1.46.10 Simple Benchmark

[TRGraphics]

Thats incrediable! I can't wait to try it!

[robing]

1.46.12 >>>>> 4.33 m




1.46.10 >>>>> 10.31 m

[Sushidelic]

same here


1.46.10


1.46.12


i'd really be interested why the images differ so much... is there anything we gotta be aware? the brightness changes can't simply be a qmc thing, so which settings are expected to lead to different results with the new build? the fine highlights on the black teapot top are much less accurate as well - what could be causing this, probably the new light map (ooops, sorry - light cache of course) option "use light cache for glossy rays " (it's not checked anyway)? i'll do some test later stressing the new features.

best regards,
michael


edit:

a version with "Use light cache for glossy rays" checked:

[dbuchhofer]

i guess we should be posting system specs too! otherwise whats the point

[egz]

I'd really be interested why the images differ so much... is there anything we gotta be aware?

Actually, the 1.46.10 images are the wrong ones, because of the glosiness with fog bug that was in that version. This has been fixed for .12 so materials that use glossiness and fog (like the wax material) will render differently.

[Gijs]

dual 2.8 xeon HT-enabled on Asus PC-DL; 1 GB ram



new store glossy rays with lightcache option:

[Sushidelic]

@egz - good point, i didn't consider that point when reading the changelog. so this part should be faster and more accurate
@dbuchhofer - specs are the same as on page one of the thread - Asus pch-dl, 2x3.54ghz xeon, 2gb ram - XP pro, SP2. i didn't mention it because i just wanted to show the version differences.

greetings,
michael

[egz]

here's a slightly different test that shows the importance of using correct geometry when using things like SSS and refraction.

The regular teapot primitive is not a solid object, but is full of holes and intersections. This makes refraction effects render incorrectly, and may lead to artifacts. I closed all the holes in the teapots and rerendered the scene:


Original version:


Don't pay attention to rendertime differences. They are both rendered in the background while I did real work

Here's the modified max7 scene if anyone wants it: http://www.egz.com/vray/speed_test_solid.max

[digitarch]

6:46,9 seconds
dual athlon 2,8
an vray 1.46.12
still seems to be the best choice regarding price/speed....

[deneb26]

comparison with older build

hi guys.

i thought to let this scene render with the newest .15 build and
came across some differences how the new build handles sss.
is there anything we should know about?
which version is the 'correct' one?
i think it doesn't matter if the older one is rendered with open cap-holes, does it?

here are the pics:

[flavrsaver]

Finally got my hands on Max 7, so I figured I would run this test. Here are my results.



Dell Precision 450
Dual 2.4 Xeon
2 GB RAM
ATI Fire XL1

[Paul Oblomov]

Finally upgraded my home PC to A64 3200 ! Damn fast bi***
Athlon 64 3200 (s939)
1gb ddr400 in dual channel mode



And with speed_test_solid

[n6]

Dual P4 XEON 2.4@3.4Ghz
1 Gb RAM

The only thing I have change is in the VRay system :
Face/Level Coef = 1
and
Render Region Divisions = 128x128





And with 100% original seetings :

[Paul Oblomov]

I've played with bucket size and got 6m 34s in speed_test_solid (bucket size was 640x480)