[w i l l]

That Cartoon Network render is ace.

Whats the difference between regular displacement and vector displacement?

http://support.nextlimit.com/display/ma ... +component

[Bubbaloo]

That Cartoon Network render is ace.

Whats the difference between regular displacement and vector displacement?

http://support.nextlimit.com/display/ma ... +component

That's our friend Herve.

[w i l l]

He needs to teach me.

That's our friend Herve.

[JamesColeman]

There are not significant changes in performance apart from massive improvements in displacement and some general improvements in noise, but definitely this version should never be slower than 2.5.

Thanks,

Juan

I might have some bad news...

http://img256.imageshack.us/img256/1741 ... chmark.jpg
http://img36.imageshack.us/img36/1607/26benchmark.jpg

I know that this Benchwell scene is really old but what else have I got as a constant to compare things with? Apart from changing the priority from low to normal this is just the ordinary Benchwell, and I promise it was left to render with no other programs working in the background.

[polynurb]

I might have some bad news...

http://img256.imageshack.us/img256/1741 ... chmark.jpg
http://img36.imageshack.us/img36/1607/26benchmark.jpg

I know that this Benchwell scene is really old but what else have I got as a constant to compare things with? Apart from changing the priority from low to normal this is just the ordinary Benchwell, and I promise it was left to render with no other programs working in the background.

although interesting to see these results, it might not be significant in terms of quality vs. time.

it could be that 2.6 produces better results "per sl" so that direct comparison is not possible.
to really see what is going on we should take a more complex scene, render it to the same time and then compare the visual quality of the images.

[JamesColeman]

it could be that 2.6 produces better results "per sl" so that direct comparison is not possible.

I know that there was some confusion when we changed from V1 to V2 that SLs no longer correlated, so some NL confirmation of this would be great. In the meantime I'll continue testing.

[Half Life]

it could be that 2.6 produces better results "per sl" so that direct comparison is not possible.

I know that there was some confusion when we changed from V1 to V2 that SLs no longer correlated, so some NL confirmation of this would be great. In the meantime I'll continue testing.

If you change the settings in Maxwell 2.6 to only save image at the end I think you'll find it will score a better bench.

Best,
Jason.

[JamesColeman]

If you change the settings in Maxwell 2.6 to only save image at the end I think you'll find it will score a better bench.

Best,
Jason.

I disabled the "Image" and "MXI" checkboxs under "Render Option > Output" in 2.6 but still got a time of 6m33s with priority = normal. Was there a different setting I was meant to change?

[Half Life]

Within the Maxwell rendering app there is a preference that will permanently disable saving -- this eliminates the gap between sample levels.

Best,
Jason.

[JamesColeman]

Within the Maxwell rendering app there is a preference that will permanently disable saving -- this eliminates the gap between sample levels.

Thanks Jason, it shaved off a couple of seconds but it's still 20 seconds (about 5%) slower than 2.5.1

I've done some more tests with more challenging scenes (ie SSS) and although I can see less chromatic noise in 2.6, it still had a lower SL (15.88) and lower benchmark (719.02) than 2.5.1 (16.08 and 779.68 respectively) when rendering for 1 hour. I'm running OSX 10.6.8 64-bit.

I've noticed the Ranch reported seeing an increase of a couple of SL in 2.6 when rendering over the same time period as 2.5.1 so I'm going to ask them for advice.

I've even tried running the programs straight from the expanded .dmg files in case there was a problem with the installation but I had the same result.

Hasn't anybody else tried the Benchwell or any scene in both 2.5.1 and 2.6? If anything just to confirm that it might be a problem with the Benchwell being so old.

[Half Life]

I guess one root issue would be the sample level -- SL15 isn't really enough because Maxwell doesn't get into solving the tougher to render issues until the higher SL anyway. And besides they can manipulate those numbers to mean anything since we don't really know exactly what they mean... the image quality/time is all you can judge.

But that said I'm not sure a couple of seconds is worth getting angsty over -- they did not tout any particular speed increases (other than displacement) and there are additional image processing calculations in place for 2.6... so it's natural you might lose a few seconds.

I would feel different if speed was the focus of this update...

Best,
Jason.

[m-Que]

Hasn't anybody else tried the Benchwell or any scene in both 2.5.1 and 2.6?

Yep, I had EXACTLY the same difference - around 5%.

[pipcleo]

great , great , great....
referenced mxs files solve all the issues with the previous inability to instance groups.
Moving , scaling and rotating a zillion trees is now a breeze.....excellent implementation !

[micheloupatrick]

Hair rendering is fast and gives great results, pretesselated displacement is an important addition as it dramatically reduces rendering time (especially when used in conjonction with other computing intensive features like SSS), the filtering improvements are real (I'll try to post an example later showing the anti-aliasing improvements), and from what I've gathered from others, particle rendering is working fine too.

Not bad for a free .1 upgrade...

[brodie_geers]

A minor gripe. I'm noticing that SL 1 seems to pretty much just be a dark gray mass now and it isn't until SL 2 that you get some light and color. I kindof miss how SL 1 used to already give you a pretty good idea of what you were going to get in case you messed up a setting somewhere. It seems now that I have to wait a bit longer.


-Brodie