Supercomputer recommendations

#1

I would like your recommendations regarding a supercomputer for Visual Effects films and Game development for next generation PS4 and XBOX

I am in the process of building my media company similar to ILM and WETA. Finance is already secured. I will be adding a games development department. It will not just be another VFX company. I am a story teller and I have a few scripts finished. I want total control over my content and freedom for my artists without corporate interference.

I have read up on ILM and WETA and I have noticed they use CPU’s and ram to render their content which can take up to 15 to 20 hours just so they can see what it looks like, because of the complexity of the scene. I want my employees not to worry about rendering but to do what they enjoy best.

I want to build a system where they can do just that. Cutting edge technology for maximum creative control without time constraints. I have contacted SGI computers twice in the last month to get advice. On both occasions they have promised to return my calls. Nothing so far. Highly unprofessional. They have just lost a contract and I am fed up of chasing them.

This is their product that I was interested in;

http://www.sgi.com/products/servers/accelerators/gpu.html

I am now stuck as to which company will help me with my business inquiries. Does anyone here know of companies that I can look into. Am I even looking in the right area?

Please guide me if I am mistaken.

Much appreciated.

#2

You’re on a wild goose chase and SGI probably realized it was a waste of time because they didn’t have what you’re looking for (nor does anybody else). Since those machines feature Tesla cards I’m guessing you’re interested in GPU accelerated renderers. While they are faster when you look at the rate in which they ray trace they are also very limited in the scope of what they can render (fur, displacements, motion blur, etc.). They are also limited by a much smaller quantity of memory compared to the main system memory. These limitations make them of little to no use in visual effects and animation production aside from preview renders and animatic type stuff where you don’t necessarily need all of the advanced rendering features like fur.

I would suggest bringing on people with rendering pipeline experience to find solutions and build the pipeline for the studio. Sure there are things that could be improved compared to what most studios are working with today but there’s no silver bullet which is what it sounds like you’re looking for.

#3

It would be worth your while taking a look at Redshift . It can render fur, displacement, deformation motion blur, multi-uvs, and much more and is not constrained by the VRAM (though it does slow down when going out-of-core). Its replaced our render pipe and imho is the best choice going forwards, given that cpu development has slowed drastically and gpus are adding more memory to cope with 4k display.

#4

It supports two applications on a single platform (only Windows) and works with GPU from only one vendor. I’m sure it’s a fine renderer and it’s nice to see that at least one renderer is getting past some of the show stopping caveats but it’s still a niche thing. It’s also not the silver bullet the original poster was looking for.

#5

Thanks for your response.

As I have mentioned before, I may have been looking into the wrong area. This is where I needed guidance from people here who have experience and what they need to speed up their work.

Which kind of components do VFX artists utilise the most? Is it the CPU and RAM, CPU, RAM and GPU? I have not yet began the recruitment process. This is coming very soon.

Should my team built it from the ground up as you have suggested to fit their needs?

I have added redshift to the list. The teams will look at all the research I have collected and decide what best suits them. I am still researching.

Your advice is greatly appreciated.

#6

There’s really little point contacting hardware manufacturers when you haven’t yet decided on what you want. If you’re going to be spending a fair bit of cash on this then it would be worth your money to hire someone for a day or so to go through your needs and properly configure you some systems which are worth your money.

If rapid feedback truly is the top of your priorities, then this will hugely shape and limit your choice of hardware and software. The more limitations you are happy to accept, the faster the render engine you can pick. You’ll need to tell us before you get any useful answers…

Mac or Windows?
Are these render nodes or workstations?
Budget?
How many do you need?
Do you have suitable Aircon for a render rack?
What high end things will you need from the render? (fur, multiple passes)
What sort of scenes will they be, architecture visuals, cartoons, product visuals, films?

#7

Well, essentially only one application now…

#8

Oi! Plenty of us still rocking the softimaaage. Actually, redshift are going to be pretty much responsible for where we go next, so I am keeping an eye on which app they pick after the 3dmax implementation.

Sorry for the slight tangent…

#9

I believe it may come down to consulting an expert for a day but I am looking into other alternatives for research purpose. I am not going to be spending anything on hardware until the team goes through the list and decides which best suites them. A supercomputer already built that they can configure and adapt or build it from the ground up.

Windows

Render nodes. Workstations are easily built I believe. I am looking to make the job of the designer and VFX artist less stressful and more creative and productive. I don’t want them to wait around for something to render all night.

Budget is dynamic. It can be adapted to invest on the hardware. This is a high priority.

I don’t know how many I would need. This is what I want to find out.

I never thought about aircon but maybe liquid cooling with aircon. Could be expensive.

High end things will depend on the artist. This is the part where I am not sure what to have because I am not an artist. I would like to have most of what VFX artists use. It’s best to have something and use it a little it than to not have it and need it badly. This kind of business always changes. Something could come up that we may need it.

It’s motion picture films and games. PS4 and XBOX one.

The reason I am spending this much focus and money into hardware and the company is because I don’t believe an actor is worth 20 million dollars. They can get money from endorsements and advertising. In fact most of them already do. That’s how they started out. By doing doing advertisements and transitioning to films. For an actor or actress to receive 20 million a film and other millions from endorsements is robbery without violence. It is absolutely ridiculous. That money should be distributed to people who spend 15-20 hours a day in front of the computer who have families to feed and a life to build.

#10

look, my opinion is just this and I’m certain many people around here don’t like it, but that’s the way it is…

…as for your staff, you can keep it fairly lean and get decent workstations for daily, hell you could probably even rent shit to get the job done… but today, in terms of short and especially long term… any serious hardware investment will gain you nothing. In two years time, you will be using old juice to charge your vision.

“if” I had the cash flow to jump start a studio, I would be seriously considering offsite rendering solutions in my pipeline for any billable work. The scalability can not be matched with an onsite farm solution.

Perhaps you could consider a hybrid leased arrangement as it grows… partial onsite, scalability offsite. “as it grows” is the key… why dump so much money upfront on hardware that you know will be outdated. Capital depreciation?.. not for that.

[edit]
…
you’re not paying an actor/actress 20$M because they pretend exceptionally well… you’re paying that to them because of all the inaccessible media coverage and non-tangibles that you and your employees at the grind stone could not possibly ever generate.

[edit2]

oh look… I hit Lord of the Posts… hahaha.

#11

I keep coming back to this.

Don’t you think ILM and Weta want the same thing?

#12

You do actually have a point there tswalk. Hardware is always outdated within two years. Having someone else render it for you and paying them is far less expensive than buying it and maintaining it. VFX will become more complex. Unless people can unleash the power of quantum computing.

Dmeyer, I believe they all want the same thing. I have yet to see Nvidia mention their products being used for such purposes. ILM tends to be quiet about the power of their computer so it’s current configuration is not known. It was originally built by SGI, the same company I was trying to contact.

According to Wikipedia, it has one of the largest renderfarms currently available.

I now have a clear idea of what my best options are thanks to the people who responded. It has answered some of the questions I had. I have taken into consideration renting a render farm rather than build one and maintain it.

#13

Cloud computing makes sense if you need a lot of computing for a small fraction of the time, like a month out of each year (online holiday shopping for example). It doesn’t make sense to maintain 1,000 computers if they will be idle 90% of the time. This convenience comes with a markup though and a pretty significant one.

It’s kind of like renting a car at the airport which is good if you’re on vacation for a few days but if you need a car everyday for the next two years then renting is a horrible idea.

For the original poster, you’ll have to hire someone for more than a day to figure these things out. You’ll need a team of full-time people to figure these things out and maintain things. Render farms don’t build themselves and pipelines don’t write themselves. People make these things happen and you’ll want the best based on what you’ve shared about your expectations.

#14

well, that’s probably a part of your problem… going to SGI. However much I would like to forget about them from their business practices in the 90’s…, I would never in todays world of multiple platform options deal with that company… ever again.

and that’s all I’ll say about it.

#15

My business plan has now somewhat changed

  • Rent a renderfarm for film VFX

  • Own a renderfarm for games design and development

The reason for owning one for game is because once the hardware maximum is reached, you can’t really upgrade the renderfarm anymore. Consoles have a lifespan of ten years. Films, however, need to be maintained every two years.

It’s just makes financial sense to rent it for films.

I will definitely be hiring people to discuss everything I have researched and received from this topic. I am very glad I have asked the professionals in this forum for advice.

Much appreciated.

By the way, this is the company currently contracted with ILM.

http://www.cirrascale.com/products_gbseries.aspx

#16

Renting render farms are good if you need that extra push for some short deadline that you simply cannot achieve with your own resources.
But for everyday use cost of renting will sky rocket.
Also don’t underestimate all the time and effort needed to prepare everything for render farm. All plugins, addons and script, all textures available etc…
It is really not simple push of a button.

Having real time rendering seems what you are looking for and tech is not there yet.
What can be done and what is actualy sane to do is to get maximum for best investment… Throwing twice as much money for improvement of 10% is BAD business decision.

Also leaving hardware decision to artist and giving them open hands… hehehe that would be like buying 1.000.000 EUR car to go to the market shopping. All they would make is nerd wet dream that doesn’t do that much better then couple times less expensive workstation.
With free option like that… do you need or do you wanna have 6 4k monitors in front of you and latests and most expensive of everything under the desk.

So making new ILM or Weta is a bit vague… check how many years are they in the field, they were not built over night.

Also even if Weta comes and gives you all their hardware at disposal, without artists and people that knows how to make things happen… you have most expensive paperweights ever.

To make things short, you are really looooong way till hardware making decisions.