The first order of business for NVIDIA is to get NV30 off the ground. Though we picked up product specifics at the launch on Monday, we had a chance to get more detail on the timing of the ramp and remaining challenges yesterday. As expected, NVDIA has chosen not to use low-k dielectric on the NV-30, a decision which involves some performance compromises but also lowers the risk of any manufacturing miscues. With the part now qualified, waiting on system integrators to finish boards is the issue – that clearly is not going to happen until the April quarter.Though the two paragraphs fall directly after each other, they seem contradictory. While the first paragraph suggests that NV30 boards won't be available until the April quarter, the second indicates that a high-end NV3X part will be padding NVIDIA's revenues until then. What gives?
It will be very important for NVIDIA to successfully ramp the lower-cost versions of NV3X in time to hit the April 2003 quarter, however. The successful launch of the high-end part will not be enough to support a revenue upturn in the April quarter. Management was not willing to answer questions about whether additional versions of NV3X have taped out, although the intention to hit the April quarter was reiterated during our conversation.
My guess would be that Merrill Lynch has part designations mixed up in the first paragraph. We'll probably first see the GeForce FX as a high-end part only, likely with the Dustbuster, in the first quarter of 2003. Several months later, right around the April quarter, mid-range and low-end versions of the GeForce FX should become available.
A staggered release schedule makes sense, but NVIDIA is playing catch-up. ATI could stand to sell a whole lot of Radeon 9700s if NVIDIA's sole first quarter GeForce FX offering is a pricey high-end part. If NVIDIA is suffering from poor yields with the GeForce FX, it may be able to use binned GPUs in a slower, cheaper version of the GeForce FX based on the same board layout as the high-end part.