Mac mini Model Performance Comparison
I'm curious what kind of performance is being seen with Mac mini's out there. The Mac mini (Late 2012) is their most powerful and customizable model so I'm sure it is the best performing, but I'm curious about the performance quality at higher quantity of cameras with different models of Mac mini.
For example, I'd be curious to see what model you're using, how many cameras (at what resolution and frame rate), and how you'd personally rate the performance.
Say you have a current model mini (Late 2014) with 8 cameras at 3MP and 15-20fps (yes, I picked a high FPS on purpose). Is the web interface still snappy for you or does it lag? Do you see CPU overload errors? Can it keep up?
I've only ever run SS on iMacs and MacBook Pros so I'm curious how the performance rates to those out there using Mac minis. Thank you!
For example, I'd be curious to see what model you're using, how many cameras (at what resolution and frame rate), and how you'd personally rate the performance.
Say you have a current model mini (Late 2014) with 8 cameras at 3MP and 15-20fps (yes, I picked a high FPS on purpose). Is the web interface still snappy for you or does it lag? Do you see CPU overload errors? Can it keep up?
I've only ever run SS on iMacs and MacBook Pros so I'm curious how the performance rates to those out there using Mac minis. Thank you!
Comments
We ourselves have a quad-core 2012 mini for our own internal CCTV system and for testing SecuritySpy, which runs about a dozen cameras at around 10fps each. The mini handles this comfortably with around 75% of the CPU remaining idle at most times. Excellent performance with no lag and responsive web server.
If anyone one there is looking for a Mac mini to run their SecuritySpy system, I would recommend trying to find a quad-core 2012 model second hand (e.g. on eBay), as these are significantly faster than the current dual-core models.
6 cameras usually 2x Foscam FI9803 (one P, one EP, both 720p 10fps), 2x Foscam C1 (720p, 10fps), 2x Foscam 8910E (640x480, 10fps) - the 2 FI9803's doing continuous capture plus motion.
everything responsive - will be upgrading cameras to higher resolution over the next year or so, and likely increasing count to 8 as well.
Security Spy itself uses 8.5-20% CPU, VTDecoderXPCService using 9-25% CPU
(at night it's closer to the 8-10% range, daytime with lots of activity/storm increases load of both processes)
I had originally settled on 10fps a while ago to reduce file sizes - but with newer Security Spy plus mostly H.264 cameras, I can likely increase fps - will do some testing this weekend maybe and post with any changes seen.
With current 10fps the 720p cameras have 1 hour continuous capture files that vary from 275 MB up to ~760MB - a full day's capture+motion (video+pictures) looks to average 12-13 GB
I have things set to keep everything for a month and it's only about 500 GB across all 6 cameras.
My mini also runs Plex media server, MacOS Server, iTunes, BTSync (backup for family computers to mine), iTunes (wifi backups, and playback to airplay speakers) and still is on average 70-80% idle CPU. Chrome across the various tabs & windows I have open uses more CPU time than Security Spy & VTDecoder
Im really happy with this system. Has been stable and I maybe reboot this Mac once a month. The box is dedicated to Security Spy 4 and running in a remote fashion housed with 2 battery backups, Really long HDMI cable and Logitech K400 wireless keyboard. The system is set up for a parking lot with foot traffic and slow moving cars. I set up a friend with a very similar system. He has 7-3MP cameras at full Res and 10 FPS with motion. He is also having the same good experience. It only take moderate intelligence to set up and there is lots of support from Ben and this forum board.
Using less than 30% of the computer's cpu power and ram. I am sure it would go to 16 without a problem, and probably would record as many as 32. The problem would be the older GPU displaying such a high number of cameras. I am holding my breath for a new Mini from Apple this month!
This setup is probably more then the typical user would spend... But this is want I knew I wanted, so I slowly worked up to this. Works flawlessly 24/7. Thank you Ben for creating such reliable software! :-)
I run a Mac mini quad core i7 2012 model, 16gb Ram, recording to 4TB USB drive.
I have on record 4 x 1080p 2MB cameras, and 2 extra 1080P 2MB cameras on view, all running 10fps (reduced from 20fps due to excessive VT usage)
The Mac runs ok, cpu usage is always around 10%, but VTDecoderXPCservice runs at 40% then slowly increases over 24hrs to 500%+ killing me machine.
I then have to stop the SS app and start again, and this brings the cpu usage back down.
I run a home security business, fitting alarms and IP network cameras, so to test and use SS has proven to be a great way to help out customers that want a better motion detection solution. I would like to understand why the VT usage increases over time, but thats a separate issue.
Like others I would like to add the extra cameras on record, but do not need a licence for 8 cameras, so would like to see some changes in the licensing structure but apart from that I highly recommend a Mac mini with SS4.
http://i.imgur.com/U7Tax6o.png
Motion Triggers Off
Audio None
Continuous Capture Off
Motion Capture Off.
One IOS app client on http://
Activity Monitor shows 50-60% of one CPU.
Doesn't seem to impact normal desktop uses (mail, web), but noticeable impact on other high-CPU apps (Minecraft, etc).
Oh, and VTEncoderXPCService and VTDecoderXPCService together are using another 40-ish percent, so my total "User" CPU is around 13%, if that helps.
Going to move SS to a dedicated Mac Mini server, but "not today".
Could you consider supporing a hardware survey tool available that can automatically gather hardware information from users that wish to take it? Valve Software (Steam App) sends these surveys out regular.
I generally have no issues at all. I do get the occasional slow-down, but it's never caused an issues with recordings or live viewing over the web interface.
Processor: 1.4Ghz Intel Core i5
Memory: 4GB 1600 MHz DDR3
Graphics: Intel HD Graphics 5000 1536MB
8 Cameras:
3 @ 6 fps, 5 @ 4 fps
5 @ 1920x1080, 2 @ 1280x720, 1@640x480
All @ H.264 format
Motion capture armed on all 24x7
CPU for all running normally is approx 40% (based on Camera Info Window)
The Mac-Mini is also running extensive Indigo Home Automation with Security Spy Plugin
All run with no issues at all, although if Apple ever bring out a new quad Mac Mini, I would be tempted.
Both Indigo and SS are rock solid and the support from Ben with the regular version improvements and fast responses really set it apart from other products. Especially like that Ben has released the iOS app to complement the main SW too.
Thanks to everyone posting in this thread, it's very interesting!
I'm using a headless 2011 Mac Mini (Server), 2.0GHz quad core i7, 16GB. I changed the system disk to an SSD, but that's only because my 5400RPM system disk failed after ~6 years. Motion capture are done to a USB-attached Drobo. I use an HDMI dongle which I find helps performance, and I recommend if you haven't added one. Aside from SecuritySpy the server runs Plex, an iOS caching server, and some lighter server processes.
I currently have four 2MP cameras and two 3MP cameras, all H.264 @ 8 FPS, motion capture armed on all 24/7. I use non-rectangular motion detection masks, which may or may not matter for CPU usage. SecuritySpy consumes under 15% CPU, and the entire server generally runs under 20% unless Plex is making it work hard. I will likely bump the frame rate at some point.
@cstout asked about the web interface; I rarely use it as I have iOS devices/apps talking to SecuritySpy (I use SecuritySpy, Spyglass, and Remote Patrol -- no real favorite).
I heartily endorse SS adding more CPU-consuming features, like person detection, fog & tree avoidance, and alerting me to cob/spider webs on the camera lens.
Mid-2010 Server with a Core 2 Duo (2 cores, 2 threads)
that's pretty much overwhelmed with 9 cameras. I could upgrade to one of my 'leftover' machines, either a
Late-2012 Server 2.6 GHz i7 (4 cores 8 threads) with Intel HD 4000 graphics
or maybe a
Late-2014 3GHz i7 (2 cores, 4 threads) with Intel Iris graphics
Any idea which would be 'better' for SecuritySpy?
Thanks!
When using multiple similar cameras with H.264 streams you will be able to tell how many are being decoded in hardware by looking at the CPU column in the Camera Info window in SecuritySpy. When sitting idle (all modes disarmed so no recording is taking place), you should see one camera that is using significantly less CPU time than the others - this indicates that it is being processed in hardware. If you see two like this, this would indicate that two camera streams are being processed in hardware.
As far as I am aware, the Mac Pro (at least with its default video hardware setup) does not support H.264 hardware encode/decode. I'm not sure why Apple chose not to add this capability, it seems like a strange decision.
It's now been two and a half years since the last Mac mini update - an unusually long time. I really do hope that Apple will stop neglecting these otherwise fantastic machines, and put out some major updates soon including a quad-core model. In the meantime, I would still advise that the 2012 quad-core mini is the best option for SecuritySpy, and you can pick them up on eBay at reasonable prices.
For all customers running SS on Macmini's headlessly (without a monitor):
As bp33 suggests above, you need to attach an Active (e.g Apple) VGA/DVI/HDMI/DP etc (not Passive PC VGA) adaptor, or HDMI monitor emulator, to restore full (video hardware) performance.
As covered in this section of SecuritySpys useful F.A.Q, entitled "Can I run SecuritySpy on a headless Mac?" here: https://bensoftware.com/securityspy/helpfaq.php#Headless
While CPU Speed and CPU cores, can always be held back by a lack of the other, e.g fast Dual core only CPU, or slow Quad-core CPU (i.e you need a balance of both):
In which case, lower cameras live streaming frame-rate, and/or Resolution (in their own web-interfaces) to compensate.
While making sure SecuritySpy own Performance is Optimised: https://bensoftware.com/securityspy/manual/#performance
When I look in activity viewer, the vtdecoderservice is still at the same CPU usage before and after I plug in the dongle.
I am sure the dongle works, as the mac mini switches to vga resolution when plugged in.
also from the Window menu, there is a "Camera Info" you can select. It shows several things including the CPU usage per camera. As Ben explained it, one of the camera's will be using the hardware decoding from the GPU and will have reduced CPU load, the rest will be using a higher CPU load - possibly the rest will have the same CPU load if the cameras are similar resolutions/frame rates/stream type.
I'm having the same issue VTD is running 12 threads and % CPU >300. CPU Temp was well over 200 Deg F. I have a headless Macmini 6,2 i7 quad core w/16Gb of RAM (the last good mini ever made! snif! ). The fan is maxed out at 5500 rpm most of the time and only 40-50 % idle CPU
Yesterday I shutdown one of my six cameras. And, added the OWC headless adapter per the recommendation above, not sure why that'd have a thing to do with it, but I added it anyway. With 5 cameras and the adapter I'm running ~170 Deg F, still with 12 VDT threads, but % CPU is down to 80 (from 300), and the CPU is 75% idle. My fan is still maxed out to keep it that low [I use TG Pro fan control] Very weird, still at the high end of where I'd like to be given that SS is the ONLY APP running other than Plex & TimeMachine Server.
Today CPU Temp was down to 161 Deg F, but with 16 VTD threads. VTD running with and 80 % CPU. The one thing I notice was the second I opened a Browser (Safari) window on a client computer with SS and all 5 cameras, the % CPU jumped to 140, and the CPU temp rose to 185. After closing the Camera browser window, CPU and Temps went back to normal; idle CPU usage went from 60% back up to 83%. I can live with that as don't continuously check the cameras from my laptop.
So here's the weird part! If I pull up the cameras in Safari from my iPad, I see almost no change (about 2%) in CPU usage.... someone explain that to me. Why would safari on a laptop cause a big jump in CPU usage and safari on an iPad not? Very strange.
Anyway, I appreciate everyone sharing. I don't know how or if the OWC headless adapter really does anything, but for $20, I'm in.
I pulled the headless adapter, didn't see and immediate bump in CPU usage.
I have a similar Mac Mini, core i7 2,6ghz, 8Gb ram.
I have 8 cameras recording non stop, and each camera also does motion detection.
My VTDecoder thing never goes above 30%, and I don't see any increase in CPU when I launch safari, ipad or multiple instances of those...
SecuritySpy is set to not recompress audio or video data from any of my cameras.
I wonder why you are seeing so high cpu usage.
My continuous recording however is a timelapse of 1frame/sec.
This way I have the better footage for motion events, and still have timelapse footage for the rest. Using the timelapse settings, this also saves space on my HD.