Hell Yeah! Direct Deepstack Integration - 5.4.0 - March 31, 2021

Believe it or not I still have 8 EVGA 1070Ti's running and mining Eth. They have been in operations since 2018! Things are bullet proof. I'll make what I paid for them when I finally sell them back on eBay.
 
Believe it or not I still have 8 EVGA 1070Ti's running and mining Eth. They have been in operations since 2018! Things are bullet proof. I'll make what I paid for them when I finally sell them back on eBay.
Awesome good to know thanks

Looks like the Evga needs 6pin AND 8pin connections if I'm not mistaken
 
Last edited:
I'd never have an issue with a used GPU. 99.9% of the equipment I run is used. A GPU doesn't really wear out, and even if ran at high load and high temp, they just thermally throttle themselves

In fact, I've had more problems with brand new GPU's being DOA out the box directly from EVGA than used ones
 
Hi, How are you finding the gtx 960 with deepstack? What speeds are you getting? What mode do you have it on?
I've been told to avoid the 960 as it uses up a lot of power and may not work with deepstack. Are you on the latest windows version of deepstack gpu?
Been looking for a gtx used around this price range, 1050 and 960's pop up.

Thanks

I've been using the GTX 960 since May of this year and it seems to work fine with deepstack.
Speeds are 85ms-160ms. As far as power goes I have not paid much attention to it. I did fire up GPU-Z and the Board Power Draw peaked in the mid 20's with my 2- 2MP LPR cameras (5231e-z12e).
Those are the only cameras I send to deepstack.
When I run test and tune deepstack on my 5MP overview cameras it will peak in the mid 40's. Idle it sits at 12.5W.
I am currently using Version 2021.02.1, but it works fine with the latest as well.
 
I've been using the GTX 960 since May of this year and it seems to work fine with deepstack.
Speeds are 85ms-160ms. As far as power goes I have not paid much attention to it. I did fire up GPU-Z and the Board Power Draw peaked in the mid 20's with my 2- 2MP LPR cameras (5231e-z12e).
Those are the only cameras I send to deepstack.
When I run test and tune deepstack on my 5MP overview cameras it will peak in the mid 40's. Idle it sits at 12.5W.
I am currently using Version 2021.02.1, but it works fine with the latest as well.

What are your LPR cams doing with Deepstack out of interest?
 

Awesome - Have to thank you all for all the tips and advice on this subject. Really appreciated.

Just waiting for a response from the Evga seller. Hopefully get back to me soon that will result in a successful purchase. (then I need to buy a new psu also - looking at a 650w + 80 bronze)
 
I live in a condo complex with lots of tree shadows and this time of year lots of blowing leaves. I use deepstack to cancel and hide the useless triggers. I use substream 2 for the actual LPR functionality, The main stream is what is recorded in blue iris.

Have you tried using zone crossings? This has saved my cpu and false alerts, done properly it won't miss positive triggers also.
I have a big tree on the street, with a big street lamp next to it.
At night and during the afternoon sun without zone crossings, my DS would be working constantly.
Cut triggers down 98%
 
Have you tried using zone crossings? This has saved my cpu and false alerts, done properly it won't miss positive triggers also.
I have a big tree on the street, with a big street lamp next to it.
At night and during the afternoon sun without zone crossings, my DS would be working constantly.
Cut triggers down 98%
I use zone crossing for my overview cameras. I've tried with my LPR cameras but the field of view is quite small and it didn't work well at all.

I forgot to mention I only use deepstack during the day.
 
Anyone know what occupied means?

1635277900244.png
 
Interesting, never seen that in my life. Guess you learn something new every day!
 
  • Like
Reactions: sebastiantombs
Since setting DS to high and adding more images, I've been getting MUCH better results. Its now always 90%+ for people, before it was sometimes as low as 60%

Still processing in 60-100ms with minimal card load

1635278272677.png
 
Awesome good to know thanks

Looks like the Evga needs 6pin AND 8pin connections if I'm not mistaken

You made me go look, crawl under the desk with a flashlight. I remember the six pin but forgot the eight pin. Yeah, it needs both.
 
  • Like
Reactions: looney2ns
Just the standard model. But I plan to look into the Dark models, I've just gotten into all the DeepStack stuff
 
Just the standard model. But I plan to look into the Dark models, I've just gotten into all the DeepStack stuff
Ok, understand, let us know times if you try ExDark which is an easy one to drop in. My times for that or custom models are approximately 10-15 times longer :wow: anything up to 4000-5000ms
 
  • Like
Reactions: beandip
Detection times will get slower when using another model. DS has to scan two databases for each detection then.
 
What's the limitation there? GPU Power?