I'm trying to modify the app_info, but I'm not finding the old page where it was hosted (it would be easier to modify it than to create it). Anybody has an estimate of how much CPU consumes the ABP2 CUDA for a 32SP and 1 Intel CPU @ 2.5 GHz? I know that it will raise the priority level of the GPU, but I want to tinker with it.
So, can anybody post his app_info for ABP2 CUDA version?
I'm trying to modify the app_info, but I'm not finding the old page where it was hosted (it would be easier to modify it than to create it). Anybody has an estimate of how much CPU consumes the ABP2 CUDA for a 32SP and 1 Intel CPU @ 2.5 GHz? I know that it will raise the priority level of the GPU, but I want to tinker with it.
So, can anybody post his app_info for ABP2 CUDA version?
If you've ever run an ABP2 task on the machine in question, you'll find what you want in an section in client_state.xml
I'm trying to modify the app_info, but I'm not finding the old page where it was hosted (it would be easier to modify it than to create it). Anybody has an estimate of how much CPU consumes the ABP2 CUDA for a 32SP and 1 Intel CPU @ 2.5 GHz? I know that it will raise the priority level of the GPU, but I want to tinker with it.
So, can anybody post his app_info for ABP2 CUDA version?
If you've ever run an ABP2 task on the machine in question, you'll find what you want in an section in client_state.xml
Thanks, I didn't knew this, but I'm still getting:
17/02/2010 13:01:56 Einstein@Home Message from server: To get more Einstein@Home work, finish current work, stop BOINC, remove app_info.xml file, and restart.
17/02/2010 13:01:56 Einstein@Home Message from server: No work sent
17/02/2010 13:01:56 Einstein@Home Message from server: Your app_info.xml file doesn't have a version of Arecibo Binary Pulsar Search (STSP).
17/02/2010 13:01:56 Einstein@Home Message from server: Your app_info.xml file doesn't have a version of Hierarchical S5 all-sky GW search #6.
... I'm still getting:
17/02/2010 13:01:56 Einstein@Home Message from server: To get more Einstein@Home work, finish current work, stop BOINC, remove app_info.xml file, and restart.
17/02/2010 13:01:56 Einstein@Home Message from server: No work sent
17/02/2010 13:01:56 Einstein@Home Message from server: Your app_info.xml file doesn't have a version of Arecibo Binary Pulsar Search (STSP).
I presume this means that your app_info.xml is declaring that your host is available to do ABP2 work but you haven't properly specified the ABP2 app and version number that you will be using - in other words you don't have a fully valid app_info.xml file, as far as handling ABP2 work is concerned.
I don't understand why you feel you need to use the AP mechanism. The current official apps, both GW and ABP2, will run quite automatically without needing to use AP. With all the recent app and app_version changes, using AP would be a bit of a nightmare if you aren't fully aware of exactly how to manage it all. It would be far easier to let BOINC automatically download new apps and versions when they are rolled out.
Quote:
17/02/2010 13:01:56 Einstein@Home Message from server: Your app_info.xml file doesn't have a version of Hierarchical S5 all-sky GW search #6.
You will always get this message if you don't include GW (S5R6) capability in your app_info.xml. There's nothing unusual about this - it's the way things are designed to work.
I don't understand why you feel you need to use the AP mechanism. The current official apps, both GW and ABP2, will run quite automatically without needing to use AP. With all the recent app and app_version changes, using AP would be a bit of a nightmare if you aren't fully aware of exactly how to manage it all. It would be far easier to let BOINC automatically download new apps and versions when they are rolled out.
Well, since the ABP2 is using 1 CPU & 1 GPU, I want just to play with the CPU % to allow another task. I didn't want to say this from the beginning because about the increasing priority if CPU or GPU is under 100% . And since I didn't want to have people telling me that my priority would go up, thus the computer will become slow, thus I will have to be careful with the app_info, I asked for the old page where the app_info for CUDA was.
Bottom line, I asked for one simple redirect link, and got two not helpful solutions.
Thanks a lot.
Though I don't suggest the use of app_info.xml to manipulate CPU % assigned to 1 GPU (if it's ever possible), I do share cristipurdel's sentiment (if that is what is intended) of our current inefficient CUDA apps which require 1 full CPU core just to run 1 GPU CUDA task.
From what I've check on Hardware Monitor on my Mac, CUDA task uses 1 CPU core while the GPU core, although uses 1 GPU, it just hovers around 28 - 32%, just shy of a third of overall GPU capabilities. I have a 8800GT, but the performance is only 30 minutes short of GT120 on a full work unit of 3.5 hours. Performance-per-watt ratio guys, 1 CPU core + (1 CPU core + 1 GPU core) is barely faster than a full 2 CPU cores.
RE: I've got three of these
)
Nobody has got any answer to the above question...?
I've also noticed that my GPUs crunched slower alongside all CPU cores loaded, as compared to just crunching on GPUs only.
RE: RE: I've got three of
)
Team Linux Users Everywhere
RE: RE: RE: I've got
)
I'm trying to modify the
)
I'm trying to modify the app_info, but I'm not finding the old page where it was hosted (it would be easier to modify it than to create it). Anybody has an estimate of how much CPU consumes the ABP2 CUDA for a 32SP and 1 Intel CPU @ 2.5 GHz? I know that it will raise the priority level of the GPU, but I want to tinker with it.
So, can anybody post his app_info for ABP2 CUDA version?
RE: I'm trying to modify
)
If you've ever run an ABP2 task on the machine in question, you'll find what you want in an section in client_state.xml
RE: RE: I'm trying to
)
Thanks, I didn't knew this, but I'm still getting:
17/02/2010 13:01:56 Einstein@Home Message from server: To get more Einstein@Home work, finish current work, stop BOINC, remove app_info.xml file, and restart.
17/02/2010 13:01:56 Einstein@Home Message from server: No work sent
17/02/2010 13:01:56 Einstein@Home Message from server: Your app_info.xml file doesn't have a version of Arecibo Binary Pulsar Search (STSP).
17/02/2010 13:01:56 Einstein@Home Message from server: Your app_info.xml file doesn't have a version of Hierarchical S5 all-sky GW search #6.
RE: Thanks, I didn't know
)
What was it again, why you need an app_info.xml for Einstein? ;-)
Gruß,
Gundolf
Computer sind nicht alles im Leben. (Kleiner Scherz)
RE: ... I'm still
)
I presume this means that your app_info.xml is declaring that your host is available to do ABP2 work but you haven't properly specified the ABP2 app and version number that you will be using - in other words you don't have a fully valid app_info.xml file, as far as handling ABP2 work is concerned.
I don't understand why you feel you need to use the AP mechanism. The current official apps, both GW and ABP2, will run quite automatically without needing to use AP. With all the recent app and app_version changes, using AP would be a bit of a nightmare if you aren't fully aware of exactly how to manage it all. It would be far easier to let BOINC automatically download new apps and versions when they are rolled out.
You will always get this message if you don't include GW (S5R6) capability in your app_info.xml. There's nothing unusual about this - it's the way things are designed to work.
Cheers,
Gary.
RE: I don't understand why
)
Well, since the ABP2 is using 1 CPU & 1 GPU, I want just to play with the CPU % to allow another task. I didn't want to say this from the beginning because about the increasing priority if CPU or GPU is under 100% . And since I didn't want to have people telling me that my priority would go up, thus the computer will become slow, thus I will have to be careful with the app_info, I asked for the old page where the app_info for CUDA was.
Bottom line, I asked for one simple redirect link, and got two not helpful solutions.
Thanks a lot.
Though I don't suggest the
)
Though I don't suggest the use of app_info.xml to manipulate CPU % assigned to 1 GPU (if it's ever possible), I do share cristipurdel's sentiment (if that is what is intended) of our current inefficient CUDA apps which require 1 full CPU core just to run 1 GPU CUDA task.
From what I've check on Hardware Monitor on my Mac, CUDA task uses 1 CPU core while the GPU core, although uses 1 GPU, it just hovers around 28 - 32%, just shy of a third of overall GPU capabilities. I have a 8800GT, but the performance is only 30 minutes short of GT120 on a full work unit of 3.5 hours. Performance-per-watt ratio guys, 1 CPU core + (1 CPU core + 1 GPU core) is barely faster than a full 2 CPU cores.