Hi All,

I have been following the posts and the guides for installing Deepstack GPU version, but it is not working for me. I was able to install Deepstack CPU version on another PC easily and that is working well.

I have followed the instructions from Nvidia well I think, until the last part of the CuDNN instructions:

https://docs.nvidia.com/deeplearning/cu ... index.html

5. Include cudnn.lib in your Visual Studio project.
a. Open the Visual Studio project and right-click on the project name.
b. Click Linker > Input > Additional Dependencies.
c. Add cudnn.lib and click OK.

I am not sure what to do here. I do have Visual Studio installed (per the instructions).

I am able to get the splash screen "DeepStack Activated" , but the log shows errors:

2021-08-02 11:11:58.783 AM Cam6 MOTION_A
2021-08-02 11:12:14.291 AM Cam6 DeepStack: Server error 100
2021-08-02 11:12:14.291 AM Cam6 DeepStack: Alert cancelled [DeepStack: 100] 15017ms
2021-08-02 11:12:16.590 AM App DeepStack has been restarted

I have read here on the forum that there could be a few reasons for these errors (which I verified), but I am wondering if my issue is due to not completing Step 5 from the CuDNN guide above?

I have asked Sam for a Deepstack GPU step-by-step installation guide for a non-programmer (like me). Hopefully he will find time.

Thanks,
Todd

Update:

My issue may be my GPU hardware and maybe it is timing out? It's a GTX 660 with 960 CUDA cores (which is not a lot).


How would one know if a particular Nvidia card had enough processing power for Deepstack? Tried to google it....

Thanks,
Todd

hi
i have the same issue with the gpu version - cant get it to work and i have one of the new Nvidia cards - which i really want to make use of!

Update:

Took out the GTX 660 and put in a GTX 1070Ti and it just started working without a fuss. Put the GTX 660 back in and it still didn't work (in case I changed something without knowing). No idea why the GTX 660 doesn't work, it has a decent amount of CUDA cores. I tried searching everywhere and there is no compatability list for NVIDIA GPUs that can run Deepstack.

I now have response times of 50-100ms....nice!

Todd wrote: ↑Thu Sep 02, 2021 12:23 am Update:

Took out the GTX 660 and put in a GTX 1070Ti and it just started working without a fuss. Put the GTX 660 back in and it still didn't work (in case I changed something without knowing). No idea why the GTX 660 doesn't work, it has a decent amount of CUDA cores. I tried searching everywhere and there is no compatability list for NVIDIA GPUs that can run Deepstack.

I now have response times of 50-100ms....nice!

That's because the GTX660 isn't supported. According to Deepstatck and its forums only the cards that support CuDNN is supported. And from what I read its the 900 series and above. Thats why your 660 won't work and your 1070 did work.

Hi.
Does deepstack gpu work with the gtx 960 to do you know?

Don't want to waste my money on a 960 if the 1050 is more suitable.

Thanks