ZoneMinder Forums

well,

i have 2 cpus, which one would be better for my ZM box ?
p4@3Ghz (2mb cache) or p4@3.2Ghz (1mb cache) ?

I installed ZM with p4@3.2Ghz (1mb cache), and cpu load average goes from 1.5 to 4 (1 usb cam, 1 remote cam, 8ch Kodicam card).

i'd say p4@3Ghz (2mb cache), but whats your opinion ?

i wouldnt expect there to b ethat much difference between them to be honest

jameswilson wrote:i wouldnt expect there to be that much difference between them to be honest

it was a great surprise to me to see a linux box with such big load, using only 30% of its memory..
ZM is cpu intensive if you use something more than a camera and something bigger than 320x240 or 5 fps..

i wonder about cpu load in windows boxes using software included with these 4ch or 8ch cards..

i will keep the 3.2 cpu for a few hours, to get load stats, and then change it..
just to see the difference..

> Zm is cpu intensive if using motion detection at high res. Phil on the todo list is looking into the possibilty of limiting the res of detection but still recording full res. If you use record mode it will massivly drop the cpu use. Havnt used the windows apps but i doubt they are stable enough for commercial use. You can look at milestone for a proper windows based app but its steep and uses ip only i think. Again im unsure of preoc req. I have succesfully used an mmx jpeg lib (see other posts) and this massivly drops the cpu use for creating jpegs but wont help with motion detection. I have moved to dual+ core processors. I think the new core duos will be awesome, but the old 820's and similar arnt much money now. Im looking forward to the dual processor AM2 boards (4 cores!) then we can see what we can do!!

jameswilson wrote:> Havnt used the windows apps but i doubt they are stable enough for commercial use.

I have succesfully used an mmx jpeg lib (see other posts) and this massivly drops the cpu use for creating jpegs but wont help with motion detection.

Im looking forward to the dual processor AM2 boards (4 cores!) then we can see what we can do!!

i've seen tenths of these (windows pcs with capture software) with 4-8-16 cameras.. Sold to the customer for thousands of euros..

i'm interested on this (mmx jpeg lib). As far as i want this for home use, i just want to capture from 1-2 cameras (the entrance, etc).. I suppose i will use the full 8 camera capture when i'll be on holidays

Isnt it funny to have a ZM computer with greater capabilities than your home-pc ?

Isnt it funny to have a ZM computer with greater capabilities than your home-pc ?

Very tempting I know

I only have 2 cams on my place so im using an old 750Meg duron with great success at 3-4 fps per cam no worries using the mmx lib. I run in mocord with no remote viwing at about 20% proc use 70% idle drops to 50% when using zm4ms to live view

Its the fps that kills yas so if you keep to 25 fps across all cams i bet both chips will be fine

Sorry to change the direction a bit - but I've a question to james about the mmx jpeg lib. You say it won't help with the motion detection, but if using remote cameras, the images come either as jpg or mjpeg - wouldn't this lib help in decoding the incoming images as well as creating new ones?

Being a bit of a wuss, my inability to find any good documentation about this lib has stopped me playing with it so far, but it certainly has my attention if it can help ease the load on my two zm boxes.

well i havnt seen that much of a benifit on motion detection as the images are converted from jpeg before being anaylsed so i suppose thats a raw horsepower thing.

Hi.

I wonder if someone has checked the motion detection speed and CPU load of the Motion project: http://www.lavrsen.dk/twiki/bin/view/Motion/WebHome.

Perhaps something can be learned from it, in order to optimize the motion detection process?


Also, I have found the following article on the subject: http://www.codeproject.com/cs/media/Mot ... ection.asp

Check the 4th image, Phil, where instead of exactly outlining the person (as in ZM), it kind of surrounds it with blocks. The writer of the article wrote that such approach is faster and has large optimization potential.

Thanks for the link. I will take a look at it. You can of course turn blobs off altogether if you don't want the overhead. However I accept that perhaps something in between might be desirable.