Animation Lab (Maya, Mudbox, C4D, PS, Illustrator)

#1

Greetings!

I’ve been looking into PC builds for an animation lab being built. Hopefully, after reading quite a few posts and trying to get up to speed on the latest hardware, I’m headed in the right direction :slight_smile:

I have a budget of approx $2500 per student machine. The computers will be running mostly Maya, as well as Mudbox, C4D, Photoshop, and Illustrator. There may not be upgrades to these computers for a few years or more so I’m hoping to put together a build that will last awhile.

I’m also thinking of building a few cpu workstations (possibly rack mounted) to be used for render nodes, as well as trying to take into consideration the possibility of GPU rendering becoming more widely used (with the likes of Octane/Vray RT becoming more mature).

For the student computer build:
Case - Fractal Design Define R4 (help keep noise down)
PSU - Seasonic X750
CPU - i7 4770k 3.5Ghz
CPU Cooler - Noctua NH-C12P SE14
MB - Asus Z87-Pro or Gigabyte Z87X-UD5H
GPU - GTX Titan, or possibly save a few bucks and do GTX 780 with possibility of SLI in future (or quadro k4000 if it’s worth considering)
RAM - Gskill 1866 Ares 16GB (2x8GB) (should I be looking at 32GB?)
HDD - Samsung 840 Pro 256GB (OS and Programs)
HDD2 - WD Black 1TB 7200 RPM (to store student data)
Monitor - HP ZR2440w 24" IPS or Dell Ultrasharp 24" IPS or Asus VS24AH-P IPS (all 16:10 IPS)

For the render nodes, I’d probably try to go with basic OC’d 3930k’s builds in rack-mounted cases.

I should mention that as an IT person I’ve been sort of pushed into supporting this 3D/animation area in the last month and have been overwhelmed by options for rendering (mental ray, vray, iray, cpu vs gpu) and ANY info from experienced folks will be greatly appreciated :slight_smile: I’m still trying to learn the workflow and how to setup bucket rendering with mental ray etc so I’m pretty green. I really want to get an animation lab up and running that the students will be both proud to use, and will facilitate a great environment to learn this stuff in. Also, if there could be things I’m totally overlooking for a lab like this, feel free to let me know!

#2

I would not take GPU renderers into consideration at this point. Most are missing many essential rendering features like motion blur, displacements, hair, instancing, and they have a relatively low memory ceiling which limits the complexity of the geometry and textures. Because of those limitations there are very few uses for GPU renderers (product renderings, architecture, stylized animations). Maybe in time the GPU renderers will be better, another 5 or 10 years perhaps, but by then the machines in the lab will need to be replaced anyway and you can reconsider again at that time.

I’d go for 32GB of memory right now. The cost is not that significant and if the university you work for is anything like the university I worked for the machines may never get upgraded until they are flat out replaced. It’s not a matter of money but the time and the red tape you have to get through to buy 50 sticks of memory. Same goes for the hard drive, why not go with 2TB disks while you’re at it. The price difference is insignificant. Good luck with the lab!

#3

I would pick a cheaper system drive, since it won’t change the student’s experience much, nor the performance critical stuff one bit (say 830, or 840 non pro, or even a spindle drive to be honest), and offset those savings to 32GB of RAM (agree with Olson).

Case you can probably cheapen out on a bit. the Titan and 780 are very, very quiet, as are most good coolers with a large footprint these days.

Do you plan on overclocking them? If not you can shave off a few more bucks by NOT buying a K i7. Stock coolers are perfectly fine btw, even with a K, unless you plan to break over 4.1-4.2 Ghz.
You can also consider the i5s if the rendering side of things will take a backseat to actual seat working. If rendering will be prominent though, then the i7 makes a large enough difference with HT to be worth the difference.

This is all for budget considerations, the config itself is practically what I run at home right now (840pro, 4770 liquid cooled at 4.7, 32GB of Ares, Titan, 750W earth PSU)

SSDs are really a “nice to have” thing unless you need to online data intensive footage or other similar things, so swapping that out for a single 2TB to deal with both system and data would be perfectly fine and probably the first thing I’d do to move the money to more RAM.
It would also make you life easier in terms of imaging/re-imaging and deployment, even if only by a small margin.

#4

I would disagree on the case, I think a good case is worth every penny in a situation like a classroom. They will get moved around, kicked on the floor, knocked over, all that fun stuff. You might also consider something like this which would make it easier to setup a whole classroom since they are half assembled (and very sturdy).

http://www.newegg.com/Product/Product.aspx?Item=N82E16816101793

The power supply is smaller though, don’t know if it would have the wattage for a huge graphics card but certainly a middle of the road graphics card. They might have options with larger power supplies if that’s a deal breaker.

#5

I really appreciate both of your responses! I’ll go with 32GB of RAM and reconsider the hard drive setup. Since this is a lab, I may not overclock these either and should perhaps rethink the processor choice and go with a stock cooler.

I’m not sure if you or someone else could respond to another question. I ended up ordering a Quadro K4000 ($754) for the test build and already have a GTX 660 I’ll throw in as well. I made sure I can return it if it’s not what we need:)

I think the graphics card choice has probably been debated to death but after reading many of those graphics card threads, I (as an IT person) am still confused about a few things which an experienced Maya person would have no trouble with:

Forgive me if this question is totally dumb, but with Maya 2014, would the DirectX VP 2.0 using something like a GTX 660 show the same viewport renders as the old viewport using opengl and a quadro k4000? Or are there things the students wouldn’t be able to see in the DirectX VP 2.0, that using the opengl viewport they could see?

My understanding (from looking at benchmarks) is that the framerates really drop with the opengl view port in Maya on the GTX cards but I guess my real question is, does this matter since the GTX cards perform well using viewport 2.0 and DirectX?

I’m not trying to start any graphics card debates(I’m ignorant!) but as an IT person, am trying to put money in the right places and need clarification. I would WAY rather go with a 780/Titan for the money and only got the K4000 so that I can find someone who has experience in Maya and test out a few options. (Our professor who teaches Maya is gone for a few months out of the country and is REALLY hard to get ahold of right now so I apologize for these ignorant questions. I’ve been placed in a rather awkward purchasing position :slight_smile:

#6

I would go with a graphics card certified by Autodesk for use with Maya if the students will be using the lab primarily for Maya. Maya is picky about graphics card and always has been, maybe a little less so with the newer viewport but I can’t personally vouch for that. I think the only gaming card they have certified is the GTX 690. Autodesk has also certified workstation graphics cards from AMD like the FirePro W7000 (better bang for the buck than Nvidia Quadro).

#7

The gtx690 is a certification anomaly, given it does nothing at all more than a 680 (690 = pacakged dual 680).
To be honest, Maya isn’t that picky with videocards these days, not with nVIDIA ones at least (while ATI can still be a roll of the dice, even in the pro line some driver cycles).
On top of that OCL still has spotty and below par support, so while AMD cards do perform way better in OCL, and have great FLOP per Buck, I honestly can’t get myself to advise people drop nVIDIA (which I have little love for as a company) and CUDA to buy into potential problems, even if at a discount.

The main detraction for 6xx and 7xx cards is the crippled DP computaiton (if you need it, which is not frequently in DCC apps), and that’s where the Titan shines as it’s the only non quadro card since the 5xx that isn’t DP crippled.

The ladder, in terms of reliability and performance slanted towards the former, goes k5k > Titan = 580 > 680 > 780. But don’t let the order fool you, if you value performance and value and don’t care about DP computation you might as well turn that around exactly.
Having a K quadro at work and a titan at home, and having had that 100% direct comparison for years and for every generation now, I firmly remain in the camp of those thinking a quadro card, for most of DCC, is frankly speaking just not worth it.

It simply isn’t true any more that Maya is picky with gaming graphic cards, no matter how much feet are stomped and breath is held.

The lack of certification is purely a matter of form factor and providers mixed with nVIDIA’s marketing lobbying. The Titan being 100% nVIDIA form factor for Asus, GB and another couple brands (Other than Palit, which skimped on a few things apparently) is for all intents and purposes as good as a Quadro, with full DP capabilities, more memory than ANY quadro out there (the fermi 6k matches it but it’s a fermi, and grossly overpriced) and a build quality and noise levels better than.

I’m two years and spare change into CUDA work that over the last few months has got fairly serious, and I’ve run that in application and in sterile tests on both, and I bought a Titan for good reasons, even if I can more than afford a k5k and would have bought on if it made any sense, but I saw no point in downgrading the memory and having a less efficient build in exchange for more money and a signature change in the drivers.

#8

Are you building these yourself, or just picking components?

I can vouch for the case, its a dream to work with and very quiet, though do consider getting a second case fan (fractal R2 140mm) to add to the front panel - connecting all the case fans to the fan controller that lives on the case itself and run at lowest speed - practically silent :slight_smile:

Also using a quadro K4000 at work and a titan at home, haven’t had any instability issues with the titan yet, though have only briefly tested maya, mostly softimage.

As for gpu rendering, I think it will mature a lot faster than 5 - 10 years. I’m testing Redshift 3D atm, which doesn’t have the traditional memory limits of gpus, does have most features ready, may well be released this year. Keep an eye out for it!

#10

i am building a pretty similar system for pretty much the exact same programs. I have went for an ATI 3gb HD 7970 GPU though, mainly as from what i can gather, the viewport performance has better write ups. Does this have any substance?

P.S i wish they had systems as good as what you are building for your class when i was studying haha :stuck_out_tongue:

#11

Thank you to all of you for taking the time to respond! I REALLY appreciate it.

I’m putting together the test machine right now. I will certainly be taking everything you guys mentioned into consideration when I order the rest of the machines. I’ll probably do a spindle drive and 780 to set aside a little more money for render nodes. Hopefully that’ll help with the render crunch time students run into towards the end of each semester :slight_smile:

#12

Yeah, I’m hoping these students will appreciate these computers. It will be a massive upgrade from what they were using last year (2011 mac mini’s…doh).

I’ve heard similar stories about the 7970 being pretty good for Maya but have a feeling Nvidia stuff still gets recommended more currently because of driver support and the fact that CUDA is still much more supported than OCL. Maybe someone experienced can verify this though. I know we did run into some Maya anomalies with the radeon’s in the mac mini’s but who knows as those computers were never purchased with the intent of running professional 3D apps on them:)

#13

FWIW that sounds like a good plan to me.
Dual monitors and off-desk resources in my experience make a large difference in working comfortably, more so than the CPU, videocard or HDD benching 5% higher.

#14

This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.