. . The project has been down for maintenance in the last 12 hours (started approx 9 hours ago not sure when it finished but has been back up for at least 3 hours) so perhaps you were trying during that time.

. . I would give it another try now. It is an often recommended practice to make sure your WU cache is filled before maintenance begins.

. . Stephen

.

Works if you get enough advance notice of the maintenance. I didn't.

One computer automatically got more; the other will take longer to be ready for more.

. . It is called Tuesday maintenance because it is performed on Tuesdays :) It happens every week. Most people don't even notice unless they are heavy hitters processing more tasks in a day than the system will let you cache. Or unless for some reason they find themselves with an abnormally low cache right when the maintenance begins. So on Tuesday mornings it is a good idea to stock your cache to the full. Just to avoid disappointment. :)

.

. . I will add those commands to the command line to be sure they are running, should I use -use_sleep_ex of 1?

.

-use_sleep has same action as -use_sleep_ex 1

. . I am impressed, it has resulted in a very optimal level of GPU utilisation, it couldn't get much better, which precludes running multiple tasks on the GPU now.

. . My command line now is :-

-high_perf_timer -use_sleep_ex 1 -sbs 384 (the card seems to like that) -tt 120 -hp

. . As I said the card could not be running any significantly better and the runtimes are now down to 38 or 39 mins, and pretty consistently. There is no issue with lag (99% of the time) so it is just about all you could ask of it.

. . Is there anything else you would like me to try or test for you? The next stage is to queue up a mass of guppis and let it go on them. But it is running well and looks good.

. . Stephen

Works if you get enough advance notice of the maintenance. I didn't.

It happens every week.
It is announced every week on the main page as News.

And the announcement has just gone up (it's triggered automatically at midnight, Berkeley time) for Wednesday 31 August. Could be Wednesday next week as well - often delayed by the Labor Day holiday on Monday - but we'll see if it returns to the normal Tuesday after that.

. . Is there anything else you would like me to try or test for you? The next stage is to queue up a mass of guppis and let it go on them. But it is running well and looks good.

. . Stephen

to watch lag on VLAR tasks with -tt 60 and -tt 120

---

SETI apps news
We're not gonna fight them. We're gonna transcend them.

. . Is there anything else you would like me to try or test for you? The next stage is to queue up a mass of guppis and let it go on them. But it is running well and looks good.

. . Stephen

to watch lag on VLAR tasks with -tt 60 and -tt 120

. . I was running -tt 60 on the Arecibo tasks with absolutley no discernible lag or issues, I am running -tt 120 now on the guppi tasks and there is no lag so far when typing with just the barest hint of stickiness on the BOINC manager slide control. I will change back to -tt 60 and see how that behaves. But I guess it would be good to have feedback by someone with an even lower order GPU than mine just to be 100% sure. Perhaps a GT710 or GT620 or such, maybe even a GT210 to really test the limits, though that may be pre-fermi.

. . But kudos to you Raistmer, even crunching guppis the GPU utilisation remains high, much better than with CUDA50. And while r3500 still loses out to CUDA50 in Arecibo runtimes (32/35 mins compared to SoG which is down to 38/39 mins with the new settings) for Guppi WUs SoG is a clear winner, down from 55 to 70 mins for CUDA50 to a very impressive 46 to 49 mins for SoG. That is a big win. Take a bow my friend. And not only much quicker but seemingly more consistent as well, I have only crunched about 10 units but they all have been in that same area, though they may be of the same batch. When I get some varied guppi WUs I will try it again. I will also change back to -tt 60 and remove -hp and see if there is much difference (if any).

. . Stephen

.

For me, -tt 60 -high_prec_timer -use_sleep for NV:

14 marked invalid, all on the GTX 560 under Windows Vista. Any suggestions for making those run better? And should I use those suggestions only for the GTX 560?

27 marked inconclusive, all but one on the GTX 560.

67 pending, mostly from the GTX 560.

58 valid.

For me, -tt 60 -high_prec_timer -use_sleep for NV:

14 marked invalid, all on the GTX 560 under Windows Vista. Any suggestions for making those run better? And should I use those suggestions only for the GTX 560?

27 marked inconclusive, all but one on the GTX 560.

67 pending, mostly from the GTX 560.

58 valid.

. . The first question that comes to mind is when did you last update your drivers? If not for a while then maybe try that.

. . Also do you have anything monitoring the GPU temps? They can chuck out errors if they are running over temp. An extra fan if the case supports one can do wonders. Otherwise any fan you can jury rig to move more air through the case. The manufacturers claim they are OK up to over 90 deg, personally I get stressed if the temps go over 65, because they will be constant sustained temps I like to keep them under 60 wherever possible.

. . I believe one of the other guys suggested trying something like Afterburner (which will also monitor the temps for you) to wind your GPU clock back a little and see if it is more stable then. Some overclocked boards will play up with crunching.

. . Your error rate sure indicates something is not right.

Stephen

to watch lag on VLAR tasks with -tt 60 and -tt 120

. . But kudos to you Raistmer, even crunching guppis the GPU utilisation remains high, much better than with CUDA50. And while r3500 still loses out to CUDA50 in Arecibo runtimes (32/35 mins compared to SoG which is down to 38/39 mins with the new settings) for Guppi WUs SoG is a clear winner, down from 55 to 70 mins for CUDA50 to a very impressive 46 to 49 mins for SoG. That is a big win. Take a bow my friend. And not only much quicker but seemingly more consistent as well, I have only crunched about 10 units but they all have been in that same area, though they may be of the same batch. When I get some varied guppi WUs I will try it again. I will also change back to -tt 60 and remove -hp and see if there is much difference (if any).

. . Stephen

.

. . HI again,

. . I changed the command line to:

-high_prec_timer -use_sleep -sbs 384 -tt 60

. . There is still no issue with lag or hesitations, to test it I watched a YouTube video while it was crunching and it played better than many I have watched on a non crunching machine. That to me is a pretty convincing test. Runtimes are only marginally affected perhaps a minute longer per WU, and that could be subjective. I even moved a couple of Arecibo VLAT tasks to the GPU queue and ran them. The first one was a disappointment and bombed out :-

http://setiathome.berkeley.edu/result.php?resultid=5129831539

. . but the second one ran AOK, so I am still impressed. But, like on the i5, runtimes were way longer than for guppis so for me they will stay on the CPU. Guppis taking about 49 mins Arecibo VLAR took over 70 mins.

. . Now I will begin trying the tuning lines to see if there is any more to be gained from r3500. If I can squeeze a little more out of it I may stay with it on the little donkey.

Stephen

For me, -tt 60 -high_prec_timer -use_sleep for NV:

14 marked invalid, all on the GTX 560 under Windows Vista. Any suggestions for making those run better? And should I use those suggestions only for the GTX 560?

27 marked inconclusive, all but one on the GTX 560.

67 pending, mostly from the GTX 560.

58 valid.

. . The first question that comes to mind is when did you last update your drivers? If not for a while then maybe try that.

. . Also do you have anything monitoring the GPU temps? They can chuck out errors if they are running over temp. An extra fan if the case supports one can do wonders. Otherwise any fan you can jury rig to move more air through the case. The manufacturers claim they are OK up to over 90 deg, personally I get stressed if the temps go over 65, because they will be constant sustained temps I like to keep them under 60 wherever possible.

. . I believe one of the other guys suggested trying something like Afterburner (which will also monitor the temps for you) to wind your GPU clock back a little and see if it is more stable then. Some overclocked boards will play up with crunching.

. . Your error rate sure indicates something is not right.

Stephen

For the 560, first to the last two drivers available for Windows Vista, then back to the last Windows Vista driver not known to have problems with OpenCL, weeks ago. Temperature monitored, currently 80 C.

For the 440, driver update today. Temperature monitor installed but the Windows 10 updates have made it stop working.

Vtune allowed me to reduce the 560 clock rates just now. Nothing similar currently installed for the 440, but does it need this?

For the 560, first to the last two drivers available for Windows Vista, then back to the last Windows Vista driver not known to have problems with OpenCL, weeks ago. Temperature monitored, currently 80 C.

That is hot. Very hot. It's within the acceptable limits, but it's still very hot.
I can't remember the exact temperatures, but when I had my GTX 560Ti (with a factory overclock) even with the ambient temperatures in the mid to high 30°c the video card was generally below 70°c (I think around 65°c with the fans at or above 75%).

---

Grant
Darwin NT

Your 560 is running far too hot for it to be reliable - it is churning out far more "signals" than it should be.
As for drivers - what source did you look at and only find two - the Nvidia archive lists twenty!
I would do a dust bunny hunt on that card, then take a look at the Nvidia site for something in the 358.xx region.

For the GTX440 windows update drivers - NEVER trust them (OK, trust them to wreck performance of whatever they are supposed to be driving). Again head off to the Nvidia site, find their driver for your particular version of Windows 10 (I just hope it isn't the anniversary downgrade).

---

Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?

For the 560, first to the last two drivers available for Windows Vista, then back to the last Windows Vista driver not known to have problems with OpenCL, weeks ago. Temperature monitored, currently 80 C.

For the 440, driver update today. Temperature monitor installed but the Windows 10 updates have made it stop working.

Vtune allowed me to reduce the 560 clock rates just now. Nothing similar currently installed for the 440, but does it need this?

. . OK that should rule out a driver issue but that si way tooooo hot in my opinion. Can you remove a cover or find a way to move a whole lot more air over that GPU card? It really needs to cool down. Winding back the GPU clock speed will help reduce temps.

.

For the 560, first to the last two drivers available for Windows Vista, then back to the last Windows Vista driver not known to have problems with OpenCL, weeks ago. Temperature monitored, currently 80 C.

That is hot. Very hot. It's within the acceptable limits, but it's still very hot.
I can't remember the exact temperatures, but when I had my GTX 560Ti (with a factory overclock) even with the ambient temperatures in the mid to high 30°c the video card was generally below 70°c (I think around 65°c with the fans at or above 75%).

. . Good point ... has he checked that the fans on the GPU are clean and working properly?

Your 560 is running far too hot for it to be reliable - it is churning out far more "signals" than it should be.
As for drivers - what source did you look at and only find two - the Nvidia archive lists twenty!
I would do a dust bunny hunt on that card, then take a look at the Nvidia site for something in the 358.xx region.

For the GTX440 windows update drivers - NEVER trust them (OK, trust them to wreck performance of whatever they are supposed to be driving). Again head off to the Nvidia site, find their driver for your particular version of Windows 10 (I just hope it isn't the anniversary downgrade).

The 560 is still running at 80 C with the lower frequencies. I suspect some software that is set to allow the temperature to reach 80 C before turning up the fan speed. A dust bunny hunt looks reasonable as soon as I can find an antistatic bag.

I get my Nvidia drivers only from the Nvidia site, but tend to use the latest versions not known to cause problems. I'm currently using 362.00 for the 560, since the later versions are either known to cause OpenCL problems or not set up to be able to install under Windows Vista. I didn't count their many versions earlier than 362.00. I'll try an earlier version if the dust bunny hunt doesn't help enough.

The 440 is under 64-bit Windows 10 Pro, with all the current updates. Is that what you call the anniversary downgrade? 372.70 driver.

Download and run either EVGA Precision X or MSI Afterburner and turn the fan/s up on that 560 so it's temps can come down and if the case is built like an oven then remove the side panel.

My GTX 560Ti will start to throw errors once it hits 74C.

Cheers.

Your 560 is running far too hot for it to be reliable - it is churning out far more "signals" than it should be.
As for drivers - what source did you look at and only find two - the Nvidia archive lists twenty!
I would do a dust bunny hunt on that card, then take a look at the Nvidia site for something in the 358.xx region.

For the GTX440 windows update drivers - NEVER trust them (OK, trust them to wreck performance of whatever they are supposed to be driving). Again head off to the Nvidia site, find their driver for your particular version of Windows 10 (I just hope it isn't the anniversary downgrade).

The 560 is still running at 80 C with the lower frequencies. I suspect some software that is set to allow the temperature to reach 80 C before turning up the fan speed. A dust bunny hunt looks reasonable as soon as I can find an antistatic bag.

I get my Nvidia drivers only from the Nvidia site, but tend to use the latest versions not known to cause problems. I'm currently using 362.00 for the 560, since the later versions are either known to cause OpenCL problems or not set up to be able to install under Windows Vista. I didn't count their many versions earlier than 362.00. I'll try an earlier version if the dust bunny hunt doesn't help enough.

The 440 is under 64-bit Windows 10 Pro, with all the current updates. Is that what you call the anniversary downgrade? 372.70 driver.

I looked inside. All fans turning, except possibly the one in the power supply, which I could not see. The main problem appeared to be dust blocking half the heatsink of the 560. Now blown away, and frequencies reset to the original. A possible second problem with the software controlling the graphics fan speed allowing too high a temperature before speeding up the fan. Let's see what that does before any other 560 change.

Not enough suggestions for the 440 yet to make any change for it.

I looked inside. All fans turning, except possibly the one in the power supply, which I could not see. The main problem appeared to be dust blocking half the heatsink of the 560. Now blown away, and frequencies reset to the original. A possible second problem with the software controlling the graphics fan speed allowing too high a temperature before speeding up the fan. Let's see what that does before any other 560 change.

Not enough suggestions for the 440 yet to make any change for it.

. . It was a matter of attacking the worst problem first.

. . Both EVGA Precision X and Afterburner will allow you to adjust your GPU fan profile. I suggest you look at that too and adjust to an aggressive curve with fan speed ramping up to at least 50% by the time the temps hit 50 C. And getting up to 100% by the time the temp hits 70 C. YOu really want that card running as cool as possible.

Stephen
.
.

I looked inside. All fans turning, except possibly the one in the power supply, which I could not see. The main problem appeared to be dust blocking half the heatsink of the 560. Now blown away, and frequencies reset to the original. A possible second problem with the software controlling the graphics fan speed allowing too high a temperature before speeding up the fan. Let's see what that does before any other 560 change.

Not enough suggestions for the 440 yet to make any change for it.

. . It was a matter of attacking the worst problem first.

. . Both EVGA Precision X and Afterburner will allow you to adjust your GPU fan profile. I suggest you look at that too and adjust to an aggressive curve with fan speed ramping up to at least 50% by the time the temps hit 50 C. And getting up to 100% by the time the temp hits 70 C. YOu really want that card running as cool as possible.

Stephen
.
.

I followed your suggestions tonight, with Afterburner. GPU temperature still at 74 C, but the fan is quieter. If that doesn't help, I'll see if I can uninstall Vtune next, in case it is responsible for the previous non-adjustable fan profile.

Vtune uninstalled, but the GPU temperature is still 72 C. I'll look for something else to try when validation catches up enough to tell if there are any invalids after this change.

the GPU temperature is still 72°C

TThrottle will let you limit the GPU temperature (by automatically making small (ms) pauses in the app computing threads if temperature reaches the limit set by you)

---

 


- ALF - "Find out what you don't do well ..... then don't do it!" :)