Seems like my default GT120 ain't supported on BOINC CUDA app. Official BOINC version 6.10.21 cannot detect usable GPU even with CUDA drivers installed...
OS X users might want to check the performance of the new CUDA app for their particular hardware. If you have a Mac with a fast CPU and a rather not so fast GPU, you will get better results with the CPU-only app, so in this case you might want to disable GPU processing.
Seems like my default GT120 ain't supported on BOINC CUDA app. Official BOINC version 6.10.21 cannot detect usable GPU even with CUDA drivers installed...
You need to be running v6.10.28 (Beta, on 'All versions' download page) to get CUDA support on Mac - they've only just given up waiting for NVidia to get 64-bit drivers working properly, and reverted to a 32-bit workround instead.
OS X users might want to check the performance of the new CUDA app for their particular hardware. If you have a Mac with a fast CPU and a rather not so fast GPU, you will get better results with the CPU-only app, so in this case you might want to disable GPU processing.
Interesting... but that still doesn't answer my question. Does the CUDA app for OS X platform requires the usage of 1 CPU core like Windows does...?
OS X users might want to check the performance of the new CUDA app for their particular hardware. If you have a Mac with a fast CPU and a rather not so fast GPU, you will get better results with the CPU-only app, so in this case you might want to disable GPU processing.
Interesting... but that still doesn't answer my question. Does the CUDA app for OS X platform requires the usage of 1 CPU core like Windows does...?
From your assertion, I assume it does...?
Yes, it dies. It's the same application, just compiled for Mac OS.
Seems like my default GT120 ain't supported on BOINC CUDA app. Official BOINC version 6.10.21 cannot detect usable GPU even with CUDA drivers installed...
Yes, it requires on CPU core to crunch a GPU-WU. However, on my MacPro 8x3.0 GHz w/GeForce 8800 GT a CPU WU grants appr. 30 credits/hour, the CUDO WUs grant appr. 40 credits/hour.
Seems like my default GT120 ain't supported on BOINC CUDA app. Official BOINC version 6.10.21 cannot detect usable GPU even with CUDA drivers installed...
Yes, it requires on CPU core to crunch a GPU-WU. However, on my MacPro 8x3.0 GHz w/GeForce 8800 GT a CPU WU grants appr. 30 credits/hour, the CUDO WUs grant appr. 40 credits/hour.
If CS is one of your goals then you would be better using the CPU only for EaH and use the GPU on Collatz so you can get two tasks done at the same time...
Seems like my default GT120 ain't supported on BOINC CUDA app. Official BOINC version 6.10.21 cannot detect usable GPU even with CUDA drivers installed...
Yes, it requires on CPU core to crunch a GPU-WU. However, on my MacPro 8x3.0 GHz w/GeForce 8800 GT a CPU WU grants appr. 30 credits/hour, the CUDO WUs grant appr. 40 credits/hour.
If CS is one of your goals then you would be better using the CPU only for EaH and use the GPU on Collatz so you can get two tasks done at the same time...
Paul,
yes, you are right. However, I am primarily here for the science but it is still nice to see that the new apps are 33% more efficient and our efforts are honored.
yes, you are right. However, I am primarily here for the science but it is still nice to see that the new apps are 33% more efficient and our efforts are honored.
My point exactly ... :)
By reallocation as I suggested you do more science. Granted not just for EaH, but for all projects...
I know testing is important, but that has been done on this application and it has been proven to work. Now the project needs to make it more efficient otherwise you are wasting resources.
By reallocation as I suggested you do more science. Granted not just for EaH, but for all projects...
...On Macbook Pros 2 CPUs + 1 GPU (preferably the faster one) is absolutely fantastic for frying eggs on the keyboard... Not sure if it contributes much to the science though you could try sending in egg baps and see if they give you credits for them... :-?
Mac OS X CUDA app 5.03 for ABP2 available and working
)
So how does this work...?
One CUDA work unit at the expense of 1 core...?
Seems like my default GT120 ain't supported on BOINC CUDA app. Official BOINC version 6.10.21 cannot detect usable GPU even with CUDA drivers installed...
OS X users might want to
)
OS X users might want to check the performance of the new CUDA app for their particular hardware. If you have a Mac with a fast CPU and a rather not so fast GPU, you will get better results with the CPU-only app, so in this case you might want to disable GPU processing.
RE: So how does this
)
You need to be running v6.10.28 (Beta, on 'All versions' download page) to get CUDA support on Mac - they've only just given up waiting for NVidia to get 64-bit drivers working properly, and reverted to a 32-bit workround instead.
RE: OS X users might want
)
Interesting... but that still doesn't answer my question. Does the CUDA app for OS X platform requires the usage of 1 CPU core like Windows does...?
From your assertion, I assume it does...?
RE: RE: OS X users might
)
Yes, it dies. It's the same application, just compiled for Mac OS.
CU
RE: So how does this
)
Yes, it requires on CPU core to crunch a GPU-WU. However, on my MacPro 8x3.0 GHz w/GeForce 8800 GT a CPU WU grants appr. 30 credits/hour, the CUDO WUs grant appr. 40 credits/hour.
RE: RE: So how does this
)
If CS is one of your goals then you would be better using the CPU only for EaH and use the GPU on Collatz so you can get two tasks done at the same time...
RE: RE: RE: So how does
)
Paul,
yes, you are right. However, I am primarily here for the science but it is still nice to see that the new apps are 33% more efficient and our efforts are honored.
RE: yes, you are right.
)
My point exactly ... :)
By reallocation as I suggested you do more science. Granted not just for EaH, but for all projects...
I know testing is important, but that has been done on this application and it has been proven to work. Now the project needs to make it more efficient otherwise you are wasting resources.
Anyway, just one man's opinion... :)
RE: By reallocation as I
)
...On Macbook Pros 2 CPUs + 1 GPU (preferably the faster one) is absolutely fantastic for frying eggs on the keyboard... Not sure if it contributes much to the science though you could try sending in egg baps and see if they give you credits for them... :-?