[Mihai]

Yup, ofcourse. But it's too early to even talk animation and Maxwell. At least you can do it for simple scenes like keytoonz have demonstrated, but honestly for a complex interior with glass, even with your 7 dedicated duals and the rest of those 150 processors, it's not doable I think. We will have to wait and see what the future baking options and mxi scene viewer and texture will allow us to do in terms of speeding up animations. Then it would be more appropriate to discuss licenses. Considering Maxwell for interior animations now as it is in it's present form is just not realistic, it's not a matter of licenses. It's a matter of paying the electricity bills for 100 dual machines that run 24/7

I'm still keeping my fingers crossed those Cell processors will be available for workstations......

[Maxer]

Yea those cell chips could really change things but I guess I could settle for a dual core system for now.

I understand this is all still very early, this may all take care of it's self and Maxwell could be 100 time faster than it is today. I just hope it's being talked about in the NL circles.

[Thomas An.]

Distributed rendering (Cooperative rendering) is used when you want more than one machine to work on a single image. Network rendering is used for instance when you want to render out an animation and each computer on your render farm is sent a different frame. These are two different operations, one is used to speed up the rendering of a single image; the other is used to speed up the rendering of a large number of images like in an animation.

Sorry to jump in. I have never done animation but this thing simply buffles me:

(Assume an image takes 120min on one machine and we need 4 frames)

1. we have four machines working on the *single* image to complete it four times faster. Therefore (120/4min)/Image * 4frames = 120min

2. We have four machines working on one image each. Therefore 120min/image with four machines working simulaneously = 120min for 4 frames.

Is there really a speed difference between methods (1) and (2) ?

[Mihai]

No, distributed rendering is for stills. Very large stills

[Maxer]

There's no difference in speed with your example, but cooperative rendering (distributed rendering) isn't typically used for doing animations. Also this is why it's important to have two different licensing schemes for cooperative rendering and network rendering. You may only need 4 computers to render out a still image, but you will never be able to render out any animation if your limited to only using 4 computers. An animation typically will require between 24 & 30 frames per second of video. At 120 minutes a 10 second animation would take 25 days of continuous rendering. This is why other render engines give you unlimited network rendering because without it animations would become impractical. This is also why if Maxwell wants to compete with these other engines they really need to think about how they will address this problem.

I guess I should add that you could render out the animation in 20 hours if you had 120 machines working on it.

[Thomas An.]

Maxer, Mihai,

I still do not understand

(Suppose we have 900 computers and need a 30sec animation @ 30frames/sec. A single frame takes 120min on one machine)

1. Distributed rendering
900 computers working on a single image --> (120min/900comps)/image * 900frames = 120min

2. Network rendering
900 computers working on a seperate frame each --> 120 min for each machine to complete a frame working simultaneously = 120 min for 900 frames.

I am not sure there is a difference on speed.
Is it just a paractical matter of a different kind ?

[DELETED]

DELETED

[Maxer]

Thomas the basic difference between the two is distributed rendering was created for still images. As you increase the number of machines your using for distributed rendering the performance each machine gives you assuming each one is of equal speed is less than the one before it. So the first one you add gives you a speed increase of 50%, the next one increases the speed only by 25% and so on. Very soon you reach a limit where adding more machines won't increase your speed and could even slow down your rendering. Since this process is much more complex than network rendering you have a large amount of overhead which is the reason for the slow down. This method is used by people who work with GI which requires large amount of system resources and long render times. By networking 10 machines together you are able to interactively render much faster than a single machine could.

Network rendering was created for animations and because of that it's more efficient at handling large numbers of frames with low overhead.

In theory these two processes are similar but in practice they really only work for the task they were designed for.

[Thomas An.]

As you increase the number of machines your using for distributed rendering the performance each machine gives you assuming each one is of equal speed is less than the one before it. So the first one you add gives you a speed increase of 50%, the next one increases the speed only by 25% and so on. Very soon you reach a limit where adding more machines won't increase your speed and could even slow down your rendering.

aah ha, that was the key that I didn't know
It changes the whole picture. I understand now.

Thank you

[Kabe]

Well, I think regardles the price tag it's not a good idea to stick with the "One license= 4 Render-CPUs" scheme.

There's no software company that does it, and this is for a reason. If you are the first with a new business model you need to ask yourself if it might be because it's stupid

Frankly I guess that if you would say to NL: "Folks, we could need licenses for 600 CPUs, but not for the price of 150 licenses", then they will seriously think about a cheaper price. If you have more machines in use, you pay less money per machine.

Disregarding that would mean to shift the break even point exspecially for big customers, which is a certified bad idea. This won't last - at some point solid business practice will enter the ship ,-)

Kabe

P.S.: At the current CPU speed, I don't see Maxell a a production renderer for animations. However, the same is true for radiosity in most cases.

[Dominic]

slightly back on topic...

i am seriously considering buying 4 (or more) sets of licenses before the alpha pricing runs out and setting up a small renderfarm.

just testing the waters to see if the interest is there.

most likely 16 dual xeon 3.4s to start and growing as the demand arises. I realize its not much compared to some of the bigger renderfarms, but if they are not willing to bite the bullet and buy the licenses for a renderer that just screams to be used on a distributed network than i guess i will.


I will also support Vray, brazil and perhaps FinalRender if i can find people still using it.


Let me know any feedback or advice that you have. Thanks!

[DELETED]

DELETED

[thomas lacroix]

same here, depends on bids, speed, nda, security...
any idea for a day rent for a 2k render for exemple?

[j_man]

Yep,

Pretty heated topic, just a couple of notes.
I'm pretty sure the limit of V-ray and FR with 10 distributed rendering machines comes from the maximum number of simultaineous IP connections allowed by windows.
When rendering animation, the frame size is much smaller than for print, (PAL DV 720 x 576) so therefor render faster. Ok so this size is still long with Maxwell, but is probably a good reason to allow more network render license's with a license. It is mazing how fast animations progress when you're using a render farm, even if the frame time is quite slow.

As for renting a farm, could be a viable option but the farm would have to be pretty well organised so I could achieve what I need to do online (like res-power).

J.

[Maxer]

Just thought I would drop a note in that this post has been going on for about two months and not a single person from NL has had a thing to say. Why is that...have they got no opinion in this matter? We've got no road map, were still in Alpha and at the end of the month there raising the price of the software. Is it to much to ask for some clarification from them on whether or not we can expect a change with this license issue! They may be developing the software but were the people who are going to be asked to buy it, don't they owe us some answers?