I've just finished building a new control room at work. It has 32 monitors and the plan was to have a single computer powering it. The old room had a few computers with odd screens keyboards/mice everywhere and decided it was time to simply things and have a single pc - with it been a single operator most of the time.
Theres not an awful lot of demanding stuff running on the machine, some scada packages, IP camera viewing software, office etc.
The issue that im having isnt down to performance. At least I don't think, the computer is of a fairly high spec. It's a HP Z840 with 2 Intel Xeon E5-2670's 4 nvidia nvs810s and 256GB of ram. The operating system is windows 10 enterprise 64bit
My slots are used as follows.
- PCIe3x4 - None
- PCIe3x16 - NVS 810 1
- PCIe3x8 - None
- PCIe3x16 - NVS 810 2
- PCIe3x8 - NVS 810 3
- PCIe3x16 - NVS 810 4
- PCIe2x4 - None
I've realised after looking at the manual]1 that I should have gpu 3 in slot 3. However the behavior of the machine is strange, I connected all 30 at first and most came on with the windows background and task bar. about 10 had no background but had the taskbar. The mouse moved at a snails pace and I was unable to postion the screens in nvidia control panel as it would crash/freeze. I unpluged the cables from gpus 1 and 2 and managed to get 16 screens on from cards 3 and 4. When I got to screen 21, 5th screen on gpu 2 the machine when crazy again. The mouse started to lag again, and some screens were showing as duplicates of each other.
I've had a look in task manager, I've not seen the cpu or ram go any higher than 4% when it locks up its just nvidia control panel that is not responding.
I'm thinking it must be some sort of bandwidth problem but not sure how to prove this or fix it.
Should I be able to get 32 1920x1200 screens of this hardware?
Is this behavior normal? I will try moving NVS 810 3 to slot 3 and see what difference that makes, any other ideas would be appreciated.
The screens are arranged in a 8 by 4 matrix.