Rendering configuration

[Boskovic]

Can anyone point some pros and cons regarding procesors and GPUs and other related stuff in order to speed up rendering process? I now about GeForce QUADRO speding up 3dMax and Maya, but since I use just Cool 3D I was wondering what would shorten rendering time in Ulead's program?

[htchien]

CPU speed: the higher the better.

GPU: ATi Radeon series or nVidia Geforce series are OK to render. If you use display card that supports PCI-E 16X would be the best. But if you have some budget issue, you still can use some AGP display card to render.

Hope this helps.

H.T.

[Boskovic]

I did some searching around the net and as I can understand so far, rendering is being done in CPU (some recent rendering plugins for max/maya utilize GPU) and GPU is being used for previewing.
Also people says that it's necessary to spend some time optimizing 3d model (less poligons, less render), and there's also driver issue (Direct 3D, OpenGL or specialized drivers like Maxtremme for GeForceQUADRO/3dMax cobmbination). Therefore I would like to know what kind of rendering engine Cool3D utilize, as well as what kind of drivers, and also what Cool3D's rendering process most benefit from (CPU, RAM, GPU or else..)

[OinkyPig]

Actually I am also thinking of this topic.

Just a quick question then.
I do notice even using the file output as DV MPEG2, it does take an extra-ordinary to do a frame by frame when compared to a video editor program rendering.

Maybe it is that the video files already have the frame data while COOL3D needs to generate frame by frame data. (Just a guess)
Could anyone advise why would COOL3D take a longer time to render?

Another item which caught my eye is that would professional AGP QUADRO or 3DLabs WildCat card do a much better job.

I have seen Canopus have their own "dedicated MPEG2 hardware encoder" card which takes the load from the CPU in rendering. Or SD Connect. Could they aid in COOL3D rendering?

Just a thought as I thought of trading in my FX5600 card for a "Professional GPU card"

[htchien]

Both COOL 3D 3.x (C3D) and COOL 3D Production Studio (C3DPS) are based on Microsoft Direct3D technology. The difference between them is, C3D supports DirectX 6 or later, C3DPS supports DirectX 8 or later. So any display card that supports DirectX 8 should be able to run both programs properly.

But to be able to run is one thing, to run fast is another.

rendering is being done in CPU (some recent rendering plugins for max/maya utilize GPU) and GPU is being used for previewing.

If the program only supports DirectX 7 or earlier, then this might be true. But in DirectX 8 or later, there are some changes. In DirectX 8, Microsoft has introduced techniques named Vertex Shader and Pixel Shader. Any display cards (and programs) that support these techniques in hardware can make the rendering mostly to be done by GPU. That's why some DirectX 8/9 display cards can render 3D graphics at the same speed no matter what CPU is used (if the CPU is fast enough).

There are several issues when you use C3D/C3DPS to render:

1. The bus interface of the display card (especially when you apply many global effects or set the rendering quality at a high level in the C3D project). To use PCI-E 16X is suggested. If you have some budget issue, you can use AGP 8X display card and mainboard. This is because when rendering with global effects or set the rendering quality to a high level, C3D/C3DPS will move the rendered frame from the video memory to system memory and then apply the global effects (or apply software anti-aliasing). In this case the bus transferring rate is the key (and the most important key).

2. CPU: the faster the better. There are some object transformations and memory operations done by CPU. So if you can have faster CPU it would be better.

3. RAM: as I mentioned above, C3D/C3DPS will move the rendered frame from video memory to the system memory, so if the system RAM can be faster it would be better.

4. When generating video file from C3D/C3DPS, not to render in Field Order A/B mode. This is because C3D/C3DPS will double the number of the rendering frames to get the field order effect. If you wish to make a field order clip, you can export the C3D project as a Frame-Based clip and import the clip into your video editor to make it.

5. Try to export your video clip to uncompressed video if you really care about performance. If you export your video with some compression CODEC, it will take some time to render the output clip with the CODEC.

6. Hard disk: the faster hard disk the better.

7. Not to set your output/display/curve quality too high. In my opinion, to set the quality to be Fair or Better is enough for generating a video clip for DVD, unless you wish to create a video clip for a film which will be showed in the theater or for a High-Definition video.

Professional display cards such as nVidia QUADRO, ATi FireGL or 3D WildCat are focused on rendering accuracy and have best use on the OpenGL technology, but they might not render as fast as nVidia Geforce 4/5/6 or ATi Radeon R/X series on Direct3D technology (as I know, some of them might be worse on Direct3D support).

Hope this helps.

H.T.

[Boskovic]

THNX! htchien

[OinkyPig]

There are several issues when you use C3D/C3DPS to render:

1. The bus interface of the display card (especially when you apply many global effects or set the rendering quality at a high level in the C3D project). To use PCI-E 16X is suggested. If you have some budget issue, you can use AGP 8X display card and mainboard. This is because when rendering with global effects or set the rendering quality to a high level, C3D/C3DPS will move the rendered frame from the video memory to system memory and then apply the global effects (or apply software anti-aliasing). In this case the bus transferring rate is the key (and the most important key).

2. CPU: the faster the better. There are some object transformations and memory operations done by CPU. So if you can have faster CPU it would be better.

3. RAM: as I mentioned above, C3D/C3DPS will move the rendered frame from video memory to the system memory, so if the system RAM can be faster it would be better.

H.T.

Hi HT, if I understand you correctly then. Just as an example.
I am running a P4 2.6C HT 512K FSB 800 Mhz, L2 Cache with 512 MB DDR Ram along a Nvidia FX 5600 256 RAM AGP 8X (128 bit.. If I read my card specs correctly).

To get COOL3D PS to run faster, probably getting an upgrade in
a) RAM (boost to > 1 GB DDR)
b) Change to AGP 8X but 256 bit bus (e.g. 9800XT or 6800)
c) CPU boost say to 3.2/3.4 GHz (highest speed for a P4 478 Pin socket)

In this order would be best? or GPU 1st, RAM, CPU?

[htchien]

Again, the faster the better.

I prefer GPU -> RAM -> CPU as you mentioned.

But if your C3D project is too complex, or you have applied some global effects in the C3D project, you will meet the bus transferring rate limit (AGP 8X: 2.1GB/s, PCI-E 16X: 4GB/s) and will not render in real-time (and still in low frame rate).

Hope this helps.

H.T.

[plasma_video]

I can add one tip to this excellent discussion.

For me, Cool 3D Studio renders MUCH faster if the display card properties are set to 16 bit high color instead of 32 bit true color. I do not know why that works, but for me it is true on both an nVidia and an ATI card. Perhaps it has something to do with either the preview display or the way the card handles the difference in bandwidth.

The speed of rendering is about 150% to 175% faster.

I found that out while trying to tweak the video card settings to improve rendering times. No other tweaks made an appreciable difference.

Tom

[htchien]

Yes, when use 16bit color mode to render, the required memory bandwidth will be less so it will render faster.

Regards,
H.T.