New AMD Ryzen

Ryzen7 2700x 8 cores/16 threads + Gtx1050 works extremely well. Much better than i7 8700 (6 cores/12 threads) alone on the same camera setup. 12 cameras 3-4 MP each. BI process at night consumes 5%, during the day max 12%, average 7-8% of the CPU. Nvidia 10%. This new AMD CPU is very good.
no its not and nvideo has been tested to be inferior and a powerhog...but nice try...most likely you didnt use the correct intel driver for the 8700...but if you like spending more on a pc and burning more power that is your choice 4.7.8 - August 28, 2018 - Support for the Nvidia CUDA hardware decoding
You dont need an 8700 to run 12 3-4mp cameras...you can run it on a much weaker older 4700 at 20-30 percent load. Much much less on a properly configured 8700.
 
Last edited:
  • Like
Reactions: bp2008
It is properly configured computer. I had parts for AMD and Intel computers and I built both, Fresh Windows 10 on both. Drivers are proper, the latest Intel and the latest Nvidia, I know how to do it. I uninstalled all pre-installed apps from W10 using powershell and I uninstalled all of them including Edge, Marketplace and all tail apps.
This week I will test what will happen when I install 1080Ti into i7-8700 system. There should be a way to make both running at the same time. Maybe I will be able to to use H.265 with cuda nad all others with intel quick sync.
Anybody tried it ?
 
  • Like
Reactions: asilva54
It is properly configured computer. I had parts for AMD and Intel computers and I built both, Fresh Windows 10 on both. Drivers are proper, the latest Intel and the latest Nvidia, I know how to do it. I uninstalled all pre-installed apps from W10 using powershell and I uninstalled all of them including Edge, Marketplace and all tail apps.
This week I will test what will happen when I install 1080Ti into i7-8700 system. There should be a way to make both running at the same time. Maybe I will be able to to use H.265 with cuda nad all others with intel quick sync.
Anybody tried it ?
You dont need the 1080ti in any of those systems with your load. its a complete waste of money and power.
 
  • Like
Reactions: bp2008
Hi guys,

Just wanted to make an update on this thread. I have been using Blue Iris for a number of years (still learning of course !!) exclusively on Intel based systems.

My latest system was an Intel based system i7-4790K, 16GB of RAM + hdds. I normally to direct-to-disk recording 24x7 and have enough hdd space to hold recordings for about 3 weeks (using 3 hdds). My cameras are:

7 x Dahua IPC-HDW5231r-z
2 x Dahua SD49225T-HN

All cameras are set at 25fps, full 1080 resolution

All worked very well on my Intel system. During the night I saw about 10% CPU usage (+ about 55% on the GPU), while during the day it was hovering around 20% with peaks to a maximum of around 30% CPU usage (+ 55% or so GPU usage).
System was very responsive even when I RDP in to it and open BI Management screen (usage will jump to about 50% +).

Now, I have a few other physical systems around the house - namely a SuperMicro based firewall and another computer used as a virtualization server.
Lately I was looking for ways of consolidating as many physical systems as I could into one single computer in the name of saving power and freeing up some space around the place.

I have decided that an AMD Threadripper will be my system of choice for this (1950x with 32 cores, lots of ram - I have 112GB in it right now - Samsung Pro NVME M2 drive for the OS components only of all the virtual servers, a few spinning disks for all the data partitions of the virtual drives/backups, a couple of NICs to pass-through to virtual servers where required).
So I built up the system using ProxMox. All my existing virtual servers were migrated to this new box plus the other few physical servers too. One day when AMD will fully fix the GPU passthrough my gaming desktop will also be migrated to this too.

The only server left was the Blue Iris server. So I decided to test. Shut down the i7 server and built a new Windows 10 virtual server on my Threadripper box. I gave it 8 cores and 16GB of RAM. I took the HDDs out of the i7 box and pass them through to the new Win10 BI virtual server. The new BI virtual server saw them as they were - intact with the existing partitions and all the data (recorded videos) still on them. I applied my exported BI settings from the i7 Blue Iris box to this new BI visualized server.
System is happily direct recording to disk 24/7 (just as before). Is not skipping anything. CPU load is consistently sitting around 20% to 30% with 6 of the assigned cores doing most of the work and the 2 remaining cores only seeing utilization over 20% when I RDP to the virtual server. Never seen the cores spike up more than 60% even with BI open watching all cameras. Alerts work just as before. Delay from live camera feeds are just they were on the i7 system. I can view recordings and alerts just fine. Memory usage is 5Gb at the very top so I might go with only 8GB instead of 16GB.
Overall I see nothing that is different than when I was using the i7 system. Nothing at all. The move to the Threadripper based virtual BI virtual server happened about 1 week ago. Yesterday I switched off my i7 BI box. I realized I would not need it anymore and from now on I will be using the new BI virtual server.

I must admit that initially I did not believe this new AMD Threadripper system will work well for this - moving my i7 BI to a virtual server was NEVER part of my plan at all especially on an AMD system!
Turns out I was wrong and will be keeping the virtualized system instead of the physical one.

I will end this by saying that I just wanted to update this with my experience. I am not saying that Intel is better than AMD or vice-versa. What I am saying is that what I have now (virtualized BI server on AMD Threadripper CPU) works just as well as my previous i7 BI physical server and I will stay with this for now. Plus with the Threadripper 1950X having 32 cores in total I have room to add more cores to my new BI virtual server if I decide to add more cameras (or higher MP cameras) in the future. This new computer It allowed me to achieve my goal - consolidate 3 physical computers into 1 so i can save power. Plus in the near future (I hope) I can add my gaming desktop to it with proper GPU passthrough.

Just as an overall reference - my Threadripper system running around 20 virtual servers in total (including the BI one) has a total overall usage of about 15% MAX across the 32 cores. So yeah plenty of room to add more virtual server or more core to the BI virtual server.

Edit - just red the post above mine - I hope I am not a complete idiot Fenderman for going with an AMD system that works for my needs and performance is on par with my previous systems :) :) :)


2 yrs (almost) update:

Only a few things have changed since my original post.
Hardware wise I am now running 2950X Threadripper CPU but same motherboard, and RAM (128GB).
The rest of the hardware is pretty much the same maybe some hdds added.

But I have a LOT more software running on this Threadripper ProxMox host - about 50 Docker containers distributed across 6 virtual Linux hosts, 4 oher Linux virtual servers, about 5 virtual Windows 10s, virtualized pfsense firewall, virtualized OpenMediaVault and of course a virtualized Blue Iris 5 Windows 10 with the Free AI Person Detection for Blue Iris I also have a couple of Samsung Andriod tablets displaying BI 24/7 through the BI client.

Just like 2 yrs ago everything is running without any kind of issues at all. Running everything I get between 17% to 25% CPU utilization.
Memory Utilization is at 80GB out of the 128GB.

BI is very responsive, running as a service with local console open (when I need to) and the 2 Samsung Tablets connected 24/7.

The Blue Iris AI person Detection is the BEST BI add on (in my opinion) I have seen.
I dont get any more false alerts at all (no tree moving, shadows, etc false alerts). These days when I get an alert I am 100% confident that is a real alert (I only alert for persons) and i have to react to it (watch it on the BI android client).

My stats from running AI Person Detection since June 2019 are:
AiStats.jpg
Next best thing apart from no more false alerts - is that AI person Detection is FREE!!!
If you have not installed it do yourself a favor and install it, configure it and get it working with BI!!!

So over all still pretty pleased with my initial investment in the Threadripper system and I still have room to grow and add more as needed.
 
  • Like
Reactions: pcunite
Great to see the diversity. More and more people have home labs, especially with plex being so popular, that we need to push for no more dedicated BI boxes, I rather have 1 modern, beefier server than 2 mid size ones because BI hasn't caught up with the times
 
2 yrs (almost) update:

Only a few things have changed since my original post.
Hardware wise I am now running 2950X Threadripper CPU but same motherboard, and RAM (128GB).
The rest of the hardware is pretty much the same maybe some hdds added.

But I have a LOT more software running on this Threadripper ProxMox host - about 50 Docker containers distributed across 6 virtual Linux hosts, 4 oher Linux virtual servers, about 5 virtual Windows 10s, virtualized pfsense firewall, virtualized OpenMediaVault and of course a virtualized Blue Iris 5 Windows 10 with the Free AI Person Detection for Blue Iris I also have a couple of Samsung Andriod tablets displaying BI 24/7 through the BI client.

Just like 2 yrs ago everything is running without any kind of issues at all. Running everything I get between 17% to 25% CPU utilization.
Memory Utilization is at 80GB out of the 128GB.

BI is very responsive, running as a service with local console open (when I need to) and the 2 Samsung Tablets connected 24/7.

The Blue Iris AI person Detection is the BEST BI add on (in my opinion) I have seen.
I dont get any more false alerts at all (no tree moving, shadows, etc false alerts). These days when I get an alert I am 100% confident that is a real alert (I only alert for persons) and i have to react to it (watch it on the BI android client).

My stats from running AI Person Detection since June 2019 are:
View attachment 51315
Next best thing apart from no more false alerts - is that AI person Detection is FREE!!!
If you have not installed it do yourself a favor and install it, configure it and get it working with BI!!!

So over all still pretty pleased with my initial investment in the Threadripper system and I still have room to grow and add more as needed.

I have Qnap server, now bought Dell Optiplex for BI (possibly to run along with Plex, Sonarr), was considering buying some hardware for pfsense but I like your solution to consolidate everything into one powerful physical server and run everything as virtual servers.

Would you mind to share what are you running on your server? Anything useful? Is pfsense act as your main firewall from outside world? I assume your other virtual servers receive filtered (firewalled) traffic (all internet traffic have to go through pfsense first?)? And you can setup up that via ProxMox?

I tried VirtualBOX in the past but it seemed pretty sluggish solution. Is ProxMox any better? How do you utilize it with docker?

Is it a rack mount server? Can you take a photo? :)

Thanks!
 
I have Qnap server, now bought Dell Optiplex for BI (possibly to run along with Plex, Sonarr), was considering buying some hardware for pfsense but I like your solution to consolidate everything into one powerful physical server and run everything as virtual servers.

Would you mind to share what are you running on your server? Anything useful? Is pfsense act as your main firewall from outside world? I assume your other virtual servers receive filtered (firewalled) traffic (all internet traffic have to go through pfsense first?)? And you can setup up that via ProxMox?

I tried VirtualBOX in the past but it seemed pretty sluggish solution. Is ProxMox any better? How do you utilize it with docker?

Is it a rack mount server? Can you take a photo? :)

Thanks!


Hi,

Will try give you some information about my system.

Gigabyte x399 Aorus Gaming 7 motherboard, with 128GB of Kingston value ram (2600MhZ) in a MountainMods Extended Ascension CYO.
Threadripper 2950x cooled by Noctua
10GB network from ProxMox to my main switch.
Quad port Intel NIC.
LSI Logic / Symbios Logic SAS2008 PCI-Express in IT Mode (passthrough to OpenMediavault) with a few hdds attached.

On the physical hardware I have installed ProxMox as the operating system using a Kingston 120GB SSD. After 2 years the wearout is at 0%
The main data drive where most of my virtual servers are created on and run from is a 1TB Samsung 960 PRO nvme drive (800 TBW) You need an m2 drive with a relatively high TBW otherwise it will die quickly. Currently at 9% wearout after 2 years.

I also have other spinning disks - some for my virtual BlueIris server, some for my virtual OpenMediaVault server and others for ProxMox (to run disk intensive applications/dockers - GreyLog for example).

pfsense (virtual) is the main firewall for all my network. All traffic goes in/out through it. 1 port of the Intel Quad NIC is passed though to pfsense for WAN connection. The rest of the NICs in pfsense are virtio NICs. Performance is the as same as when I had pfsense running on physical server. Cant tell any difference for better or worst.

I have 6 minimal Ubuntu 18.04.3 virtual server. Docker-CE + Docker Compose installed on all 6 of them. In total 60+ docker containers (just counted them) - way too many to list here. Not using docker swarm. Each docker image is pinned to a particular Ubuntu virtual server.

In addition to the docker images above I also have a few virtual servers.

LAMP server (Linux Apache MySql PHP)
PMG (proxmox mail gateway)
2 x Debian 9
OpenMediaVault
BlueIris
5 x Win10
KodiHeadless
and pfsense

Nothing is ever sluggish in the system at all, everything performs very well.
 
Last edited:
  • Like
Reactions: bp2008 and visata
Noticed in another thread that to totals (on the bottom of the Status > Cameras tab) are the best way to enable comprisons between systems. Below is mine.

Totals.jpg

Edit: just to clarify - only have 9 Cameras. The other 9 are clones for AI Tool.
 
Last edited:
that is wrong, sharp motion is entirely the shutter speed its running at, FPS has absolutely nothing to do with clarity in motion.. I can run 60FPS and get a totally blurry image with a low shutter speed, or I can run 2FPS and get a nice crisp image with a high shutter speed.

Know I'm quoting a very old post here but just to clear this up,I agree with Nayr. In a simple camera such as a point and shoot or a dashcamera, the shutter speed is often fixed by the frame rate as it's simply a function of 1 second divided by the number of frames (minus the lag times for shutter operation etc). So setting 60fps with one of those simple cameras would give a shutter speed of around 1/60th. However, CCTV and most more advanced video cameras don't operate that way. They operate so the frame rate and shutter speed can be set independently. So you could have 2fps but a 1/60th shutter speed. All it would mean is 2/60th's of the second were captured on video and 58/60th's of the second weren't. What that leads to is jerky movement where objects have moved large distances between frame captures , not blur. With such a ridiculous example, it might also lead to a flashing image due to the period of black between the images, however to a more reasonable example, that won't be the case. Blur is entirely a product of shutter speed (although the motion of the object does dictate the shutter speed required). Smooth motion is a product of the frame rate and it's relation to the shutter speed in so far as it's possible to have a low frame rate and slow shutter producing a smooth blurry motion. However, realistically, with a sharp image from a sufficiently fast shutter speed for the object at hand to be captured sharply, it's likely to be the frame rate that dictates the smoothness not the shutter, as it's the frame rate that dictates how far an object may appear to have moved between frame captures and whether it's noticeable. It's only when you fix frame rate and shutter speed inextricably together in a way where one dictates the other, such as a dashcam where the frame rate dictates the shutter speed, do the lines for the cause of blur become crossed. Even here though, it's the shutter speed that's responsible for the blur, it's just the frame rate is dictating the shutter speed because of the simple way such a camera is set up.
 
Last edited:
With such a ridiculous example, it might also lead to a flashing image due to the period of black between the images,

Only if the video player used "black frame insertion", which is only really viable when frame rates are high enough that you won't notice the black frames being inserted.
 
  • Like
Reactions: CCTVCam