Message boards :
Number crunching :
help: New Quad core PC with GTX460
Message board moderation
Author | Message |
---|---|
Sean Arrowsmith Send message Joined: 17 Feb 01 Posts: 34 Credit: 4,955,471 RAC: 0 |
Hi, just got a new PC, a 4Gb Quad Core AMD II x4 with a Nvidia GTX460 I down loaded BOINC, installed it, then attached to SETI, SETI download the core apps and stuff. When finished I then installed the latest Lunatics Unified Installer from his site selecing X64 plus the Cuda option Now on my other Quad core (No GPU) I see in task view 4 tasks each being worked on at the same time (Same with my dual core) but on the new PC only 1 tasks is running and it's just using the lunaticCuda app and in tasks view it show for the only running task 0.04CPU + 1 NVIDA GPU IS this correct or should the apps be running on 4 CPU's plus the GPU? Any help would be great. |
Gundolf Jahn Send message Joined: 19 Sep 00 Posts: 3184 Credit: 446,358 RAC: 0 |
All the CPU tasks on that host error out with: Incorrect function. (0x1) - exit code 1 (0x1) Gruß, Gundolf Computer sind nicht alles im Leben. (Kleiner Scherz) SETI@home classic workunits 3,758 SETI@home classic CPU time 66,520 hours |
Claggy Send message Joined: 5 Jul 99 Posts: 4654 Credit: 47,537,079 RAC: 4 |
Which CPU MB app did you install? Claggy |
Jim Wright Send message Joined: 3 Sep 99 Posts: 36 Credit: 54,761,763 RAC: 17 |
I have a AMD Phenom(tm) II X6 1055T Processor plus NVIDIA GeForce GTX 465. The tasks page shows six tasks running simultaneously plus "0.67 CPUs+ 1.00 NVIDIA GPU". Check the Advanced/Preferences/Processor Usage tab and the percent to be used on multi processor systems. Set it to 100%. Also check your computing preferences at set@home. I do have an observation that may need some fixing: Workunit delivery on such a system seems to be severely constrained. I get only about as much work as can be processed in 24 hours rather than the three days I specified in my Computing Preferences. The amount of work delivered is insufficient to keep the machine busy over any extended (24 hours plus) outage. It seems that a count of all workunits (both cuda and non-cuda) delivered and stored on my system is the constraining measure used whereas there should be separate totals maintained for non-cuda and cuda workunits. I'm seeing 7-8 hours each for non-cuda workunits and 15-20 MINUTES each for cuda workunits. Therefore I would need 72 non-cuda workunits available to cover a 3 day outage. AND ALSO ... 288 cuda workunits to keep the cuda processors busy during the same 3 days. When long Seti@home outages occur, I have to do work from other sites just to keep the machines busy. Luckily, Einstein@home is also of interest to me. Jim Wright |
Fred J. Verster Send message Joined: 21 Apr 04 Posts: 3252 Credit: 31,903,643 RAC: 0 |
Yeah, but the problem is your C.P.U. as you can clearly see here Results . Or is this your host? Use the UNIFIED Installer V0.37 from the LUNATICs. Lunatics. And GPUz, also CPUz gives usefull info about your CPU & GPU. |
Sean Arrowsmith Send message Joined: 17 Feb 01 Posts: 34 Credit: 4,955,471 RAC: 0 |
Yes, my PC is http://setiathome.berkeley.edu/show_host_detail.php?hostid=5883516 I can now see clearly that something is wrong with the CPU from your help. I used Lunitics latest 037 installer. Unchecked AP but selected AMD SSE3 and also the CUDA during the installer. Done this twice before origianlly posting in case I had done it incorrectly. At least I know that it just does not work one one WU at a time but should be one per CPU/GPU The app_info.xml is the one from the lunatic package,I'll check it all over again to see what is what. thanks |
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13736 Credit: 208,696,464 RAC: 304 |
Workunit delivery on such a system seems to be severely constrained. I get only about as much work as can be processed in 24 hours rather than the three days I specified in my Computing Preferences. The amount of work delivered is insufficient to keep the machine busy over any extended (24 hours plus) outage. Did you manually derive & enter the FLOPS value in to the app_info file? If not, that's probably why your cache won't fill up due to the Result Duration Correction Factor going all over the place. Even with that value set, when i've 4 days of CPU work it's about 3.75 days of GPU work. Grant Darwin NT |
Sean Arrowsmith Send message Joined: 17 Feb 01 Posts: 34 Credit: 4,955,471 RAC: 0 |
Sorry, I do not understand anything about FLOPS etc. Could you explain more please about FLOPS. Also, how do I know that the GTX460 isbeing maxed out or not as the cuda app is running but the GPU temp is constant at 68Deg C which suggests that it's not working to capacity. It this right? Thanks. Ps I have now got the CPU's to crunch, I changed the app_info.xml file to make sure the CPU were poiniting to the corrct application AK_v8b_win_x64_SSE3 as for some reason they were pointing to a different application that did not exist. |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
I believe you should be using "AK_v8b_win_SSE3_AMD.exe". As the 64-bit lunatics installer indicates "SSE3_AMD (x86, faster on sse3 amd)". SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13736 Credit: 208,696,464 RAC: 304 |
Sorry, I do not understand anything about FLOPS etc. Could you explain more please about FLOPS. Someone else could explain it better than i can, roughly the difference in processsing ability between the CPU & GPU causes the Result Duration Correction Factor (RDCF) to move all over the place (it is used when determining how much work to download to maintain the cache setting). By entering a FLOPS (Floating Operations Per Second) value in the app_info file you're giving a reference that helps keep the RDCF for the CPU & GPU very close to each other. That allows the system to actually download enough work to maintain your cache setting. NB- make sure to finish of all your current work before setting the FLOPS value, otherwise all existing work will error out. Info on FLOPS value Also, how do I know that the GTX460 isbeing maxed out or not as the cuda app is running but the GPU temp is constant at 68Deg C which suggests that it's not working to capacity. It this right? GPUZ will show you how hard the GPU is working. On my GTX460, i set the app_info value so that the video card runs 2 WUs at the same time, resulting in the GPU loading being around 96%. Grant Darwin NT |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14650 Credit: 200,643,578 RAC: 874 |
Someone else could explain it better than i can, roughly the difference in processsing ability between the CPU & GPU causes the Result Duration Correction Factor (RDCF) to move all over the place (it is used when determining how much work to download to maintain the cache setting). That's actually very old advice, dating back to before the new server-based DCF adjustments were introduced in June 2010. For most users, and general purposes, I'd just forget flops nowadays, and let the server handle it. Don't set your cache too high on a new host until after ten tasks on each type have validated, and you should be fine. |
Jim Wright Send message Joined: 3 Sep 99 Posts: 36 Credit: 54,761,763 RAC: 17 |
??? I cannot locate the app_info.xml file on my Windows 7 system. Where is it? Thanks |
OzzFan Send message Joined: 9 Apr 02 Posts: 15691 Credit: 84,761,841 RAC: 28 |
You will only have an app_info.xml file if you're using optimized applications. If you are, the location varies depending on which OS you're using. Windows 2000/XP: C:\Documents and Settings\Application Data\BOINC\projects\setiathome.berkeley.edu Windows Vista/7: C:\ProgramData\BOINC\projects\setiathome.berkeley.edu Note that the "Application Data" and "ProgramData" folders are hidden by default. If you type them into the Windows Explorer address bar, you can still navigate to them without having to fiddle with your hide/unhide settings. |
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13736 Credit: 208,696,464 RAC: 304 |
Someone else could explain it better than i can, roughly the difference in processsing ability between the CPU & GPU causes the Result Duration Correction Factor (RDCF) to move all over the place (it is used when determining how much work to download to maintain the cache setting). Ah, i've always had a 4 day cache. When i got my new system i installed BOINC, then used the Lunatics installer to install the optimised apps, then connected to the net & Seti. Even after a day or so the RDCF was going up & down like a yoyo with no end in sight, that's when i manually entered the FLOPS value. Grant Darwin NT |
Sean Arrowsmith Send message Joined: 17 Feb 01 Posts: 34 Credit: 4,955,471 RAC: 0 |
Do you mean I would be better off running the x86 app instead of the x64 app on my x64 Win7 system? |
Sean Arrowsmith Send message Joined: 17 Feb 01 Posts: 34 Credit: 4,955,471 RAC: 0 |
[quote GPUZ will show you how hard the GPU is working. On my GTX460, i set the app_info value so that the video card runs 2 WUs at the same time, resulting in the GPU loading being around 96%.[/quote] OK, GPUZ shows an average of 65% loading over the course of an hours worth of SETI GPU crunching with an average temp of 68Deg c. What is the app_info for 2 WU at the same time please. Is there an actual page some where showing what tweaks can be made to app_info etc as I remember one to increase the number of days to show your graph stats over. Thanks |
Gundolf Jahn Send message Joined: 19 Sep 00 Posts: 3184 Credit: 446,358 RAC: 0 |
Is there an actual page some where showing what tweaks can be made to app_info etc as I remember one to increase the number of days to show your graph stats over. That is in the cc_config.xml, not the app_info.xml. ;-) For running two tasks concurrently on your GPU, you'll have to go anonymous platform, which prevents your computers from downloading the newest version of the stock application on its own. There have been several threads around on that topic recently. Gruß, Gundolf |
Sean Arrowsmith Send message Joined: 17 Feb 01 Posts: 34 Credit: 4,955,471 RAC: 0 |
That's me being a dumb. Thanks
I am on anonymous as running lunatic 0.37 |
perryjay Send message Joined: 20 Aug 02 Posts: 3377 Credit: 20,676,751 RAC: 0 |
Sean, Go into your app_info and find every instance where it says <count> To run two units at a time change it to .5 if you can run three units at a time change it to .33. In 64 bit I believe the installer put in six instances of <count> so you would have to change all of them. This is what one of mine looks like... <coproc> edit: The normal caveats apply.. shut down SETI and BOINC first before making any changes to app_info. Use notepad to make changes ect. PROUD MEMBER OF Team Starfire World BOINC |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
That is what the Lunatics guys put in the .37 installer GUI. Since they developed the software I would guess they know what they are doing. SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
©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.