Does appleseed support multithreaded rendering? And if so, up to how many cores?
Yes it does, up to as many cores as you have.
I was just wondering how does that work. I ran appleseed studio on my windows PC on appleseed studio which has 4 cores 8GB RAM + SSD harddrive and I ran it on a linux server with 32 cores 32GB of RAM on appleseed cli (same scene) and my windows PC was faster by 2.5 . Do you have an idea of why this would happen?
What scene did you render? What was the resolution and the number of samples/pixel?
Thanks. The number of spp is unfortunately not displayed in the logs but by default the sample scene uses 64 spp.
Was the Linux server a physical machine you had exclusive access to, or an AWS instance? Did you check the renders produced on both machines, and did they match?
The Linux server is a cloud instance, although not AWS. My windows render matches the gallery render. I have not yet got a final linux renders since I keep underestimating how much time I would need, so I request cloud resources for not enough time and my render gets cut off by the cloud system. I tried bumping my core count to 64 and memory to 128 GB, but I don’t see any increase in speed.
I’m terribly sorry, I think I fixed it. It was an issue with the cloud scheduling policies. Thanks for answer my questions. Sorry for wasting your time. Although, it helped me that you said that appleseed supports multithreading, even it is almost painfully obvious that a renderer should support this, it could be nice to have it mentioned explicitly in the features.
You don’t need much memory for this scene, 8 GB is absolutely plenty, 4 GB would probably do just as well.
I was wrong on the number of spp. By default the scene is setup to render 64 passes of 16 spp each for a total of 1024 spp. No kidding it’s taking ages
You can set both the numbers of passes and spp on appleseed.cli’s command line:
./appleseed.cli emily.appleseed --output emily.png --samples 16 16 --passes 4
--samples 16 16, the first value is the minimum number of spp and the second value is the maximum number of spp. However this scene is not using the adaptive sampler so only the second value is taken into account.
Here it takes about 1 min 10.
Ahah, no problem at all! Glad that you got it working.
Note however that it’s likely that appleseed doesn’t scale linearly to 32 cores. It may be that this is only marginally faster than, say, 16 cores, due to various factors (memory bandwidth, overhead, number of spp, resolution, tile size…).
Oh I see, thanks again franz. Sorry again.
Yes, it is.