The Go 7800 GTX vectors
So, let's consider the hardware that's in the notebook, by looking at the vectors it's designed around:
Performance per watt:
NVIDIA claim that this 7800 architecture has been designed to deliver the most performance per watt. This means focusing on efficiency - it's no good having an awesome amount of power if there's no way to cool the chip in a small form factor. NVIDIA have employed some technology to keep things regulated internally to make sure this hits a good ratio of performance and power.
Obviously, the chip couldn't bear the 7800 name if it didn't deliver somewhere in the same range of performance as the 7800 desktop chip. Previous mobile chips have not been as fast as their desktop counterparts - NVIDIA have worked to ensure, they hope, that this one is.
PureVideo was introduced with the 6-series, and it's something that is now being rolled out across all of NVIDIA's product lines. With DVD, HDTV and Media Center becoming so important to both home and mobile users (especially with 17" notebook screens) having great video quality is a must.
To hit those performance vectors, NVIDIA is using these architectures:
This new technology has a few different tricks to keep battery consumption down. Clock speeds are managed dynamically - if speed isn't being used, it can be throttled back. Additionally, entire sections of the GPU can be turned completely off to save battery. Less clock speed also means less voltage, which means less power. Where applications don't need the full 16 lanes of PCI Express bandwidth for graphics, this can be clocked down to 1x, for applications like 2D.
That's right, the internal architecture is almost exactly the same as the desktop 7800 GTX. We covered the GeForce 7-series architecture in detail back in June. You can read about that here
For those who don't have the time to read our five page article discussing the architectural changes, we'll cover it in brief here. The GeForce Go 7800 GTX has the same 24 pixel pipelines, 8 vertex shaders and 16 pixel output engines as the desktop version and comes clocked at 400/1100MHz - slightly slower than the desktop version.
There is of course, Shader Model 3.0 support, improved FP16 blend performance - most commonly used with High Dynamic Range Rendering. There's also the new Transparency Adaptive Anti-Aliasing technique that means that there are even more edges smoothed out in the scene. This is because the alpha components (or transparent textures) are not picked up by conventional MultiSampling techniques.
Finally, the pixel shader efficiency is improved quite massively with popular shader instructions being processed much more efficiently. You can read more about the improved pixel shader on page 3
of our piece back in June.
We've talked previously about the video quality offered by PureVideo. It offers all sorts of gubbins to make sure the video you get on your display is as good as possible: WMV HD and MPEG 2 are both hardware accelerated, for starters. Video can be scaled and filtered in hardware, and the chip also provides for motion adaptive de-interlacing and NTSC pulldown correction.