I’m not bragging, but only using 16 cores you can now compile all the shader binaries in 32s. If I used all 32 it would be a waste of power bill.
Follow
@thendrix Not to be that guy but if you were using a VLIW architecture, you might get better performance.
But jokes aside, I'd suspect that 8 cores might be the fastest because of the distribution across the die and memory bandwidth.
Also I just realized that might have been a jab at the scalability.
The reason it’s so slow is that I’m having to use I/O to work around a broken upstream library. Basically the NVMe is the bottleneck. Once they fix the upstream library I can do most everything in memory.
That said for UE4/UE5 this would take 30-40 min.