Everything related to Maxwell network rendering systems.
User avatar
By lsega77
#119816
I really don't know what the hec I'm talking about but here's my stabbed at it.

Feature 1: (again speculation)
You know the first thing I noticed about this video is the once the files were merged the noise in the image was significantly reduced (smoothed). I can only speculate that this means that in a distributed rendering setup each node would/could be set to render at a reduced SL value. Then in the composite render the image is in effect 'sampled up' to achieve the much smoother result. Therefore renderings would be sped up such that the more nodes you have the lower SL you can achieve per each node to achieve the 'sampled up' version.

Feature 2: (here we go again)
I noticed that there are four lights yet there were five files merged together. Maybe each mxi is actually a type of rendering pass. So output0.mxi is a type of material/geometry/raw GI pass. Then output1-4.mxi are passes which contain the information relative to each light (i.e. light location, falloff, relative location to geometry, effect on materiality). So when the images are 'composited' together each light can therefore be effected/manipulated relative to parameters. In a way this sound very similar to an HDRI with an infinate amount of exposures per light source.

Well might be worth a good laugh but that my guess

:D :D :D
User avatar
By dutch_designer
#119818
Ok so basically you render out an mxi image for each light seperately which allows you to combine the .mxi's and adjust their intensity in relation to the other ones?? I'm very confused... perhaps because I haven't messed with .mxi images myself yet and don't quite understand what info they hold besides luminance as with hdri's.
By giacob
#119819
mverta said one dont need separate passes but can adjust lighting of different emitters separately in a single file at pleasure... that is astounding at least
Last edited by giacob on Tue Feb 14, 2006 11:54 pm, edited 2 times in total.
User avatar
By Fernando Tella
#119821
Maxer, at the begining you see a noisy image which I guess it's the result of one of the nodes; then five mxi's (from five different nodes i guess) are merged in one file; then the merged mxi is opened and you see the same image without noise.

Great job NL!

One question: when you are rendering in coop. can you see the final merged result as it is updating? Would be nice.
User avatar
By dutch_designer
#119822
mverta wrote:That's not what's happening. The combination of separate .mxi's rendered cooperatively into one file is one thing. The realtime interactive lighting is a separate thing. You don't need separate passes for separate lights. You can just do a regular old render, only control your lights separately during, and after the render, in realtime. Just re-read that a few times.

_Mike
Ok got it, so one .mxi with seperate lights in there would do just as well for this demo.. it's just that he combined seperate .mxi images of the same .mxs file rendered on different machines?
User avatar
By lsega77
#119823
lsega77 wrote:I really don't know what the hec I'm talking about but here's my stabbed at it.

Feature 1: (again speculation)
You know the first thing ... blah blah blah blah blah
Well I guess that's what happens when you think too much about what you're writing. You get beaten to the punch. Oh well.

Cool stuff though.
User avatar
By tlass
#119824
well I would like to know the following...

When it first came up... the idea of a cooperative rendering
didn't looked that good to me ... because too many people
had problems at the point where MR should merge the results...

Thinking about production with deadlines i didn't liked the idea
to wait till the end of rendertime, hoping that the merging will
be done succesfully...

So is their any advantage to merge mxi-files instantly instead
via network on the fly???

I mean do i have a second try if the merging fails???
Or is maybe simply the concept more stable???

btw... the light feature is fantastic! :wink:

cheers

Torsten
By giacob
#119825
since u got at this point ..... why dont do the same with phisycall sky or\and sun....or is it already possible?
no rethinking... it can work just with emitters...
Last edited by giacob on Wed Feb 15, 2006 12:01 am, edited 2 times in total.
User avatar
By dutch_designer
#119826
Fernando Tella wrote:Maxer, at the begining you see a noisy image which I guess it's the result of one of the nodes; then five mxi's (from five different nodes i guess) are merged in one file; then the merged mxi is opened and you see the same image without noise.
I'm gobsmacked on how that would work.. one noisy image + another noisy image = cleaner image??
Again, no experience with network/distributed rendering.. so err yeah, newbie talking here perhaps..
Last edited by dutch_designer on Tue Feb 14, 2006 11:58 pm, edited 1 time in total.
User avatar
By Maxer
#119827
So you’re saying that we will be able to interactively adjust light levels as well as f-stop and shutter speed to give us any combination of light levels. Does this include daylight?
Last edited by Maxer on Tue Feb 14, 2006 11:58 pm, edited 1 time in total.
By resu
#119828
So is the reneder engine now calculating all light intensities of the lights in a single pass?

resu
User avatar
By tom
#119830
bla :o bla bla :o :roll: Oh someone please kill me right now :D
User avatar
By mverta
#119832
Maxer wrote:So you’re saying that we will be able to interactively adjust light levels as well as f-stop and shutter speed to give us any combination of light levels. Does this include daylight?
Yes. Yes.

_Mike
User avatar
By aitraaz
#119833
mverta wrote:That's not what's happening. The combination of separate .mxi's rendered cooperatively into one file is one thing. The realtime interactive lighting is a separate thing. You don't need separate passes for separate lights. You can just do a regular old render, only control your lights separately during, and after the render, in realtime. Just re-read that a few times.

_Mike
Ok Mike I've done 5 re-read's thanks for the tip. I'll try and rephrase -

I understand that for the interactive lighting there is *no* need to render seperate passes. (Picked this up on the third re-read 8) )

But,

The thing unclear to me was regarding *solely* the cooperative rendering feature (Something entirely seperate from the *other* interactive lighting feature.)

I was curious to know *if* the video demonstrates *how the cooperative rendering will function in maxwell*. Will single images be split and be rendered individually on each individual node and then be merged afterwards (similar to the cooperative rendering in beta/alpha), or can we expect a distributed rendering system in which all nodes work together on a single image (thus no merge of .mxi's afterwards)?

Just curious eh it all looks pretty cool to me...

No re-reads please... :)
Last edited by aitraaz on Wed Feb 15, 2006 12:22 am, edited 3 times in total.
  • 1
  • 2
  • 3
  • 4
  • 5
  • 13
render engines and Maxwell

I'm talking about arch-viz and architecture as tho[…]

When wanting to select a material with File > O[…]

> .\maxwell.exe -benchwell -nowait -priority:[…]