Message boards :
Number crunching :
Perhaps some tweaks needed to newly installed GTX650ti cards...
Message board moderation
Previous · 1 · 2 · 3 · 4 · 5 · Next
Author | Message |
---|---|
_ Send message Joined: 15 Nov 12 Posts: 299 Credit: 9,037,618 RAC: 0 |
Actually, after looking over a bunch of my inconclusive tasks, it seems pretty clear that both cards are producing the inconclusive tasks. Not just one. Now I am not sure if I am going to switch the cords or not. Will keep everyone posted... |
juan BFP Send message Joined: 16 Mar 07 Posts: 9786 Credit: 572,710,851 RAC: 3,799 |
Inconclusives are normal we all get one from time to time, invalid on the other hand if validated on other host if another thing, we are talking about invalids not inconclusives and by one of the earlier post in this thread all your invalids comes from the GPU 2 only. |
_ Send message Joined: 15 Nov 12 Posts: 299 Credit: 9,037,618 RAC: 0 |
Inconclusives are normal we all get one from time to time, invalid on the other hand if validated on other host if another thing, we are talking about invalid not inconclusives and by one of the earlier post in this thread all your invalids comes from the GPU 2 only. Arg. Somewhere along the lines I got mixed up and thought we were talking about inconclusives. Sorry Juan, this has my head spinning. I was more worried about inconclusives than invalids due to the fact that I was getting almost 50 inconclusives in 24 hours. I have only gotten 3-4 invalids, which is less of a concern as I know they happen sometimes. I'm going to let this sit for a little while and see if the situation improves. I've changed the GPU's to only do 1 WU at a time for now. |
juan BFP Send message Joined: 16 Mar 07 Posts: 9786 Credit: 572,710,851 RAC: 3,799 |
No problem, on other hand 50 inconclusives is a high number for a single host for example in all my hosts i have now 110 inconclusives and not a single invalid. That high number "could" point something but is to early to say with confidence. Leave running with 1 WU at a time and look. BTW did you check your AV software? |
_ Send message Joined: 15 Nov 12 Posts: 299 Credit: 9,037,618 RAC: 0 |
No problem, on other hand 50 inconclusives is a high number for a single host for example in all my hosts i have now 110 inconclusives and not a single invalid. That high number "could" point something but is to early to say with confidence. Leave running with 1 WU at a time and look. BTW did you check your AV software? Antivirus software? No I haven't. Not a bad idea just to rule it out. |
Jeff Buck Send message Joined: 11 Feb 00 Posts: 1441 Credit: 148,764,870 RAC: 0 |
Inconclusives are normal we all get one from time to time, invalid on the other hand if validated on other host if another thing, we are talking about invalid not inconclusives and by one of the earlier post in this thread all your invalids comes from the GPU 2 only. In my first post, I mentioned that all of your Invalids seemed to be coming from Device 2, and those had wildly different counts from your wingmen. My later post was about what I thought was the odd Spike counts I saw in the handful of Inconclusives I sampled, where your Spike counts were always just slightly higher than your wingman's. Whether these will be validated or not, I don't really know. I just looked at a few more and it appears that all of those with just slightly elevated counts are coming from Device 1. On the other hand, I also noticed you have an Inconclusive, Task 3339789622, which overflowed with a Spike count of 30. This one is from Device 2 and fits the pattern of your earlier Invalids, so I think it's highly likely that this will eventually end up in the Invalid column also. You might want to take the time to look through all your Inconclusives and see if the Device 1/Device 2 pattern holds all the way through. You'll need to do your own research to see if that holds true for all of your Inconclusives. By the way, I also noticed that one of your newest Inconclusives, Task 3339800599 actually has an empty Stderr. Truncated Stderr files are being discussed in another thread, but all the examples so far posted there have to do with "-9 overflow" tasks. |
Philhnnss Send message Joined: 22 Feb 08 Posts: 63 Credit: 30,694,327 RAC: 162 |
Using that ASUS, "GPU Tweek" program, that should be on the disk that came with your cards, bump up the voltage on both cards until you get a flat line going on the moniter graph. For me I had to bump it up from 1100 to 1125. Not claiming to be an expert but it seems logical that a video card would like it better being fed a smooth consistant amount of power. V/S the peaks and valley's shown on that graph at 1100. In my case it also made the cards run a small amount cooler running them at a consitant voltage. I also run my fans at 75% all the time. Once you find the sweet spot's you just save it as your profile. Then every time you start your computer it will load. Or you can go in and set it every time if you want. The only other thing I can ask is you did install CUDA 5.5 or 5.035 from NVIDA's site right? I ask because it's not on your list of things you did. |
_ Send message Joined: 15 Nov 12 Posts: 299 Credit: 9,037,618 RAC: 0 |
Using that ASUS, "GPU Tweek" program, that should be on the disk that came with Thanks Phil for your input on this. In general, the voltage does not jump around much at all. While running two WU's yesterday while I was at work, it seemed that the voltages changed some over the course of the day. But from one minute to the next, the voltages stayed smooth. I am only running 1 WU at a time now until I can see whats going on, and this morning the voltages are running smooth again too. I am going to leave GPU Tweak going all day to see if the voltages change as the day goes on. What was happening to your WUs before you adjusted the voltage? A lot of invalids, or inconclusives? As for CUDA 5.5 vs 5.035... I am not sure. All I can remember doing is getting the latest drivers from their site, the same drivers you have. Can you provide a link in regards to the 5.5 vs 5.035? I simply can't remember making any choice in this department. Is there a way to tell? Which did you end up picking? |
juan BFP Send message Joined: 16 Mar 07 Posts: 9786 Credit: 572,710,851 RAC: 3,799 |
Did you switch Power cables? Did you try to switch the GPUs from the PCI-e slots? Look your inconclusives on the GPU2 most of them has a large number os spikes and on the GPU1 none. You need to know if is not a hardware related problem, related to an specific GPU or power rail. |
_ Send message Joined: 15 Nov 12 Posts: 299 Credit: 9,037,618 RAC: 0 |
Did you switch Power cables? Did you try to switch the GPUs from the PCI-e slots? Look your inconclusives on the GPU2 most of them has a large number os spikes and on the GPU1 none. You need to know if is not a hardware related problem, related to an specific GPU or power rail. I have not switched any of those things yet. Perhaps I should, but I am going maybe a bit too slow in this process trying to understand what is happening :) Out of the last ten inconclusive WUs, 6 were from Device 2, and 4 were from Device 1. It doesn't overly seem to be ONE card or ONE power cable that is causing the problems. This makes me hesitate to switch things around. |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14650 Credit: 200,643,578 RAC: 874 |
Did you switch Power cables? Did you try to switch the GPUs from the PCI-e slots? Look your inconclusives on the GPU2 most of them has a large number os spikes and on the GPU1 none. You need to know if is not a hardware related problem, related to an specific GPU or power rail. An 'inconclusive' task merely represents a difference between the result you reported, and the result the other user (your 'wingmate') reported. It doesn't - at this stage - mean there is any problem with your GPU or computer. You can often make an informed guess as to which of the two results is going to be marked invalid in the end. If your task runs for roughly the length of time that it was estimated to need at the beginning, and your wingmate only runs it for 20 seconds or less, then most likely it's your wingmate that has the problem. Conversely, if your wingmate runs the task to full term, and yours quits early, then it's likely that you have the problem. With time and experience, you can perform a deeper analysis than that, but a simple early check like that can help to put your mind at rest. |
Wiggo Send message Joined: 24 Jan 00 Posts: 34744 Credit: 261,360,520 RAC: 489 |
mherr170, what is the make and model of your PSU? Cheers. |
_ Send message Joined: 15 Nov 12 Posts: 299 Credit: 9,037,618 RAC: 0 |
mherr170, what is the make and model of your PSU? Hi Wiggo! It is a Dell D750E-00 (I think!) |
_ Send message Joined: 15 Nov 12 Posts: 299 Credit: 9,037,618 RAC: 0 |
Did you switch Power cables? Did you try to switch the GPUs from the PCI-e slots? Look your inconclusives on the GPU2 most of them has a large number os spikes and on the GPU1 none. You need to know if is not a hardware related problem, related to an specific GPU or power rail. Thanks for that explanation Richard. Perhaps I am jumping the gun on the worrying, but I had never had so many inconclusives so soon on any of my hosts before. Right now I only have two invalids... so maybe it is just a matter of time to see where those inconclusives land... |
Philhnnss Send message Joined: 22 Feb 08 Posts: 63 Credit: 30,694,327 RAC: 162 |
https://developer.nvidia.com/cuda-toolkit-archive 5.5 is the newest release but when I went to install it, my computer would not run it. I do not remember what reason it gave. I then downloaded 5.035, the OCT 2013 release from the list in the above link, and it ran like it should. You will probably have to re-install you video driver after you install CUDA. CUDA comes with a video driver but it will probably be an old one. Do a clean install both times. Adjusting the voltage is just something I have done since my Nvidia 250 days. Did everything I could to make those cards run as cool as possible. So now it is just an automatic thing I do. My voltage seems to stutter at 1100. But at 1125 it is a flat line. |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14650 Credit: 200,643,578 RAC: 874 |
https://developer.nvidia.com/cuda-toolkit-archive Excuse me, but may I ask why you're suggesting that regular users install developer tools? (you've done it four posts in a row now) If all you want to do is to run somebody else's programs (like 99.9% of the readers here), everything you need - including the cuda runtime - is included in the regular driver package downloadable from http://www.nvidia.com/, http://www.nvidia.co.uk/, or your local equivalent. The only place they might be missing is the driver upgrades offered via Microsoft Update. Developer resources are only needed by, err, program developers. Programmers, in other words. |
_ Send message Joined: 15 Nov 12 Posts: 299 Credit: 9,037,618 RAC: 0 |
I kind of had the same thought as Richard. Is there something indeed worthwhile in the developer's package? |
Philhnnss Send message Joined: 22 Feb 08 Posts: 63 Credit: 30,694,327 RAC: 162 |
OK, never mind. It is just something I have always done. |
_ Send message Joined: 15 Nov 12 Posts: 299 Credit: 9,037,618 RAC: 0 |
OK, never mind. It is just something I have always done. No worries, I appreciate the input! |
Batter Up Send message Joined: 5 May 99 Posts: 1946 Credit: 24,860,347 RAC: 0 |
I remember I had to install a "developers package" to get some software to work, not SETI but isn't SETI and BOINC always in development? I will say though less is more when it comes crunching, one will get more if one only installs what is needed to crunch. |
©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.