Not bad thanks for the data. There's a lot of talk about the 12th gen, wonder how much faster that is.
Probably cheaper buying a low-range GPU than a 12gen i7 or i9 though.
DeepStack Case Study: Performance from CPU to GPU version
- Thread starter jaydeel
- Start date
Probably cheaper buying a low-range GPU than a 12gen i7 or i9 though.
Are these on a sub-stream?
I think the 50ms times from the cheapest 30 series RTX are hard to beat for sub-streams for the money. I have it sample 3 images, once every 750ms. I was worried if the system got motion alerts on several cameras at once. A 250ms time could get backed up with 4-5 cameras triggering. That shouldn't be problem with 50ms times.
Yeah, substream @ 1912x1080, default DS model only and Medium mode
Sorry if this has already been answered but, I have a Radeon RX560. Will DS GPU version work or does it only work with Nvidia?
MikeLud1
IPCT Contributor
Well, it's the season for tinkering and maybe some Black Friday deals.
I'm running 12 or 13 cameras all running deepstack on a Dell Optiplex 7050 Micro i7-7700/16GB
Catching up on this and other threads, it would seem like the GTX 1050 ti would be a good choice. Although I'm sure I have a more powerful PSU in the garage, I'm also sure it's many years old.
Any recommendations on a particular brand or other cards before I squeeze the trigger?
Thanks in advance for all of the helpful knowlege!
I'm running 12 or 13 cameras all running deepstack on a Dell Optiplex 7050 Micro i7-7700/16GB
Catching up on this and other threads, it would seem like the GTX 1050 ti would be a good choice. Although I'm sure I have a more powerful PSU in the garage, I'm also sure it's many years old.
Any recommendations on a particular brand or other cards before I squeeze the trigger?
Thanks in advance for all of the helpful knowlege!
A question about cuDNN version (for a P400). All the docs on the Deepstack site, and references here, point to v10.1. But the Nvidia site gives v11.5 as the most recent available. For a new install, is it recommended to stay back on v10.1 or keep up with the latest? Thank you.
IReallyLikePizza2
Known around here
IReallyLikePizza2
Known around here
Apparently I'm less of a person than my trash can. Sometimes I wonder just how smart DeedStack is...
MikeLud1
IPCT Contributor
DeepStack has not updated the torch and torchvision that supports 11.X yet so it only works with 10.X
MikeLud1
IPCT Contributor
You probably are masking the upper part of the image so DeepStack is only see half of you.
IReallyLikePizza2
Known around here
But I've never seen a human with wheels. How it thinks thats a human is strange
Crap, I went with 11.5. It seems to be working so far for the last week on my GTX 1060. What kind of problems would running the newer version cause?
According to NVIDIA you should be fine:
"since CUDA is backward compatible, existing CUDA applications can continue to be used with newer CUDA versions"
Okay, picked up a MSI GeForce GTX 1050 Ti 4GT LP Graphics Board VD6238 4GB GDDR5 for $275 on ebay. Looking forward to seeing what this does to the response times. I should get it and installed within a week.
As an eBay Associate IPCamTalk earns from qualifying purchases.
Hello everyone !
I'm new in this world, i came from Surveillance Station from Synology to Blue Iris/DeepStack.
I bought an entirly new server for this : a Zotac ECM73070C with an i7-10700 and RTX3070.
I also installed a M.2 Google Coral TPU but it seem to not be supported by DS so...
I Installed DS GPU version for Windows, lastest nVidia drivers, cuda and cudnn, but my CPU seem to be under heavy load and my GPU doing nothing...
Is there a way to check that ? Time respond, which componant is using, etc ?
Thanks !
I'm new in this world, i came from Surveillance Station from Synology to Blue Iris/DeepStack.
I bought an entirly new server for this : a Zotac ECM73070C with an i7-10700 and RTX3070.
I also installed a M.2 Google Coral TPU but it seem to not be supported by DS so...
I Installed DS GPU version for Windows, lastest nVidia drivers, cuda and cudnn, but my CPU seem to be under heavy load and my GPU doing nothing...
Is there a way to check that ? Time respond, which componant is using, etc ?
Thanks !
Have you done EVERY optimization in the wiki? Substreams is a must.
In Task Manager you can see if the GPU is being utilized.
MikeLud1
IPCT Contributor
DeepStack does not support RTX3000 series GPUs yet without some hacking see steps in the below post