[T0M0]

Those render times

Could someone from NL explain why is V4 slower than V3 (CPU engine) ?
There weren't announced any new features/changes for CPU version and I expected at least same, maybe faster, results. Not slow like this with such a easy and simple scene.
I don't want to imagine, if I render for 4-5 hours in V3, in V4 it will take what, 6-8 hours... ?

Thanks seghier for your tests.

[seghier]

thanks T0M0
i am not sure but maybe V4 at the same SL is more cleaner ? maxwell team can confirm that
anther tests :
Colors
V3 jpg size : 367 ko

V4 jpg size : 364 ko


Render time
V3

V4

[Mihai]

Thanks for checking with the color checker. So it seems an issue with a change in contrast, not hues shifting.

[seghier]

thanks
maxwell V3 with low priority faster than V4 with normal priority (normal always activated by default )

[jwiede]

thanks
maxwell V3 with low priority faster than V4 with normal priority (normal always activated by default )

Ugh, so sounds like V4 is notably slower than V3 on CPU, if V4 even had OS scheduling advantage over V3 and V3 still won.

WTH?!

[luis.hijarrubia]

Those render times

Could someone from NL explain why is V4 slower than V3 (CPU engine) ?
There weren't announced any new features/changes for CPU version and I expected at least same, maybe faster, results. Not slow like this with such a easy and simple scene.
I don't want to imagine, if I render for 4-5 hours in V3, in V4 it will take what, 6-8 hours... ?

Thanks seghier for your tests.

Yes, no mayor changes on CPU engine. The performance issue if a deploy issue. We are working on it.

[tom]

Have you tested using Constant Dome only? I mean the contrast difference... Looks like gamma to me, we will check that.

[seghier]

With sun enabled or disabled ?
I don't test that
In the render you can see that the histogram is different
And if you open any mxi created by V3 in V4 the histogram changed

[tom]

Oh cool! It's about the buffer then. Something easier to fix

[seghier]

Oh cool! It's about the buffer then. Something easier to fix

yes they fixed it
i don't test a render just compare the mxi file and the histogram is now the same