Refurbished PC (CPU), return or not?

[karlocikovic]

Hello,

I ordered some refurbished HP Z230 workstation (for approx. 170€) with Xeon E3-1245 V3 (4th gen, 4/8 cores/threads) which has integrated graphic and QuickSync. Instead of that I got Z220 with Xeon E3-1240 v2 (4th gen, 4/8 cores/threads) which doesn't have integrated graphic card, but with that configuration I got nVidia Quadro K620 1GB, and I also have and prepared for that configuration nVidia GT1030 2GB.

I have 7-8 cams (4MP and 8MP), using substream and Deepstack for analysis...

Can I force decoding, streaming, deepstack all through GT1030? I have 2 slots for graphic cards so theoreticaly I can put inside both GT1030 + Quadro K620 ....
Return that or not?

 

[iwanttosee]

I had my lenovo M93p with an i5-4570 cpu with quicksync and 8GB of ram running five 5mp cameras with dual stream, one 2 mp camera with dual stream, and three 3 mp cameras with single stream. CPU usage was 15-25%, keep in mind the same PC also hosted two linux virtual machines. I was using the cpu only for deepstack, every time deep stack is required, CPU cycle goes to 100% for a bit and the result is around 1300ms (average) to analyze the photos.

I found a p620 (512 cuda cores, low profile, and 35 watt max) for $60 so I added it to my pc. I uninstall deepstack and reinstall the GPU dependent deepstack, 8GB of ram suddenly is not enough and I had to get 16GB. CPU cycles no longer maxed out and 300ms is the average to analyze the picture. I tried to force BI to use my P620 to decode, but it only stayed on for a while and then it will quit so I just use quicksync now. I'm not chasing the numbers and my p620 made the cpu usable during deepstack analysis so I'm good with it.

What's the return period? I'd set it up with BI demo and see if you like how it runs before deciding. Take maybe 1 hour if you have windows installed.
GT1030 and K620 both have 384 cuda cores, I'd use what it has and see if you like the result before adding another card.

 

[karlocikovic]

I had my lenovo M93p with an i5-4570 cpu with quicksync and 8GB of ram running five 5mp cameras with dual stream, one 2 mp camera with dual stream, and three 3 mp cameras with single stream. CPU usage was 15-25%, keep in mind the same PC also hosted two linux virtual machines. I was using the cpu only for deepstack, every time deep stack is required, CPU cycle goes to 100% for a bit and the result is around 1300ms (average) to analyze the photos.

I found a p620 (512 cuda cores, low profile, and 35 watt max) for $60 so I added it to my pc. I uninstall deepstack and reinstall the GPU dependent deepstack, 8GB of ram suddenly is not enough and I had to get 16GB. CPU cycles no longer maxed out and 300ms is the average to analyze the picture. I tried to force BI to use my P620 to decode, but it only stayed on for a while and then it will quit so I just use quicksync now. I'm not chasing the numbers and my p620 made the cpu usable during deepstack analysis so I'm good with it.

What's the return period? I'd set it up with BI demo and see if you like how it runs before deciding. Take maybe 1 hour if you have windows installed.
GT1030 and K620 both have 384 cuda cores, I'd use what it has and see if you like the result before adding another card.

Thanks for sharing your opinion! I have 16GB of RAM in that machine. I must decide today!

However, I guess we missunderstood - I will use GPU Deepstack (using CUDA) in every case, but would CPU without Quicksync be enough for handling streeaming, decoding?

 

[iwanttosee]

but would CPU without Quicksync be enough for handling streeaming, decoding?

With my i5 4570, I saw very little difference if any with quicksync enable or disable with my 9 cameras and with the optimal setup listed here. Optimizing Blue Iris's CPU Usage

 

[wittaj]

Now with substreams, the dependence of hardware acceleration is greatly diminished, and in some instances has been problematic with recent BI updates.

 

[karlocikovic]

Now with substreams, the dependence of hardware acceleration is greatly diminished, and in some instances has been problematic with recent BI updates.

Wittaj, so this 8 thread Xeon without integrated graphic and gt1030 would be ok for <10 cams?

 

[wittaj]

I don't know much about the Xeon and how they compare to the Intel iX XXXX CPUs, but I know that with doing every optimization in the wiki that the integrated graphic card isn't as important as they were before.

 

[Flintstone61]

I ran 15 cams on a Core i7 - 3770s (3RD generation 4 core with 8 threads) with 32GB ram, using quicksync, Dual streams H264 / direct to Disk .BVR files, but, as BI versions changed, I found I could run with or without Quicksync enabled.
I even had a GT710 Nvidia in there as an experiment for awhile. But didn't see much difference.
10 Cams shouldn't be a problem, Never have used Deepstack.

 

[Flintstone61]

I see the HP Z230 workstation listed on ebay with xeon's and i7 3770's so they are about the same architechure.
22nm - and FCLGA1155 socket.

 

[karlocikovic]

It is running great for now.

 

[karlocikovic]

I would like to hear your thoughts: Recently I bought some 3gen Xeon e3-1240 v2 (8 threads at 3.4GhZ), 16GB DDR3... I have option to buy cheap i7 6700 + 16GB DD4 configuration. I know after that I would like to have nvme ssd and maybe some more RAM what I can't have with my current configuration.. I have 6x8MP cams + 4x4MP cams, using GPU Deepstack with GT1030.

Would I see some benefit from that upgrade, or better save money for something smarter?

 

[mat200]

I would like to hear your thoughts: Recently I bought some 3gen Xeon e3-1240 v2 (8 threads at 3.4GhZ), 16GB DDR3... I have option to buy cheap i7 6700 + 16GB DD4 configuration. I know after that I would like to have nvme ssd and maybe some more RAM what I can't have with my current configuration.. I have 6x8MP cams + 4x4MP cams, using GPU Deepstack with GT1030.

Would I see some benefit from that upgrade, or better save money for something smarter?

Hi @karlocikovic

I know a number running used i7-6700 here have seemed to be happy with it..

i7 6700 + 16GB DD4