I am running PreForm in a Win 10 VirtualBox. All my computers run Linux natively.
I was wondering, roughly how many cores and how much memory should we give PreForm in order for it to perform reasonably well?
I’m currently giving it 4 cores and 12Gb of RAM. PreForm seems a bit sluggish. For example, generating print files for a fairly simple dodecahedron model is taking over 10 minutes.
Is this normal? I imagined it would be snappier than this. It’s possible I haven’t set up my Virtual Machine as well as I should have. I will try doing the same from home tonight. I don’t often use VM Windows machines and I frequently forget how to make the Virtual Windows machine happy.
I have tried rebooting the virtual machine a few times, giving it additional resources. It seems every time I give the machine more ram or cores, they are used.
I haven’t tested on a virtual box on Linux, and it’s not a supported configuration, but that sounds too slow. Can you attach a form file of the 10-minute dodecahedron?
Most of our algorithms try to use all available cores.
I suspect the slowdown might have been due to the graphics card virtualization – I might not have that setup correctly. On my home computer I turned the GPU virtualization off, and gave PreForm 12 cores and 24 Gb of RAM and it was quite a bit more responsive.
I will attach the form file next time I’m in my office.
I will upload the form file next week, unfortunately I forgot to grab it last time I was in the office.
Today I went in and gave my office computer 8 cores and 20Gb of RAM. Preform did not need all the memory, but it did use all the cores. With 8 cores it wasn’t exactly snappy, but it was fast-enough.
So my impression is Pre-form seems to be reasonably snappy provided it has 8-12 cores and 12Gb of RAM.
Will Pre-form use NVIDIA-CUDA processors if they are made available? I have not tried to use them in a virtual machine before.
It’ll be even less supported, but I’ve been running Perform in WINE and it’s performing perfectly well for my needs.
I’ve noticed the model export is slow too (yet strangely seemingly not proportional to layer count nor model complexity), but I’ve got a headless machine with a GPU where I can copy my .form files and leave them to churn (and start printing automatically on a primed, remote print enabled printer when they’re done) while I work on something else on my laptop.
Quitting and relaunching it after each job seems to help, it seems to slow down if you leave it too long, but compared to the pain of running a VM it seems fine.