Bitmap vs. VRayHDRI (VRayBitmap) performance?

[thwalker]

Hi all!

At my work we're looking to ensure a standard of using VRayHDRI/Bitmap as the standard input node, because everything currently uses the native max Bitmap node. Prior to jumping into extensive comparisons of quality/performance, does anyone have any stats on comparison? Is this a negligible concern in the first place?
Our concerns are mainly in regards to render time changes, filtering quality, and blurring quality.

Thanks!

[^Lele^]

It's a nuanced answer.
In short: it ought to be better across the board.

Specifically:
a) Bitmap loading is done differently: they are loaded as required by the traced rays, rather than at render start. This has a number of benefits (somewhat more staggered data access, kinder on most infrastructure), but works best with b)
b) VrayBitmap loaders support tiled, mip-mapped .tx files (the conversion is entirely optional, but *warmly* suggested, via the provided script, or binary tools.). Meaning load on demand will happen on a tile (f.e. 64x64px) of the right resolution for the given camera view. This also has a specifiable maximum RAM usage value, and if that is reached the old data will be unloaded in favour of the new. These features would allow you to convert, say, 16k originals to .tx (which will make RAM usage drop, compared to loading originals.), and forget about memory management.
c) VRayBitmap loaders offer a number of sharp filter methods: Elliptical, and Sharp Isotropic are both much sharper than the max's own area filtering, while being comparable or quicker for speed.
d) Speed is much better across the board (as we control the loaders' code, it can be optimised more thoroughly.) in loading and render speed (mileage varies, but i've seen up to 30% increase over max, and even more over OSL bitmap loaders.). There could be cases where the sharper filtering will produce more work for the anti-aliaser, and so to reach a given noise threshold it'll require more rays to be traced. It generally more than works out as faster overall, anyways.

You may get more or less benefits from them depending from the kind of work you do.
However, in general they ought to prove quite nicer to work with than max bitmap loaders, particularly if you decided to adopt the .tx workflow.
Test it on a few scenes which you find representative, the provided conversion script (in the right-mouse click menu) should take care of it all without issue, with and without .tx conversion.