GPU rendering is not the magic bullet it was promised to be

[Garrick]

i dont know, I think the luxology guys may want to reconsider the whole gpu thing

http://vimeo.com/14013808

(14 minutes in the hardware stuff is enabled)

cheers

cp

It says this is a private video.

[cpember]

odd, i dont understand how they do that then, because it's still live on the original page...

http://shop.vray.info/siggraph-2010-cha ... demos.html

3rd one down

cheers

cp

[Greg]

People...it's not about modo vs. Octane. Let me say that one more time so we can all get over it. The video was not about modo vs. Octane. It was a message to people who had already bought modo (posted on a private forum so only people with valid modo licenses could view it) about whether or not it was time to invest heavily in GPU rendering. The people viewing the video had already bought modo..many of them (like me) had also already bought Octane...and Maxwell..and Keyshot...you get the picture.

They posted Octane results because Octane gave good results worthy of comparison and it wasn't a hybrid renderer. It wasn't a value proposition either. Octane is a hundred bucks. You can run it fast on a $500 video card - it's an incredible value. Nobody was trying to compare the value. modo users wanted some info on the GPU 'situation', and Luxology tried to give us some data in addition to their conclusions. The conclusion? The GPU will certainly have a place in the pipeline at some point, they won't sacrifice features to run on a GPU and they won't tie themselves to one platform (CUDA) - and (most importantly) the GPU isn't some magical solution to rendering. And they're right - it's NOT. It doesn't mean it's not totally usable right now (as many people around here have proved) or not full of promise. It's just not something you flip a switch on for a magical speed increase. The gains of GPU rendering have to be worth the dev. time, effort and monetary investment without a sacrifice in features. Whatever gain is there (even if the Lux test results were off) justifies more research but not a complete change of plans.

If the video lights a fire under the Octane devs and makes it run faster (like Radience was saying by computing a preview in RGB space) then we all benefit. If it pushes the Octane devs to prove everybody wrong and release a good and fast SSS solution for Octane then we all benefit. modo is trying to find it's place in a world dominated by Autodesk...not go after little companies like Refractive.

-Greg

[Jaberwocky]

Just a thought.Should Octane have Export Plug ins for cad products like Autocad/Inventor and Turbocad.That would widen the consumer base quite a lot.In terms of speed.Its quite simple really.Wait till V1.0 hits the road and magazines like PCPro and other web sites do a round up test of the many renderers out there along with cost comparisons.It it then that you will all truly know how good Octane is against the others in the current line up.

[radiance]

People...it's not about modo vs. Octane. Let me say that one more time so we can all get over it. The video was not about modo vs. Octane. It was a message to people who had already bought modo (posted on a private forum so only people with valid modo licenses could view it) about whether or not it was time to invest heavily in GPU rendering. The people viewing the video had already bought modo..many of them (like me) had also already bought Octane...and Maxwell..and Keyshot...you get the picture.

They posted Octane results because Octane gave good results worthy of comparison and it wasn't a hybrid renderer. It wasn't a value proposition either. Octane is a hundred bucks. You can run it fast on a $500 video card - it's an incredible value. Nobody was trying to compare the value. modo users wanted some info on the GPU 'situation', and Luxology tried to give us some data in addition to their conclusions. The conclusion? The GPU will certainly have a place in the pipeline at some point, they won't sacrifice features to run on a GPU and they won't tie themselves to one platform (CUDA) - and (most importantly) the GPU isn't some magical solution to rendering. And they're right - it's NOT. It doesn't mean it's not totally usable right now (as many people around here have proved) or not full of promise. It's just not something you flip a switch on for a magical speed increase. The gains of GPU rendering have to be worth the dev. time, effort and monetary investment without a sacrifice in features. Whatever gain is there (even if the Lux test results were off) justifies more research but not a complete change of plans.

If the video lights a fire under the Octane devs and makes it run faster (like Radience was saying by computing a preview in RGB space) then we all benefit. If it pushes the Octane devs to prove everybody wrong and release a good and fast SSS solution for Octane then we all benefit. modo is trying to find it's place in a world dominated by Autodesk...not go after little companies like Refractive.

-Greg

Sorry but i have to dissagree.
The comparisons made between both engines are completely unfair and handicapped.

Interpolating irradiance cached GI computed in traditional RGB colours will always render faster than a spectral unbiased engine.
Put an experienced vray user behind a machine and give him a day to finetune the vray engine settings, and compare the rendertime with the output of a zero configuration maxwell image.
Offcourse it's 100x faster, it's simply calculating 100x less information and interpolating it into an incorrect solution.\

Octane is an spectral unbiased renderer, not a classic raytracer with biased GI options.

Add to that the extreme hardware handicap, eg 2 old quadro cards against dual 6core i7 xeons.
For the price of that xeon machine, you can stack 4 GTX480 cards in your machine and octane will blow away the modo irradiance cached render.

We actually invested in developing a plugin for modo. what we got in return was a deliberate, staged, handicapped comparison of a large company with a lot of money and staff against a small group of enthusiastic developers with limited resources.

'not the magic bullet it was promised to be' ?
what about 'a comparison between modo's render engine and refractive software's octane render',
would'nt that have been a more friendly title ?

This video has polarised my feelings about luxology.
It's videos like this that harm you're image more than 'competing against autodesk'.

Yours,
Radiance

[andrian]

I have to agree with Radiance here, I was pretty pissed off when I saw and read trough their forum a cheap (in terms of hardware used - VGA) comparison between engines, witch only interest was to show what great a put-up render engine of Modo is.... but that just me...

[Greg]

Why would Luxology release an 'anti' Octane marketing video on a private message board accessible only to people who have **already** bought modo? That's just weird. It was a general statement about GPU rendering - Octane was the program they compared against. The other GPU renderes that I know of (Arion -- never used it) and bunkspeed (own the CPU version, tried the GPU demo) are hybrid. I'm not sure about the V-Ray one, I've never used V-Ray. Luxology (like Next Limit, if you've ever seen their boards) are bombarded by 'GPU' questions and that video was posted in direct response to them on a special forum that only customers of modo can access. Read whatever you want into it - but do you really think Luxology is going out of their way to attack Octane for an audience of people who have already bought modo? It doesn't really add up.

I just popped a GTX 480 in my workstation and Octane runs nicely on it - I plan on putting another one in there. I have no idea what kind of machine can hold 4 480's (you need 4 pci-e 16x slots and a power supply capable of supplying 1 8pin and 1 6pin power cable per card) - so I'm not sure that kind of machine + $2,000 in 480's is much cheaper than a $5K 12 core Mac Pro..and it really isn't the point. The main point of the video is you don't get a **massive** speed increase (and you don't...and from what I've been reading on the Arion forums it's the same deal in regards to Arion vs. Fry) and Luxology won't sacrifice features (like all the features modo has that Octane doesn't...displacement, fur, SSS, transparency absorption..etc.).

Yes - I know Octane is unbiased...as is Maxwell which I also own. Nobody it seems will let you forget that - it's the biggest marketing buzzword since 'uncompressed'. Not a single one of my clients will ever give a crap if an image is from an unbiased renderer or not - they will only care what it looks like and how much it costs them....in that sense I think it's an apples to apples comparison (although it's not in the technical sense). Anyway - if you feel attacked then take it up with Luxology...they're a pretty open company. In your position I don't think you can really help but to feel like they put you in a negative light, so I understand where you're coming from.... but I really don't think that's the case and if you look at the circumstances it doesn't really make sense...nor does it make sense to be aggressive towards Octane (they're just not the same kind of program...most people buy modo to model.)

You could always do your own comparisons and post them, I'd be interested in seeing that. There was a long thread on the lux forums by some users comparing modo, Maxwell and Octane way before Brad made that video. Furthermore (and take this for what it's worth) I know Brad - he was trying to get people off of his back about GPU development and provide some reasons for doing so (lack of features, memory...fairly negligible speed increase). To think that he made a staged and calculated sneak attack against Octane directed at people who already own modo (most of whom probably couldn't care less about Octane but just want GPU rendering inside of modo) and then followed it up with a modcast where he explained the whole thing (I'm assuming you think that was staged too?) doesn't make sense to me. The comparison may not have been fair regarding hardware and biased vs. unbiased - but it's not about Octane...it's about the future of GPU rendering in modo. Many of the people who where invited to see the video don't really care about Octane (no offense), they're modo customers.


-Greg

[radiance]

hmmm....

400.000 samples/sec on a i7 vs 4.000.000 samples/sec on a GTX470 = 10x, is that neglibable ?

Radiance

[gristle]

Taking earlier brute force comments into account as well?

[8Eggar8]

Invading this post to break up the fight.

What ever happened with this technology. Has it gone anywhere as of Late?

http://www.youtube.com/watch?v=THaam5mw ... re=related