that’s par for the course on AMD cards, if you’re running one.
The reason I ask is that my actual, real installation of blender renders the the same bmw .blend much faster, and I’m sure that’d be the case if I loaded up the other source .blends as well.
the data may be taking into account GPU renders in that data
a core i3 with a 1070 will render something much faster on the cuda backend than just a cpu render on a ryzen system will.
I ran Blender workloads on Vega just fine before.
last i checked AMD support is in the experimental feature set.
that doesn’t matter. If this is based on the stable branch of 2.79, and they’ve disabled experimental features to make sure it works, there may be AMD compute conflicts.
try running it from the cli after passing the CYCLES_OPENCL_TEST-ALL env variable
Are you talking in Blender? Because I don’t have it installed anymore.
no with the benchmark
open up cmd, navigate to the benchmark folder with cd, type set CYCLES_OPENCL_TEST=all then execute blender-benchmark.exe from the same command prompt
if that doesn’t work try set CYCLES_OPENCL_TEST=true
crap that was supposed to be an = not a -
my bad
so CYCLES_OPENCL_TEST=all
Well, running Full Benchmark. Pretty sure it was just running Quick CPU last time. CPU is still at stock.
Edit: taking full CPU load but no GPU.
you have to do a gpu run for it to use the GPU. That’s how cycles works.
did you try passing the (corrected from typo) env variable?
I tried?
Honestly I’m about to give up, I have other things to do this evening.
you just do exactly what you did before but without the typo. the gpu test should run.
It doesn’t.
It spikes and then drops off to nothing again.
does the gpu show up as an option for the render test?
Yes and I chose it. I give up for now, need to eat dinner and spend time with the wife. I’ll do my best to try more tomorrow evening.
then my guess would be that
