NEW HARDWARE SUPPORT XEON PHI

[zacuttialberto]

HI to everyone,

I would like to ask to have a serious explanation regarding possibility to use XEON PHI with Corona Render.
Actually i moved from Octane Render to Corona Render because i dident like the result of corona with artificial lith and the material system but the speed of rednering and the easy way to update the  hardware with GPU render it's mazing.
So Actually As i understud corona render is Embree based render so has to be easy to make it work on XEON PHI.
Let me know please because i'm plannning to set up a new rendering and modelling workstation and this will be a base of future configuration.

Best Regards

Alberto

[PROH]

I don't get it: You moved FROM Octane TO Corona because you didn't like the results of Corona?

[FrostKiwi]

has to be easy to make it work on XEON PHI.

Ondra has stated before, that it's not as simple as it might seem.
Yes, it's possible to port the code over to XeonPhi, however it would not make the best out of the performance the phi may offer.
It's one thing to port it, it's another thing to make it go fast and optimize for that given plattform. Due to the current memory limitation it's out of the game anyways.

Don't take my word for it, but there are no plans to either port or optimize for Phi, because it's a substantially different architecture. It's x86, but it still requires to be coded in a way for it to utilize all of dem power.

tl;dr: Achiving or surpasing the performance per $ compared to the CPU, requires so much recoding, it's not viable.

I don't get it: You moved FROM Octane TO Corona because you didn't like the results of Corona?

I'm sure it's just a typo.

[Juraj]

There aren't plans to use current (2nd generation) XeonsPhis.

As was stated above, it's 'pseudo'-x86 platform, easier to code than for fully proprietary CUDA, but still not easy.
Other things are, current XeonPhi are already getting outdated, they have limited memory (nVidia GTX have already more. 6gb for Phi vs 12 for GTX Titan, and the Titan is much more powerful).

If you're interested, Vlado and his team built internal prototype for Phi, but they weren't impressed much. Although, I think the results weren't bad at all, the card equalling to middle-class dual Xeon build they had in comparison.

But perhaps, wait for future :- ) Maybe something will eventually come up from this AMD collaboration ? Or simply the next generation of GPUs, or mixed hybrids, will already be too lucrative to pass by (very possible, TitanX is already lucrative deal).

If you want to get the best out of Corona today, buy the best CPU performance you can in simple i7, or dual-Xeon unit.

[zacuttialberto]

Ti Was just to understand, plans to but i7 5960 from silicon lottery..

Ciao

Alberto

[Ondra]

This would be a huge effort. We are observing hardware development and we may make some strategic decision later