#Render / Video Farm
21 messages in this thread
Is there a high demand for sending rendering jobs to outside company's? The
reason I'm asking this is for the simple reason that my company is about to
purchase 130+ P100's all attached to a Netware 100BASET network; and durring
the night I figure I could have them rendering rather then sleeping! I could
send the output to either DAT tape or to video (PVR). Let me know if there is
a demand for this… Thanx
Bob-
I do a fair amount of long distance rendering for others here on the forum with
my small network system. While it would appear that "the computers do all the
work" to the average computer illiterate on the street, the hard facts are it
takes special knowledge of the process, 3DS networking operations, animation
rendering idiosyncrancies, and UPS power systems, not to mention the video
side of things. People with budgets to farm out rendering are usually
demanding on video quality and up against a deadline. If you think running a
render farm of 6-10 machines is just a simple matter of throwing a big switch
at 5 PM when the boss goes home, plugging in a DAT tape with the project files
from the client, then you go home… return the next day to a completed video
of the animation you obviously have no concept of this business. And 100
machines??? Well this is just fantasyland stuff. Others have announced this
same proprosal as yours ever since the network rendering of r3 came on the
scene but to my knowledge none have ever successfully got these super farms
operational to a full service business.
The reality of the situation is they soon realized that running a sideline
business like this is #1. not profitable; 2. difficult to make the boss
understand why they were so tired the next day after fighting problems on the
rendering job all night; 3. The first job they were sent was a total nightmare
of screwups requiring a lot of fixing prior to rendering; 4. They had to go out
and purchase 80% of the budget of IPAS routines to complete the job. 5. The
boss found out about it and wanted some of the gross money; too much so that it
reduced the wage to do this to 5 cents per hour if that.
What is realistic is if you have a moderate sized network say 10-15 machines
and already have a fair investment in IPAS routines, broadcast Video equipment
and have the engineering know how to pull it off you can start your own
business as a video post operation with animation services. Then to fill the
void between major animation projects, take on selected jobs for this output
service. You only use this service offering as a profit center that otherwise
fills hours on the expensive equipment you already purchased. Again this
equipment must include video output to broadcast because it's been my
experience that the only jobs that get farmed out are video output combined
with rendering. ie these jobs are either just video output or a combo of the
two. Every potential client I spoke with almost always wants to do his own
rendering. Also realistic is that other's jobs will always require your full
attention during the rendering process and they all have crazy deadlines. Just
forget about sleeping on the job and leaving the computers to do all the work.
That kind of thinking is usually the figment of the Harvard Business School
professors not successful entrepreneurs.
Don,
Right on! We have had up to a 30 P100 64MB Farm at one time and the project was
a nightmare! From the time limit imposed to networking considerations and even
tougher – figuring out the network rendering bug between R3 and R4 machines.
And, then the laying to tape of 17,000 frames with the matrox board – Yeah
there is a market but few can afford to pay you for what it will cost you in
time, hardware, and sweat & blood.
Matt Armstrong
TRG / DFW Multimedia
Don,
that was a great reality check. The more machines you add the more admin you
have to do. Give me one or two super fast machines any day, versus 100
moderately fast.
I don't like the idea of sending out rendering. Projects need a lot of baby
sitting and they rarely (if ever) are right on the first render.
I think your right on some of your comments, while I have been a network
administrator for over six years at a 1.2 billion company and working with 3ds
for a little over one year, I sometimes still think of myself as that "computer
illiterate on the street"… not too much any more! 🙂
All I plan to do is get the jobs (if any) done and asses wether it is worth
continuing. Thanx for your input!
Bob-
Robert,
I would really like to have a shot at selling your company (130) P100 systems.
We specialize in high-end custom systems.
Matt Armstrong
TRG / DFW Multimedia
Hi Matt, sorry AT&T GIS beat you to it. Contract is already signed!
Bob-
Bob,
Well, can't win em' all… Just let me know if you need any help setting up
your farm. One thing, make sure you have a nice BIG fileserver to handle your
network overhead…
Matt
Yes there is a demand for this. I would need to be sure that you could work
with deadlines, etc. but this would be attractive to me on some jobs
Well, I figure it will be another 30-45 days before I have the hardware in
place. Once this is done, I would be ready to begin. I figure I can output
about 2.5 minutes of rendering per day (15 minutes per week). This is on the
safe side; as I am basing this figure on 80 computers. (not the 100+ that are
to be installed)
I would need to know what type of media you need this output to, and what kind
(if any) plug ins' you require to get the job done. (lenzFX, bones pro etc)
Bob-
Bob:
You never did say how much memory you were going to have on your systems…
Hi Greg, well its going to vary from about 10 with 64 meg (back office) to the
rest of them at 16 meg. I figure that with the average render time of 15 min.
per frame, I could finish 4,400+ frames per day (about 2.5 min worth of video).
I know that this is not a super setup, but hey! I did not have to buy the
computers either (grin) What do you think of this?
Bob-
Bob:
>> its going to vary from about 10 with 64 meg (back office) to the rest
>> of them at 16 meg.
We have found that 16 megs isn't nearly enough for most of our projects, you'll
be beating up your hard disks pretty bad. I think you'll find that most people
will only send you their biggest, hairiest projects (if they were small they'd
do them themselves!) Could you average out the memory to give the machines
all 32 megs while rendering?
Greg Pyros
Yes, I can add an additional 16 meg. (32 total). I'll have to see how much
this is going to raise the total cost of the project! If its not too much I
should have no problem adding it in.
Bob-
Bob:
>> 32 megs
I think it will help you out, and get you more clients! What are you doing for
tape output once you have the frames rendered?
Greg Pyros
Greg:
>> What are you doing for tape output once you have the frames rendered?
The client would have an option for either 2gig DAT, 1.8gig Optical disk, or
for video tape, I can rent pro equiptment for a local video house; and using my
(soon to be had) DPS Perception card output to tape. What is the preferred
method for a client?
Bob-
Bob:
>> What is the preferred method for a client?
Depends on the client! Some will be happy with a flic, others want D-1 or
they'll go somewhere else!
>> option for either 2gig DAT, 1.8gig Optical disk…
You might want to include Exabyte 8mm also, that is what all the big post
houses use for Abekas or Accom output. (Does that make me a big post house?)
Greg Pyros
Bob,
<< the rest of them at 16 meg >>
I really don't think 16mb of ram is enough. If your running r4 I would say
32mb would be a minimum. And if you plan on upgrading to r5 for NT the first
16mb would be used just to load Windows NT.
Mark
I will have to see what features the new 3ds "r5" has and if it will be worth
it. If it is, will you still be able to use the slave feature with 100+ pc's
with one copy of 3ds? I already have 8 pc's running NT (64 meg); I could test
it out on them when r5 comes out and see if it will do what I need.
Bob-
Robert,
I doubt you will be able to use 3DS r4 slave machines with 3DS r5 – We
encountered a nightmare rendering a 7000 frame job when our lead animator
decided to upgrade to r4 from r3 during the middle of this job. It turned out
that r4 did some different things with the set files and we experienced all
sorts of goodies like field reversal, dropped fields, etc. Message is: Get
ready for some real fun…
What are you going to do when it's 7:00 in the morning and you've contracted a
job to be finished and shipped out by 12:00 and your employees start coming in
to work on their computers? Not a pretty scenario.
PS: 64MB per machine is the minimum. Most jobs we have contracted for run
between 50MB and 80MB frame size. Render time per frame is between 10 sec. and
2 minutes depending upon complexity.
Anyway, Good Luck…
Matt Armstrong
TRG / DFW Multimedia
Robert: I've been out of the loop on this conversation for two weeks and want
to jump in with some comments. I agree with the guys who are arguing for 64K
plus. For complex projects the time swapping out to disk is going to kill you.
If I were to use your facilities I would want to be able to bring you project
files on an Exabyte and receive them back on the same medium. I also wonder
what kind of personel you plan to have on site while the render is going. Feel
free to call me at 213-957-6868 to discuss your plans further.