We're in the process of migrating our VDI enviroment from Windows 7 to Windows 10. Our current config is 2 sockets x 1 core = 2 CPU total and 8 GB of RAM. We currently only have about 20 people testing. Of those there are a small handful of people complaining about slow performance. Mostly when using Webex and IE. I myself am in the same test group, using the same config and I haven't encountered this issue. I'm skeptical that adding more CPU is the answer.
I realize that there is not a "One size fits all" configuration.
I would love to hear from folks going through this and what your recommendations are.
The correct answer here is: It depends.
I have some customers happy with 2 vcpus, others with 3 or 4 (sockets or cores doesn't matter). Windows 10 is however not happy with 1 vcpu.
You should monitor your environment, talk with your heaviest users and test to see what you do.
I am working with a customer that is running 1 socket, 2 cores, and 4 gig of RAM. The image is stripped down to the bare necessities and profile / applications are delivered by UEM / App Volumes. Performance does not seem to be an issue for the customer.
We are currently doing 3 vCPU and 8GB of memory. The image is very well optimized but still seeing some performance hits in our heaviest users. Unfortunately there are always two things: Windows is a resource hog and there are some users that are never satisfied with the performance.
We started on LTSB 1607 and 2 vCPU was fine, but later versions of Windows pretty much required 4 vCPU. I remember reading somewhere that 1809 and higher should be 4 vCPU, but I can't find where I saw that
No, the VMWare guide has been by far the most comprehensive one. I've also found that a slow client computer can significantly impact VDI performance. Also screen resolution plays a big factor in VDI responsiveness, like if you're using 4K monitors.
We were originally running 3vCPUs with LTSC 1809 but after adjusting the scheduler to get around all of the Intel cross thread compromises I realized it was just going to waste 1 thread per machine anyway so I increased it to 4. We didn't really notice any difference between 3 and 4 vCPUs. We DID notice a huge difference when upgrading them from 4GB RAM to 6GB, and another smaller difference when upgrading them again to 8GB. With the exception of 2 apps everything is now running great. (And those 2 apps are just bad apps, they run terribly on dedicated hardware as well)
Yes we do have Stratusphere and vROps both monitoring the environment and we can't seem to be able to find what is truly the cause of our issues.
I've done quite a bit of digging over the past few months to try and find out really what is causing our performance hits.