singe core speed vs core count

Post Reply
mkw
Posts: 5
Joined: Fri Jul 09, 2021 3:31 pm

singe core speed vs core count

Post by mkw »

Hello

I am new user of BI. It is running for about two months and I decided to migrate. Now when I started to move cameras over I reached max of crappy HW I have tested it on. I need to move to something more powerful.
I have 2011 v3 based server but it is quite power hungry beast with e5-2680 v3. I am thinking about swapping out just CPU to save $$ and go down with power consumption. I had in mind moving to low power e5-2608l v4 that is 8core 16 thread chip with 1.6 G base frequency.
I have read through this forum but couldn`t really find an answer - how much cores/threads can BI utilize and how important single core speed is ?? I`d be really grateful for hints as testing this myself is about 2 days work - I have to migrate stuff that I have on this server right now just to see.

Regards, Mark
Matts1984
Posts: 496
Joined: Fri Apr 10, 2020 1:12 pm
Location: Maryland, USA

Re: singe core speed vs core count

Post by Matts1984 »

I have not seen or am aware of a limit to the number of cores. AFAIK, the more the better and I'm already using quite a few - albeit on pretty old processors.

As for clock speed vs number of cores and which being better.... thats a hard call. This site has samples of real-world BI installs that you should be able to extrapolate what you're looking for. https://biupdatehelper.hopto.org/default.html#stats
Blue Iris 5.8.8.x | Server 2022 VM | Xeon X5650 @ 2.67GHz - 12 Cores | 12GB RAM | 7TB RAID | Sophos UTM WAF | 4x SV3C 5MP Bullet A | 1x SV3C 5MP PTZ HX | 1x SV3C 5MP Bullet HX | 1x SV3C 5MP Dome HX | 2x Amcrest 5MP Bullet
mkw
Posts: 5
Joined: Fri Jul 09, 2021 3:31 pm

Re: singe core speed vs core count

Post by mkw »

Thank you very much for the link. It gives some general idea what to expect.
I was thinking about this core speed issue and come up with very easy to perform test scenario - I am ashamed of myself I didn`t think about it earlier ! What I did:
- I measured long enough average utilization of the CPU clocked at 4.1 GHz and it was between 68-70% on average. Depending on idle or triggered state.
- clocked exporting sample clip to mp4
- Downclocked CPU to 2.2 GHz and measured the same average utilization to be 70-73%
- exported the same clip
- checked overall responsiveness of the UI, logs for deepstack timings etc.
Observations - when running as a background service I don`t see any negative impact after reducing CPU frequency by almost half. While working with the UI I didn`t notice major impact. It is a little bit more sluggish, but it is not painful. With deepstack running on GPU I did not see any impact in average processing times (some operations have to be CPU dependent). Reencoding took longer - this is probably done exclusively on CPU.
Conclusion - single core speed do not matter that much, with CPU`s with higher IPC than my ancient amd fx4100 I tested it on it will matter even less. I decided to go for low power consumption, low clock high core count Xeon.

Shared my experiment as maybe someone in the future will find it useful.
User avatar
TimG
Posts: 2084
Joined: Tue Jun 18, 2019 10:45 am
Location: Nottinghamshire, UK.

Re: singe core speed vs core count

Post by TimG »

Nice one. Thanks for the info.
Forum Moderator.
Problem ? Ask and we will try to assist, but please check the Help file.
Post Reply