Search Issue Tracker
Fixed in 2022.1.X
GPU lightmapper isn't detecting the discrete Intel IRIS Xe MAX GPU with 4GB memory
Open a project using the GPU lightmapper on a machine with a discrete Intel® IRIS Xᵉ MAX GPU with 4GB dedicated memory.
The OpenCL driver is reporting just over 3 gigabytes of global memory but I would expect the full 4GB to be available.
Both the integrated Xe GPU and the discrete one are showing up when enumerating the devices, but the Iris Xe max is filtered out.
The Iris Xe max is available for baking.
The amount of memory we reserve for the driver is empirical, based on our experiments with AMD and Nvidia drivers and is based just on the type of workloads the Unity editor generate. Getting a number from the driver is of course ideal but is not part of the OpenCL spec. The way we identify compatible devices is based on global memory size so one option would be to allow 3GB+ devices if they are from Intel and don't subtract the kOpenCLDriverReservedMemory chunk later. I think I have convinced myself that this is the best solution as it will work on current drivers too.
All about bugs
View bugs we have successfully reproduced, and vote for the bugs you want to see fixed most urgently.
- Light's Culling Mask does not work with Scriptable Renderer Feature when having only one Culling Mask
- "TouchScreenKeyboard.hideInput" triggers numerous errors & breaks function when built for Android
- Meta Quest performance loss between URP versions when built
- Lowering Cloud Layer channel opacity beyond a certain threshold results in graphical artifact
- [Android 5.1] Player is stuck on splash screen when opening a specific project