the new i7 imacs and autodesk smoke?

[smok3]

What i imagine is to build 3 new editing workstations which will roughly be:
- fast imac with fusion internal drive and external thunderbolt raid array for video
- aja io xt for preview / capturing
- software: autodesk smoke & adobe cs6 collection

then either;
- gigabit networking for getting the material out of the fiber-channel network (dedicated server with fiber-channel connection just for that)
- getting some sort of fiber-channel to thunderbolt converter for each machine and connecting directly (problem is this is somehow driver related = all the nodes must have the same version as well ...)

This is strictly theoretical, i haven't actually have a chance to test a system like that, basically I'am a bit scared at the thought of giving imacs 100% CPU usage for 20 hours/day...?

Thoughts? (mac pro's are not allowed into EU anymore for some reason...) > I'd like to keep all the systems as simple as possible.
Are you running a system like that? (What are the pros/cons, is it working as you expected > fast!)

[DanielSB]

To be perfectly honest, I'd be scared about that, too. They do tend to get quite hot with even moderate continuous load.

I would also think twice about tethering them via thunderbolt, considering you're doing your RAID on that bus. Bandwidth is far from unlimited on these buses, though pretty awesome compared to older technologies. Perhaps waiting for the new Mac Pro could be an idea? It has Thunderbolt 2.

Or, you might consider making some Hackintoshes. The new Pro is neat to look at, but personally I really don't see why they went that route. The primary point of getting a Pro in the first place is the expansion possibilites, and even if Thunderbolt 2 is better than Thunderbolt 1, nothing beats having a bunch of PCIe slots for those SDI cards and/or pro audio cards. I seriously think Apple misjudged the market here — even if it's still a very nice piece of engineering.

Anyway, I'm writing this on a Hackintosh, and to be perfectly honest, it's the nicest setup I've ever had, and I have more expansion and upgrade options than I'd have with any Apple-branded system. Plus, it's much cheaper. Plus, it's much faster.

Just some thoughts :)

EDIT: did I just answer an almost four months old post? Sorry if that's not comme il faut! But it is quite a the top of the list :)

[smok3]

@DanielSB, it is still actual (for me), basically i'll wait for the new mac-pros (however getting into doubts there as well, due to missing expansions). Also decided on the software more or less (smoke 2013 did not made it, although i can still be overvoted ).

[Blue_MiSfit]

No hackintosh for pro stuff. You don't want to deal with that, trust me. It's also not exactly legal AFAIK, which is a no-no for pro work.

What kind of SAN are you working with, StorNext or some derivative like XSan? If you're using StorNext you could totally just do DLC, which will run data and metadata all over ethernet. The problem is, you'll be limited to 1 GigE, and best practice is to segregate the two types of traffic across two interfaces if at all possible. If you need more than 1 GigE, I'd get a Thunderbolt Fibre Channel interface, though I haven't personally played with this.

The new Mac Pro looks great, but it sounds like a lot of breakout boxes, since everything connects via Thunderbolt. I'm wondering who will come out with a clever solution for managing all that!

[smok3]

No hackitonshes for sure, now about xyzSAN, i was thinking about some sort of nix or osx server bundled with all the possible cards and they can install whatever ****** networking software there, mac's will just see nice "netatalk" folders and can stay pretty much default. I think that will simplify things at the cost of speed.

[Blue_MiSfit]

I don't know enough about the IO demands of smoke / flame, but I imagine you'll be working with high end media, right? Certainly HD ProRes or similar on the low end and maybe 4k DPX on the high end? In either case, I wouldn't try working with any of that stuff over NAS protocols. Inconsistent latency and throughput will probably be dealbreakers. The advantage of a fibre san (or a hybrid fibre / ethernet san) is predictable latency, and extreme flexibility for lun / filesystem configuration.

You can probably do most of this with a RAID + software (or pure software ZFS type thing) backend, and an AFP front-end, but I don't know about the latency.

[smok3]

The practice shows that "best" way is to have media "locally" stored (some sort of thunderbolt 2 external array on each machine?), so network would be just a download protocol. However something faster could be interesting when sharing stuff from machine to machine (most interesting for me & always wins is 0 administration, I just don't want to deal with that anymore - It tends to force me to use specific software or specific organization for editing which is a no no).

That way the server works as a separator and different admins can sit there and bastardize it as they see fit, instead on my editing machine...

p.s. Most stuff so far is prores 4:2:2 or 4:4:4 internally, material comes in all kind of bs codecs, lots of it in different mxf variations.

p.s.2. smoke 2013 will not happen as i said (demo was tested on 3 different configurations and the ui moves like a snail, basically 2012 version was way better for the stuff we do, but due to low info values provided, the reasons are unknown - again something i just don't want to waste time with).