The FGRP searh works fine here, approx same crunch time,as Binary Radio Pulsar Search on my GPU,
Great! That's what we are testing here.
Quote:
Cred payment is to low on this project!!!
This is a test project that may even be used to test crediting, which might or might not work as you expect.
What else can we do other than to write
Quote:
If you care about credit, this project is certainly not the right one for you.
on the front page (and AFAIK even in the welcome message in external account managers)?
BM
Hi again
I am aware of the cred info, even : Most of the time Albert@Home will have no work to run at all, and when it has, the applcations are experimental, might be unstable, unreliable and may even damage your computer..-)
I ment to say; when having Fix cred payment! why so low ?
You can set it to 0, but then Testing out the project will take ages..
Cred are a ´bonus´ for me while helping out project with testing !
I´ve been participated in several Alpha and Beta testing project, so this is not something new to me
I've got to say I'm quite impressed with the Nvidia OpenCL app! If I compare the times I get here to the times the CPU-app produce over at Einstein the result is:
The FGRP CPU-app over at Einstein takes about 5700-6000s per task.
The Nvidia OpenCL app here takes about 1300s per task.
I also believe this app has the potential to be much better as the current version use a full CPU core and only loads my GTX660Ti to about 30% and that load comes in bursts. Hopefully this can be improved in future releases.
FGRP search CPU only (via Einstein): 3930K 4.4 GHz No HT, 2850 sec
So far there have been no failed tasks due to the new application. I did have one task error out but that was due to something missing in my Linux image. The FGRP tasks take around 35% longer than the BRP4 tasks to complete but complete much quicker than FGRP search on CPUs alone. I did not have GPU usage via Linux NVIDIA as nvidia-smi does not report usage for Geforce cards.
Should Nvidia cards with 512Mb RAM receive work with this app? My ATI host has done quite a few units, but so far I haven't seen any on my Nvidia host (8800GT + 9600GT, each with 512Mb RAM, nvidia driver 295.75). The systems are essentially identical except for the video card.
2013-03-27 05:29:44.3316 [PID=32301] [version] Checking plan class 'FGRPopencl-nvidia'
2013-03-27 05:29:44.3316 [PID=32301] [version] NVidia device (or driver) doesn't support OpenCL
In principle, the devices support OpenCL, so I suspect the difficulty lies in the driver.
I didn't want to spam the boards with my stats - just milestone theads - but apparently signatures are no longer optional. Follow the link if you're interested.
I also believe this app has the potential to be much better as the current version use a full CPU core and only loads my GTX660Ti to about 30% and that load comes in bursts. Hopefully this can be improved in future releases.
Keep up the good work!
I doubt there will much improvement for Nvidia until they come up with better OpenCl driver support.
If I understand that correctly, the app is requesting RAM of at least 512Mb but the card is reporting only 511.8125Mb available. The app only needs "300+MB Free" (post #1 in this thread) so is there any scope for a bit of tolerance here? :)
RE: RE: The FGRP searh
Hi again
I am aware of the cred info, even : Most of the time Albert@Home will have no work to run at all, and when it has, the applcations are experimental, might be unstable, unreliable and may even damage your computer..-)
I ment to say; when having Fix cred payment! why so low ?
You can set it to 0, but then Testing out the project will take ages..
Cred are a ´bonus´ for me while helping out project with testing !
I´ve been participated in several Alpha and Beta testing project, so this is not something new to me
PS Bernd, like your haircut :)
I've got to say I'm quite
I've got to say I'm quite impressed with the Nvidia OpenCL app! If I compare the times I get here to the times the CPU-app produce over at Einstein the result is:
The FGRP CPU-app over at Einstein takes about 5700-6000s per task.
The Nvidia OpenCL app here takes about 1300s per task.
I also believe this app has the potential to be much better as the current version use a full CPU core and only loads my GTX660Ti to about 30% and that load comes in bursts. Hopefully this can be improved in future releases.
Keep up the good work!
Hi gpu1 0-30% gpu2 0% load.
Hi gpu1 0-30% gpu2 0% load. Suspend gpu1 still 0-30% load.It seems that gpu2 working on gpu1.
I am very glad to see that
I am very glad to see that FGRP search has been ported to GPUs. Thanks for developing the GPU applications for FGRP search.
Here are some results I have seen on different hardware and OS.
Tasks running per GPU: 1
OS, Driver, Card, CPU Frequency, GPU Usage, Runtime, PCI-E
Linux, 12.11 Beta 11, 7970, 3930K 4.4 GHz, 55%, 955 sec, x16 3.0
Linux, 295.33, GTX 580, 920 4.2 GHz, NA, 1180 sec, x16 2.0
Linux, 295.33, GTX 580, 920 4.2 GHz, NA, 1280 sec, x8 2.0
XP 64, 314.07, GTX 680, 920 4.2 GHz, 0-30%, 1065-1157 sec, x16 2.0
FGRP search CPU only (via Einstein): 3930K 4.4 GHz No HT, 2850 sec
So far there have been no failed tasks due to the new application. I did have one task error out but that was due to something missing in my Linux image. The FGRP tasks take around 35% longer than the BRP4 tasks to complete but complete much quicker than FGRP search on CPUs alone. I did not have GPU usage via Linux NVIDIA as nvidia-smi does not report usage for Geforce cards.
Should Nvidia cards with
Should Nvidia cards with 512Mb RAM receive work with this app? My ATI host has done quite a few units, but so far I haven't seen any on my Nvidia host (8800GT + 9600GT, each with 512Mb RAM, nvidia driver 295.75). The systems are essentially identical except for the video card.
From the server log for your
From the server log for your Nvidia host http://albert.phys.uwm.edu/host_sched_logs/5/5805:
In principle, the devices support OpenCL, so I suspect the difficulty lies in the driver.
I didn't want to spam the boards with my stats - just milestone theads - but apparently signatures are no longer optional. Follow the link if you're interested.
http://www.boincsynergy.com/images/stats/comb-3475.jpg
Fab, thanks! (and I'll know
Fab, thanks! (and I'll know where to look in future).
The drivers are claimed to support OpenCL but maybe a more recent version is needed - I'll update the host and see if that fixes it.
RE: I also believe this
I doubt there will much improvement for Nvidia until they come up with better OpenCl driver support.
RE: From the server log for
After upgrading to 310.40 (latest), the server log is now showing this:-
If I understand that correctly, the app is requesting RAM of at least 512Mb but the card is reporting only 511.8125Mb available. The app only needs "300+MB Free" (post #1 in this thread) so is there any scope for a bit of tolerance here? :)
RE: Hi gpu1 0-30% gpu2 0%
same thing for me. HD7870s in crossfire
13.3 BETA 3