This question was asked over on the RepRap forums but I figured it would get a more direct answer here.
If someone wants to build a new machine on which to run OpenSCAD, what would have greater benefit: faster CPU or more GPUs?
As far as I can tell from empirical testing and reading old threads in various places, OpenSCAD doesn't take advantage of multiple cores or use extra GPUs for rendering so is it true that the fastest processor (in terms of raw MHz) will always win out?
When and how this will happen is hard to say. Both are proposed as topics for http://semesterofcode.com/ so if anybody knows students/universities who might be interested, point them there :). The application phase is currently open.
(I think it's a European project, so only universities in Europe can take part)
Re: What improves rendering speed most? CPU or GPU?
On Tue, 21 Oct 2014 08:19:25 -0700 (PDT)
Bananapeel <[hidden email]> wrote:
> High single thread speed is the important factor. However, not all processors
> deliver the same performance per mhz, so don't look just at the specs.
> Go here: http://www.cpubenchmark.net/cpu_list.php >
> Click on one (or select more than one to compare) and look at the number
> "single thread rating". Don't be fooled by the big red number, which is the
> multi thread rating.
Your CPU performance is also dependant upon the thermal properties and
power envelope of whatever its stuck in. For continuous power draw like
openscad that'll matter.
The real fix though is to sort out OpenSCAD - even the serial rendering
can probably be fixed to be at least 100 times faster just by moving to
fixed point integer maths.