Possible TG3 (Free) bucket size & GI render problem

[Bjur]

Encountered another strange behaviour of the TG 3 Free version.

I just wanted to render and compare a TG 2.5 test scene of mine in TG 3 Free (V.30040) this night.

Resolution, Detail and AA were adapted (1280 pix max, Detail 0.6, AA 4). GISD, RTA and the new DOF = were uninvolved:

The scene got nevertheless rendered in smaller than set bucket size and without SSP on it seems, GI settings weren't adapted proper from the original file..

TG 3 needed also more time to render this scene.

[attach=1]

[Matt]

The algorithm that chooses bucket sizes has changed in TG3. This is deliberate.

If you simply loaded the TG2.5 project into TG3 and didn't change anything, the dots should be spaced fairly similarly. However, I can tell from the GI detail of 3.2 in TG3 that your render detail is 0.8. Is that correct? Was it 0.8 in both tests? (My calculations are off here, nevermind)

Ah, I realise what it is. You're running the free version which is limited to a detail of 0.6. This could produce a lower density of GI samples if you render it before editing the GI settings.

Matt

[Matt]

I'm not sure what's going on now. Your render detail is 0.6 so that is within the limits of the Free edition. However, when I load it into TG3 it says the GI cache detail is 4.8, not the 3.2 you saw. But if you rendered at GI cache detail 3.2 then that would explain why the dot density is lower. Somehow this got changed to 3.2 - perhaps there's a bug there?

Matt

[Bjur]

That's what i mean, the GI density gets higher, even without any action or wanted readjustments..

Something seems to go on/wrong again (maybe, i hope not).

My original TG 2.5 file has settings like GI 2, 2, 8, SSP on, rest unchecked, detail @ 0.4, AA 4. @ a resolution of 2880 x 2010 pix.

The upper render picture is the original TG 2.5 file opened in TG 2.5 -> with equal settings to the TG 3 Free max settings.
The picture below is the same file rendered in TG 3 Free with all needed settings adjusted like resolution and detail, AA, GI and stuff.

GI settings don't get converted correctly it seems - even the bucket size got some love/readjustments.
Supersample prepass for example is shown checked in TG 3 like in the original file, but in the render,
it seems the SSP didn't work when watching the GI pre/cache render (far less GI caching points or whatever).

And the original "GI relative detail" of 2 gets set to the value of "3.2" (in my case) instead to stay at "2"..
Of course i could readjust the GI relative level on my own, but what's up with the already checked Supersample prepass for example?

I don't believe that's all intended in all.

Or is it?

BTW., I'm rly sry for my "ebil" English Matt, i just want to help.. 

[Matt]

Hi,

The first thing to note is that we've changed the GI settings for TG3. The detail in the GI prepass (and therefore the GI cache) is controlled with a new setting called GI cache detail. This is not the same as the old GI relative detail from TG2. Old projects are converted to use the new setting. The number is different. It's normal for GI relative detail 2 to be converted to GI cache detail 3.2. Some rules are applied for the conversion, but I would prefer people not to focus on how the conversion happens, but rather to become acquainted with the new settings in TG3. The main thing that's different about the GI detail in TG3 is that the spacing of the dots doesn't change when you change the renderer's main detail setting. It is not relative to the detail now.

With detail at 0.4, GI relative detail 2 should convert to GI cache detail 3.2, so that's working as intended. But the spacing of the dots is different in your tests, and they should be the same. Did you leave GI cache detail at 3.2 (which it should be) or did you change it to 2?

If super-sample prepass were turned off, you would see much fewer dots. The change between your two renders is much smaller than that.

Your English is fine

Matt

[Bjur]

No, i didn't change anything in both versions beside i set the detail from 0.4 up to 0.6 (letting AA @ 4) and the resolution down to 1280 pix wide (to match both versions with the max. settings of TG 3 Free).

In TG 3 i let the GI cache detail @ 3.2 (where the old/original GI relative detail was @ 2).
The Supersample prepass is active in both versions and GISD is unchecked as it is in the original TG 2.5 file too.

Far less "GI points" seems to be calculated in TG 3 Free - like the Supersample prepass isn't working or is not checked (but it is checked and was rechecked several times)!

TG 2.5 was also calculating the GI faster, even calculating more visible GI points as you can see in the picture in the first post.
It shows the difference between both versions when the GI gets calculated (missing GI points).

That's why i'm wondering.

Auto reduction of the bucket render size was active in both.
It's ok if TG 3 just handles the given start size of 256 in a different way now than TG 2.5 did, i wasn't aware of this..

Alex

[Matt]

When you changed the detail from 0.4 to 0.6, that would increase the GI dot density in TG2.5 but not in TG3. In TG3 you would have to increase the GI cache detail to 4.8 to produce the same result.

If you simply load the file in both versions and don't change anything, they should produce a similar spacing of GI dots.

The 256 for bucket size is just the maximum bucket size. Terragen usually reduces this automatically if you have "allow auto reduction" turne on. TG3 reduces using a different algorithm.

Matt

[Bjur]

When you changed the detail from 0.4 to 0.6, that would increase the GI dot density in TG2.5 but not in TG3. In TG3 you would have to increase the GI cache detail to 4.8 to produce the same result.

If you simply load the file in both versions and don't change anything, they should produce a similar spacing of GI dots.

Hi Matt.

At first i want to thank you for your rock solid patience trying to help me not to falling into madness by testing/discovering the new TG 3 and it's differences.

Indeed, the different results i got again and again in my GI renders (GI dot spacing) was because i was using every time my original .tdg file
for both versions and not the already for TG 3 Free max. settings adjusted TG 2.5 .tdg file i attached for you in my 1st post. 

Never thought that the difference between the old GI relative detail and the new GI cache detail would be so different in it's values.

With my for TG 3 already adjusted and attached TG 2.5 .tdg file (@ detail 0.6, AA 4, 1280 pix.) i get the same GI cache detail value of 4.8 like you did (not the value of 3.2)
and the GI dot spacing is the same now like in TG 2.5..

Out of curiosity:

Is the value of 4.8 now generally too high for any/my old simple TG 2.5 scene as the virgin TG 3 opening scene just offers a GI cache detail of 2 right from the start?

Aren't be able to compare both outcomes at the moment but will do a new comparison soon as possible..

Thx again for your help!

Alex