Should you buy a 3990X CPU for rendering?
Jamesy writes:
When the 3990X dropped in February of this year, it certainly landed with a thud! But is it a good purchase for rendering? Is that even a sane question when its clear its faster than anything else out there CPU-wise right now?
If you’re thinking about purchasing either a new workstation, or perhaps even building a render farm. After crunching numbers, I have some insights I’d love to share to answer these questions :)
My answer to that question – is no.
The newer Redenengines were all GPU renderer – they need a fast graphiccard and not a fast and extrem expensive CPU. There are Renderfarms with GPU only servers with 10 and more Titans in it. Those GPU render “workstations” are a lot cheaper then the CPU ones
It all depends on your other requirements I guess.
Rendering / comping all day, the answer probably is yes.
Like with simulations, you want more more cores…. So yes in that case too.
Video-editing, not really or maybe.
And in the end it all depends on your budget. Look at the the new Mac Pro for instance… ;)
But great article, thanks for the testing :)
It’ll be great for after effects rendering
Not necessarily, AE doesn’t scale very well to multiple cores
True, but multiple processes on multiple machines rendering individual frames do.
If your rendering needs surpass the memory limits of a GPU (as is often the case with archviz), then yes, a high-core CPU is worth it. Otherwise you get more bang for your buck with GPU. Personally, I’d go with the cheaper 32 core threadripper and pair it with a good gfx card or two. With hybrid rendering you can render with both at the same time.
La respuesta es pues claro!! Rizen rize!!!!, Sin lugar a dudas, quien tiene como costearlo go for it pa quien no, esperar que baje en el mercado. Pero excelente adquisición para renderizado, v Ray, blender, maya, unreal 4
When buying these machines the value is more than what’s spent. It’s also what’s earned.
If I render 3D videos at $1000 per output minute and one machine takes 1 hour to render a video and the other takes 40 minutes then my income is greatly increased if we are able to sustain 100% production. Additionally if we are able to produce faster we may be able to charge more.
That’s a dumb metric. “Closest to origin” is not useful, even after “normalization.”
“Closest to origin” is not a valid metric here, because the function is not linear (planar.)
If my time is worth 100 bucks an hour and I only buy one CPU, then the faster one will almost always win, unless I render very little.
If I can buy more than one CPU, then the cost of system overhead also matters — Mobo, ram, disk, power. And I still can’t buy a fractional CPU.
So, let’s say I wait X hours in a year for rendering, and value my time at Y dollars per hour, and only use one CPU, then a much better graph can be drawn, of actual “total time plus dollars” cost.
I’m all for using data, but not abusing data.
Hells yes you should! The 3900x is the best power for dollar ever. We are running them slightly overclocked at 4200 and they are faster than our stock threadripper 2950x at nearly half the cost. Replacing our quad core i7 5820s is a 75% speed increase per machine. Do it.