vray dr

[Tim Saunders]

if a few of you wouldn't mind responding. so far i am only farmed out to one additional pc. my i.t. guy is real nervous about setting up very many systems for dr for fear of the network being bogged down. anyone have any problems with rendering over 5-10 systems and the network slowing down. we have an office of 60+ computers that are always accessing the network. another concern of mine, does anyone farm the dr to computers that are used for drafting and stuff while dr is in proccess, or does dr pretty much tie up that station?

[DavidR]

The drafters won't be happy if you're doing dr while they work -try doing a 96x96 region render -Vray still uses 100% CPU until it's done, then the next bucket will start on the machine, so it's not a good idea. How about doing it at night? We rarely need dr for A3/A4 300dpi; are you doing monstrously huge renders? We find ours render in 12 hours on dual Xeon 2.8 and in 3.5-6 hours on dual Opteron 275. A bunch of older machines (you're limited to 10) mifght do the same in dr or backburner strips -I love Backburner.

[C.Vestal]

David is correct. The processor at each workstation will be at 100%.

 

Your network itself will see a very small load unless you are using a lot of maps. Then when you will see an increase of network traffic when all the computers access (or receive maps if you choose include maps) the files. After that is just the transfer of the individual frames when each workstation completes. A uncompressed 720x480 tiff with alpha is approximately 1.2MB.

[DavidR]

David is correct. The processor at each workstation will be at 100%.

Your network itself will see a very small load unless you are using a lot of maps. Then when you will see an increase of network traffic when all the computers access (or receive maps if you choose include maps) the files. After that is just the transfer of the individual frames when each workstation completes. A uncompressed 720x480 tiff with alpha is approximately 1.2MB.

Right, but you're talking about 1 frame per node, not distributed rendering(dr). He means splitting 1 frame among multiple machines; in that case, the machines need to communicate constantly to update who's doing what until the image is finished. You can increase bucket size in Vray to maybe compensate. I did this to see if we could force Vray to use more memory (it works) but I didn't check to see if the network traffic was less -intuition says yes, but I'm not 100% sure. We increased bucket size from 64/64 to 128, 256, 512, and 1000 (square buckets) pixels. Memory useage went up with size, as expected, and 1000x1000 (Vray max) spun up to 4.01GB (nodes have 4GB)and crashed. 256 was our sweet spot but yours will probably differ. I'd say you'll need to render at night, but maybe you have a couple of people at meetings/offsite, so you'll have extra machines during the day. If you figure out who's machine is who's, you can manage it without making enemies. If it's unworkable, try proposing getting 1 or more dedicated machines. They could be nodes (small, integrated graphics) in a server room or workstations that could enter drafting service after a year of rendering, eg. If you get a new machine, get dual Opteron 265-280 -they are freaking incredible -2x faster than Intel's best with Vray, draw much less power/less cooling/less noise -can't be beat.

[Tim Saunders]

thanks guys. david, your right on. and that is what i planned. aproaching a few who do periodically step out for meetings. plus we have 2 more systems in my same cubicle that are for burning and photoshop use for the rest of the office who don't have those features on their machines, so they could be utilized too. night rendering is tyical for me, however i have that once or twice a month deal when someone wants something that day, and my scene suffers to make rendering faster.