The example I posted (one of my various app_config.xml files) was quite specific to the particular hardware it was designed for and was not a 'one size fits all' type of example. The only reason I posted it was to act as an example of how you could include an xml text file inside BBCode 'code' tags to preserve the entire formatting of the file rather than have the 'line-by-line' problem that Zalster was experiencing when he was posting his example.
Yes it was obvious what you said. The only interesting thing was how 0.3 CPU worked with 0.5 GPU when the default an E@H WU arrives with is 1 CPU + 1 GPU. I thought it may have been a made up example so I gave it a try.
Gary Roberts wrote:
There is documentation on the Berkeley website that covers the app_config mechanism. You also need to be aware of the major differences in CPU support requirements between NVIDIA and AMD GPUs. If you don't plan for these things you are likely to have performance problems.
I have cc_config & app_config files on all of my rigs & projects. I've been over that web page countless times using trial & error to flesh out the ambiguities. E.g.,
<no_alt_platform>0|1</no_alt_platform>If enabled, the client will run applications only for its primary platform. For example, a Win64 machine will run only Win64 apps, and not Win32.
<no_gpus>0|1</no_gpus>If 1, don't use GPUs even if they're present. Requires a client restart.
Does "if enabled" = 0 or 1 ??? Also, doesn't "require a client restart" is obsolete as a Options/Read config files works. Be nice if the BOINC boyz would go through and update, define, and clarify exactly what's meant.
Gary Roberts wrote:
You also need to be aware of the major differences in CPU support requirements between NVIDIA and AMD GPUs. If you don't plan for these things you are likely to have performance problems.
Not one word on that web page about "CPU support requirements." I've been doing this for years & it's trial & error and sharing on the forums.
Gary Roberts wrote:
If you mention what hardware you have and what you are trying to achieve, there would be quite a few people who could give you sound advice. The best place to do that would be a new thread in Cruncher's Corner. The thread here was about GPUGrid needing CPU volunteers and how to use the app_config mechanism to limit the number of CPU cores being used when running CPU tasks at GPUGrid. As this is unrelated to what you are talking about, it's best to allow this old thread to 'RIP' :-).
I recently left 23 Linux rigs requesting Quantum Chemistry CPU WUs from GPUGRID for weeks and did not get a single one. I gave up. Now I'm just lusting after Zalster's secrets as his rigs get higher scores than mine: http://www.gpugrid.net/top_hosts.php
I recently left 23 Linux rigs requesting Quantum Chemistry CPU WUs from GPUGRID for weeks and did not get a single one. I gave up.
Were they also running the GPU work units? The BOINC scheduler does not do that well on the same project, for obscure reasons that someone here can probably explain. I use a second BOINC instance for that, or else use different machines for CPU and GPU from a given project.
Gary Roberts wrote:The
)
Yes it was obvious what you said. The only interesting thing was how 0.3 CPU worked with 0.5 GPU when the default an E@H WU arrives with is 1 CPU + 1 GPU. I thought it may have been a made up example so I gave it a try.
I have cc_config & app_config files on all of my rigs & projects. I've been over that web page countless times using trial & error to flesh out the ambiguities. E.g.,
Does "if enabled" = 0 or 1 ??? Also, doesn't "require a client restart" is obsolete as a Options/Read config files works. Be nice if the BOINC boyz would go through and update, define, and clarify exactly what's meant.
Not one word on that web page about "CPU support requirements." I've been doing this for years & it's trial & error and sharing on the forums.
I recently left 23 Linux rigs requesting Quantum Chemistry CPU WUs from GPUGRID for weeks and did not get a single one. I gave up. Now I'm just lusting after Zalster's secrets as his rigs get higher scores than mine: http://www.gpugrid.net/top_hosts.php
Aurum wrote:I recently left
)
Were they also running the GPU work units? The BOINC scheduler does not do that well on the same project, for obscure reasons that someone here can probably explain. I use a second BOINC instance for that, or else use different machines for CPU and GPU from a given project.