NVIDIA's NV17M: A Brief Update

Since the press launch of NVIDIA's NV17M took place at Comdex we were unable to bring you the usual full set of details about the chip immediately upon its launch. And although we did put together a couple of pages of information regarding the new mobile chip and its potential role as a desktop part as well, there were still a few questions that remained unanswered. The purpose of this quick update is to shed light on some of the ambiguous issues surrounding the NV17M and a potential issue that NVIDIA may face in the naming and marketing of the solution.

A GeForce2 or GeForce3?

Ask yourself, what makes a GeForce3 any different from a GeForce2? Is it the addition of DirectX 8.1 compliant programmable pixel and vertex shaders? Is it just the use of NVIDIA's Lightspeed Memory Architecture? Or is it a combination of the two that truly make a chip deserving of the GeForce3 moniker?

Upon first sight, the most interesting thing about NVIDIA's NV17M launch at Comdex was that there was no actual name for the part announced along with its specs. The reason behind this is that there is actually a very fine line NVIDIA is walking with the NV17M part. The GPU itself has a lot of potential in the market to be a very low cost, high performance solution both on the mobile and on the desktop sides but it achieves its price point through a sacrifice that would cause many to question whether it's a GeForce2, a GeForce3 or neither.

We mentioned in our initial article that the NV17M might not have a full implementation of a DirectX 8.1 compliant programmable T&L pipeline. Since the publication of that article we've received confirmation from NVIDIA as to exactly what the NV17M does and does not support. Here's a quick breakdown of the specs:

• 0.15-micron GPU w/ Mobile AGP Package (see our original article for more information)
• GeForce3 Memory Controller w/ Lightspeed Memory Architecture
• New Video Processing Engine with IDCT support
• Quincunx AA support
• No Pixel Shader support
• Limited Vertex Shader support

The last two bullets are the most important; The lack of any pixel shader support prevents the NV17M from being a true DirectX 8.1 part. We were also told that the limited vertex shader capabilities of the GPU are specifically in the removal of two vertex shader "techniques" that are not supported by the NV17M. Our guess is that the most die consuming vertex shader elements were removed and the rest left intact.

The end result of this hybrid design is a GPU that will perform perfectly fine in all of today's games yet won't boast support for future titles that fully utilize pixel and vertex shader programs. This won't matter as much on the mobile side but it has much more meaning on the desktop end of things.

There's nothing wrong with this design choice but where the controversy does lie is in the naming of the chip.

Here's where NVIDIA can actually learn from ATI; more specifically in a mistake ATI made. Many of you will remember the release of the Radeon VE as a heavily crippled Radeon core basically without any T&L engine. The controversy surrounding this product was mostly on the part of developers who complained that they are no longer able to recommend that end users buy any Radeon or GeForce based card for their games if they require T&L support. The same potential lies in the naming of this NV17M part.

From a marketing standpoint, it's clear that granting the chip the GeForce3 title would help drive more sales. But from a developer standpoint it's another VE situation just waiting to happen. Hopefully the fact that NVIDIA didn't launch the chip with an official title means that somewhere in their massive Santa Clara offices they are having this very same debate.