IMHO, if the time for MP support is not now, then its already passed…
I don’t quite understand “whether MP should be implemented or not” is even a discussion, sorry, but i think this should be a high priority, the “small” features can wait..
I don’t think its a question between whether MP should be adopted or not… it will be adopted. The question is when. EI must juggle between a number of decisions on how it wants to expand its product. MP is a much larger task that will absorb many resources...potentially more than what EI has to give and so the Igors’ position of waiting on true MP implementation is wise though the average user doesn’t want to hear it. This is why the Rama question came up. If Rama could be improved to better handle distributed rendering within multiproc machines and address its shortcomings, we could “shore up” the existing method of accessing multi-processors until such a time that MP can be adequately and properly implemented into Camera.
I know I’ll sound a bit like a kill-joy. But I don’t mind he current Renderama way of working. But then again, I’m mainly interested in animation. It would be nice to have Renderama look at multiple cores in a more transparent way though.
As far as MP support within a single frame is concerned. Are there actually any 3D apps out there where the renderspeed scales up in an almost linear way with the amount of cores? According to the rendertest Dave referred to, even C4D gets less than a 50% increase with a doubling up of the cores. Is Camera’s method that much slower?
I don’t think its a question between whether MP should be adopted or not… it will be adopted. The question is when. EI must juggle between a number of decisions on how it wants to expand its product. MP is a much larger task that will absorb many resources...potentially more than what EI has to give and so the Igors’ position of waiting on true MP implementation is wise though the average user doesn’t want to hear it. This is why the Rama question came up. If Rama could be improved to better handle distributed rendering within multiproc machines and address its shortcomings, we could “shore up” the existing method of accessing multi-processors until such a time that MP can be adequately and properly implemented into Camera.
Having a ‘practical’ POV is essential in getting anything done but being overly-practicality will usually kill the ‘fun’ in life. How many kids would have not gotten into trouble and had a great time if they always listened to every word their parents had to offer.
Practical common-sense and living one’s dreams have to be balanced...too much of one and life is dead and boring....too much of the other and there is foundation missing.
This thread reminds me of these ideas.
I think we have to keep in mind that many existing users want useful features BUT EI also needs to attract new blood/customers to survive.
So what’s going to attract new users......exciting features.
Not necessarily common-sense practical plans for the future but ‘right-now’ cool and useful features which will cause the cards and cash to be removed from the wallet. There is a certain amount of ‘excitement’ which has to go in marketing and talking about a product’s future...possibly development and marketing need to get their message aligned?
Hearing about the practical importances and reasons for not getting your new toys (features) just doesn’t produce a real excitement......and there is a reason why marketing needs to have exciting, fresh news.
I’m not trying to be critical of anyone as I can see the practical POV here......but there is a big BUT here.
I know I’ll sound a bit like a kill-joy. But I don’t mind he current Renderama way of working. But then again, I’m mainly interested in animation. It would be nice to have Renderama look at multiple cores in a more transparent way though.
As far as MP support within a single frame is concerned. Are there actually any 3D apps out there where the renderspeed scales up in an almost linear way with the amount of cores? According to the rendertest Dave referred to, even C4D gets less than a 50% increase with a doubling up of the cores. Is Camera’s method that much slower?
Based on other EIAS benchmarks (blikjes GI animation test) Renderama seems to be nearly linearly faster per processor for animation. I would guess that means that EIAS will stay competitive in the animation arena with the renderama solution. If other apps are relying on MP which gives a theoretical 50% increase in speed with each doubling of cores, then randerama SHOULD be FASTER. At least in theory. So Brian’s concern that MP might be slower has some merit.
My work is in visualization for new products and environments. That means LOTS of stills.
For stills, 4 cameras can give less than 40% speed boost for single image rendering over a single camera.
That’s almost not worth it. Renderama IS useful to queue up mutiple images, but sometimes you just need to crank through ONE high res image.
One area that could be improved with renderama is it’s inefficiency in “striping” an image.
Currently this has to be done manually, and leads to a number of inefficiencies namely:
1- the law of diminishing returns sets in quickly when increasing the number of strips due to the increased overhead of communication between renderama, the slaves, and slave cameras.
2-often one camera hits a particularly cpu intensive spot and that one camera will take much longer than the others to complete it’s portion of the image, effectively negating most of the performance gained through striping.
Perhaps adding “dynamic” striping (try the cinebench demo on an MP machine sometime to see this in action) and reducing the communication overhead within the suite of Rama apps could greatly increase the effectiveness of the existing system for still rendering.
If this requires less programming resources, then EITG should go for it.
From this discussion, it sounds like creating an MP version of camera will take a long time, and require most or all of EITG’s programming resources.
I feel confident in EITG’s current direction of opening up EIAS on the front and back ends (integration with other apps, and multilayer rendering). This should make EIAS much more attractive for animation, film and TV.
So- can Renderama’s performance be improved? Is this a “small” programming task vs. MP Camera’s “monumental” one?
The Multilayer camera sound great. But the post processing effects sounds a bit gimmicky. Infini-D 4.5 had it and I used it zero times.
I don’t know individual workflows, but I’ve never seen 3D that went straight from Camera to the final project. It either goes through Photoshop or After Effects (or other compositor).
My 2¢—don’t spend too much time trying to put AE functionality when we already have AE.
. Are there actually any 3D apps out there where the renderspeed scales up in an almost linear way with the amount of cores? According to the rendertest Dave referred to, even C4D gets less than a 50% increase with a doubling up of the cores. Is Camera’s method that much slower?
It doesen’t matter, the only way we’re gonna get faster preview snapshots is either new settings and options that will probably give a lower render quallity, or an MP camera that will result in fast and bettter quality snapshots.
Of cause, the developers might have some “clever tricks” in the pipeline, if so, thats great.
Talking about new users or ever just demo testers, they probably won’t bother looking at renderama at first, simply just hitting the snapshot button, the more that CPU’s evolve on a multi-core basis the slower EI camera will “appear” to be - in comparison to other renders that do use MP
Of cause renderama is a solution for final renders, but how much time is spent previewing beforehand…
Seventy-five percent of any of my given projects is spent doing snap-shots - 20% modeling and 5% final rendering.
I have a 2 year old 2Ghz dual G5 in which when using EI, one processor sits idle - I do illustrations mostly. I have been considering getting one of the new quads or 8 cores when my lease on this one is up in a few months. That’s a lot of wasted horse-power when using EI....I guess I shouldn’t get too excited about speed when my chosen tool is not up to performing efficiently on current hardware...or even last year’s hardware.
Give me something which will speed up snaps and set-up like FPrime, IPR or take advantage of current hardware technology.
I also agree...this idea of outputting layers is nice but that’s not what I personally need. My old copy of C4D 7.3 has output to PSD with everything to it’s own layer if I want...and I hardly ever use that.
I think this is a very good point. Getting MP working is going to be OK, but if an IPR feature were included, it would be very impressive. These features should be developed together.
I think the current plans are a good for the health of the company, but speaking selfishly, I would also rather see MP and IPR instead of the multilayer output, because that is what I need, too. I guess I’ll have to try extra hard to find a use for the multi-layer features.
The FPrime programmers know something that we don´t know yet. It´s amazingly fast. How they do it?
I would kill for something like that.
FelixCat
There are also some really incredible shots of Pixar’s real-time solution here, and it’s hard to tell the difference between the “preview” and the final.
According the link, Pixar is using “estimations” and “approximations”. These kinds of tricks should be incorporated into EIAS for sure. These previews look “good enough” for a lot of final client work.
“Real Time” and “interactive” are going to be big industry buzzwords from here on… “
Yes, I’m sure there will be much a-buzz with these buzzwords but.......right now, I think most EI users would settle for last year’s (or even a couple of year’s ago) buzz.
These approximations have to have some pretty spiffy algorythms driving them - somehow I don’t think the man-power or resources are at hand for EI to undertake some sort of development like this.
Anyway, I think I’ve said my say - I need speed from EI...that is it’s heritage, along with quality output.
I can’t agree more, Multi-Layer will be HUGE for us and many of the other EIAS users I know in my field.
As for the ability to have post-processing effects being gimmicky .. This feature was actually required in order for the multi-layer solution to work, it wasn’t additional work, it was a happy side effect. My favourite kind of side effect
Seventy-five percent of any of my given projects is spent doing snap-shots - 20% modeling and 5% final rendering.
Fair point. I don’t know enough about other people’s workflow. I was only commenting on my own situation.
80% of my time is spent staring at the F-curve windowand the Project window. So I probably could tell you a few things about those.
