Blender with Cloud Power! - Gaming & Rendering

Brought this up months back on here when talking with Brecht about Blenders future, Got slapped down hard when i said Blender should start planing now for the cloud. Today we find out why dev’s with access to next gen consoles are saying they piss on the best current gaming pc. All i can say is Told ya!, Now we need to work out the catch up plan.

Most people in the world do not posses bandwith needed for seamless game streaming. My guess is that its just marketing right now and well be able to talk in detail about this after seeing the thing in action instead of MS backed articles.

I also do not get how Blender is involved in all this. Do you propose to implement some kind of streaming feature into Blender Engine?

EDIT: As far as I know, games are not the focus of Blender Foundation right now anyway.

As far as cloud processing on xbone goes, I’ll believe it when I see it.
What all that has to do with Blender, I have no idea. The only thing that Blender needs to be able to do on remote servers is render. And that’s covered already.

I remember reading 3-4 months backs Sony had purchased a big cloud vendor for the PS4 use, It makes sense that both machines will take this approach. Stream processing is unlikely, it is possible but like you say only with a stupid fast broadband pipe. But pushing quite a big chunk of stuff could be done even on your normal 12MB broadband connection. My point about Blender was in the same way Blender modeller evolved, needs to evolve, Nvidia showed an impressive Cloud based Path tracer example through the cloud using multiple GPU’s of Titan from thousands of miles away. Blender provides software for free but who said that a good funding resource would be powered Cloud. Pay a small fee to rent time for cycles rendering projects to help fund the main app, Also like i said as time moves on giving Blender game engine access for pc gaming (opengl, Catch all PC market) makes sense. Guess where just have to see how things evolve here.

Blender on the cloud would be great but to sustain it all and also help fund development I think the fee would be quite excessive.

Also how would BF afford to build those massive computer farms? They barely even have enough to have 4 active developers on a minimum wage.

This was my point, Blender was promised to always be free for the future but cloud is something could charge for. In terms of setup costs, Targeted partnerships with other company’s would be the way forward to start with. Approach a current cloud provider (there are many now) and work out a scheme for usage, thats the beauty of cloud (mate of mine works for Avnet, Within cloud you can set multiple operating systems through VMware etc that change on the fly for what current users want to do). Also i don’t think it would be at a big push to get a crowd funding project going to cover Blender clouds setup cost if a partner can be worked out (or even many Cloud partners who have spare capacity as many do).

@MadMinstrel, Yep blender can do renderfarm. But it’s not intercative realtime which means spotting issues, one of the biggest issues even for top company’s like pixar is animating etc but not quite getting what you thought you would so the scene would be sent back to artist’s fixed then sent back for rendering. This is not effiecient, especialy if your a small company. This demo shows the power of Cloud.

It doesnt take a geniuse after watching the video, that the setup costs to start with could be very reasonable, and as time goes on gpu will get faster and cheaper (and also once AMD sort there opencl out this market will become very interesting). For example even at current priceing from above:

16 Gpu Box $39.900 x 12 = $478,800 = 192 GPU’s power (in realterms very low setup cost)

As Blender is free you dont have the autodesk license costs, You wouldnt have the license cost of otoys software (unless that what you wanted to use).

The system could be setup that it’s utilised 24 hors a day, people book a time slot (say 60 minutes) and for example get charged a flat hour fee of $100 as shown above could be enough to render out a full 45-55 minute holywood quality film render (good deal by anyone’s scale).

$100 x 24 hours a day = $2400
$2400 x 365 days = £876,000, $876,000 - $478,000 purchase cost for kit = $398,000 profit before running costs. even after running cost’s with such a small production volume of kit you would get a very nice little profit. Scale this up 20x and blender never has funding issues ever again.

If BF could find a service that let’s them use the servers for such a low cost that they can offer the service cheap and still get enough revenue to support development then I’d say go for it. Personally I think that’s impossible though.

12MB connection is not the norm. Even in a country that is in the top 5 list of the fastest internet connections in the whole world the norm is barely 1MB DSL connection. It`ll take years for fiber-optic network to expand for it to be considered the norm. And a lot of the world is lagging behind in that regard, USA included.

@Bloodwork, For gaming you might be right but most people have at least 2mb connection min. But for Blender cloud rendering that wouldnt be a big issue, you can stream video on slow connections even. All this system would really be doing is doing all the leg work on the cloud and the connection would only be needed to send the video results and give and receive mouse control etc, as part of the service you could maybe pay a few extra POUNDS (im british) and be sent the render results by Blueray. This does make sense, if i dont get much response from the dev guys ill have to get in touch for a conversation, If theyt could just make sure Blender can run in the cloud Blender cloud could even be setup by private company (e.g ME!). Like i said i dont think crowd funding $500,000 would be that difficult once all Blender users see the possibilities.

Just watched that Nvidia video and it seems that Ocatane guys will provide exactly what you ask for, they even mentioned Blender support and the price seems more than ok - “a maximum of $1/GPU hour”.

I am not exactly sure if streaming from the cloud is a viable option for 3d CG tasks where visual quality matters.
Render-equation converges in milliseconds in the cloud but can`t say if its noise or streaming compression artifacts? :wink:
Having a hard time texture painting due to the latency?
Problems selecting the vertex due to compression because you lack bandwidth?
Sluggish sculpting due to latency?

@Bloodwork, Well then you’d have to buy octane for blender (not sure how much it costs yet)and pay the license for every team member on your project,also at 1$ per gpu/hour on above set-up spec would cost $192 an hour. Ive shown above even charging a flat fee $100 hour for 192 gpu’s still creates a nice profit. Using Blender cycles (which is what im saying) would also give Blender access to gaming path’s in the future with no extra costs and give devs much better control over feature enhancements than relying on another company (otoy) for bug fixes etc etc.

@arexma, This would be for pure rendering not using blender as a full app in the cloud (i guess you could but why would you want to at $100 or $192 an hour just to do sculpting or texture painting) :confused:

Practically anybody with the required skills can set up a EC2-based cloud service for rendering or simulation with no upfront investment in hardware. Cycles could be patched to work over the network in the viewport, just like Octane Render demoed above. The question is whether that is really a viable market, as far as Blender users are concerned.

@Zalamander, Thank god someone who understands what im saying. That basic premise was what im putting forward. As i said above by strategic partnerships with other cloud providers even with current gen server tech it makes sense. The example i gave above about nvidia was how this can be leveraged even for GPU based setups (hence future game engine links) and agian people look at the simple math i laid out, After the first year for the setup of the 192 GPU devices think about year 2. $876,000 profit before costs because tech cost were factored in the first year profits, but none of this is possible without Blender being updated to take advantage of the cloud full stop. The beuty of opencl (even though nvidia are trying there best to kill it off because it potentially hurts CUDA, by making it slow as S*hit on nvidia hardware) is that even cpu devices can now take advantage of the benefits (unified CPU, GPU processing architectures). But who as a Blender user would not want the ability to pay $300-400 (not even pounds) to create there own Hollywood quality movie (emphasis lies then with artists to create).

I would love a script or something to facilitate this. I doubt it would come from the BF because that would be endorsing Amazon but I know there is a script like this for Mitsuba. Maybe someone with the required (python I believe?) skills could get it to work with blender?

I really don’t understand what you’re trying to convince people of. Do you want Blender to start offering a cloud rendering service? If so, why? The video you embedded above shows that other companies are already doing this. BF doesn’t have to. The guy from Otoy even mentioned Blender as one of the applications that they’ve got this working with.

If you think this is such a great idea, put a business plan together, get some funding and start up your own cloud render service…

I think the Blender Foundation should concentrate on making Blender as great a piece of software as they can. Devoting expense and time into a render farm isn’t such a good move… There are companies out there with more money than the Blender Foundation that can afford to buy thousands of GPUs in bulk and would be able to undercut the cost of what the Blender Foundation would be able to offer. Would Blender users use the Blender Foundation render farm just because it’s coming from the Blender Foundation? Maybe some will, but any production looking to keep costs low will shop around and go for the service that provides best value for money.

Such a service would also be a bit silly to just focus on one application, in my opinion. Would the Blender Cloud just work for Blender? Sorry to say this, but compared to the rest of the market, Blender is quite a niche application. A render farm just focussing on Blender wouldn’t do as well as one that renders from many different applications. A Blender Cloud would be crippled right from the start mainly because of this. The competition have a bigger slice of the market. Unless the whole of the CG industry switches to Blender, that is never going to change.

If Brecht did slap you down before for this idea, it’s probably because he doesn’t see why the Blender Foundation should offer such a service.

@blurredmotion, Yes Blender to start offering this. Like i said many times Blender had pledged to be free full all time, but a cloud render service would allow profits for development of the main engine. As i also said above, Otoy are doing this but your have to purchase Octane for blender and pay $1 per gpu (which isnt as cheap as i pointed out above as what could be designed with very little start up capital (Crowd funding perfect situation here). Well that could be easyly avoided by Blender cloud only working with cloud services (they said Blender app would be free they never said they have to help other company’s make money). And yes there is no reason why cloud services couldn’t allow other engines and apps, but as being aimed at Blender dev’s, Blender dev’s should always be given priority. Aim full stop to create large scale Blender dev funding. PS no Brecht didn’t slap me down but the other forum members who have no imagination or ambition for where Blender should be in 2-3 years time.

It’s an interesting idea, definitely, is all down to if the BF thinks it would be worth it. I suppose Blender does have similarities with Otoy in that they develop render engines, so it’s not out of the question to suggest that they offer a similar cloud rendering service.

@blurredmotion, Well Blender has a big advantage. Blender is a full featured modelling, rendering, gam engine, physics, fluids app. Octane is not. Blender has all the spunk to get this done, question is do the dev’s. I might not be there favourite person (who knows) but id fight tooth and nail for blender pushing the boundaries in it’s own advantage rather than others. Also as im developing anew HDRI Opengl renderer to plug in blender mixed with path tracing realtime, with Blender as my level editor,Physics editor,cut scene editor etc it’s in my advantage to push this through. :slight_smile: