this post was submitted on 17 Nov 2023
0 points (50.0% liked)
Framework
76 readers
1 users here now
Discussion around the Framework mission of building products that last longer by making them upgradeable, customizable, and repairable. Consumer electronics can be better for you and for the environment.
founded 1 year ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
Have you actually measured any performance difference?
According to my understanding, there should be no performance difference at all (hardware should not care). If there is, it would be caused by games & applications treating the iGPU wrongly, as if it was a dGPU and hence using that value in wrong ways that improve or worsen the performance.
No it's on my to-do list now but I would imagine having pre allocated the ram for the GPU would reduce memory management overhead.
It's anecdotal but (on Linux at least) Steam in 4K locks up and the driver appears to crash with Auto but does just fine with the reserved 4GB.
I'd think you'd have a very hard time to get reports other than anecdotal ones about this specific topic, so totally fine. As I have no AMD iGPU available to test, I have no idea how common that kind of problem is.
But sounds a lot like either the AMD Linux driver or the Steam Big Picture implementation for Linux does things wrong. I believe I explained it in other posts, that I would not expect it to directly impact performance, but any program that treats this like a dGPU and tries to manage memory on a low level will probably handle things very wrong when only tiny amounts of memory is allocated.
And while I have no deep knowledge of GPU APIs, which API is used should influence greatly, how likely it is for an application to bake in a way that only works for dGPUs or high amounts of UMA.