PC and Mac Mixed Render Nodes

[xfon5168]

I'm currently daring myself to go for a new MacPro and would like to use my pair of dual core AMD64 PCs for rendering. Now I would need to transfer my license to Mac (lot of other vendors don't make a charge for this :hmm: ) but I guess I'm technically supposed to give up my PC license at the same time. So how is it possible to NetRender with mixed PC and Mac? Well I guess I could cheat and just install the server/clients and I understand it does work but I would be stuck when R10 comes out as I could then only upgrade my Mac license but couldn't upgrade my PC NetRender clients.


Acctually, I think it won't matter. If you uninstall it from your PC, and just install only the Clients I think you could still do it because I dont think NET Render is tied to being a PC or a MAC. So you could get the platform transfer but still have net installed on your PC's...I'd ask MAXON though, but I think that's the case.

[Srek]

The net render clients are not tied to a specific platform. You can install them on OS X or Windows as you like it, the installation disks provide both installers.
Only those parts of CINEMA that require a serial number (CINEMA 4D itself and Net Server) are licensed platform specific. The Net server can adress clients on OS X or Windows likewise.

Cheers
Björn

[philhoole]

Thanks Srek, I hadn't appreciated that both installers were on the CD and the licensing was different. Makes my decision a little easier to go for a MacPro now but it's still a lot of money for me.

[leigh]

Just a possible caution though... I am not 100% sure if this is an issue with C4D rendering, but with other renderers I've used (eg Mental Ray, Maya Software, LightWave), you can get slightly different results when rendering on different CPU types (Intel and AMD), and even though these differences are in things you may think are subtle, like motion blur, procedural textures, etc, they can actually be really noticeable when watched as an entire clip.

[Srek]

Yes, Intel and AMD can vary in floating point results in theory, but i failed to see any differences in rendering in CINEMA 4D. The differences between a PowerPC and a x86 CPU are very noticable though.
Differences will be most noticable if they come from accumulated calculations, like GI or unbacked Dynamics.

Cheers
Björn

[imashination]

From my experiences, you will have problems with mixed G5 and P4 networks, especially with the brightness of the final rendered image. You could always do what we did and segregate them into a p4 and g5 net render, and send of different projects on each.

[zoetropeuk]

Does anyone know what the latest version of NetRender is and is there a universial version available for Intel Macs ?

Matt

[Jorge Arango]

Does anyone know what the latest version of NetRender is and is there a universial version available for Intel Macs ?

Matt

9.603.

Yes, there are PPC and UB versions for server and client.


Jorge Arango

[Simtub]

From my experiences, you will have problems with mixed G5 and P4 networks, especially with the brightness of the final rendered image. You could always do what we did and segregate them into a p4 and g5 net render, and send of different projects on each.

I'm wondering if there would be problems if using intel based imacs and mac pro's mixed with P4's and PC's in general??

[dann_stubbs]

I'm wondering if there would be problems if using intel based imacs and mac pro's mixed with P4's and PC's in general??

the OS won't make the difference - the processors do

so there could be differences between AMD and INTEL - but may be less noticible or less of an issue then mixing either of those with a G5/G4/G3 etc chip

for 100% crucial work it really is safest to stick to one chip type - or as mash says separate into groups of processor type

dann

[FredSpeaks]

Actually, the OS does make a difference. I just completed a project that was rendering on mostly AMD machines (there was also a centrino in there). I added in a Core duo Mac Mini and almost got hosed. The frames rendered on the Mac were darker than those rendered on WinXP. So I had to turn of the mini and render just with the WinXP boxes.

I also let my G5 render out a few frames and it matched the brightness of the Mac Mini.

[dann_stubbs]

Actually, the OS does make a difference. I just completed a project that was rendering on mostly AMD machines (there was also a centrino in there). I added in a Core duo Mac Mini and almost got hosed. The frames rendered on the Mac were darker than those rendered on WinXP. So I had to turn of the mini and render just with the WinXP boxes.

I also let my G5 render out a few frames and it matched the brightness of the Mac Mini.

i'm not sure...

the main differences between mac and pc OS are the operating systems themselves use different gamma.

real rendering differences in chips (rendering) are from math errors or how the chips process - so it would be unusual for the same chip to process math differently due to the OS. (big endian and little endian rounding primarily as one)

what you mention - rendering an intel core duo in a mostly amd farm is what i would be suspect of causing differences. but - i have seen a difference in brightness in one version of C4D between platforms - i think that was in 9.5 - but the user who had the issue did eventually find the setting in the file that was causing it (it was rendering darker on one then the other) but that was probably a year ago and they never mentioned "what" the issue actually was...

srek has also agreed (which is from what i know a very common agreement on this issue)

see post #14 by srek on

http://forums.cgsociety.org/showthread.php?t=351180

dann

[FredSpeaks]

Dann,

I know that it has been said that the OS should make no difference, but I am just relating my most recent headache. From my cursory investigation it seems that it was the Mac OS that was the culprit. But I guess it needs further investigating. I hope to get to the bottom of this soon('cause I would like to add a couple more mac minis to render with).

On the job that gave me problems, I had no powerpc chips rendering. There were 3 intel (one being the Core Duo) and 5 AMDs (semprons). There were 2 chunks of darker images and then a handfull of darker images scattered (where net starts breaking up the remaining frames to more machines) It was a pain to go frame by frame and compare the darker frames and then send out multiple jobs to fix the frames.

I updated everything to 9.6 when I added the intel mini. One machine is refusing to run with the 9.6 update so I think I am going to take everything back to 9.5. Ernest has mentioned a problem with the brightness of multipass images and the 9.6 update. I have my suspicions that this may be part of the problem. Hopefully this week I will take my machines back to 9.5 and see if the problem still exists.

Previously (back at 8.5 or 9) I was able to render to AMDs, Intels, PowerPC G4s and G5s without any headaches. I am just doing "simple" (used loosely) architectural animations. Just an animated camera, a few moving cars and some moving planes with quicktime movies of people on them. I avoid using procedurals as well to avoid calculating differences in textures. It seems that with the most recent updates, multi-chip environments nolonger work.

[Simtub]

I avoid using procedurals as well to avoid calculating differences in textures. It seems that with the most recent updates, multi-chip environments nolonger work.

That is the issue I am wondering about because right now at the place I work, having upgraded a lot of the MAC machines to be intel based cpu's, We were wondering if it was good to just bring in a load of intel based PC's which would mean both windows and mac systems would run from the same processor type. Could this possibly work if it was just the cpus that made differences to rendering and not the OS itself?

[Troyan]

Interesting. We just finished using NetRender 9.6 across a myriad of machines (G4, G5, Intel Macs, Amd PC and Intel PC's, all of them dual processors (half of the machines were in different locations over the web)) and perhaps we're just lucky, but there were no problems. We used Mograph, glows, procedural textures and even imported some liquid simulation data from Realflow and everything was smooth as silk. Well, we did have one problem where there were 2 machines without the Mograph plugin installed and those frames had to be re-rendered. Which brings me to a digression that Net Render REALLY needs to alert the server or the browser interface that there is a plugin issue. The only notification was a one line alert in the client windows and the frames continued to render.

[dann_stubbs]

Interesting. We just finished using NetRender 9.6 across a myriad of machines (G4, G5, Intel Macs, Amd PC and Intel PC's, all of them dual processors (half of the machines were in different locations over the web)) and perhaps we're just lucky, but there were no problems. We used Mograph, glows, procedural textures and even imported some liquid simulation data from Realflow and everything was smooth as silk. Well, we did have one problem where there were 2 machines without the Mograph plugin installed and those frames had to be re-rendered. Which brings me to a digression that Net Render REALLY needs to alert the server or the browser interface that there is a plugin issue. The only notification was a one line alert in the client windows and the frames continued to render.

my experience is the same for the most part - C4D doesn't seem to have too many issues with cross platform - some settings can cause issues like TP with a random function etc... but not too many. procedurals can show some small variation too...

i'm mostly interested in your use of realflow - maybe it is with a PC server that it works ok - because with an OSX server and PC clients it just won't work at all due to the path naming issues - i would think that the mac clients wouldn't work on a PC server either due to the same issue - so unless you didn't mix OS platforms for those particular RF renders, would you care to elaborate?

dann

[Troyan]

my experience is the same for the most part - C4D doesn't seem to have too many issues with cross platform - some settings can cause issues like TP with a random function etc... but not too many. procedurals can show some small variation too...

i'm mostly interested in your use of realflow - maybe it is with a PC server that it works ok - because with an OSX server and PC clients it just won't work at all due to the path naming issues - i would think that the mac clients wouldn't work on a PC server either due to the same issue - so unless you didn't mix OS platforms for those particular RF renders, would you care to elaborate?

dann

I'll have to get my cohort to elaborate on Realflow. What we did was imported a mesh from Cinema into Realflow, created a simulation on one seat (it was just of blood oozing from a vessel into an incision), then imported a series of .obj files back into Cinema. He is Le Coder so he did some Xpresso trickery with the Obj. files in Cinema and the results were spectacular. I'm a little dismayed that a seat of Realflow only works with 2 processors, especially with my Quad 3 Xenon system on the way.

[Srek]

On the job that gave me problems, I had no powerpc chips rendering. There were 3 intel (one being the Core Duo) and 5 AMDs (semprons). There were 2 chunks of darker images and then a handfull of darker images scattered (where net starts breaking up the remaining frames to more machines) It was a pain to go frame by frame and compare the darker frames and then send out multiple jobs to fix the frames
I have no real idea why this happened. The rendering algorithms etc. are absolutely identical on Mac and PC. The only thing i can think of that might make a difference are certain codecs that might differ between systems, but then even those can only make a difference for single materials in wich textures with that specific codec are used.
If you are able to reproduce this i would be thankfull if you can send me the scene with a bit of background information on the clients to b_marl (at) maxon (dot) net.
Cheers
Björn

[FredSpeaks]

I reinstalled Cinema to 9.5.2 today on all the clients. All the clients have Quicktime 7.1 as well. Here are the results. The G5 was rendered on Cinema, not client...http://atechstudio.com/clients/cgtalk/Pentium4.jpghttp://atechstudio.com/clients/cgtalk/Sempron.jpghttp://atechstudio.com/clients/cgtalk/CoreDuo.jpghttp://atechstudio.com/clients/cgtalk/G5.jpg

[dann_stubbs]

I'll have to get my cohort to elaborate on Realflow. What we did was imported a mesh from Cinema into Realflow, created a simulation on one seat (it was just of blood oozing from a vessel into an incision), then imported a series of .obj files back into Cinema. He is Le Coder so he did some Xpresso trickery with the Obj. files in Cinema and the results were spectacular. I'm a little dismayed that a seat of Realflow only works with 2 processors, especially with my Quad 3 Xenon system on the way.

ahh.. well if you imported them as obj into cinema that explains it - it is using the RF loader plugin to load the .bin files that won't work.

dann

[unseenthings]

ahh.. well if you imported them as obj into cinema that explains it - it is using the RF loader plugin to load the .bin files that won't work.

dann

That's what we did... we imported the .obj sequence and used xpresso (that I found here on cgtalk and another very similar one, I think from frenchcinema4d.com) that did "sequential visibility" -- making one child visible per frame, in order. All of the obj's under that xpresso null worked just wonderfully. We were originally trying to use blender fluids, but it was just a little too difficult to tweak the fluid parameters with any kind of consistency... and since that (obj sequence) was the method I was using for the blender fluids we'd imported, I just used it for the realflow ones as well. I think the files ended up being about 9mb even with about 150 frames of blood animation in there, so it wasn't too bad, even across the network.

[FredSpeaks]

Srek,

I know you have been busy with R10 coming out. Just wondering if you had a chance to look at the file. Any knews on why the slight difference between Mac and PC?

Thanks.

[Carl007]

Sorry to bring up this old thread.

But, I have a mixed OSX PPC and intel render-farm with the latest build of C4D R10.111.

My problem is with a sequence where I have a baked particle stream where the rotation values of the particles seems to be inverted or something related to the different chip.
This problem have occurred in several different sequences with baked and unbaked particle streams.
Even baked streams with 4 samples per frame shows this error...

Do anyone have any workaround or other tip, I would be grateful.

/Carl

[CGTalk Moderation]

This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.