Page 1 of 1

iGPU benchmarking

Posted: Fri Apr 14, 2017 3:07 pm
by bruce
In response to another topic about support for Intel iGPUs, I decided to do a little testing. At the present time, Intel GPUs are not supported, but I recently acquired a new Windows10 computer with AMD Radeon A8-5500 APU which contains an integrated GPU that is supported... (called a Radeon HD 7560D) Before installing a "real" GPU, I did some testing.

I don't contend that this iGPU is equivalent to any Intel iGPU. I simply present the data that I've acquired.

Project: 9414 Run 1519 Clone 0 Gen 6 was downloaded and started processing. That project has a timeout of 3 days and an expiration of 5 days. The GPU completed the WU in 3.45 days. Since the result was returned after the Timeout, it received no bonus points. Since it was returned before the Expiration, it did receive baseline points only.

I recommend that FAH should drop support for this iGPU. It's simply too slow to be useful even though it's doing a good job of displaying my desktop. I intend to install a "real" GPU and would be perfectly happy if I could continue to use the iGPU to drive my screen. (In the past, installing a dGPU has disabled the iGPU but that's another thing I need to test.

I encourage others to do additional testing on any similar hardware if it is supported and report their results.

Re: iGPU benchmarking

Posted: Fri Apr 14, 2017 4:35 pm
by JimboPalmer
Did you test the WU speed with and without CPU WUs running at the same time? Did you track temps with and without CPU WUs?

Re: iGPU benchmarking

Posted: Fri Apr 14, 2017 4:55 pm
by bruce
Yes. With the default of 3 CPUs processing, the GPU frame times didn't change when I suspended the CPU slot.

Also, in light of that information, I'll bet that CPU:4 would give the same result, but FAHClient mandates that one CPU be reserved to process the PCIe data even though that's really not applicable for an iGPU. The CPU ran several core_a7 AVX assignments during that time.

No, I didn't track temperature.