Optimizing Computer/BI Server

[srvfan]

My current BI server is a Dell Optiplex 9020 tower. The motherboard is a Dell 0N4YC8 with an Intel i5-4590 CPU, 4 Cores, 4 Processors; there is no GPU attached to this computer. I am currently running BI version 5.5.6.2.1, and have 12 cameras. I have cloned 5 of those cameras in order to run DeepStack (CPU version). I should also add that I plan on adding other cameras to include some PTZs in the future.

With BI pulled up, I am running between 29-43% on CPU just sitting here. I have gone over the Wiki page for optimization several times and am following the tips/tricks. BI is running as a service, and once I close the program window out, I notice the CPU drops down to about 25%. Even with that, I have a hunch that I may need to update some of my hardware, but am trying to find the most cost-effective path.

Would a GPU make sense/be a good investment? I figured this would allow me to install the GPU version of DeepStack to take some of the load off of the CPU. If this is a good path towards optimization, are there any recommendations as to what GPU would fit my needs and be compatible with my old motherboard and CPU?

Would I be better off switching out the CPU for a better ROI? In reading the Wiki, I see that the i5-12600K would be a good solution for performance. I’ve found a motherboard/CPU bundle on Amazon featuring this CPU, but shuddered at the price ($474 presently). I figured if I performed that kind of upgrade, I would have to replace the motherboard to accommodate the new CPU. Kind of makes me wonder if I would be better off replacing the entire computer altogether.

Any constructive thoughts/ideas would be greatly appreciated as I just want to make sure my server runs efficiently and is “future-proof” for quite some time.

 

[wittaj]

That is still a capable system.

Getting a cheap GPU like the P400 or GTX1030 (usually around $130ish) or so can take the load off of the CPU for Deepstack.

Just remember to turn off hardware acceleration in BI at the global and camera level.

I wouldn't switch out the motherboard. Too many bad things can happen. It would be cheaper to buy a 7th or 8th generation that you know works than take a chance swapping out the motherboard.

As long as the CPU isn't maxing out at 100% and causes alerts or triggers or recording problems, no need to buy new unless you have the money burning a hole in your pocket LOL.

 

[srvfan]

That is still a capable system.

Getting a cheap GPU like the P400 or GTX1030 (usually around $130ish) or so can take the load off of the CPU for Deepstack.

Just remember to turn off hardware acceleration in BI at the global and camera level.

I wouldn't switch out the motherboard. Too many bad things can happen. It would be cheaper to buy a 7th or 8th generation that you know works than take a chance swapping out the motherboard.

As long as the CPU isn't maxing out at 100% and causes alerts or triggers or recording problems, no need to buy new unless you have the money burning a hole in your pocket LOL.

No burning at this time, LOL. I was just worried about the motherboard being compatible with newer CPUs. I'm not well-versed in screwing around with the hardware, so I figured that replacing the motherboard with a pre-installed CPU might be the easiest route. With my admission of being less than knowledgeable with hardware, is there any particular CPU you would recommend? I was looking on Amazon and found

https://www.amazon.com/Intel-i5-9400-Processor-Processors-984507/dp/B07MGZ9FJZ/ref=sr_1_3_mod_primary_new?keywords=intel+8th+gen+processor&qid=1662746298&sbo=RZvfv//HxDF+O5021pAnSA==&sr=8-3

as well as

https://www.amazon.com/Intel-i5-8400-Desktop-Processor-Cores/dp/B0759FGJ3Q/ref=sr_1_4?keywords=intel+8th+gen+processor&qid=1662746298&sr=8-4

, but I was unsure if either would work with my system. Currently, the only time I max out at 100% is when I'm replaying an alert clip and trying to use the tuning with DeepStack option. Speaking of DeepStack, I'm probably going to have to start another thread asking about the training, custom models, etc.

I will definitely look into the GPU first. I can handle the $130 or so, but when I saw the bundles, I cringed, lol. Now once I install a card, is there anything else I should do upon starting? Any software to add, or do I just restart and download the DeepStack GPU version?

Thanks

 

[srvfan]

One other question; in each of my cameras, the Hardware Decode option is set to Default. In the Global BI settings, Cameras tab, under Hardware Accelerated Decode, I show just Intel. Once a GPU is installed, I can just change that single setting to NO and be done, correct?

 

[wittaj]

There is a lot more involved with swapping out the intel chip - power supply, etc. that unless you do it a lot, you run the risk of completely FUBAR your system. Either make do with what you got or by a refurb.

A refurb 8th gen computer won't cost you much more than just that $200ish chip and will save you a lot of frustration trying to swap out a chip or motherboard.

Dell Inspiron 3670 Desktop - IntelCorei5-8400@2.8 12GB RAM 1TB HDD WiFi Win11PRO | eBay

For sale is (1) Dell Inspiron 3670 Desktop Computer. STORAGE: 1TB HDD. This computer is “plug and play” ready to go out of the box. The machines are all hand built, and tested by a Microsoft Certified System Engineer.

www.ebay.com

If you go the GPU route for Deepstack, simply follow the directions here:

Using DeepStack with Windows 10 (CPU and GPU)

docs.deepstack.cc

The GPU itself is plug-n-play.

I would suggest turning off hardware acceleration now.

Around the time DeepStack was introduced, many here had their system become unstable with hardware acceleration on (even if not using DeepStack). Some have also been fine.

This hits everyone at a different point. Some had their system go wonky immediately, some it was after a specific update, and some still don't have a problem, yet the trend is showing running hardware acceleration will result in a problem at some point.

However, with substreams being introduced, the CPU% needed to offload video to a GPU is more than the CPU% savings seen by offloading to a GPU. Especially after about 12 cameras, the CPU goes up by using a GPU and hardware acceleration.

You now only would use a graphics card if you are using Deepstack and use the GPU to process the DeepStack to improve processing time over the DeepStack CPU version, but do not use it for hardware acceleration within BI.

 

[looney2ns]

You can purchase refurbished I5-8500 systems for under $200 on Ebay.

elitedesk i5-8500 sff in PC Desktops and All-In-Ones: Search Result | eBay

Get the best deals for elitedesk i5-8500 sff at eBay.com. We have a great online selection at the lowest prices with Fast & Free shipping on many items!

www.ebay.com

 

[srvfan]

Thanks @wittaj for all your help, not just in this post. I have already turned off Hardware Acceleration at the global level as per your post. I have bookmarked the link you provided for DeepStack GPU and will be looking at purchasing a P400 or GTX1030. I'd rather the money for a potential motherboard/CPU bundle go towards adding another surveillance HDD to my setup and MORE CAMERAS!! LOL.
Thanks again

 

[srvfan]

You can purchase refurbished I5-8500 systems for under $200 on Ebay.

elitedesk i5-8500 sff in PC Desktops and All-In-Ones: Search Result | eBay

Get the best deals for elitedesk i5-8500 sff at eBay.com. We have a great online selection at the lowest prices with Fast & Free shipping on many items!

www.ebay.com

Thank you @looney2ns; I might wind up purchasing one of those in order to have a backup on hand, especially if I screw up my existing setup, lol. It might be nice to find a rackmount version of one of these options, but I'm sure that would probably be near impossible?

 

[srvfan]

Getting a cheap GPU like the P400 or GTX1030 (usually around $130ish) or so can take the load off of the CPU for Deepstack.

Okay, again, forgive my stupidity. I’ve viewed some GPUs on Amazon based upon your recommendations. Would either one of these work? Is one a better option over the other or are they both on the same plain? Should I look for low profile or anything specific? Would 2g of RAM be sufficient or would it be better to shoot at around 4? Thanks

https://www.amazon.com/PNY-Quadro-P400-Graphic-Card/dp/B08463LPY3/ref=mp_s_a_1_1?crid=30BXP64PORZZQ&keywords=nvidia+p400&qid=1662769784&refinements=p_n_condition-type%3A2224371011%2Cp_85%3A2470955011&rnid=2470954011&rps=1&s=electronics&sprefix=nvidia+p400%2Caps%2C122&sr=1-1

https://www.amazon.com/Gigabyte-GV-N1030OC-2GI-Nvidia-GeForce-Graphics/dp/B07646VQ6T/ref=mp_s_a_1_3?crid=3CL4KCURRRVE5&keywords=nvidia+gtx1030+gigabyte&qid=1662770167&sprefix=nvidia+gtx1030+gigabyte%2Caps%2C164&sr=8-3

 

[wittaj]

Obviously more RAM is better.

The 1030 has more CUDA than the P400, but not enough to make a difference and you should get whichever one is cheaper at the moment.

I have the the 1030 in my 4th gen for Deepstack and my times are under 100ms, usually around 55ms or so.

You need a low profile for that machine, so you can't get too much beefier in the GPU or it won't work.

 

[srvfan]

@wittaj,

I just received my GT1030 and installed. Following the DeepStack GPU link you listed above, I installed CUDA 11.3.1. When I try to install cuDNN, it states that I have to have an account to proceed. Would you happen to know if downloading cuDNN is a necessity? If so, is there a workaround so I will not have to create another account?

Also, once I download/install DeepStack GPU, that should automatically overwrite and take over my existing CPU download, correct? Thanks

 

[wittaj]

Yeah I created an account and did it, but I don't think that last step is necessary. It is the CUDA it needs.

Some have had it successfully by just downloading the GPU and others have had trouble.

At a minimum, shut down BI and the service before you do it.

I would suggest uninstall of the CPU version but you can always try it first without LOL.

Whichever way you do it, then reboot the computer so everything takes affect.

 

[srvfan]

Yeah I created an account and did it, but I don't think that last step is necessary. It is the CUDA it needs.

Some have had it successfully by just downloading the GPU and others have had trouble.

At a minimum, shut down BI and the service before you do it.

I would suggest uninstall of the CPU version but you can always try it first without LOL.

Whichever way you do it, then reboot the computer so everything takes affect.

Thanks; it looks like it worked. I also downloaded the license plate model I found in another thread, so lots of toys to play with now!

 

[clowg]

Has anyone got a copy of "DeepStack-Installer-GPU-2022.01.1.exe" that I could download? The official download link on Using DeepStack with Windows 10 (CPU and GPU) seems to have been suspended:

 

[MikeLud1]

Has anyone got a copy of "DeepStack-Installer-GPU-2022.01.1.exe" that I could download? The official download link on Using DeepStack with Windows 10 (CPU and GPU) seems to have been suspended:

View attachment 141099

Try the below link

Release DeepStack 2022.01.1 Release · johnolafenwa/DeepStack

This releases fixes a number of bugs and introduces new features and performance enhancements. Highlights Include: New Image Super Resolution Endpoint Multithreaded Inference providing significant...

github.com

 

[clowg]

Thanks!
Perfect

Go GitHub!!