PDA

View Full Version : FPrime and LW



mccabejc
11-04-2007, 10:54 PM
How can it be that the FPrime renderer performs so much faster than the LW renderer? I mean, FPrime has been out for years as I recall... could it be that the LW team can't figure out how to duplicate the FPrime performance? It just amazes me that there's such a huge difference. I'm considering buying FPrime, but still holding out hope that the LW team will figure it out soon...

Taran-Q
11-05-2007, 01:06 AM
What about LWCAD and other premium plugins (pictrix). As many already say, LW seems to have become just a platform for running plugins.....
Seems that many of us are waiting for a long time for LW to build those interactive features into LW ....

brian.coates
11-05-2007, 01:55 AM
How can it be that the FPrime renderer performs so much faster than the LW renderer?
Personally I've found that LW's current render speed is approaching that of FPrime, in that the time required to produce the same image at the same quality is very similar. A lot of FPrime's percieved speed comes from the fact that the interactive window can show you the full size image as it's working on it and reflects any mesh and/or texture changes straight away.

Since switching to the Mac UB version of LW I've had to live without FPrime, (no UB version available), and frankly I haven't missed it.


As many already say, LW seems to have become just a platform for running plugins.
Well, that makes it very similar to a lot of high-end 3D apps. I've been told that MAX is pretty well useless without spending the same amount again on third party plug-ins. From Newtek's point of view, why should they spend time and resources on integrating features that other people are developing just as effectively, if not better.

Yes, it would be good to not have to spend the extra money for the really nice stuff but that applies to a lot of things in life.

Pavlov
11-05-2007, 02:39 AM
Agreed, direction for the future 3D apps is mostly offering a strong architecture for 3rd party tools. Market is becoming extremely segmented and is offering very dedicated solutions for everything, so how a tool cooperates with other is a keypoint.
I'd never invest on a closed software, even if it offers a killer toolset (a near friend of LW, as example). It will be very probably old in few months, market runs much faster than any single app today.

Paolo

Pavlov
11-05-2007, 05:48 PM
Megalodon, i'm curious about your settings... can i see a frame from a movie where you can say LW performs better than Fprime AND it has same quality ?
As said, usually i can make run LW even faster than Fprime, but always at the price of a quality loss... btw i can miss something.

Paolo

Pavlov
11-06-2007, 04:30 AM
wow... which lighting rig are you using, and which machine ?
there's surely something odd in Fprime ones. anyway for this kind of stuff it's probaḅe that LW is faster, since tere's a few stuff around.
in much larger scenes with reflections and heavy AA, i couldnt match Frpime speed in LW.

Paolo

pva
11-06-2007, 08:55 AM
For me LW render works better and faster than Fprime for stills.

We render our scenes @ 4724x3543 (40x30cm print resolution) and, after making a lot of testing, the native render is 3x faster than Fprime is with the same or more quality rendering.

Try to render with Fprime a scene with more than 1 million polygons, with complex materials like metals, glasses with refractions, realistic wood, quite a few bump maps, etc... It just can't make it fast. I rendered an scene like that with 3 radiosity bounces (I know, 2 is enough, but I wanted 3 for testing purposes) and 26h before pressing "OK" in the Fprime Render (not in the Viewer of course) the image still had a very visible noise. It was in the overall finished, but with noise.

Then we rendered the same scene using Final Gather Interpolated, 3 bounces, MES 3.5mm, 250 RPE, MPS 0, Tolerance 0, Ray Recursion level 10 (we use such a higher RPE and such a lower MES because we want small lightning detail and avoiding splotches in untextured surfaces). The result was the Radiosity Calculation finished in 4 hours, and the antialiasing and adaptative sampling passes in 5 hours (using a real lens camera, Antialiasing 4, adaptative sampling 0.0095 / oversampling 0.12), total render time 9 hours, same quality than the 26 hours Fprime render, but without any noise in it.

So, in my experience, if you know how to setup the original LW render, it is much faster than Fprime.

That said I have to say too I love Fprime, and I use it in my everyday work for make fast previews of textures, lights, etc.. but not for final rendering.

pooby
11-06-2007, 09:22 AM
Well, I can't get a solid GI render out of LW. I've tried, and it seems that if you have animated elements, you get FIZZ or boiling. With Fprime, I never get fizzing on the GI.
Of course LW will be faster if you put all the interpolating cheats on, Fprime is doing it without those, but until they are stabilised from frame to frame then it's useless for animation.
Print work is of course, FAR easier to work with.

Captain Obvious
11-06-2007, 09:56 AM
Try to render with Fprime a scene with more than 1 million polygons, with complex materials like metals, glasses with refractions, realistic wood, quite a few bump maps, etc... It just can't make it fast. I rendered an scene like that with 3 radiosity bounces (I know, 2 is enough, but I wanted 3 for testing purposes) and 26h before pressing "OK" in the Fprime Render (not in the Viewer of course) the image still had a very visible noise. It was in the overall finished, but with noise.
I normally render to 4992x3382, and FPrime usually finishes frames in less than half an hour, even with 2+ million polygons, 2-3 bounces, area lights, blurred reflections, etc.

That's using split frame rendering, mind you, but still.

pva
11-06-2007, 10:03 AM
I normally render to 4992x3382, and FPrime usually finishes frames in less than half an hour, even with 2+ million polygons, 2-3 bounces, area lights, blurred reflections, etc.

That's using split frame rendering, mind you, but still.

Well, you'll probably have a render farm, because less than half an hour is an impressive render time even using split framing.

Anyway any test I did rendering the whole frame in a single computer showed me that the LW render is faster than FPrime, I'm just sharing my experiences :)

Captain Obvious
11-06-2007, 10:15 AM
300 gigahertz worth of Core 2 Quad :D

pva
11-06-2007, 10:17 AM
300 gigahertz worth of Core 2 Quad :D

HOLY S***!! :boogiedow

MacDoggie
11-06-2007, 11:51 AM
For me LW render works better and faster than Fprime for stills.

We render our scenes @ 4724x3543 (40x30cm print resolution) and, after making a lot of testing, the native render is 3x faster than Fprime is with the same or more quality rendering.

Try to render with Fprime a scene with more than 1 million polygons, with complex materials like metals, glasses with refractions, realistic wood, quite a few bump maps, etc... It just can't make it fast. I rendered an scene like that with 3 radiosity bounces (I know, 2 is enough, but I wanted 3 for testing purposes) and 26h before pressing "OK" in the Fprime Render (not in the Viewer of course) the image still had a very visible noise. It was in the overall finished, but with noise.

Then we rendered the same scene using Final Gather Interpolated, 3 bounces, MES 3.5mm, 250 RPE, MPS 0, Tolerance 0, Ray Recursion level 10 (we use such a higher RPE and such a lower MES because we want small lightning detail and avoiding splotches in untextured surfaces). The result was the Radiosity Calculation finished in 4 hours, and the antialiasing and adaptative sampling passes in 5 hours (using a real lens camera, Antialiasing 4, adaptative sampling 0.0095 / oversampling 0.12), total render time 9 hours, same quality than the 26 hours Fprime render, but without any noise in it.

So, in my experience, if you know how to setup the original LW render, it is much faster than Fprime.

That said I have to say too I love Fprime, and I use it in my everyday work for make fast previews of textures, lights, etc.. but not for final rendering.

I have to agree about F-Prime. I just attempted to render a Radiosity render (Final Gather) fully tweeked at 2560 x 2048 nothing special and 6 hours later the calculation process finally finished and this was in the UB build. Sorry, but that doesn't cut it:twak: .

I reverted back to the 9.2 CSM build and booted up F-Prime and in an hour I had an images to send to the client as a proof while I am waiting for the final render which would be pretty much close to being finished in 6 hours. The new radiosity render is sweet don't get me wrong but when the ball is dropped by LW, F-Prime somehow manages to get things back on track, even in it's crippled state.

Even when I am put in critical situation by the LW renderer. F-Prime has just insured that I WILL make my deadline. In this case FPrime has just paid for it's self ...again! I look forward to the day when F-Prime is working in the Mac UB builds as well as the 64 bit builds. I'd like to say that I don't miss F-Prime in the UB builds but that would just be plain not so...

Cheers