DEEPSTACK GPU-ESXi6.0U3-UBUNTU18.04-QUADRO_K600

Prepared a fresh setup today:

1.ESXi on Dell Precision T3610 XEON E5-1650V2
2.NVIDIA QUADRO K600
3.UBUNTU 18LTS with NVIDIA Driver Version: 440.33.01 CUDA Version: 10.2
Installed following the description:
https://docs.deepstack.cc/using-deepstack-with-nvidia-gpus/index.html#step-2-setup-nvidia-drivers

I can see the GPU inside the docker container, but get timeout error (500). I’ve checked :gpu-x5-beta
and the latest official gpu version.
I’ve ran many tests on different HW and the only gpu version I got running is on a WIN10 baremetal install.
any ideas what I’m doing wrong here?

I’m running now the cpu version in the docker and facing this performance which seems to be decent:

I wonder what processing times I can expect with the gpu version and the Quadro K600?
Processing time on an XEON X5560 2.8GHz cpu was 3 sec per face.
300-500ms is similar to the processing time with a GTX 1060 on windows gpu version

I suspect it is pytorch / gpu unsupported as below…


Thank you. I will not waste more time in investigation on the wrong HW. I’ve ordered a K20 Card which supports cuda 3.5.

https://www.geeks3d.com/20120517/nvidia-kepler-gk110-architecture-whitepaper-2880-cuda-cores-and-compute-capability-3-5/

Coming back to the performance I can expect:
Is it likely to get faster processing times than 300ms. per face with the gpu version of deepstack?

not sure about times for face detection, but if i had to guess I’d say that any GPU that is functional with deepstack would get at least 5x faster responses than an old xeon CPU, or something along those lines.