then they need to rethink what they are doing and need to build their own super computers and use their own power to crunch these numbers. The only way this works is by cooperation not through the alienation of users and equipment who are trying to assist in a project they in essence get nothing for other than a higher power bill, wear and tear on equipment and a warm feeling of helping these projects. Yes and no, the projects are run mostly by people that still have high ideals and think 'if they come and help fine, if not that's okay too'. The advantage for us users is that there are SOOO many Boinc projects to chose from when one project gets too ornery, or outside our comfort zone, we just move on to another project. I have personally dropped several and will never go back for different reasons. As soon as more projects starting acknowledging us crunchers in their Scientific Papers though that pendulum will swing more to the crunchers side and we will be more respected. A few projects have acknowledged the help of the crunchers, but not many and not in the ways you might expect them to have. It's more of a 'thanks to the Boinc project and it's crunchers for compiling the data, but WE did all the work making it useful'. The original Distributed Computing Model was based on getting x number of people, using their home pc's, working together on a single project, Seti, and it worked beyond anyones wildest dreams!! By making Boinc not just Seti related, there are now more then a dozen active projects, the original idea is still alive and well. |
Send message Joined: 2 Aug 08 Posts: 24 Credit: 374,440,641 RAC: 0 |
I do have the same problem. My 5870 + 6950 in one system would crunch MW fine if they would send out the ati14 app. But they send the OpenCL app. This crashes after the WU reaches 100%. The cards are not occed and crunch collatz, seti, albert etc fine. Link to a wu: http://milkyway.cs.rpi.edu/milkyway/result.php?resultid=590603344 Greetz to all |
Send message Joined: 4 Sep 12 Posts: 219 Credit: 456,474 RAC: 0 |
There were similar problems reported over the summer for the Einstein OpenCL apps. They were resolved by building against a newer set of BOINC API libraries. See the Einstein discussion (and testing process) at BRP5 Version 1.36 not running with Boinc 7.0.64 |
Send message Joined: 5 Nov 10 Posts: 69 Credit: 15,064,831 RAC: 0 |
I have HD3850's (AGP) and an HD2400 (PCI not -E) crunching very nicely ... check my credits :-D My advice is to get the BOINC version and Catalyst version that works then stick to it. HTH, Ray FWIW(1) it took me HOURS over DAYS to get the right mix. FWIW(2) ignore those irritating pleas from BOINC to use the latest version. FWIW(3) ignore the temptation to update drivers to the latest & greatest. If you do/there are other reasons why you need to update your drivers and then BOINC suddenly can't crunch don't blame BOINC/Project(s). |
Send message Joined: 5 Nov 10 Posts: 69 Credit: 15,064,831 RAC: 0 |
Slight correction:- the HD2400 is crunching very nicely for Moo! and Collatz. Not for MW@H ... no tasks supplied. So if your project does not supply tasks that your otherwise OK setup crunches OK - complain to that project admin. Ergo - this is a complaint to the MW@H project admin ... MW@H does not give my HD2400 any tasks. Specifically MilkyWay@Home 0.82 (ati14) FWIW it used to, many moons ago - but not anymore. |
Send message Joined: 14 Feb 09 Posts: 999 Credit: 74,932,619 RAC: 0 |
Slight correction:- the HD2400 is crunching very nicely for Moo! and Collatz. Nope, it should have never given tasks to the HD2400 as that GPU does not support double precision. The minimum GPU for MW is the HD3850, and then most of X800 or higher variants from there |
©2024 Astroinformatics Group