Message boards :
Number crunching :
GTX 590 questions
Message board moderation
Previous · 1 · 2
Author | Message |
---|---|
juan BFP Send message Joined: 16 Mar 07 Posts: 9786 Credit: 572,710,851 RAC: 3,799 |
Keep in mind GTX690 is a PCI-e 3.0 board, MB like the ASUS Crosshair II Formula have 2 PCIe 2.0 or more slots. Most of the cheap 3 PCIe capable boards slow down the PCIe (from 16x to 8x or even 4x or less) when you use more than one GPU. Go for a real double PCI-e 3.0x16 MB if you want to get all the power the 690 have. It´something like you have a V8 with twin turbo engine and try to run it with common gas, it will work but if you want to realy deploy all it´s capacity you need a high grade gas (or even a NOX). Ok SETI could not use all that power now but who knows the future? |
bluestar Send message Joined: 5 Sep 12 Posts: 7015 Credit: 2,084,789 RAC: 3 |
I have two GTX 680 4 GB Gainward cards installed in my computer. There is no SLI connector between them yet. Would there be any advantage of having such a connector installed? I have four such cables lying on the table in front of me. Two (the longest ones) are having similar length but different labeling. One reads NVIDIA SLI (ASUS), the other one just ASUS. The third one reads ASUS as well. The fourth which is unpacked reads CROSSFIRE. The question is as follows: Do I need one or two such cables in order to connect two GTX 680 graphics cards? By the way, I noticed Juan BFB's advice regarding optimized applications. I will gladly download and run these applications in due order when I get my discs up and running under Windows 7 Ultimate and Windows 8. Also, at least when PrimeGrid is concerned, definitely 64 bits applications are having shorter run times than 32 bits applications when a processor with a 64 bits capability is having a 64 bits operating system installed. |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14650 Credit: 200,643,578 RAC: 874 |
I have two GTX 680 4 GB Gainward cards installed in my computer. You might need the SLI cables if you are an extreme gamer. But for BOINC crunching - including, but not limited to, SETI - you don't need them at all. |
Bernie Vine Send message Joined: 26 May 99 Posts: 9954 Credit: 103,452,613 RAC: 328 |
The fourth which is unpacked reads CROSSFIRE Crossfire is the AMD/ATI equivalent of SLI so I would discard that one! |
juan BFP Send message Joined: 16 Mar 07 Posts: 9786 Credit: 572,710,851 RAC: 3,799 |
Optimized apps work with all windows (XP, 7 and 8) with out any modification. Expect no less than 20% of more RAC (even more with kepplers) when you switch to them, you just need to choose what is the best for your GPU. But that is an easy task. |
David Steel Send message Joined: 12 Jul 01 Posts: 52 Credit: 17,713,386 RAC: 0 |
Ok, second GTX590 in second PC, updated Boinc from 6.10.58 to 7.0.28 and driver 311.06 to 314.07. Now only one GPU is used... any suggestions? 14/03/2013 1:58:15 PM | | NVIDIA GPU 0: GeForce GTX 590 (driver version 314.7, CUDA version 5.0, compute capability 2.0, 1536MB, 8381786MB available, 1244 GFLOPS peak) 14/03/2013 1:58:15 PM | | NVIDIA GPU 1 (not used): GeForce GTX 590 (driver version 314.7, CUDA version 5.0, compute capability 2.0, 1536MB, 1370MB available, 1244 GFLOPS peak) 14/03/2013 1:58:15 PM | | OpenCL: NVIDIA GPU 0: GeForce GTX 590 (driver version 314.07, device version OpenCL 1.1 CUDA, 1536MB, 8381786MB available) 14/03/2013 1:58:15 PM | | OpenCL: NVIDIA GPU 1 (not used): GeForce GTX 590 (driver version 314.07, device version OpenCL 1.1 CUDA, 1536MB, 1370MB available) |
Tim Send message Joined: 19 May 99 Posts: 211 Credit: 278,575,259 RAC: 0 |
|
David Steel Send message Joined: 12 Jul 01 Posts: 52 Credit: 17,713,386 RAC: 0 |
Why does the 'recommended' version have this problem?, they should tell us about it when we download or install it, there is nothing in the release notes about the Nvidia GPU memory reporting Bug... Anyway, that link suggests version 7.0.44, can't see it on the Boinc download page, only 7.0.28 or 7.0.56. Is it better to go with the cc_config.xml fix or version 7.0.56? Thanks |
Wiggo Send message Joined: 24 Jan 00 Posts: 34744 Credit: 261,360,520 RAC: 489 |
The reason for this happening now is due to the newer drivers reporting less memory on the 2nd card so BOINC chooses the best hardware meaning the 2nd card doesn't get used. This is not a BOINC problem really but a driver thing. Cheers. |
Tim Send message Joined: 19 May 99 Posts: 211 Credit: 278,575,259 RAC: 0 |
Why does the 'recommended' version have this problem?, they should tell us about it when we download or install it, there is nothing in the release notes about the Nvidia GPU memory reporting Bug... Go with 7.0.28 AND cc_config.xml fix. Tim |
William Send message Joined: 14 Feb 13 Posts: 2037 Credit: 17,689,662 RAC: 0 |
Why does the 'recommended' version have this problem?, they should tell us about it when we download or install it, there is nothing in the release notes about the Nvidia GPU memory reporting Bug... Pushing Boinc 7 out too fast was one of the worst misjudements made by the BOINC dev team in a long time. The small but dedicated team of BOINC alpha testers has been trying to push for a new recommended version for months - 7.0.28 is so buggy, that we've been advising to use latest Alpha for quite some time now. It may have a few bugs left, but nothing compared to 7.0.28. The current 7.0.56 is a release candidate - that means it is currently being tested thoroughly for any show-stoppers, but should be OK. A few minor issues probably remain, but they are limited to very special running conditions and micromanaging, so won't affect the normal cruncher. Edit: I'd use 7.0.56 - you will need the cc_config.xml setting 'use all GPUs' whenever you have different GPUs in the system and want them all to crunch. Else BOINC will only use the 'best' card - regardless of BOINC version. The dual 590 and 690 might be handled slightly differently depending on version. Anyway I'd use latest alpha. A person who won't read has no advantage over one who can't read. (Mark Twain) |
David Steel Send message Joined: 12 Jul 01 Posts: 52 Credit: 17,713,386 RAC: 0 |
Thanks, I already had 7.0.28 installed so I added the cc_config.xml fix, now crunching 2 WU's per GPU. Now for the next problem, is 7.0.28 a pig of a thing or is it just me? My other PC with 6.10.58 has no problems getting work units, but this PC with 7.0.28 can only get a few WU's at a time and always has 'download pending - project backoff' Can I roll back to 6.10.58 or has 7.0.56 fixed this? |
Wiggo Send message Joined: 24 Jan 00 Posts: 34744 Credit: 261,360,520 RAC: 489 |
To roll back you'll have to run your current work out or lose it as you'll have to do a clean install. I've havn't heard of any real good reasons for me yet to go beyond 6.10.60 but I've heard plenty of real bad ones that I want to avoid. Cheers. |
rob smith Send message Joined: 7 Mar 03 Posts: 22188 Credit: 416,307,556 RAC: 380 |
While 7.0.28 may have it problems, moving on to another version can be problematic. I just tried on one of my crunchers, and trashed about a dozen tasks because I didn't plan the move properly. All the tasks on the GPU were re-estimated to unrealistic short run times (Astopulse in 1minute:30seconds - I would love, but with the current setup impossible). So rolled back to 7.0.28. Bob Smith Member of Seti PIPPS (Pluto is a Planet Protest Society) Somewhere in the (un)known Universe? |
Vipin Palazhi Send message Joined: 29 Feb 08 Posts: 286 Credit: 167,386,578 RAC: 0 |
Thanks, I already had 7.0.28 installed so I added the cc_config.xml fix, now crunching 2 WU's per GPU. You can rollback to an older version, but make sure to copy the BOINC folders both in program files (C:\Program Files\BOINC) as well as program data (C:\ProgramData\BOINC) or (C:\Documents and Settings\All Users\Application Data\BOINC) if you are using WinXP. After performing an uninstall of boinc, put the two folders back in their original location and then install the older version. You should be able to run boinc again, I have done this countless times without having to finish the existing cahce. And the best version according to me is still 6.10.60, although I have now upgraded all my rigs to 7.0.28. ______________ |
Claggy Send message Joined: 5 Jul 99 Posts: 4654 Credit: 47,537,079 RAC: 4 |
Thanks, I already had 7.0.28 installed so I added the cc_config.xml fix, now crunching 2 WU's per GPU. What cache settings are you using for that host? Claggy |
David Steel Send message Joined: 12 Jul 01 Posts: 52 Credit: 17,713,386 RAC: 0 |
Is that Boinc - Preferences: network usage? If so it's set to Minimum work buffer 0.10 days and Max work buffer 10.00 days. It's the same for both PC's |
Claggy Send message Joined: 5 Jul 99 Posts: 4654 Credit: 47,537,079 RAC: 4 |
Is that Boinc - Preferences: network usage? Boinc 7 has a totally new scheduler, where Boinc 6 would cache 10 days work with that setting, Boinc 7 uses the 1st setting as a Low water mark, and the 2nd as a High water mark, it'll fill to the low and high water marks, then wait for the amount of work to drop below the low water mark before asking for work again, Best to swap the preferences for the Boinc host around, but you don't want a too high a 1st setting as this also says how many days offline you'll be, something like 6 + 0.01 works well. BOINC 7.0 In finer detail Claggy |
David Steel Send message Joined: 12 Jul 01 Posts: 52 Credit: 17,713,386 RAC: 0 |
Ok thanks ill try it. What about my network usage settings on the Seti website, does one over ride the other? 'Maintain enough tasks to keep busy for at least' has a blank field... ... 'and up to an additional' is 10 days. Also, why did they remove the messages tab for 7.0.28??? Having a seperate 'event log' page is crap! 6.10.58 looks better and better all the time... |
Claggy Send message Joined: 5 Jul 99 Posts: 4654 Credit: 47,537,079 RAC: 4 |
Ok thanks ill try it. The local preferences in Boinc Manager over-ride the Web preferences. Claggy |
©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.