There is, sadly, no "one fits all" solution... Between dGPUs with proper OpenGL VRAM reports (NVIDIA's under Linux and Windows, AMD with Mesa drivers under Linux), dGPUs with bogus such reports (AMD under Windows), and iGPUs/APUs without any VRAM and totally silly "VRAM" reports (AMD/Intel), there is simply no
reliable way for the viewer to guess which is right...
You currently have two ways to override the bogus VRAM reports by OpenGL drivers: either set the LL_VRAM_MB environment variable (and yes, it should also work under Windows) or the "VRAMOverride" debug setting to whatever amount (counted in MB) is suitable for your system (*)... On such systems, disabling the "VRAM checks" in the Graphics preferences may also help.
This said, if you got a better algorithm than the one implemented, please be my guest and contribute it !
![Very Happy :D](./images/smilies/icon_e_biggrin.gif)
(*) I will expose this setting as a spinner, for next release, in the Graphics settings: at least, this way, it will be easier for people to fix VRAM-related issues...