PDA

View Full Version : Brazil and Vray side by side comparison


visualboo
08-21-2002, 03:05 PM
I've been trying to get decent results out of Vray but I just cant get them to look like brazil. There's just wayy more detail in the brazil renders.

Here's a render of a model I did for sythetic light studios a while back. Very simple scene.... object>plane>1 direct light. I wasn't going for a perfectly awsome and artistic render, rather testing a bunch of different render settings to look at quality stuff :)

Here's the Vray version>
http://www.visualboo.com/images/clients/synthlight/vray%20small.jpg

Here's the Brazil version>
http://www.visualboo.com/images/clients/synthlight/brazil%20small.jpg

See all the detail in the brazil version? hmmmm.... juicy
Oh, dont mind the missing batteries in the bottom one, they had Vray materials applied and didn't render ;)

Anyone know any good tricks with Vray?

Das
08-21-2002, 03:58 PM
I notice a few causes for the VRay image not looking up to snuff:

- the contrast is quite a bit lower than in the Brazil version. It looks like the ambient is too high. This is causing the shadows to not look deep enough.

- there aren't enough samples on the GI. This is especially obvious along the inside 'ledge' at the base. The shadow there is splotchy. I think that's also causing VRay to miss some of the detail shadows, like the one under the curved tank at the upper left.

The Brazil render seems slightly blurry along the labeled tank. Odd.

visualboo
08-21-2002, 04:15 PM
- the contrast is quite a bit lower than in the Brazil version. It looks like the ambient is too high. This is causing the shadows to not look deep enough.
There is no ambient at all :shrug:

- there aren't enough samples on the GI. This is especially obvious along the inside 'ledge' at the base. The shadow there is splotchy. I think that's also causing VRay to miss some of the detail shadows, like the one under the curved tank at the upper left.
Yeah, I should have upped the samples a bit. I'll do that and post an updated version. I've tried upping the samples on this model before but it still doesn't bring out that much detail. It cleans up the splotchy parts though.

The Brazil render seems slightly blurry along the labeled tank. Odd.
yes quite odd. Might be jpg compression :shrug:

CHRiTTeR
08-21-2002, 04:32 PM
Verry handy to know: How long did those take to render?

Das
08-21-2002, 04:53 PM
No ambient? Hmm.

Ah, I see. I misread the cause of the lack of contrast. The VRay scene is missing the highs, not the lows.

Here's the VRay histogram:

Das
08-21-2002, 04:55 PM
And here's the Brazil one:

Das
08-21-2002, 04:58 PM
The Brazil image isn't using the lows all that well, but most monitors don't show the lows that well anyway. The VRay image is using only half the available range, for the most part. Makes it look washed out, and reduces the visible detail.

visualboo
08-21-2002, 05:02 PM
Chritter: Not very long :D hehe, honestly I wasn't paying much attention to exact render times as I was comparing render quality vs speed.

Vray is deffinitly faster but comes with a price. At least that's what I'm trying to get to the bottom of now. :)

visualboo
08-21-2002, 05:10 PM
Das: ahh, I never thought about using PS on renders like that. Very cool.

Yeah, washed out renders from Vray are my biggest complaint.

Das
08-21-2002, 05:25 PM
I just got finalRender recently, and I had some troubles with washed out images the other way. Too much light, not enough darks.

finalRender has pretty easy settings for fixing it, though. One thing finalRender does not lack is controls; way too easy to get lost in the panels, lol

visualboo
08-21-2002, 05:36 PM
Actually my vray renders have too much light :shrug:

JuRrAsStOiL
08-22-2002, 02:20 AM
the pictures are down :(

visualboo
08-22-2002, 05:31 AM
Who's? Are you sure? I just cleared my cache and hit refresh and they worked :shrug:

JuRrAsStOiL
08-22-2002, 03:41 PM
well, now they are up, yesterday they were down,
i did try copy&paste and got 404 ... so ...

yeah the contrast in the Brazil image is a lot better,
I really dislike VRay's color bleeding. Could you
perhaps post the scene so we could compare those
with fR?

egz
08-29-2002, 09:13 PM
According to a message posted by Vlado on the VRay forum, the default values for the GI multipliers are set too high in the current versions. Try lowering both multipliers to 0.5. This default value will most likely be changed in future versions.

This will take care of the color bleeding too.

Also, the new version 1.09.01 which was just released for public testing (registered advanced users only) adds a lot of detail in renders without a speed penalty.

visualboo
08-29-2002, 09:39 PM
Originally posted by JuRrAsStOiL
yeah the contrast in the Brazil image is a lot better,
I really dislike VRay's color bleeding. Could you
perhaps post the scene so we could compare those
with fR?

Sorry JuRrAsStOiL, wish I could... but I can't.

Could have been a nice old fashioned showdown :)

Wiro
09-01-2002, 03:26 PM
Hm, I'd be interested in seeing if reducing the multiplier to 0.5 is the solution. I've always thought that VRay renders lacked shadow detail in corners but also had too little color bleeding.

I know VRay is really fast and I've assumed this speed came from alot of undersampling (you get fast renders and less shadowdetail in fR this way too) but if reducing the multiplier is all, we should see quite some improvements soon!

Wiro

egz
09-01-2002, 05:19 PM
Expect people to start posting renders with the new 1.09.01 version soon. Looks good, I've been too busy to render much in it yet though.

Also the normal complaint has been that VRay has too much color bleeding.

ilasolomon
09-01-2002, 10:01 PM
hello
i did a test with brazil buplic beta 0.4.53 & vray 1.0.9
(i'm supposed to finish my A&B's cockpit, but this thread is so interesting!)
i tried to keep the rendering circumstances as same as possible.
the rendering machine spec:
AMD Athlon XP 1700+, 768 mb sdram, IDE Hdd, Windows XP pro, 3ds max 4.25
well, these are the results:


http://www.topleftpixel.com/ila/temp/gi_test/gi_brazil_22min50sec.jpg

http://www.topleftpixel.com/ila/temp/gi_test/gi_vray_04min53sec.jpg

. i know that the final beta version of brazil is faster,
consider it's 2 times faster, but still too slow, compared to vray)
. changing the shade rate from 7 to 15 will give less noises
but longer rendering time.

. if you notice you find out that vray's output is sharper &
has no noises compare tobrazil's one, & it's 4 times faster!

. vray image look a bit lighter. if i change the bounce multip
from 0.9 to 0.8 or 0.75 it will be same as [b]brazil[b]'s.

visualboo
09-01-2002, 11:02 PM
Cool...

I've been so damn busy lately between work, projects, and stuff I haven't had any time to re-render these.

I'll do it tonight and post the pics :)

ilasolomon
09-02-2002, 09:51 PM
...& this is the Final Render version.
the quality is not bad (compare to VRAY) but you can see some
artifacts, & the render time is longer.
http://www.topleftpixel.com/ila/temp/gi_test/gi_frender_15min42sec.jpg
still VRay is on TOP of the GI sulotions!

Wiro
09-02-2002, 10:16 PM
The VRay render looks very nice here. Great contrast and the speed is considerable!

I must note though that you shouldn't do a final render in, err, finalRender with a prepass of 1/4. Always 1/1 for best quality.
Also fR St-1 is irritatingly dependant on scene size for the sample setup so if you haven't Reset-XFormed your objects it might get confused and miss detail.

Are there any differences in the demo of VRay and the full release speedwise?
I was wondering about doing some speedtests between VRay, fR St-0 and St-1 out of curiosity.

Wiro

ilasolomon
09-02-2002, 10:41 PM
well,wiro, i'm not very familliar with FRender yet, but i
think changing the prepass from 1/4 to 1/1 will increase the
rendering time a lot more, no?...i should test it.
Vray's demo & full are the same in everything, just demo version
watermarks the image & the setting couldn't be saved on file.
anyway, you was one of the Ghost beta testers, if i
remember correctly, do you have Brazil ver. 1.0? i'm curios about
speed/quality performance from public test to beta 0.4.53 or 1.0
without that infos my benchmarking is not complete!

drewbie
09-02-2002, 10:50 PM
Are all the techniques being used by the three renderers the same? I know that you're using the QMC for Brazil, but the irradiance cache stuff for Vray isn't the same thing, is it? I'm not sure what you used for finalRender either, but I think judging them on (possibly) different techniques is not going to really tell you anything, other than what technique works fastest. Also note that each renderer is not limited to just that technique.

ilasolomon
09-02-2002, 11:02 PM
you are right about that, but since this test is about
SPEED/QUALITY performance,i tried to use the fastest algorithm
for each one, i know that direct computation of VRay is as slow as
brazil or fRender, but i saw too that irradiance map is much faster
& has better quality rather than other renderers/methods.
& i didn't try global photon mapping that came with VRay 1.0.9
(the readme file says that it's good for interior scenes)
another thing, the public test brazil only has one method for GI
so i didn't have another choice! ;)
(i did some tests with MentalRay too, but it was slow & ugly!)

egz
09-02-2002, 11:02 PM
The irridiance maps in VRay use QMC sampling as well. But does the technique really matter at all? IMHO the final result is all that counts.

ilasolomon
09-02-2002, 11:07 PM
IMHO the final result is all that counts.

agreed!

Wiro
09-02-2002, 11:15 PM
Final result and speed :)

ila_solomon, no I was never a Brazil tester. In fact I have the feeling they hate me because I'm using fR :shrug:
Ghost was free to the public, hence why one of my pics is in their gallery.
And yes, using 1/1 will render even slower, hehe.


Damn, the VRay demo ran out yesterday. Hm, maybe someone I know uses VRay.

Wiro

drewbie
09-02-2002, 11:26 PM
Wiro, the Brazil folks don't hate you. ;) I had your snowtrapper image as my background for a while. :)

Back to discussion...

Anyway, yeah, the speed/quality is the main factor, but making a comparison when you can't have the best of each of the different softwares, it's a moot point. The best way to make a comparison is to make the scene available (which is usually not an option) or to get a hold of the software. :)

Anyway, I'm sure there will be plenty more of these "contests" especially now that Brazil is released and fR stage 1 is coming out soon. Use what works best for you though, cause that's all that matters.

-Drew

Wiro
09-02-2002, 11:36 PM
Drewbie: well, maybe not ALL of you, lol!
Cool you liked my snowtrooper though, hehe.

Yeah, that's why I won't be comparing with Brazil. Not that it's a real comparison anyway, it's one thing to render a skydome scene and another entirely to render a fully detailed one, especially indoors and to make such a scene work for all renderers (including making equivalents of each one's propriotary materials) is more work than it's worth.

Wiro

Da_elf
09-03-2002, 12:42 AM
I agree about the final results are what matters. Someone asked if the method were the same. which the answer would be no. Most GI systems use methods for faking it to have a higher speed. However i assume the direct comp method in vray is more or less the same as the brute force method in fR and i think Brazil has a simular function.

Flyby
09-03-2002, 10:06 AM
Funny how some people can get almost religious about render X or Y. In most cases discusions always ends up with a heated debate which renderengine is "best".:rolleyes:
But very few ever bother to define the createria used to qualify it as "best"...
As I see it, people have different needs: some prefer quality and detail as main criteria, other focus on pure render speed, others a balanced mix or the overall production evalution (economical aspects)...

There is no point in chasing a winner as long the software you use is adaquate for the creteria you find important.

In my personal case, I have both fR and Vray. My criteria is focused on acceptable results in a very short timeframe. fR was unable to deliver that, Vray however is excellent for those criteria. As simple as that....:beer:

Wiro
09-03-2002, 10:18 AM
Not religious, just curious :)

Wiro

derelict
09-03-2002, 10:41 AM
Ahhh! but Flyby, most people would not have your money to buy BOTH of the renderers yes. So thanks to threads like this, one would have a better investment yes. ;)

Money does not grow on trees unless one download it from the net for free... now thats piracy if you ask me!:scream:

Flyby
09-03-2002, 11:42 AM
Oh, but I'm not contesting the value of debating the differences between renders.Sorry If ppl perceive it that way...
Only... one should have a clear understanding of his needs.
What is "best" for one person isn't the best solution for another, so I was only cautioning not get carried away...:hmm:

As for money, I too have to work for my money, you know, but i make a living on it, so I always make sure my clients pay for what i need... :)
That won't say, for exemple, that I'm able to buy 10 dual nodes without bracking my back...:D
I'm still a small one-man shop...

ilasolomon
09-03-2002, 02:51 PM
Flyby, please note that this was in term of GI, not others
abilities of the renderers, maybe we dicuss them here later
i did some test in term of Reflections & Vray is in middle of the
chart! (guess what's the fastest raytracer?...it's Mental Ray!)
nobody (atleast me) said X is better than Y @all. so let this be continued! ;)

derelict
09-03-2002, 04:25 PM
I hope there's no hard feelings there my fellow Flyby?:)

Please try to understand in any given thread, there are bound to be jerks around who would pick a fight :shrug: , but if one is able to look between the post one WILL get what one wants.

So...

Hear Hear! I concur with fellow Solomon!:thumbsup:

Let the test begin! Wait let me get my pen please...

... hmm, my popcorn is near me... and juice in on my right, ok I'm ready!:p


So, what is the next test?

I'm One of the fellowship of the thread,

Dere is the name.

Ps: U C I'm planning to spend a wollup on a renderer and hope it is the 'Bestest' and 'goodest' of them all. ;)

egz
09-03-2002, 05:27 PM
Originally posted by Wiro
Damn, the VRay demo ran out yesterday. Hm, maybe someone I know uses VRay.

Wiro

New demo with the latest features is out now.

CHRiTTeR
09-03-2002, 09:02 PM
If anyone wishes to know more about Brazil version 1.0 go here:
CGTalk Brazil 1.0 is released thread (http://www.cgtalk.com/showthread.php?s=&threadid=19462)


By the way, In this thread is also mentioned that the commercial version of Brazil is a lot faster than the public version. If it's still slower than Vray? I don't know. Is there someone who can compare the commercial version of Brazil and Vray?


Thanks,
CHRiTTeR

Refracted
09-11-2002, 01:45 PM
well, i cant say for sure what the difference in speed between vray and brazil is, but 22 minutes for that image brazil rendered seems alot.. i've done more complicated stuff, at a larger res than that that have rendered quicker for sure (in brazil v1).. for my TV sized video (NTSC-DV), i cant render anything much more than 6-7 minutes per frame because of my limited machines and the time it would require to render that.., and i can easily use brazil to do this (in fact, my non-GI scenes will often supprise me with sub-minute rendertimes :P)..

this "comparison" isn't very fair since it's using the final version of both vray and fR, and a still experimental build of brazil.. .. maybe if the scene was available, someone could render it in brazil v1..

well, i gotta get back to class.. but i just wanted to mention that we shouldn't shoot down brazil too quickly :P

CHRiTTeR
09-11-2002, 03:06 PM
Refracted:
Totaly agree!

visualboo
09-11-2002, 04:28 PM
on a slightly different note.... I've been messing with some hdr stuff in vray and noticed that the shadows don't come from the light source. Check the atached image to see what I mean.

It's like they just come from a generic skydome. The grey objects are actually completely white and there's no light in the scene, just the hdr plane.

egz
09-11-2002, 04:35 PM
Your environment background is grey. It emits light.

To make your HDR image your only lightsource, either set the max background to black, or turn on max environment override in VRay, and set that to black.

visualboo
09-11-2002, 04:46 PM
ahh cool. Yeah, I'm kinda hdr stoopid :) All the renderers handle stuff so differently, some stuff effects renders in one renderer but not in another. It's weird.

Thanks tho

visualboo
09-11-2002, 04:58 PM
:thumbsup:

visualboo
09-11-2002, 07:17 PM
I've been tweaking some settings.... These test renders only have one direct light. Not meant to be super stylish or anything.

I still need to up the samples a bit though. There's some minor blotches.

http://www.visualboo.com/images/clients/synthlight/render-17-vRay-W-settings.jpg

I'm working on a brazil version right now..... update soon.

egz
09-11-2002, 09:02 PM
I can tell from your screenshot that you're not using the latest version (you're using 1.09.01a or b). I suggest you try the test with the latest build since it's faster and includes a few bugfixes as well.

egz
09-11-2002, 09:05 PM
Also, I notice that your irridiance map sampling starts at -1. You can start a lot lower (like -4 or -3) and still get as good results. That's what the threshold settings are for.

The latest build has some very useful, built in presets you can look at.

visualboo
09-11-2002, 09:41 PM
Yup, using b. I'll get the newer version and re-render (tweak a couple settings too).

Dang that brazil render was taking forever. I'll render that fully tonight. I love brazil but I wish t was just a tad quicker. Hopefully version 1 will pull through.

visualboo
09-11-2002, 11:27 PM
This is still 1.09.01b btw.

So I got the render time down to 11 minutes. That's a lot better. I tried lowering the irradiance map before but it rendered pretty blotchy. Not sure why it doesn't do that now. :shrug:

http://www.visualboo.com/images/clients/synthlight/render-17_2-vRay-W-settings.jpg

Tomorrow I'll post the Brazil version (4.53 wishin it was 1 :) )

Flyby
09-13-2002, 02:06 PM
I think you can even find faster settings.

Try min=-3 max =-1

-> If you really want high quality , you can't beat direct computation, however, using IR maps is a great way to have GI approximations. But you'll only get the real speed advantage if you use rather low values.
so, using max=0 is almost the same as doing direct computation in some cases and is not a good idea if you want speed.

decrease the Hsph subdivs to 15
increase interp samples to 30

->From testing I noticed that the "Hsph subdivs" parameter has a very big impact on rendertimes. "Interpolated samples" has far less negative impact on the rendertime.
So what I suggest is to compensate the decrease in subdivs with a higher sampling. You'll loose a little bit of detail in the shadows, but it will be hardly noticable.
The moment you get splotches on the floor, you know your "Hsph subdivs" are too low...;)

egz
09-13-2002, 02:59 PM
That was true of the 1.08.0x builds,

But with the new builds the time penalty is really minimal even when using supersampling, mainly due to a much smarter sampling algorithm.

In the 1.09.0x versions you will get about the same time with -3 1 (gi supersampling) as with -3 -1 in the old builds. The key here is to use good threshold settings.

CGTalk Moderation
01-13-2006, 02:00 PM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.