• 8-32MB of SDR or DDR SDRAM
• 166MHz memory clock
• 32/64/128-bit memory bus
• 2.6GBps on 128-bit bus with SDR SDRAM

DDR memory will likely only be used with a 64-bit bus, just like the GeForce2 MX. Narrower memory busses will likely be matched to 8MB of SDRAM. The Toshiba system NVIDIA used for demonstration had 16MB of SDRAM. The core clock will limit high-resolution performance as will the memory bandwidth, but we expect to see, and did see in NVIDIA's presentation, quite good performance.

• TwinView support for two displays
• Two integrated LVDS transmitters capable of 1600x1200
• One integrated 350MHz RAMDAC capable of 2048x1536@60Hz
• Interfaces for TMDS and TV-out off-chip solutions

The GeForce2 Go has a full range of outputs, making it extremely versatile for system designers. NVIDIA said that their design could display video to both the LCD and an external display, and that other solutions could not. We did not have a chance to test this assertion on other displays.

(Note: all numbers, for both NVIDIA GeForce2 Go and ATI Rage Mobility 128, are NVIDIA's claims measured in Watts)

NVIDIA claims that, by offloading T&L processing from the CPU, the GeForce2 Go can save overall system power since, they say, the processor uses more power than the GeForce2 Go for T&L processing. NVIDIA claims a 2W overall savings, though the savings would likely depend on the system. In comparison, the GeForce2 MX takes about ~4W of power. The GeForce2 Go uses a lower core clock and a lower input voltage as well as internal power saving measures including selectively powering down currently unused parts of the GPU in order to cut power usage.