Message boards : Number crunching : Long run Santi worse then Noelia
Author | Message |
---|---|
I have 4 LR from Santi erroring in a row, after 32000, 34000 and 40600 seconds. | |
ID: 31723 | Rating: 0 | rate: / Reply Quote | |
Hello: I see your GTX770 performance is very low, it took more than twice mine. | |
ID: 31724 | Rating: 0 | rate: / Reply Quote | |
You are correct, looking at my BoincTasks history log for all my machines combined, there have been quite a few (at least 27) SANTI errored in the past few days. Except all of my errored units are the short SANTI runs. Not just a single machine, it happened on all machines and windows xp as well as windows 7. | |
ID: 31725 | Rating: 0 | rate: / Reply Quote | |
So it's the short ones that are the problem? Because I'm not experiencing any issues wight the long runs. | |
ID: 31726 | Rating: 0 | rate: / Reply Quote | |
Dont have issues too on both short and long santis | |
ID: 31727 | Rating: 0 | rate: / Reply Quote | |
My last LR that finished is from 21 July. I see that all LR from Santi and Nathan only use around 25% of memory, while that was 48% a few weeks ago. | |
ID: 31728 | Rating: 0 | rate: / Reply Quote | |
I have 4 LR from Santi erroring in a row, after 32000, 34000 and 40600 seconds. Same problem with SANTI_RAP74. http://www.gpugrid.net/forum_thread.php?id=3426 | |
ID: 31733 | Rating: 0 | rate: / Reply Quote | |
Looking close at the systems I see that after a Santi SR finished without error or a Santi SR errored out, that the GPU clock is down clocked and I need to reboot to get the clock running at around 1024MHz again. That's no problem if I am with the PC's but at home I cannot boot my rigs at work. | |
ID: 31738 | Rating: 0 | rate: / Reply Quote | |
To provide some counterpoint, my machines have completed 18 (and 3 more nearing completion) SANTI long WUs with no failures, no problems at all. This includes GTX 670, GTX 560, GTX 650 Ti and even GTX 460/768mb GPUs. The ones that are showstoppers here are the NOELIA RUN WUs which won't RUN properly on anything but the GTX 670. | |
ID: 31739 | Rating: 0 | rate: / Reply Quote | |
To provide some counterpoint, my machines have completed 18 (and 3 more nearing completion) SANTI long WUs with no failures, no problems at all. My hosts did it the same way. This includes GTX 670, GTX 560, GTX 650 Ti and even GTX 460/768mb GPUs. My hosts have GTX670s and GTX680s at the moment. The ones that are showstoppers here are the NOELIA RUN WUs which won't RUN properly on anything but the GTX 670. I've crunched a couple of that NOELIAs on my GTX480 properly. I'm quite sure, that the GTX 660 Ti is capable of crunching these workunits, but it needs an older driver (v307.90~v314.22) to do that. | |
ID: 31740 | Rating: 0 | rate: / Reply Quote | |
The ones that are showstoppers here are the NOELIA RUN WUs which won't RUN properly on anything but the GTX 670. Also > 1GB GPU ram to do them at reasonable efficiency. Luckily the SANTI WUs run fine in 768MB. | |
ID: 31741 | Rating: 0 | rate: / Reply Quote | |
I've crunched a couple of that NOELIAs on my GTX480 properly. Strange how different hardware/OS systems handle things. My 660Ti crunches NOELIIA and NATHAN tasks without any problems with the 320:18 drivers and XP but it chokes on the SANTI_RAP74 tasks. Maybe I should drop down to the 314:22 driver and see if the SANTI_RAP74 tasks play nicer. ____________ | |
ID: 31745 | Rating: 0 | rate: / Reply Quote | |
Strange how different hardware/OS systems handle things. My 660Ti crunches NOELIIA and NATHAN tasks without any problems with the 320:18 drivers and XP but it chokes on the SANTI_RAP74 tasks. Maybe I should drop down to the 314:22 driver and see if the SANTI_RAP74 tasks play nicer. MY GTX 670, GTX 560, GTX 650 Ti and GTX 460/768mb GPUs are now at 22 SANTI_RAP74 WUs completed with no errors. I'm using 310.90 drivers in Win7-64. Tried a newer driver a while back on 1 box and it started erroring some NOELIA WUs, Switched back to 310.90 and the errors went away. The only WUs that are a problem are the NOELIA_RUN, but that's because they need > 1GB memory or they slow to a crawl. | |
ID: 31750 | Rating: 0 | rate: / Reply Quote | |
I have again 3 Santi errors in a row on my GTX660 with 314.22 drivers after 60-70% of the WU finished. I have set this host to SR´s in the hope to finish WU´s without error. All the errors are not helping science. | |
ID: 31794 | Rating: 0 | rate: / Reply Quote | |
I have again 3 Santi errors in a row on my GTX660 with 314.22 drivers after 60-70% of the WU finished. I have set this host to SR´s in the hope to finish WU´s without error. All the errors are not helping science. Have you tried the 310.90 drivers? MY GTX 670, GTX 560, GTX 650 Ti and GTX 460/768mb GPUs are now at 44 SANTI_RAP74 WUs completed with no errors. I'm using 310.90 drivers (like your machine, running Win7-64). Looking at your GTX 660: http://www.gpugrid.net/results.php?hostid=153284&offset=0&show_names=1&state=0&appid= It looks like a lot of NATHANs and NOELIAs are erroring too. If it's not OCed try slowing it down below stock specs or move it to a known good machine. If it still errors, RMA it. | |
ID: 31797 | Rating: 0 | rate: / Reply Quote | |
I have again 3 Santi errors in a row on my GTX660 with 314.22 drivers after 60-70% of the WU finished. I have set this host to SR´s in the hope to finish WU´s without error. All the errors are not helping science. No I haven't tried the 310.90 driver, will do that. Not a lot of Noelia's or Nathan's with error, most are Santi's and they error after a long run time. The rig where it is in now is the best option, its in the coldest room and did rather well a few weeks ago. Could be the heat, but card is at 64°C. Its not OCed and I have it at clock speed as with the specs on the box. ____________ Greetings from TJ | |
ID: 31798 | Rating: 0 | rate: / Reply Quote | |
The specs on the box are not always good for all tasks from GPUGrid. None of my cards would run @ boxedspecs errorfree ^^ Since Fermi&Kepler changes are needed sometimes. 2xx and lower run all at original specs longer time ago ^^ | |
ID: 31799 | Rating: 0 | rate: / Reply Quote | |
Looks like switching to the 310.90 drivers may have solved my error issue with the SANTI_RAP74 WUs. Just completed 1 with no errors. Have 1 now running and 1 more in cache. Fingers crossed. | |
ID: 31813 | Rating: 0 | rate: / Reply Quote | |
Looks like switching to the 310.90 drivers may have solved my error issue with the SANTI_RAP74 WUs. Just completed 1 with no errors. Have 1 now running and 1 more in cache. Fingers crossed. I went back to 310.90 drivers to for the 660, but still errors on the SR. One LR Santi did finish same did Nathan and Noelia WU. My AMD PC with a 770 and driver 320.49 had nil errors yet and did all types of tasks. It looks to me that the 660 and Santi WU's need a special set up. The PC is now running a Santi_bax1 and I see the Kernel times slightly rising. When a Noelia ran this morning, the kernel times where very flat, almost not visible. ____________ Greetings from TJ | |
ID: 31821 | Rating: 0 | rate: / Reply Quote | |
Looks like switching to the 310.90 drivers may have solved my error issue with the WUs. Just completed 1 with no errors. Have 1 now running and 1 more in cache. Fingers crossed. My card is a 660TI and it finished a second SANTI_RAP74 task without error. One thing I've noticed is that with the 320 driver my GPU load was at 97%. With the 310 driver it's at 95%. Maybe that extra 2% load had something to do with the errors on the 320 driver. Just guessing. ____________ | |
ID: 31822 | Rating: 0 | rate: / Reply Quote | |
TJ, have you ever tried installing the drivers without the stuff you don't need; sound, 3D vision, updater? | |
ID: 31824 | Rating: 0 | rate: / Reply Quote | |
TJ, have you ever tried installing the drivers without the stuff you don't need; sound, 3D vision, updater? Yes, I did with no results. I also tried to run GPUGRID without anything crunching on the CPU and had no difference. Overnight two SR from Santi finished okay, and yesterday one LR. So it goes better. Seeing errors at wingman too, gives me confidence that the card is not faulty, but I will swap it when it become less warm here. But on this particular rig I had more error with Santi then Noelia, that's why I started this thread. ____________ Greetings from TJ | |
ID: 31825 | Rating: 0 | rate: / Reply Quote | |
My AMD PC with a 770 and driver 320.49 had nil errors yet and did all types of tasks. Since both are CC 3.0 Keplers I'm pretty sure the difference lies elsewhere than the cards in general. If you're talking about a specific card (which is OC'ed almost too far, too warm or whatever) on the other hand, this could well be. MrS ____________ Scanning for our furry friends since Jan 2002 | |
ID: 31830 | Rating: 0 | rate: / Reply Quote | |
There is definitely something going on. :) ----------------------------- Run time: 55,832s ----------------------------- CPU time: 39,996s Approximate elapsed time for entire WU: 28,272s 30x6-SANTI_RAP74wt-11-34-RND0448_0 ----------------------------- Run time: 55,403s ----------------------------- CPU time: 50,072s Approximate elapsed time for entire WU: 28,634s I've double cheched that the GPUs didn't downclocked. The three different runtimes should be nearly the same, like this: 47x19-SANTI_RAP74wt-6-34-RND8040_1 ----------------------------- Run time: 27,474s ----------------------------- CPU time: 27,362s Approximate elapsed time for entire WU: 27,483s These two strange SANTI LRs were acting like they were crunching: 95% GPU usage, progress indicator shows some progress, but the running time was abnormally high. I thought that my GPUs are downclocked, but they didn't. So I've restarted the host expecting the workunits will fail, but to my surprise they didn't. Instead they've started from 0% progress, while their elapsed time was ~9 hours, and they've finished successfully, with these strange run times. | |
ID: 31832 | Rating: 0 | rate: / Reply Quote | |
One LR Santi errored after 21000 seconds, but one finished okay. But I had a system crash overnight as well. The course seems to be a Noelia. After a restart I got a Santi SR and that failed after 1419 seconds. Thankfully I have a Noelia LR now! However I had a Noelia failed and caused the nVidia driver to crash, they do that within seconds. | |
ID: 31838 | Rating: 0 | rate: / Reply Quote | |
MY GTX 670, GTX 560, GTX 650 Ti and GTX 460/768mb GPUs are now at 84 SANTI_RAP74 WUs completed with with only 1 WU that errored after a bit over an hour. Again, I'm using 310.90 drivers in Win7-64. The only WUs that are a problem here are the NOELIA_RUN and the new 290... NOELIAs, but that's because they need > 1GB memory or they slow to a crawl. Personally I wish all the WUs were SANTIs. | |
ID: 31852 | Rating: 0 | rate: / Reply Quote | |
My own recent experience is that Santis are fine if you don't overclock. But they are very particular. I have had to recently reduce the clocks on my GTX 660s from 993 MHz (a factory overclock of less than 1 percent) to 980 MHz, the Nvidia specified rate. I have also tweaked up the core voltage a small amount for some additional overhead. That had not been necessary on any of the other longs, but with the Santis they work fine, until you get a hard one, and then they fail. | |
ID: 31853 | Rating: 0 | rate: / Reply Quote | |
Hi, Beyond: MY GTX 670, GTX 560, GTX 650 Ti and GTX 460/768mb GPUs are now at 84 SANTI_RAP74 WUs completed with with only 1 WU that errored after a bit over an hour. Again, I'm using 310.90 drivers in Win7-64. The only WUs that are a problem here are the NOELIA_RUN and the new 290... NOELIAs, but that's because they need > 1GB memory or they slow to a crawl. Personally I wish all the WUs were SANTIs. | |
ID: 31855 | Rating: 0 | rate: / Reply Quote | |
I don't want Santi's any more again 4 fail in a row LR and SR on the 660. I have also 310.19 drivers, in fact I have tried all drivers from 3xx.xx until latest beta, and no OC. With Noelia and Nathan with every driver good results. | |
ID: 31859 | Rating: 0 | rate: / Reply Quote | |
with GTX770 and the latest official driver, does all WU, including Santi okay. So I will only buy 690 and higher cards. They are more expensive, but 30 or more errors after 60% running costs a lot of energy that was complete spoiled. TJ, the SANTIs run better on low end cards than any other WUs. They even run on my GTX 460/768MB. There's something wrong either with your system setup or your GTX 660. Looking at your computer it reports: NVIDIA GeForce GTX 660 (2048MB) driver: 311.6 The 311.6 driver is bad. When I tested it I had many errors. Returned to 310.90 and the errors disappeared. I think you need to use driver sweeper to totally remove the traces of all the drivers you've tried. Then install 310.90 and see what happens. If you swap the 660 into a known good system and it still fails, return it or send it to the manufacturer for repair. | |
ID: 31865 | Rating: 0 | rate: / Reply Quote | |
Strange times... | |
ID: 31867 | Rating: 0 | rate: / Reply Quote | |
with GTX770 and the latest official driver, does all WU, including Santi okay. So I will only buy 690 and higher cards. They are more expensive, but 30 or more errors after 60% running costs a lot of energy that was complete spoiled. Your right Beyond, if I check BOINC first page I see indeed 311.6 however I am absolutely sure I downloaded and installed 310.90. I have even saved the file for just in case, here it is: 310.90-desktop-win8-win7-winvista-64bit-english-whql. After a clean install, first remove nVidia software, reboot, ccleaner, reboot, install new driver (as clean install), reboot, ccleaner, reboot and than BOINC, it seems to be the 311 one. I have it from nVidia US site. ____________ Greetings from TJ | |
ID: 31869 | Rating: 0 | rate: / Reply Quote | |
Your right Beyond, if I check BOINC first page I see indeed 311.6 however I am absolutely sure I downloaded and installed 310.90. I have even saved the file for just in case, here it is: 310.90-desktop-win8-win7-winvista-64bit-english-whql. It is because Windows automatic update gives you the new one as soon as you install 310.90. In fact, it might have even found it on your system from a previous download, and installed it without even having to download it again. (It is another feature from the geniuses at MS). Try the following: Control Panel/System/Advanced System Settings => Hardware Tab/Device Installation Settings "No, let me choose what to do" => Never install driver software from Windows Update However, to get rid of the ones already downloaded, I think you will need to uninstall them manually yourself. First, get rid of whatever you can from Control Panel/Programs and Features, rebooting as necessary. Then go into Device Manager and remove and delete the drivers from the computer. (I also run Driver Sweeper, or the newer version Driver Fusion at that point too.) Then, when you get back to the default Microsoft VGA drivers, you can start over again. | |
ID: 31875 | Rating: 0 | rate: / Reply Quote | |
Thanks Jim, | |
ID: 31878 | Rating: 0 | rate: / Reply Quote | |
I think the 314 drivers are fine, but it doesn't really matter in my case, since I use a dedicated PC for GPUGrid and don't need the card for display purposes. My only suggestion is if you do upgrade, do a clean uninstall with Driver Sweeper or Driver Fusion; it solves a lot of problems. | |
ID: 31882 | Rating: 0 | rate: / Reply Quote | |
My only suggestion is if you do upgrade, do a clean uninstall with Driver Sweeper or Driver Fusion; it solves a lot of problems. TJ, we keep saying this. Please do it. | |
ID: 31885 | Rating: 0 | rate: / Reply Quote | |
Okay, 310.90 drivers are in affect now and system was cleaned with driver sweeper as well. To test it I let it now do SR and no tasks at first on the CPU. | |
ID: 31895 | Rating: 0 | rate: / Reply Quote | |
Message boards : Number crunching : Long run Santi worse then Noelia