With 3 the load is lower. And doesn't PCIE x16 give plenty of bandwidth? It's not like I'm running 2 GPU's and the PCIE drops to 8x. I've been told on GPUGRID that win7 is about 11% slower than winXP and linux with their CUDA tasks because of the way win7 handles it. I think it's likely that's the cause here.
to my experience this is mostly due to all that nifty crap W7 has by default.
as for einstein - yes, the linux-app running full CPU is a lot faster. on other projects like PG, collatz, GPUgrid it's just one or the other.
Quote:
It depends a bit on what else I'm doing with the PC or the kind of CPU tasks that Im running. But if I leave it alone 4 WU's running together will complete in about 90-100 minutes.
that's probably as fast as you can get right now. of course it would be much better if the app would be improved and simply running a single WU would fully utilize the GPU.
Quote:
Read somewhere it could be caused by a lack of power but my PSU is 850 watt which should be plenty for my system with just 1 GPU, and besides the GPU regularly gets stressed more while gaming (especially the witcher 2 but damn it looks awesome!).
that's silly! if the PSU would not be able to feed you host, you'd have real crashes.
Quote:
I also want to thank everyone for their feedback and advice, much appreciated!
Hi - How do you find out your GPU load?
I have a GTX 460 1GB videa card and I want to see how much it's being used.
I also don't know how to change how many WU's it's doing at one time.
I was trying to figure that out from your app_info file.
Where is the manual/docs for the app_info file?
I'm running windows 7 X64 on Intel Core 2 Q9300, 4 MB memory,
Got the GPU-Z and GPU Monitor tools and can see what's going on now, though I don't really understand what's happening in a good vs. bad sense, "looks" good since it's running 3 GPU tasks instead of 1, but not sure how to tell until they finish if they are overall more throughput.
The GPU did seem to crash on me once so far, then went comp error one of the tasks.
FYI - I lost a lot of work when I did this fooling around with the app+info file, since I didn't know what I was doing. Sort of wiped out what I did and downloaded new stuff a bunch of times, but I think I have the hang of it as I'm changing things now in the app_info.xml file...
Why does BOINC delete a lot of the files/executables when it either doesn't think there's work etc.? Not sure why I keep having to copy stuff back. Not sure if it's because I have things in my app_info file that aren't currently being worked on??
So now what I have is: an Intel X9300 Quad core on GByte Mobo,at approx 2.5 Ghz, DDR2 PC2-6400 Mem (4 GBytes), NVIDIA GTX 460 1 GB overclocked quite a bit (ASUS card).
I am running 4 tasks on the CPU cores, and 3 cuda tasks on the single GTX 460. I have 1.05 and 1.07 BPS apps running simultaneously!! (.2 CPU + .33 NVIDIA GPU each says, 2 of the 3 are high priority.)
Just as an info, I run now 3 WUs at a time on a GTX 460, 4 on CPU cores:
I see all 4 CPU cores at 100%, and 2.2 of 4 GB Main memory being used.
GPU Monitor gives me temp of 97 deg F, GPU load of approx 92-98%, Vmem is approx 55-58%.
Ah! Using GPU-Z:
- GPU Load: +/- 91-99%
- Memory used: +/- 993 MB MAX
- Mem Ctrl: 59% MAX
- T°: 36°C
- Fan speed : 20%/1380 RPM
Should I be trying to get another task on the GPU to get it to 100% utilization?
I think the answer is NO, because not enough memory, right, already close to 100% GPU use as well?
How do I know if my CPUs being pegged are hurting feeding the GPU. I guess right now that, and knowing if it's truly faster are my issues??
I can wait to see how much work it's doing say tomorrow and how fast they're going.
Do you guys have some way you're benchmarking because just counting how many CUDA tasks finished in a certain amount of time manually?
I keep getting msgs about the one of the CUDA tasks not being able to run (hasn't really started yet) say every 30 secs or so that says not enough CUDU (3 others are already trying to run which I think is why). Not sure why it does this.
Why does BOINC delete a lot of the files/executables when it either doesn't think there's work etc.? Not sure why I keep having to copy stuff back. Not sure if it's because I have things in my app_info file that aren't currently being worked on??
This is 'locality scheduling' in action ( our server is chatting with your rig upon contact and deciding what to do next, preferably with what data files you already have ). See how it goes : I'd suggest it is just settling down with your new setup and BOINC specifically has no work ( project wide ) in the part of the parameter space that you held files for, alternatively it could be that others in the same part of the search space as yourself are chewing through quickly too .... Richard?
Quote:
Should I be trying to get another task on the GPU to get it to 100% utilization? I think the answer is NO, because not enough memory, right, already close to 100% GPU use as well?
How do I know if my CPUs being pegged are hurting feeding the GPU. I guess right now that, and knowing if it's truly faster are my issues??
Your GPU is as fully utilised as possible, the CPUs are fine ( no harm ) and would be doing things other than feeding the GPU.
Quote:
Do you guys have some way you're benchmarking because just counting how many CUDA tasks finished in a certain amount of time manually?
That's the way I do it, by checking the runtimes via BOINC 'Projects' tab -> select E@H -> hit 'Your computers' -> on the web page that appears in your browser select 'tasks' for the relevant rig -> the subsequent web page shows the 'Run time' ( wall clock ) and 'CPU time' ( thread time ).
Quote:
I keep getting msgs about the one of the CUDA tasks not being able to run (hasn't really started yet) say every 30 secs or so that says not enough CUDU (3 others are already trying to run which I think is why). Not sure why it does this.
In the queue waiting their turn. After a while BOINC will adjust WU requests per your demonstrated load & this will settle down. You don't need to manage that.
Good to see you're having fun! :-)
Cheers, Mike.
I have made this letter longer than usual because I lack the time to make it shorter ...
... and my other CPU is a Ryzen 5950X :-) Blaise Pascal
I can't get any of the app's to work on my GTX 580 & Win XP to try and run 4 Wu's at once, all I keep getting is the following message:
[Einstein@Home] [error] State file error: missing application einsteinbinary_BRP4
[Einstein@Home] [error] Can't handle workunit in state file
Once I start BOINC the Einstein GPU Wu's get wiped out and no more is sent no matter how I configure the app file or what dll's I put in the Einstein project folder.
I've been trying to use the one claggy posted with no success:
EDIT: I didn't lose any Wu's though as I copied the entire BOINC Folder before messing around with an app file, I also disabled the Network activity for BOINC ... :)
einstein_S5GC1HF
Global Correlations S5 HF search #1
RE: With 3 the load is
)
to my experience this is mostly due to all that nifty crap W7 has by default.
as for einstein - yes, the linux-app running full CPU is a lot faster. on other projects like PG, collatz, GPUgrid it's just one or the other.
that's probably as fast as you can get right now. of course it would be much better if the app would be improved and simply running a single WU would fully utilize the GPU.
that's silly! if the PSU would not be able to feed you host, you'd have real crashes.
HTH!
Hi - How do you find out your
)
Hi - How do you find out your GPU load?
I have a GTX 460 1GB videa card and I want to see how much it's being used.
I also don't know how to change how many WU's it's doing at one time.
I was trying to figure that out from your app_info file.
Where is the manual/docs for the app_info file?
I'm running windows 7 X64 on Intel Core 2 Q9300, 4 MB memory,
RE: Where is the
)
Anonymous platform
RE: Where is the
)
The app_info.xml file is part of the anonymous platform mechanism.
BOINC FAQ Service
Official BOINC wiki
Installing BOINC on Linux
RE: Hi - How do you find
)
run GPU Z, download it from HERE
You guys are AWESOME! Thanks
)
You guys are AWESOME! Thanks so much!!
Got the GPU-Z and GPU Monitor tools and can see what's going on now, though I don't really understand what's happening in a good vs. bad sense, "looks" good since it's running 3 GPU tasks instead of 1, but not sure how to tell until they finish if they are overall more throughput.
The GPU did seem to crash on me once so far, then went comp error one of the tasks.
FYI - I lost a lot of work when I did this fooling around with the app+info file, since I didn't know what I was doing. Sort of wiped out what I did and downloaded new stuff a bunch of times, but I think I have the hang of it as I'm changing things now in the app_info.xml file...
Why does BOINC delete a lot of the files/executables when it either doesn't think there's work etc.? Not sure why I keep having to copy stuff back. Not sure if it's because I have things in my app_info file that aren't currently being worked on??
So now what I have is: an Intel X9300 Quad core on GByte Mobo,at approx 2.5 Ghz, DDR2 PC2-6400 Mem (4 GBytes), NVIDIA GTX 460 1 GB overclocked quite a bit (ASUS card).
I am running 4 tasks on the CPU cores, and 3 cuda tasks on the single GTX 460. I have 1.05 and 1.07 BPS apps running simultaneously!! (.2 CPU + .33 NVIDIA GPU each says, 2 of the 3 are high priority.)
Just as an info, I run now 3 WUs at a time on a GTX 460, 4 on CPU cores:
I see all 4 CPU cores at 100%, and 2.2 of 4 GB Main memory being used.
GPU Monitor gives me temp of 97 deg F, GPU load of approx 92-98%, Vmem is approx 55-58%.
Ah! Using GPU-Z:
- GPU Load: +/- 91-99%
- Memory used: +/- 993 MB MAX
- Mem Ctrl: 59% MAX
- T°: 36°C
- Fan speed : 20%/1380 RPM
Should I be trying to get another task on the GPU to get it to 100% utilization?
I think the answer is NO, because not enough memory, right, already close to 100% GPU use as well?
How do I know if my CPUs being pegged are hurting feeding the GPU. I guess right now that, and knowing if it's truly faster are my issues??
I can wait to see how much work it's doing say tomorrow and how fast they're going.
Do you guys have some way you're benchmarking because just counting how many CUDA tasks finished in a certain amount of time manually?
I keep getting msgs about the one of the CUDA tasks not being able to run (hasn't really started yet) say every 30 secs or so that says not enough CUDU (3 others are already trying to run which I think is why). Not sure why it does this.
RE: Why does BOINC delete a
)
This is 'locality scheduling' in action ( our server is chatting with your rig upon contact and deciding what to do next, preferably with what data files you already have ). See how it goes : I'd suggest it is just settling down with your new setup and BOINC specifically has no work ( project wide ) in the part of the parameter space that you held files for, alternatively it could be that others in the same part of the search space as yourself are chewing through quickly too .... Richard?
Your GPU is as fully utilised as possible, the CPUs are fine ( no harm ) and would be doing things other than feeding the GPU.
That's the way I do it, by checking the runtimes via BOINC 'Projects' tab -> select E@H -> hit 'Your computers' -> on the web page that appears in your browser select 'tasks' for the relevant rig -> the subsequent web page shows the 'Run time' ( wall clock ) and 'CPU time' ( thread time ).
In the queue waiting their turn. After a while BOINC will adjust WU requests per your demonstrated load & this will settle down. You don't need to manage that.
Good to see you're having fun! :-)
Cheers, Mike.
I have made this letter longer than usual because I lack the time to make it shorter ...
... and my other CPU is a Ryzen 5950X :-) Blaise Pascal
Oh nvm, it was F and not C.
)
Oh nvm, it was F and not C. Ignore my post.
I can't get any of the app's
)
I can't get any of the app's to work on my GTX 580 & Win XP to try and run 4 Wu's at once, all I keep getting is the following message:
[Einstein@Home] [error] State file error: missing application einsteinbinary_BRP4
[Einstein@Home] [error] Can't handle workunit in state file
Once I start BOINC the Einstein GPU Wu's get wiped out and no more is sent no matter how I configure the app file or what dll's I put in the Einstein project folder.
I've been trying to use the one claggy posted with no success:
EDIT: I didn't lose any Wu's though as I copied the entire BOINC Folder before messing around with an app file, I also disabled the Network activity for BOINC ... :)
einstein_S5GC1HF
Global Correlations S5 HF search #1
einstein_S5GC1HF_3.06_windows_intelx86__S5GCESSE2.exe
einstein_S5R6_3.01_graphics_windows_intelx86.exe
einstein_S5GC1HF
306
windows_intelx86
1.000000
1.000000
S5GCESSE2
6.13.0
einstein_S5GC1HF_3.06_windows_intelx86__S5GCESSE2.exe
einstein_S5R6_3.01_graphics_windows_intelx86.exe
graphics_app
einsteinbinary_BRP3
Binary Radio Pulsar Search
einsteinbinary_BRP3_1.05_windows_intelx86__BRP3SSE.exe
einsteinbinary_BRP3_1.07_windows_intelx86__BRP3cuda32.exe
einsteinbinary_BRP3_1.00_graphics_windows_intelx86.exe
cudart_xp32_32_16.dll
cufft_xp32_32_16.dll
db.dev.win.3d35195e
dbhs.dev.win.3d35195e
einsteinbinary_BRP3
105
windows_intelx86
1.000000
1.000000
BRP3SSE
6.13.0
einsteinbinary_BRP3_1.05_windows_intelx86__BRP3SSE.exe
einsteinbinary_BRP3_1.00_graphics_windows_intelx86.exe
graphics_app
einsteinbinary_BRP3
107
windows_intelx86
0.200000
1.000000
BRP3cuda32
6.13.0
einsteinbinary_BRP3_1.07_windows_intelx86__BRP3cuda32.exe
cudart_xp32_32_16.dll
cudart32_32_16.dll
cufft_xp32_32_16.dll
cufft32_32_16.dll
einsteinbinary_BRP3_1.00_graphics_windows_intelx86.exe
graphics_app
db.dev.win.3d35195e
db.dev
dbhs.dev.win.3d35195e
dbhs.dev
CUDA
1.000000
314572800.000000
einstein_S6Bucket
Gravitational Wave S6 GC search
einstein_S6Bucket_1.01_windows_intelx86__SSE2.exe
einstein_S5R6_3.01_graphics_windows_intelx86.exe
einstein_S6Bucket
101
windows_intelx86
1.000000
1.000000
SSE2
6.13.0
einstein_S6Bucket_1.01_windows_intelx86__SSE2.exe
einstein_S5R6_3.01_graphics_windows_intelx86.exe
graphics_app
I finally got the app file
)
I finally got the app file figured out, running 6 @ a time on a Dual GTX 580 Setup, will have to see what the times are later ...