building a rendering machine

[Howard]

Originally posted by: Randybob
Quadros in SLI will support multiple monitors

You mean the SLI Multi-View? I'm not sure if this helps, but:

SLI Multiview is only enabled with Linux by using the Xinerama option in the X11 config file's ServerLayout section. Windows is too restrictive to the NVidia driver to enable true SLI Multiview. Instead Windows will affect something like SLI Multiview (extend desktop to additional displays), but the NVidia driver will take a fairly big performance hit, especially if the graphics window spans displays.

http://forums1.itrc.hp.com/service/foru...dmit=-682735245+1165096173506+28353475

 

[Ilikepiedoyou]

wow thats very interesting, What would happen if you were only rendering in one window, but had your desktop accross multiple windows behind it?

 

[Ilikepiedoyou]

has anyone else had any experience with this? Also any critisism on my computer choices above?

 

[Randybob]

Originally posted by: Howard

Originally posted by: Randybob
Quadros in SLI will support multiple monitors

You mean the SLI Multi-View? I'm not sure if this helps, but:

SLI Multiview is only enabled with Linux by using the Xinerama option in the X11 config file's ServerLayout section. Windows is too restrictive to the NVidia driver to enable true SLI Multiview. Instead Windows will affect something like SLI Multiview (extend desktop to additional displays), but the NVidia driver will take a fairly big performance hit, especially if the graphics window spans displays.

http://forums1.itrc.hp.com/service/foru...dmit=-682735245+1165096173506+28353475

That's the first I've heard of that. I'd want NVIDIA's tech support to confirm or deny that though. It seems really wierd that SLI Multiview wouldn't work as advertised in Windows, especially since NVIDIA knows that the majority of Quadro buyers would be using Windows. Oh well, if it is true, he can do it the old fashioned way and have 1 graphics card for each monitor, without enabling SLI.

 

[Ilikepiedoyou]

if you have two cars with a monitor hooked to each, can you still span one dektop across them?

 

[Randybob]

Ilikepiedoyou:

You could get what you have, or you could get a Core 2 Duo and this motherboard will offset the cost. The only real difference is some of its expandability, and those differences are minor. I don't foresee a need for a PCI-Express x4 slot, unless you want to spend money on a RAID array or possibly a PURE card (since those cost as much as a new system, and they're not out for PCI-Express yet anyway). Either way, you may run into issues with upgrading to a quad core chip on a budget motherboard. Just because a chip fits in the socket doesn't mean that the mobo will meet its power requirements.

As for that free PSU, you might be getting what you're paying for. Don't be surprised if you have to plunk down the cash for a good one if this dies on you. Since it is free after rebate, you don't have much to lose. Just make sure it will support PCI-Express and SLI.

For your hard drive, get a good SATA hard drive from Seagate or Western Digital with a large cache (8-16 mb). You'd also be surprised how much space you'll need if you want to render out to frames all the time. 100+ gigs is decent, but for the cost, 200-300 gigs will give you lots of headroom.

Google for some reviews for good cases. After all, you're going to have your hands in there putting this thing together. You might as well get one that is decent to work in. You don't want cramped space, sharp edges, or the covers for your spare slots to be cheap punch-outs (they can damage your system when you are upgrading if you are unlucky). You also want good cooling since this is supposed to be a stable workstation. I wouldn't get one that is super expensive unless it was BTX-upgradable. But you were looking to save money anyway.

 

[Randybob]

Originally posted by: Ilikepiedoyou
if you have two cars with a monitor hooked to each, can you still span one dektop across them?

Yep, we did that with the old rig I worked with at my internship. That hardware was pre-SLI.

 

[Howard]

Originally posted by: Randybob
Oh well, if it is true, he can do it the old fashioned way and have 1 graphics card for each monitor, without enabling SLI.

In which case any old PCIe/PCI card (assuming it can support the desired resolution/refresh rate) would do fine, right?

 

[Randybob]

Originally posted by: Howard

Originally posted by: Randybob
Oh well, if it is true, he can do it the old fashioned way and have 1 graphics card for each monitor, without enabling SLI.

In which case any old PCIe/PCI card (assuming it can support the desired resolution/refresh rate) would do fine, right?

Yes, but with today's 3D apps, I wouldn't trust a regular PCI card. We got away with it at my internship, but we were using Lightwave 7.0 which isn't terribly picky about which graphics card you use. These days, you want to have at least 128 mb of graphics memory available per display, and that number is only going to go up while the number of plain-jane PCI cards continues to shrink.

 

[Ilikepiedoyou]

hmm, thanks for showing me this randybob, the only difference I see between the two boards (besides the nearly 1/3 price difference) is that the 570 can run both card but only at 8x and not 16x, and opposed to running just one at 16x. What I don't know is whether other items in the line will provide barries before the 8x barrier is hit. the tests which I have been able to uncover show that the e6400 performs much better in rendering applicatoins as oppsed to a similar level athlon, they don't mention much about the opteron, but the from what I understand the chips are very similar just different sockets.

 

[Ilikepiedoyou]

Howard - I have been re-reviewing what you have been posting about people getting sli to function, this has me a bit worried.

 

[Randybob]

Originally posted by: Ilikepiedoyou
hmm, thanks for showing me this randybob, the only difference I see between the two boards (besides the nearly 1/3 price difference) is that the 570 can run both card but only at 8x and not 16x, and opposed to running just one at 16x. What I don't know is whether other items in the line will provide barries before the 8x barrier is hit. the tests which I have been able to uncover show that the e6400 performs much better in rendering applicatoins as oppsed to a similar level athlon, they don't mention much about the opteron, but the from what I understand the chips are very similar just different sockets.

There isn't really any difference to speak of between the Opteron and the Athlon 64. The Opteron has twice the L2 cache, but but that's it. That extra 512 K is not going to catch it up with a Core 2 Duo.

Unless I'm mistaken, the 8X barrier is only going to affect you if you play games. Besides, if you get a second card and don't want to do SLI, you can get a better one than what you have and the FX1400 can be your secondary display card so all you'll see is improvement.

 

[Ilikepiedoyou]

in 3ds max or other open gl 3d programs, is most of the rendering done by the cpu or gpu? I guess I should have asked that a long time ago. Also I would think (but what do I know, I'm a noob) that the 16x and 8x issues affect every use for the card, if you are moving that enough data on the lanes to reach that speed limitation ,regardless of whether you are playing a game or rendering, but I really don't know.

 

[Randybob]

Like I said, rendering does not use the graphics card. The graphics card will affect modeling, surfacing, and setting up your scene. OpenGL stuff. OpenGL is just a serviceable, interactive preview of your work because rendering in real time would require a supercomputer. In fact, professional render nodes only have the absolute minimum integrated graphics and that is just so they can run Windows. Their power comes from the multiple Xeons/Opterons and RAM. The finished work is uploaded to them for final rendering, so there is no point in having an OpenGL card in there.

 

[themisfit610]

performance differences with x16 vs x8 are nonexistent. Honestly, PCIe is such overkill it's not even funny Maybe with a new GeForce 8800gtx you might have some difference, but I really really doubt it.

When you are setting things up in your scene, having a fast video card is nice, because you can get a rough preview - much better than a simple CPU-driven wireframe.

When you actually go to render the scene / animation, and take a coffee break to wait for the system to chug out the frame(s), you are not using your video card at all. True rendering (raytracing) is CPU only. This may change in the future, but from now, it's all CPU.

 

[Ilikepiedoyou]

does any member in the nvidia 500 series support intel quad core? And if it does, would it have what is nessesary to run it effectively

 

[Ilikepiedoyou]

bump

 

[Randybob]

Originally posted by: Ilikepiedoyou
does any member in the nvidia 500 series support intel quad core? And if it does, would it have what is nessesary to run it effectively

Fraid not.

 

[Ilikepiedoyou]

If processing power is so important, a have a few old computers lying around, all amds ranging from 700mhz to about 1.3 I beleive. Is there any way that I can add there processing power effectively to that of my new computer to speed up rendering time?

 

[Randybob]

Originally posted by: Ilikepiedoyou
If processing power is so important, a have a few old computers lying around, all amds ranging from 700mhz to about 1.3 I beleive. Is there any way that I can add there processing power effectively to that of my new computer to speed up rendering time?

Network rendering is possible, but it's a pain in the butt to set up. I don't know how to set it up. I'm pretty sure your machines need to be good enough to run Max or whatever you plan to use. That 700 Mhz machine probably won't cut it.

 

[Ilikepiedoyou]

The 700mhz runs max 7, just don't make things to complicated. Where would be a good place to find out more about setting something like that up?

 

[Ilikepiedoyou]

bump

 

[Ilikepiedoyou]

bump

 

[gsellis]

Originally posted by: themisfit610
For the applications I use, performance differences with x16 vs x8 are nonexistent. Honestly, PCIe is such overkill it's not even funny Maybe with a new GeForce 8800gtx you might have some difference, but I really really doubt it.

Fixed this for you.

My NLE is restricted on AGP and I will be experiencing it at the end of the month doing 1080i edits. Vista's virtualization of graphics will better use the extra bandwidth too.

Ilikepiedoyou - What are the outstanding questions that require the bump? Since you now have your hardware, the only performance extras would be buying extra network licenses and building some extra sets with a KVM switch and a single head (a rack of machines, 1 keyboard, 1 mouse, 1 monitor) to split rendering. You have big, full pockets? This is overkill unless you have a business and business for it.