There was a bit on FX PHD, where The Creator of Arnold talked about scalability in threading and how more cores doesn't nessesarily mean faster renders
*threads* pixel rendering time *speedup* 1 18.94s 1x 2 11.91s 1.6x 4 7.23s 2.6x 8 9.44s 2.0x 16 12.37s 1.5x 32 13.39s 1.4x 64 14.65 1.3x This is with renderman, which he says has fairly poor scalability. https://www.fxguide.com/featured/the-state-of-rendering-part-2/ So it's not so much the power of the motor but how it uses it. On 23 March 2014 22:10, Mirko Jankovic <[email protected]> wrote: > actually not really.. > you know every rendering takes same amount of time approx. > before we had scan line and software render only. it took like 10 minutes > per frame. couple years later got Pentiums at 200 Mhz and after that 400Mhz > and more.. it still needed almost same amount of time to render frame. > But as CPU got faster also requirements on render got higher as well then > GI and ray tracing etc.. so.. > who say it wont be some new gimmick that will utilize all those cores but > still rendering frames will take standard amount of time :) > > > On Sun, Mar 23, 2014 at 11:00 PM, Ed Manning <[email protected]> wrote: > >> In all seriousness, though. This means that rendering power is going to >> be effectively free very soon. By which I mean that the capital costs of >> any given amount of render power will be insignificant compared to the >> non-rendering production costs for any given project. Which means that >> render times (which will I'm sure remain constant per frame as they have >> for 30 years) will no longer be an issue so much as quickly getting assets >> finished to the point of being able to render them. >> >> Sorta like disk storage was in about 2002, or cloud server availability >> in 2008. >> > >
