Do me a favour (timings) ?

Question

We need to find a way through the DVD encoding bottleneck, so we're thinking of building another system (Windows 7) pretty much for this purpose. 
Before we spend the extra cash on an nVidia graphics cards I'm looking for some indication of how much time the GPU may help. We typically don't use the GPU assisted effects, so those don't come in to play at all.   It's all about the time saved on resizing 1080p to DVD frame sizes.
What I'm looking for is help with this:
1) Create a 1920x1080 25p sequence and fill with 1 minute of video (you can choose your own length from a few seconds upwards).
2) Disable GPU acceleration
3) Export to MPEG DVD, PAL Progressive Widescreen Highest Quality
4) Change the Bitrate settings to VBR 2 Pass 2.8, 6, 8  (just so we're on the same settings)
5) Set audio to PCM (just so we're on the same settings)
6) Check the Use Maximum Render Quality
7) Uncheck the Use Previews (so it's not using previews)
Now export this and time it.
8) Repeat the above but with GPU enabled.
How much (if any) difference is there?    I'm not worried about the absolute 'real time' scores, more about how much the GPU assist helps
Also, if you have time, add the export to the Queue instead of direct export (with GPU enabled) and see how long that takes via Media Encoder.
I don't mind if it's only 10 seconds or 30 seconds of video source, I'm really only interested in the difference the GPU makes.
For reference, my 8 core mac (2.8Ghz) would take 3m 35s to export directly (no GPU).   Via Media Encoder it takes 38 seconds to feed the project to Media Encoder (which was not running) and 5m29s for Media Encoder to run (with all 8 cores at 100% - why the difference in time?  Argghh!!).
So, if any one is willing to run this comparison of GPU vs non GPU for me I would REALLY appreciate it!   I'm really not worried how long your systems take, just whether it's faster or not with the GPU enabled.
Thanks!

Answer 1

I think I pointed you to this blog post by Todd before, but most of it is explained here: some details about scaling in Premiere Pro CS5 and CS5.5 « Premiere Pro work area
Excerpt from that post:
Moving a lot of processing to the GPU can also make things better, not just faster.
A good example is scaling. There are lots of different scaling
algorithms, and they each have their pros and cons. Some are better for
scaling things up, some are better for scaling things down; some are
better for sharp graphics, and some are better for gradual changes in
color across an image. The real tradeoff, though, is that the
high-quality algorithms are also—in general—the slow algorithms.
However, these higher-quality algorithms are only really slow if you
are forced to execute them serially, but they are relatively fast when
you can run them in parallel. One of the huge advantages of GPU
processing is that GPUs are massively parallel, with hundreds of
parallel processing units. There are a lot of pixel operations that are very
amenable to parallel processing, since you don’t need to know the
result of the operation on one pixel to do the same operation on its
neighbor in the same image. Scaling is just such an operation. When you
move scaling operations to the GPU, you get to take advantage of scaling
algorithms that were just plain unfeasible on the CPU.
So, scaling using CUDA can be better. And faster. In some tests done here, scaling was more than 40 times faster on the GPU than on the CPU at maximum quality.
When Premiere Pro is just using the CPU for the processing of scaling operations, it uses the following scaling methods:
playback: bilinear
paused: Gaussian low-pass sampled with bilinear
high-quality export (Maximum Render Quality off): Gaussian low-pass sampled with bilinear
Maximum Render Quality export: variable-radius bicubic
The variable-radius bicubic scaling done on the CPU is very similar
to the standard bicubic mode in Photoshop, though the Premiere Pro
version is multi-threaded and optimized with some SSE instructions. Even
with these optimizations, it is still extremely slow. For high-quality
scaling at faster-than-real-time processing, you need to use a CUDA
card.
When Premiere Pro is using CUDA on the GPU to accelerate the
processing of scaling operations, it uses the following scaling methods:
playback: bilinear
paused: Lanczos 2 low-pass sampled with bicubic
export: Lanczos 2 low-pass sampled with bicubic
For export, scaling with CUDA is always at maximum quality,
regardless of quality settings. (This only applies to scaling done on
the GPU.) Maximum Render Quality can still make a difference with
CUDA-accelerated exports for any parts of the render that are processed
on the CPU.
I added the emphasis at a couple of points to highlight what I was referring to. Basically, the export quality (when it comes to scaling) tiers out like this, from highest to lowest:
GPU on (Lanczos 2 low-pass sampled with bicubic) [typically fastest]
GPU off, MRQ on (variable radius bicubic) [typically slowest]
GPU off, MRQ off (Gaussian low-pass sampled with bilinear) [somewhere in between]
GPU off with MRQ on can--practically speaking--get close to GPU on, but may still be a bit lower quality, and it will take many times longer to export. GPU on with MRQ on may or may not be any different from GPU on with MRQ off; it depends on whether the segments of the sequence are being processed by CUDA or not (yellow vs. red bar).
Clear as mud?