Virtual Machine Video Playback Issues In Premiere

With COVID-19 impacting our ability to work in-office, my company has set up Microsoft Azure virtual machines, allowing us to continue our work in Premiere and After Effects remotely. We connect to this through the Teradici PCoIP client. This is all great and I feel fortunate that I'm able to continue working from home. The problem is that it just doesn't work well. On the plus side, our IT dept is open to scaling the power of these machines however we need them to.

Here's the current setup:

- Windows 10 Enterprise, 64-bit

- Premiere Pro v14.0 (Magician)

- CPU: Intel Xeon E5-2690 v4 @2.60GHz 

- RAM: 112GB (I don't have more detailed specs on DDR speed or architecture unfortunately)

- GRAPHICS: NVIDIA Tesla M60

- Virtual machine connects remotely via a 1GB hardline to our AVID Nexis media storage at the office.

Issues:

Ruling out internet speed/connection issues, I'd appreciate some assistance troubleshooting the hardware to give us a more usable workflow. Export and rendering speeds seem to be fine, but the timeline playback issues are a nightmare. Constant lagging and freezing. Nearly all video needs to be pre-rendered before any significant playback can be done. Going frame by frame to make precise edits is also inconsistent and laggy. Sometimes even pre-rendered "green bar" sections don't play back clean.Often get "unknown render error" prompts when timeline rendering. Premiere & After Effects will both crash on me at least once a day (if not more).

Troubleshooting:

- Was originally connected through VPN for company security requirements, but have been testing through a direct public connection and have had no noticeable improvements.

- I've tested with my cache & preview files stored on both the C drive and our dedicated "cache" drive on Nexis. Both options have plenty of space.

- Have 106gb of RAM allocated for both programs

- Tested editing with media stored both directly on the virtual machine and our Nexis server, with no noticable difference.

- My first thought is the graphics card is the bottleneck. I've read about Tesla cards being used as secondary cards, but have had trouble finding any information on using them as a standalone. Are they able to  make use of any CUDA acceleration within Adobe or their Mercury Playback? Within Adobe, I'm able to select the Mercury Playback GPU Acceleration option, but I don't know if that reflects whether or not the card is optimized for it.

To be honest, I'm a bit out of my depth trying to troubleshoot a remote virtual machine that's part of a larger server farm. I don't know what graphics card options are available in this scenario to optimize our workflow (if the card is the bottlneck). I realize this is a complicated issue with a lot of variables, but would appreciate any insight you guys could give me.