multi GPU and GPU memory reading

hello,

Using a system with 2 Gpu ( P4000 + P400) set in affinity mode.
Investigating why TD is dropping frame after a while using a log and perform monitor recorder . It seems that the perform chop count only the total mem of the P400 reading 1981mo (a bit less that the 2048 given) whereas the process is running on the P4000. Affinity setting is correct , I was expecting to see the total gpu mem of the P4000.

It is correct ?

Thanks

xavier

It’s just being incorrectly reported. It shouldn’t be the cause of the dropped frames. We query GPU memory using a DirectX call, which doesn’t map directly to the OpenGL GPU affinity that is used for rendering.

yes, that shouldn’t be the origin of my problem . but as the perform chop was reporting 2500 Mo used on a total of 1800 Mo … I was just wondering if I got something wrong in my set up.
The system start at 60fps and fall to 20-14fps slowly after a few hours. only clue is the amount of Cpu memory increasing at the same time. Suspecting the drive read speed . But with 2 x ultra fast Samsung 960 pro NVMe . maybe as the integrator put those in mirror mode via windows 10 raid management . just wondering if it could be the problem.

thanks