Ok, I borrowed a nasty machine to test, dual xeon 2.8ghz (24 core 48 thread), titan x card, 64 gb of RAM a promise raid and a few SSDs. (it was a video editing box in a past life)
I’m seeing one thing I think is odd when trying out Infiltrator and other large demos. A first-time load of infiltrator takes about 15 minutes. The odd thing is that none of the parts of the computer ever seem to max out during the build, even when it launches a huge number of shader compiles at once. The overall CPU never goes above 40% and is usually less, it never uses more than half the ram or a few percent of the disk bandwidth.
So it’s not constrained by disk, ram or CPU, so why doesn’t something go up to 100%? I’m wondering if there is something configured wrong on the machine, if something on unreal editor needs to be configured for more threads or if an unreal build/load just isn’t parallel enough to keep 48 threads busy?
I’m using the binary download of UnrealEngine, haven’t made any mods to the settings. The machine is a fairly standard Dell 7910 workstation box. It doesn’t seem that much faster than my Dell 7810 which is 4 core, 8 thread 3.6ghz.
I’m wondering if I should stick to machines with a smaller number of faster cores for unreal work if builds won’t keep a 24 core machine busy?
Any ideas?