NVIDIA GeForce Go 7800 GTX Preview
Intro and Specifications
In the world of the mobile GPU, if there are two things that are a constant it is delays and more delays. First mobile users have to wait quite a while between the release of a desktop product and its corresponding mobile part, and then users have to wait again in order for OEMs to implement the design and get it out for sale. In some cases this can take anywhere between 6-9 months from initial desktop product release usually leaving mobile systems one generation or more behind in GPU technology.
NVIDIA has plans to change this problem with the release of the mobile G70 product as they did with the 7800 GTX desktop part.
The GeForce Go 7800 GTX
The NVIDIA GeForce Go 7800 GTX is being announced today and will take the place of the GeForce Go 6800 as the flagship mobile graphics processor. As far as the architecture of this new GPU is concerned, everything from the desktop release of the product has been carried over to the mobile part. That means SM3.0 support, HDR rendering, PureVideo technology and anything else you can read about in our 7800 GTX preview in June.
Below is a table that looks at a quick comparison between the 7800 and 6800 mobile GPUs.
From this table we can get a lot of interesting information such as shader and geometry performance increases over the previous generation. The new chip does have over 100 million more transistors than the 6800 Ultra mobile chip. Probably most important though are the next two lines: power consumption and the package is the same as the previous generation. Because the power consumption and the package is the same, notebook designers don't have to change their thermal designs much, if at all. As you might imagine this makes the design process infinitely easier and much faster.
The clock frequencies shown here (400MHz core and 550MHz memory) do run a bit slower than the desktop parts (430MHz core and 600MHz memory), but only by a slight margin.
While the features in the GeForce Go 7800 GTX remain mostly unchanged from the desktop part, there are some special optimizations that were done to the mobile GPUs in order to make them more adapted to the mobile market. NVIDIA has gone through some great lengths to improve the power management of the Go GPU so that mobile users can still get great performance, lower heat and a longer batter life.
First, NVIDIA implemented dynamic clock scaling on the Go 7800 GTX which allows it to change its clock rate based on the amount of work the GPU is being given. This GPU can run at speeds as low as 16 MHz if needed. The engine, memory and pixel clocks can all be adjusted in this way.
In addition to scaling, clock gate is heavily used in the mobile product, even more so than in the desktop parts for obvious reasons. Gating allows the GPU to turn off certain parts of the GPU (such as the video system) when it is not in use, effectively lowering the clock frequency to 0 for that section resulting in more power savings.
Voltage scaling was the third feature that NVIDIA implemented and was made possible by the use of the 0.09 micron process. The Go 7800 can run at voltages lower than any other mobile performance part.
In addition, NVIDIA is using a link management system that allows it to switch between x1 and x16 lanes of PCI Express for the GPU depending on the bandwidth required.