Jim and the crew make it possible for us to save something like a few hundred thousand dollars on a camera body, the hardware makers send us boxes that are twice as fast every few seasons, and we just want more....
Ralph, for you it may be an option save a few hundred thousand, I never had that option in the first place. Jim is creating tools for me as an independent film maker that I never would have access to.
I'm just looking for solutions that meet the same new price brackets he now markets to
I've done about 20-30 large, paying, FCP jobs, using about 8-9 FCP systems. Always had crashes, gamma & render inconsistencies. Not a big problem (except when it screws up the online for a Lexus promo you're finishing on a Quantel at $$$/hour) but a minor irritation to be sure.
Anyway, just saying... it'd be awesome for a RED render farm. And you'd have cash left over to buy a used copy of Avid ;) Just need redrushes for PC, please... or some way to create .RLX looks on PC for use with MetaFuze.
Don't forgot HD's Apple solutions have had such a long history with Avid too!!!
I just got my new mac pro 2.93 and it's twice as fast as my previous machine 2.66 running compressor! Looking good folks!! Now where's FCP3?
It's so I don't have this!!
"...A new wing was built to house Illiac IV It took one month to disassemble the unit from our testbed in Paoli, which had 100 tons of air conditioning built into it.
The computer totalled 53' in length, and took 11 40' vans to house it, weighing 99 tons. One truck alone had only power supplies in it. Illiac IV had a total of 11,739 pc boards......."
It's beginning to seem pretty obvious to me that the render farm is something that absolutely must be done to keep pace with the enormous files that the Red cameras will be producing. I double-dog-dare you to render anything from that monster 6 x 17cm sensor using an off-the-shelf workstation and meet any deadlines.
That said, I am imagining a rack of commodity linux boxes all connected to your workstation via gibabit ethernet and possibly eSATA as well. Your workstation can delegate a huge rendering job to the various machines by simply chopping up the source file and handing pieces off to the various machines which would devour it like hungry sharks. Adding 8 extra cores to the task might be as easy as popping in one of these --a nehalem equivalent of which is no doubt due shortly.
AMD also offers good price/performance. You can get comparable clock speeds to the Intel chips for half the price. Or you can get a full 16-core AMD monster, something like this:
4 x AMD Opteron 8380 2.5GHz 75w CPU
Tyan 4-socket mobo
32 GB of G-Skill RAM
6 x Western Digital Caviar 640GB Drives in RAID 0 config.
550 Power Supply, 4u rackmount case
approx $6700 including tax and shipping. Compare to a 8-core Mac with RAID card, 4 drives, and 32GB for $12,100.
This is very different from a mac but would probably offer similar performance. I'd love to try and benchmark one. Some notable differences:
* AMD box has 4 processors and 16 full cores.
* Mac box has 2 processors and 8 cores but each is hyperthreaded...results in 16 threads under optimum circumstances
* Mac has slightly higher clock speed, 2.66Ghz vs. 2.5Ghz
* Mac memory runs faster (1066 vs. 667). However, the chip-to-memory bandwidth advantage of the Mac may be mitigated by fact that it has half as many CPU sockets. The existence of 4 separate sockets in the AMD box means that you have twice as many sockets per thread.
* AMD machine has 6 drives to Mac's 4. Also room for 3 more internal drives with 9 internal drive slots.
* AMD has no optical drive and inferior video card but it wouldn't be running as a desktop machine. For optical drive, add $25.
* AHEM - $6000 price difference
Heck you could install windows on it but by using linux, you might gain several advantages:
1) no mac tax
2) no windows tax
3) no need to port software to windows if software can run on linux*
*Given that OSX is really BSD Unix, I would imagine that these rendering algorithms already exist in a code format that is quite ready for export to a linux box.
Does that sound ridiculous? Seems to me all that remains is the development of a bit of network code and a bit of code to let your workstation break the job out to the various machines. If done right, this render farm approach could support really amazing scalability. Just plug in another machine.
|« Previous Thread | Next Thread »|