Hi everyone – we’re excited to have released CloudXR v3.2 this week! This is a big update and it includes several highly requested features from our partners and customers.
Key updates include:
Support for pre-configuring the remote server with a client profile
Networking-focused client APIs to get and set
Swift support for iOS development
Oculus Quest 90hz support
For more information about these new features, visit the latest technical blog. We’re always looking for feedback and can’t wait to see what you build.
We are testing cloudXR 3.2 and the client crashes or we lose connection after a short period of time.
Client Device
Oculus Quest 2
Server :
Steam VR : 1.15.12 and it was the same with the latest version
Windows 10 Pro Version 21H2 19044.1826
Windows Feature Experience Pack 120.2212.4180.0
Processor Intel Xeon Processor (Skylake, IBRS) 2.99 GHz
RAM 16.0 GB
Nvidia driver version : 472.98
GPU model : Nvidia Grid V100D 8Q
Steps to reproduce
Launch the cloud xr server
Launch steam vr
Connect the client
Launch vr application on the server
wait and it will crash after a short period of time
Here are the logs on the server (-v -t -tle) and client (-v -t -tle -tqs) + client application server_logs.7z (72.6 KB)
Tested with a customer.
Public IP with port forward Fail with 3.2 (all needed port opened and tested with Cxr 3.11)
Public IP with DMZ OK with 3.2
seems that there is a problem on some ports
Edit: problem appear on Oculus Quest2 - Not on Android Smartphone
We’ve noticed some symptoms of connection problems that we’re still tracking down concerning vGPU VMs. Is this session hosted within a VM? If so, is it using vGPU or passthrough? Which version of vGPU are you running?
We’ve discovered a bug in vGPU that prevents CloudXR 3.2 (and CloudXR 3.1) clients to connect properly/at all to servers hosted on vGPU. A workaround is to enable pass-through on the VM if possible or revert back to CloudXR 3.0.
Regarding cloud providers, this will impact Azure instances which leverage vGPU “out of the box” to support fractionalized workloads. This does not appear to impact AWS instances.
Big issue, we are working on R&D project for the two biggest Ita Telco company
and all the hypervisor are VmWare in Telco Datacenter, with shared Tesla T4 (Two Vm’s for GPU).