Discussing multi-user systems involves understanding how multiple users access and share resources simultaneously.
Discussing multi-user systems involves understanding how multiple users access and share resources simultaneously.
Okay sorry about not being very clear at first. This is my first ever forum post btw. This is the scenario. My friend has a makespace with things like 3d printers, laser cutters, can Mills and green screens. People who work on these machines often have to bring design files from their homes because we don't have any decent PCs at the makerspace. I thought we could save money by having a really powerful PC with several peripheral stations. This would allow people to work on their individual projects without the need of several workstation grade PCs. Windows is the only option because we use things like Corel, Autodesk & Adobe. Pls keep your answers coming. If it is cost effective we might actually do this. Thanks for the help so far.
You need to evaluate how these designs compare. How challenging are they? Could they run on something like the 860K from AMD with a GTX 960 or another budget card? I’m not very familiar with the exact hardware needed, but it’s feasible to have several video cards in one system and create different workstations from that setup.
In general, you can achieve this using either VMware ESXi or Linux with KVM. I’m not comfortable with the ESXi approach. Check out the referenced document. For KVM, the most reliable guide is likely the one you mentioned. Start with the initial post to grasp the concept, and review the final 10-20 pages for newer techniques. You should consider using libvirt with OVMF. The optimal hardware setup would include an E5 or higher Xeon processor, a server motherboard, and AMD graphics cards. You won’t need any additional kernel updates or software patches. This was my temporary configuration and performed well: E5-2630 v2 Supermicro X9SRL-F GTX 780ti
I understand there are two options: 1. Utilize an older driver version (340.xx perhaps), which might not work with the 970. 2. Disable hv_time and other hyper-v flags, as newer QEMU versions could support it automatically. You may lose about 10% performance but at least it will run.
It's feasible through simpler methods using a Windows MultiPoint server on the primary overkill machine. Connecting a service named Zero Clients would yield the desired configuration. You can achieve this with various models—some more affordable, like around $20 each, while others offer four USB ports without PS/2 or an Ethernet jack. The more costly option isn't necessary unless you don't require PS/2 connections or additional ports. These thin/zero clients only use standard USB ports. Despite the software cost, individual CALs or user codes are likely required. EDIT: This appears to be the most economical choice.
A 10% improvement on performance is okay using VMs. I may need to check if disabling hyper-V flags helps. Or I should test if Xen works in this situation.
I think the issue stems from NVIDIA intentionally disabling drivers when virtualization is detected, especially on lower-end hardware. It seems the setup might depend on specific features like <features>, <kvm>, and a hidden state enabled.