RC5 25% faster than SP2 and little better quality!

[RErender]

If you watch the areas I've circled nobody thinks the SP2 look clearer and better defined?

[vlado]

Hard to tell really I think that the overall noise level is about the same. It might be a little smoother in SP2...

Best regards,
Vlado

[stilgarna]

Hard to tell really I think that the overall noise level is about the same. It might be a little smoother in SP2...

Best regards,
Vlado

I agree !! very hard to tell...

But now we have to deal with Sp2... The difference between 1.5 Rcs and SP2 is so huge than I can live with slower renderer...

But would be nice to know why ?

Best regards.

[cecofuli]

and what is the last image you posted? Says SP2 and 13:56?

Hi RErender

is a network Octa 64bit VRay 64bit SP2 + 3 Prescott 32bit SP2. I want to see the difference. If you scoll with full Octa SP2 64bit (17min) with Network render SP2 32bit (13min 56 sec) , you can recognize the buckets by 32bit machine.

And i agree with all. The noise level is +- about the same. And yes, i can live with 25% but with all features in SP2, but why this slowdonwn? Is better -25%, not +25% rendertime I don't know if the fault is Windows, SP or 32 or 64bit, but this difference is a little to high IMO. Maybe, in the next SP, Chaosgroup found the solution

Bye

cecofuli

[ene.xis]

set up some test with caustics high reflections and low glossies, etc, etc
that would be some serious comparison imo.
Cause from my experience vray has become faster comparing the same noise levels and AA sampling, but as the sampler changed so did the values used imo.

[^Lele^]

I don't understand something: did you just open the rcX scene in sp2 and pressed Render?
If that is the case, it's very likely to be slower, as the sp2 sampler is a lot more sensitive, and it's years apart from the values (and dare i say it, even methods) we all used to use in rcX.
So as mentioned you'd need some tweaking in the noise thresholds, and possibly some tweaks in the AA settings.

[m_hinks]

I was just thinking the same thing! It does make comparing the different version hard though.

[cecofuli]

Yes, but with 25% time render i hope in better noise quality. And, like you can see, me and Morbid Angel prefer RC5, Vlado and RErender SP2. So the difference is not so clear.

This scene is not so simple (QMC+QMC, all DMC, DOF), but maybe with MBlur, glossy, Area shadows the difference from this two version could be much more visible.

Bye

cecofuli

[ene.xis]

that makes no sense whatsoever...2 me both look equivalent, BUT given the diferent rendertimes i m guessing from my personal experience that the sp2 is oversampling, which is leading to the higher rendertime!

[Morbid Angel]

cecofuli can you add a glossy refleciton to this scene and render with it?

[cecofuli]

As Lele says, maybe lincreasing noise threshold in DMC sampler, in the AA or twatking the SP2 version, we can get 13 min 10 sec with no difference.
Morbid, if you want, the scene is freeware.

DOWNLOAD

Now i'm rendering some frames for tomorrow, so my octa is busy

bye

cecofuli

[Morbid Angel]

ok...how do i install both RC5 and SP2 on the same machine?

[cecofuli]

Hi Morbid,

RC5 come with software license. Do you remember?

I'm going to sleep now. Bye

cecofuli

[Morbid Angel]

So, there was a number of problems with your scene. Logarythmic exposure, projected gradient as screen mapping for gi.
I have stripped down everything. For a clean test you must have the basics and only basics. Here below I have recrated your scene with max 2008 RC5 and max 2009 SP2. Sun/sky physical cam.
For RC5 I have set adaptive qmc 1/64 at 0.01, gi qmc/qmc 8/3
For SP2 I have reset vray settings to default, and setup manually the same settings: adaptive dmc 1/64 at 0.01, gi brute force/brute force 8/3.
Both scenes use gamma correction with gamma 2.2 (sampling at gamma image). Lastly a simple area filter is applied...
More tests are coming...

[percydaman]

just letting you know dmitri, but your sky is different between your two images and the RC5 is slightly warmer on the direct lit geometry, also probably due to the difference in the sky.