unclefester
06-10-2006, 01:45 PM
Good morning Ladies and Gents,
First off - i'm NOT a max8 user so when you this post please bear that in mind. I'm an IT guy for a firm that uses 3dsmax8 amongst other things for their visual work and i'm getting more interested in network rendering and how it works or in this case doesn't work.
I've been looking at the backburner software which seems to work ok but it has a major shortcoming that i'm wondering if there are work arounds for.
1. When you complete a render scene, the client computer closes max8 down and then reopens for the new scene, is there a way to force the client machine to keep max8 open between rendered scenes? Almost 40% of the render time is taken up with closing and opening max8! If it's 1000 renders per job that's 999 unescessary opens and 999 unescessary closes just for that one job.
2. My next question is with specific regard to render performance. I've got two Dell quad xeon dual core (woodcrest core) servers that i've been testing out with rendering and they seem to be working about 30% faster than the P4 dual cores we have ( not dual core duo ). From this and my admittedly limited understanding, i'm torn between running 4 of these as render machines rather than say 8 p4 dual core 2 duo machines. Is there a 'one way' rule for the best way to acheve render performance? I know this will depend on the complexity of the render scene but my interest is in the best way to build a render farm for coping with complex scenes. I'm probably asking the impossible but i read that the new dual core 2 P4 chips are about on a par with the xeon woodcrest chips, the only difference is that you can get a server with 4 dual core xeon chips and you can only get single socket P4 dual core 2 machines. I'm trying to get my head around whether it's best to have one machine with a lot of high spec processors or lots of machines with high spec single processors.
Cheers for any help you can give!
Patrick
First off - i'm NOT a max8 user so when you this post please bear that in mind. I'm an IT guy for a firm that uses 3dsmax8 amongst other things for their visual work and i'm getting more interested in network rendering and how it works or in this case doesn't work.
I've been looking at the backburner software which seems to work ok but it has a major shortcoming that i'm wondering if there are work arounds for.
1. When you complete a render scene, the client computer closes max8 down and then reopens for the new scene, is there a way to force the client machine to keep max8 open between rendered scenes? Almost 40% of the render time is taken up with closing and opening max8! If it's 1000 renders per job that's 999 unescessary opens and 999 unescessary closes just for that one job.
2. My next question is with specific regard to render performance. I've got two Dell quad xeon dual core (woodcrest core) servers that i've been testing out with rendering and they seem to be working about 30% faster than the P4 dual cores we have ( not dual core duo ). From this and my admittedly limited understanding, i'm torn between running 4 of these as render machines rather than say 8 p4 dual core 2 duo machines. Is there a 'one way' rule for the best way to acheve render performance? I know this will depend on the complexity of the render scene but my interest is in the best way to build a render farm for coping with complex scenes. I'm probably asking the impossible but i read that the new dual core 2 P4 chips are about on a par with the xeon woodcrest chips, the only difference is that you can get a server with 4 dual core xeon chips and you can only get single socket P4 dual core 2 machines. I'm trying to get my head around whether it's best to have one machine with a lot of high spec processors or lots of machines with high spec single processors.
Cheers for any help you can give!
Patrick