Blender and My Hardware...

I have been using an HP Pavillion with a Pentium D 2.8Ghz Dual Core, 3Gigs of RAM, and an nVidia 8600GT Graphics Card w/ 256MB of GDDR3 Mem.

With this computer, Blender ran just fine. High Res/High Quality Renders took a while, but it worked great.

I just purchase a new system, as well as integrated some hardware that I had lying around. And to understand the problem that I am having, you will probable need to understand how my system is set up. My set up is a bit complicated so I will try to explain it in the most basic way that I can.

List of hardware…how it is arranged will be explained after this list.

Motherboard: nVidia nForce 680i SLI Motherboard
Processor: Intel Core Duo Quad
Ram: 4GB
Harddrive: 300GB OS: Windows Vista
Graphics: nVidia 8800 GTX (x2 - set up in SLI configuration)
768MB GDDR3 Mem (x2)
Over clocked to 580Mhz on the core and 1396Mhz on the shader clock

In addition to my main computer, I have 4 other motherboards set up in a processing farm. (I will explain this in a moment)

Motherboard #'s 1 and 2:
MSI Industrial 945GM1 Mini-ITX
Processors for #'s 1 and 2: Core 2 Duo Mobile

Motherboard #'s 3 and 4:
Intel D201GLY Mini-ITX
Processors for #'s 3 and 4: 1.33 Ghz Celeron 2xx processor

Ram for each: 512MB
Main harddrive: NAS Networked 80GB - OS: Linux Fedora Core 5

Essentially my system is set up as a farm. I have my main computer with the specs listed above, then using que management software I route half of the processing load to the ‘farm’ which is ran by one harddrive w/ one operating system. Once the ‘farm’ gets it’s load, it divides the chunk that the main computer sent over into 4 equal data blocks for each of the 4 motherboards (or separate computers) to process. After all the processors are done with their data block, it gets sent back as a completed file.

The easiest way to describe it is that I have ten processors (5 not including separate cores, by this I mean that my que management only divides the data blocks up five ways. I could do a data block for each core, but that’s a lot more work.:rolleyes:) working a processing load simultaneously. Only the load is first divided into two ‘blocks’, then the second OS splits it again into 4 ‘blocks’. The reason for this is that I have a quad core (4 processors) on the main computer, and 4 separate motherboards in the farm. Two of my motherboards in the farm have core 2 duo’s, but for convenience sake, I am only running one OS for all four, so I am not able to divide it for dual core’s. Though, the core 2 duo’s and the quad have hyperthreading which will slice their work load in two/four anyways.

Now after all of that, I have a couple questions about running Blender 2.44 w/ YafRay.

First, I have ample amounts of graphics power. Not enough to make another Shrek or anything, but plenty just the same. I am not seeing any graphics enhancement within Blender. No anti-aliasing, no shader help, etc. Are there settings in Blender that I need to change to get the software to utilize the graphics card set up, or does Blender have a user interface graphics limit?

Second, my rendering times have decreased dramatically, but I am wondering if I am getting the most out of my system. In YafRay, there is a setting for Number of Processors. Does this number act kind of like my que management and break up the work load into that number of data blocks? And if so, should I even mess with it? My quad core uses hyperthreading, so it knows how to split and reconstruct the data very quickly. And, my que management sends half of that load over to the ‘farm’. And, I highly doubt that Blender will recognize that I there are 4 more processors on the other end of my Gb Ethernet. So, (I may be wrong) I don’t think I should bother Blender with those, just wondering if it should be set to four

Or, if I set it to 2/5/8/10 (numbers depending on how you define “separate” processors in my system…again, a little complicated), will my processor and que management recognize that it has already been divided up? …or, does this feature do something else entirely?

All help will be greatly appreciated! I know this is more of an advanced question, but I thought I would ask anyway. (I may have to head over to my SLIZone forum and ask those guys…they are freakishly good at what they do!:eek:)

Anyway, thanks! :yes:

Thats quite a computer you’ve got there! I would recommend playing with the setting and see what happens to your render times. The “processors” setting simply tells yafray how many threads to launch. A thread is a process that is normally used by 1 processor, so increasing it lets more processors in on the action. On any normal multi-core computer, increasing the “processors” setting to your number of cores drastically cuts rendering time. However, your software might be smart enough to do this automagically, so I think you’ll just have to do some test renders and see.

Please let us know how this turns out.

Thanks…I’ve already started in on the endless task of trial and error…so hopefully I’ll find my sweet spot soon. And more importantly, why it is the sweet spot. (so I know how to make adjustments when I add more motherboards.)

Thanks again!

Generally, launching too many threads doesn’t hurt you at all. I would pick a small scene to test render, something that would take maybe a couple of minutes so that it doesn’t take you days to figure it out. Good luck!