haze wrote:Notiusweb wrote:Q - Without denoiser - Is Octane V4 measurably faster than Octane V3?
1 - If YES, why the heck even bother with development on V3?
2 - if NO, what the heck!?
PS - heck
We are working on speed in two ways:
1. increasing the number of samples that arrive for tonemapping in the same delta time
2. increasing the quality of those samples in the same time
Our focus currently is #2, meaning that in general in V4, you don't need as many samples for the same level of noise/error (actually that means that the MS/sec measurement as speed becomes less important)
With XB4 there will be some more improvements, and you should be able to see a good improvement across most of your scenes.
That sounds very cool. I know that Tonemapping is high-cost item for the GPU, in early V3 testing it was a cause of crashes for cards running on a 1x PCIE connection, whereas an 8 or 16x connection could handle the cost better.
(Abstrax gave us the tonemapper checkbox to selectively exclude and include GPUs those GPUs which were connected 1x vs 8 or 16x, which resolved crashes)
Not sure if this is applicable, but will be interested to see if it whams the selected-for-tonemapping card harder due to what sounds like a higher compressed #samples charged per delta-time unit...but if it runs faster, really we won't care!
So coming is the V4 of V4!
Congratulations to the team for the job thus far.
And also good work to the users for always pushing the product forward. A lot of other forums fuss, but no muss.
This has more muss than fuss, and by muss I mean 'muscle', as in power.