dburrell asks: When will you add hardware FSAA, and Matrox style bump mapping? These features have been requested in the past but it seems NVidia and ATI have taken the position of "the consumer won't miss it." I do miss it, and this impacts the amount of dollars I am willing to spend on Video Card updates.

NVIDIA: NVIDIA implemented FSAA with the GeForce2 Family of chipsets. We are always working to improve our performance and feature set so you can expect to see improvements in this area. Environment Mapped Bump Mapping, is inferior compared to the effects you get with Per-pixel bump-mapping so we see no need to implement that feature at this time.

rabbitfighter asks: Could you explain the problems with S3TC on the GeForce 2 architecture? NVIDIA has never publicly acknowledged a problem, yet clearly there is a severe one and it is limited to NVIDIA chips. The ATI Radeon and even the S3 Savage 2000 have no problems with S3TC causing severe image quality problems. Since this issue was never addressed in a driver update, one can only assume that the problem existed in the hardware. Is this the case?

NVIDIA: This applies specifically to the DXT1 format of DXTC. NVIDIA implements the DXT1 mode in 16-bits as that mode is a 16-bit format. Quality can be improved if you expand it to 24-bits and then interpolate, but the products functions exactly as it was designed and to the specification. All other DXT modes function properly as well.

Arcadian asks: Currently your video accelerators take advantage of high speed DDR SDRAM memory. Some of your newer designs are further taking advantage of a direct connection increasing the memory clock speed and connection width to allow for greater bandwidths. I was wondering if you have ever thought of using RDRAM instead of DDR SDRAM as a direct memory interface? RDRAM clearly has an advantage of being able to clock higher and because of the lower pin count, you can interleave with RDRAM in multiple channels, thus providing much more bandwidth than the DDR SDRAM alternative.

NVIDIA: Actually, GeForce2 Ultra delivers a higher memory bandwidth, at a lower cost than a 4-channel DRDRAM solution, making it the best choice for high performance. A 4-channel DRDRAM solution delivers 6.4GB/sec of peak memory bandwidth, while GeForce2 Ultra's 128-bit 230 MHz DDR (460 MHz data rate) memory interface delivers 7.4GB/sec of memory bandwidth, while costing less per bit of storage.

Arcadian asks: If you were thinking of using RDRAM in future products, I was looking for a rough release date as to when we can expect such a product? If you are not planning to use RDRAM at this time, I was wondering if you can run down some of the possible disadvantages to using such a technology, and why you do not think it could improve the speeds of your products over the already expensive DDR SDRAM?

NVIDIA: DRDRAM is more expensive per bit, and slower than DDR SDRAM. While DDR is more expensive than SDR, it is less expensive than DRDRAM by a substantial amount. The major memory manufacturers would much rather manufacture DDR memory as it is much more cost effective from a manufacturing perspective, and delivers substantially better price/performance.