MLT working - 9000 series or higher only

[pixym]

You'd better wrote the contrary beacause FPrime is out since more than Five year (Vray RT since two years?)
Oh that really does not matter
It is a good start for comparison as you wrote.

I know FPrime was there way before other competitors
I use FPrime with lightwave 9.4

An interior scene is good
As long as there's no light portals used or other cheats

LW 9.4??? I guess you mean LW 9.6 because LW 9.4 has never been released.

[Sam]

I don't remember, its 9.5 maybe or 9.3
Are you sure 9.4 was not released ? Because its the number I remember
I never used the 9.6, I switched to Cinema 4D before it

Oh I see
I wrote "I use FPrime with Lightwave"
Instead of "I used FPrime with Lightwave"

[radiance]

Hi,

MLT will probably make it into the 2.1 bugfix update.
i have a working version but it needs more optimization.

it's not suitable at all for navigating/editing,
but it renders faster than pathtracing, explores caustics, and give more speed due to the fact that it's not MLT but it's some new MLT like algorithm i'm developing.

it's basically importance sampling, eg sampling more in areas where it's needed and less in areas where it's not needed,
and bright contributions, eg fireflies/hot pixels, are explored to equalize them onto the render film.

EDIT: you can also finetune the amount of work per cuda pass, so you can manually set the FPS of the updates,
and if you set it high, like 50 or 60 FPS, you can minimize and work in other apps with no glitches (on single GPU systems)

one thing to note is that MLT and related algorithms usually are slower than pathtracing on very simple directly lit scenes (like exteriors or objects with an HDR map environment)
but there's ways to optimize that too...

Radiance

[pixym]

Rad, I really want to thank you for your availability and the way you communicate with us.
Thank you very very much for these informations.

[pixelrush]

so you have MTL working but its not MTL
well as long as it does the job...

btw are we likely to see the situation where Octane only works with fairly recent cards and drivers as it progresses?
fast forward to Octane 2 and it needs cuda 5 + wont run on nvidia series 2xx?
I am a bit hesitant to commit to multiple expensive cards now if not too far down the track things have changed again..
have you basically got your approach sorted now so that the requirements will be consistant into the future?

[gristle]

That is a good question. At least with CPUs you can still use them if you get newer/quicker systems.

[radiance]

That is a good question. At least with CPUs you can still use them if you get newer/quicker systems.

that's not entirely true, many applications nowadays require for example SSE2 which a pentium IV CPU does'nt have...

there aren't going to be continous issues like this.
the issue is just with 8000 cards, they lack a few options that are needed, and these are the only feats that we need for complex algorithms, so you can be sure you can use a 9000+ (eg compute model 1.1 or higher) for a long time.

Radiance

[gristle]

True, I was not thinking back as far as a pentium 4 spaceheater.

[andrian]

EDIT: you can also finetune the amount of work per cuda pass, so you can manually set the FPS of the updates,
and if you set it high, like 50 or 60 FPS, you can minimize and work in other apps with no glitches (on single GPU systems)


Radiance

The higher the FPS is set, the higher rendertimes to expect and vice versa or its not related?

[Krisonrik]

btw,

the new MLT algorithm won't be MLT, i'm developing a new algorithm that's much better.
it also does'nt work with passes anymore, so you don't have any more UI lag when using octane on single GPU systems.

users can render finals with it, much more efficient, with pronounced caustics, and windows / GPU will respond as normal.

Radiance

Please tell us you are not a god decent from heaven to us graphics nerds. You are too cool, should be illegal!