Nvidia
Taking it on the road
One of the worst things that you can do in Vegas is gamble for prolonged periods of time and expect to win. Coming in a close second? Planning a press release breakfast in a city that lets you walk the streets with a jug of grandpappy's cough syrup. The NVIDIA press event kicked off on the first day of Comdex at eight in the morning. If you ever want to showcase to a groggy, decidedly un-pumped up crowd, this was the perfect time. We pulled ourselves out of our comas, cursing all the while. Coming down from our hotel room, we were confused as to whether the people in the casinos had been there the entire night, or whether they had actually woken up that early to gamble - either way, not a good sign.
NVIDIA's big release during this Comdex has little to do with desktop users. This time around, the NV17M was released, NVIDIA's new laptop graphics processor. The NV17M is to replace the GeForce2Go that came out last year. As you can tell by the name, this chip is neither a GeForce2, nor a GeForce3, landing somewhere in between the NV15 and the NV20. NV17M will run at about 250MHz as opposed to GF2G's 143MHz main clock, and will more than double the previous generation's 2.6GB/sec memory bandwidth, topping out at 8GB/sec.
Not just more speed
The new chip comes with a number of feature improvements. The addition of FSAA to the mobile scene should prove rather interesting. Integrated in the core is the Lightspeed memory architecture found in the GeForce3. To reduce the space consumed by their solution, NVIDIA managed to put the entire graphics package onto one chip, this is dubbed Mobile AGP Package (MAP). Another feature called PowerMizer helps to reduce power consumption. Depending on how the system is being used at the time, the core will dynamically adjust by lowering voltages, reducing clock speed, and shutting of various parts of the chip.
With NV17M, Nvidia has finally taken upon itself to update one of the most lacking areas of its graphics processors (and no, we're not harping on 2D quality here): hardware acceleration of video/DVD, handled by the new Videp Processing Engine (VPE). Through the VPE, NV17M can now handle inverse quantization in addition to inverse discrete cosine transform, motion compensation, and YUB-RGB color space conversion, all in hardware. This one-ups ATI's entire Radeon line, which has long held the crown for hardware-accelerated MPEG2 decoding. With the less-demanding notebook market as a testbed, we can safely assume that some of this technology will make it into Nvidia's next desktop performance GPU, giving it yet another notch to match or beat the Radeon's multimedia reign.
NV17M is scheduled to be released in early 2002.