Huge CPU difference between two i7 computers

[erkme73]

So I'm running my BI (4.0.1.1x64) on my i7 (4770) with 32GB of ram. My brother is running his same version on an i7 (4770R) with 16GB of ram.

Both systems are naked and headless - with no other applications, and no monitors connected.

My system has about 30 cameras - all of them running a 1280x720 h.264 (the majority are Hikvision).

My brother's system has 18 cameras - all of them 1280x720 h264 with 6 of them Hikvision.

His CPU usage is between 2% and 8%. It never crosses above 10% - even on windy days. He's set up to use h264 BVR encoding.

I've duplicated his settings, and even dropped down to 18 cameras, and I can't get below 40%.

When I have all 30 on, I'm running around 80%, with wind (or night time rain driving it to 100%). To avoid that, I've started using D2D, which keeps it about 60% independent of rain/wind.

Still, I can't help but wonder what it is about his system that is keeping his CPU so low, I'd even question whether he's running BI at all...


In fact, early on, I had similar experiences where I was hovering around 10 to 15% CPU with 25 cameras. I even took screen shots and sent them to my brother and father to brag about the low CPU. But I simply can't get there from here any more.

 

[Zxel]

There have been reports of high CPU usage on some systems with BI v4.0.1.0 - check out this thread.

 

[erkme73]

His video is the Intel iris Pro 5200 - where mine is the Intel Graphics 4600. Does the display adapter have that much of an affect on the CPU usage?

 

[erkme73]

Thanks Zxel - this discrepancy has been observed well before V4 - when we were both running 3.XX.

 

[Zxel]

There are many settings in BI, the cameras and of course the system itself that can effect CPU settings. As for the CPU usage you're reporting - are you looking at the computer itself or the percentage at the bottom of the BI interface? In other words how much is BI alone taking.

You should also take a look at the network bandwidth being used, is he using the same amount as you are?

 

[erkme73]

Interesting point. I'm having him check to see what his NIC bandwidth is. I know mine sits somewhere between 45 and 70 mbps. He did mention something about having all of his cameras on a different subnet - 192.168.2.x where the rest of the network (including the BI server) is on 192.168.1.x. I can't imagine that this would make a difference, but at this point I'm game.

 

[Zxel]

Interesting point. I'm having him check to see what his NIC bandwidth is. I know mine sits somewhere between 45 and 70 mbps. He did mention something about having all of his cameras on a different subnet - 192.168.2.x where the rest of the network (including the BI server) is on 192.168.1.x. I can't imagine that this would make a difference, but at this point I'm game.

In general it wouldn't - but it certainly could depending on how the whole network is layed out. Having the cameras on an isolated subnet (brilliant - fyi) the cameras would not have to contend with any other traffic, an excellent way to help latency and trafic contention. I'd be curious how he handles the flow of trafffic between the 2 subnets, i.e. is he using a bridge?

 

[bp2008]

Something has to be different between your systems. I think the 4770R is supposed to be slower than the 4770. Maybe his cams are running a really low frame rate? LOL Even with a really low frame rate I would expect higher CPU usage than that.

 

[Zxel]

Something has to be different between your systems. I think the 4770R is supposed to be slower than the 4770. Maybe his cams are running a really low frame rate? LOL Even with a really low frame rate I would expect higher CPU usage than that.

Ya, I was thinking the same thing, which is why the traffic check - I think there will be a huge difference.

 

[Overcon]

Some things to check are the camera settings, like:

MTU
Bitrate Type
Video Quality
Frame Rate
Max. Bitrate
I Frame Interval

All those things can impact performance. Not sure about HD speed, but if there is conjestion writing the streams to the HD. I would imagine that would cause more usage, since the data has to be cached to be written, which uses CPU cycles.

What I would do is take each system down to one camera of the same model and set all the settings to the same thing, including the same settings in BI. That way each system is identical or as much as you can get and then see how much each system is using via the task manager and BI with just one camera with all the same settings. Then you can go to five if you like and see if anything changes et cetera.

Best way to troubleshoot is to simplify and then expend.

 

[tjl]

18 cams at 8% hmm. Start recording and see what the numbers are.