I just checked 10 bit log dpx files, it is flying! Just some basic operations afterwards, blur, grade, roto shape premulted, transform, plus. Also I tried a gpu accelerated denoise. Pretty impressive. Maybe we all need to go back to 10 bit log files for storage :) 16 bit dpx was also very fast.
I On 27 March 2013 23:52, Michael Bogen <[email protected]> wrote: > Setting your disk cache and your local cache to the card makes nuke very > fast. Mostly dpx files. Thanks Deke for the tip about the Exr. > > michaelb > mbfx.me > > > On Mar 27, 2013, at 7:16 PM, Michael Garrett <[email protected]> > wrote: > > Hi Deke, > > That all makes sense, I figured I should try Cineon/DPX out of curiosity > for the exact reasons you state, although it's not part of our pipeline. > It'll be interesting to compare notes about the card performance as a > whole, though. > > I wonder if it's possible to at some point get exr's to decompress on the > gpu? > > Michael > > > On 27 March 2013 18:47, Deke Kincaid <[email protected]> wrote: > >> Hi Michael >> >> I'm actually testing this right now as Fusionio just gave us a bunch of >> them. Early tests reveal that with dpx it's awesome but with openexr zip >> compressed file it it is spending more time with compression, not sure if >> it is cpu bound or what(needs more study but its slower). Openexr >> uncompressed files though are considerably superfast but of course the >> issue is that it is 18 meg a frame. These are single layer rgba exr files. >> >> ----- >> Deke Kincaid >> Creative Specialist >> The Foundry >> Mobile: (310) 883 4313 >> Tel: (310) 399 4555 - Fax: (310) 450 4516 >> >> The Foundry Visionmongers Ltd. >> Registered in England and Wales No: 4642027 >> >> >> On Wed, Mar 27, 2013 at 3:26 PM, Michael Garrett >> <[email protected]>wrote: >> >>> I'm evaluating one of these at the moment and am interested to know if >>> others have got it working with Nuke nicely, meaning, have you been able to >>> really utilise the insane bandwidth of this card to massively accelerate >>> any part of your day to day compositing? >>> >>> So far, I've found it has no benefit when localising all Reads in a >>> somewhat heavy comp, or even playing back a sequence of exr's or deep >>> files, compared to localised sequences on a 10K Raptor drive also in my >>> workstation - hopefully I'm missing something big though, this is day one >>> after all. >>> >>> There may be real tangible benefits to putting the Nuke cache on it >>> though - I'll see how it goes. >>> >>> I'm also guessing that as gpu processing becomes more prevalent in Nuke >>> that we will see a real speed advantage handing data from a card like this >>> straight to the gpu. >>> >>> Thanks, >>> Michael >>> >>> _______________________________________________ >>> Nuke-users mailing list >>> [email protected], http://forums.thefoundry.co.uk/ >>> http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users >>> >> >> >> _______________________________________________ >> Nuke-users mailing list >> [email protected], http://forums.thefoundry.co.uk/ >> http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users >> > > _______________________________________________ > Nuke-users mailing list > [email protected], http://forums.thefoundry.co.uk/ > http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users > > > _______________________________________________ > Nuke-users mailing list > [email protected], http://forums.thefoundry.co.uk/ > http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users >
_______________________________________________ Nuke-users mailing list [email protected], http://forums.thefoundry.co.uk/ http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
