Joined: 5 Aug 99
I noticed the tendency that the calculation of the package takes longer (on the same computer). Gradually, the computer needs more and more time to process the data from the package.
For example, currently my computer is processing package for over 4 days and it still needs almost 55 days to finish it (using Intel GPU!). It is slim change that this package will be calculated before deadline, even though I'm calculatin on a quite powerful computer.
By creating such large packages you will be able to earn points rarely (compared to other projects) and there is a high risk that after two months of work I will not get any points at all, because I will not be able to complete the calculation on time. This discourages participation in this project.
I have been a member of this project for many years, but in this situation I am more and more often wondering whether I should not devote the power of my computer to projects with better results. :-/
Joined: 9 Jun 99
Your built-in Intel GPU is not strong enough on its own. It may be, if you allow a full CPU core to help it run. But as long as you use all other cores for the other projects you run, it'll go nowhere.
The tasks aren't that big and they have very generous deadlines. They'll just not run very quickly if you run work at full blast on all CPU cores. The Intel HD Graphics 4600 is only the second generation OpenCL capable GPGPU, its speed is slow when compared to the eight generation in today's Intel CPU.
If you want to speed up thing, try setting "Use at most N% of the CPUs" in your computing preferences to 99%. Since this value is an integer, it means that instead of 8 CPU cores, your CPU will only use 7. The 7th will be free and is then capable of helping out GPUs, which includes your Nvidia Quadro as well. Perhaps it's even better to free two CPU cores (set value to 75%), so that one CPU core caters for the Nvidia and the other for the built-in Intel GPU.
Remember that work done on the GPU runs primarily on the CPU which runs the actual science application and does all the translation of task data into kernels that the GPU can read, transport those kernels to the GPU's memory, wait for the GPU to do the calculations, then transport the result back to disk and where necessary translate that back into something the scientists understand.
When all CPU cores are constantly busy with other calculations, you can imagine that this transporting is going slow, because it will only happen at times when the CPU is doing less intense work. By freeing one or more cores, you make it easier for this data transport to and fro to happen, thus calculations on the GPU will speed up.
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.