by Anna Filatova
08/19/2002 | 09:06 AM
The arguments about the due date for the new NVIDIA NV30 solution seem to have broken out anew, especially taking into account the problems NVIDIA faces. Another portion of fuel was added to the today’s fire by Anandtech site. In this respect, I would like to sum up everything we already know on the schedule for the upcoming NV30 chip.
So, at first NVIDIA wanted to release NV30 together with DirectX9 in October. However, about a month ago some info about the possible delay started cropping up here and there.<%BANNER[article]%>
According to the initial plans, NVIDIA wanted to start making the first working silicon of the notorious NV30 in May already. But the company president confirmed officially that two weeks ago this process wasn’t even close to start. All the existing data on NV30 including screenshots, performance, etc. were obtained from software simulators of the new chip.
Nevertheless, some time ago we got a word that last week the manufacturing processor of the new chip started successfully. This way, NVIDIA is evidently falling at least 3 months behind the schedule with its NV30.
On the other hand, it usually takes about 90-120 days from the beginning of the first silicon production until the mass manufacturing of the final chips. As a result, we can theoretically expect the graphics cards based on NVIDIA NV30 to start selling in November at the latest. However, NV30 has already been delayed once, so there is no certainty that it will not be delayed again. When the company finally gets the first silicon versions, they may discover some bugs in the chip, which may in their turn require immediate correction or even chip redesign. So, if worst comes to worst NV30 may not be launched this year at all.
This way, ATI and Matrox are in very favourable situation now, as there are no competitors from NVIDIA in the market. So, they can easily earn some really good money in Q3 if they take advantage of the situation smartly.
One more problem for NVIDIA is the fact that NV30 chips may turn out pretty expensive to produce, so that the company will not be able to make the graphics cards based on these chips sell at competitive prices. The thing is that unlike ATI, which uses a well-polished 0.15micron technology of TSMC to manufacture its R300, NVIDIA pinned its hopes upon the 0.13micron TSMC technology. However, so far TSMC has very limited production capacities to work with this fine technology, which make around 1% only of the entire production. Moreover, the yields are very low and make only 10-20%. For a better comparison take Intel, which starts the mass production of new processors only when they manage to achieve at least 50% yields. This way, even if NVIDIA manages to recommend selling its NV30 based graphics cards for around $400, which is required to make these solutions competitive to ATI’s, the supplies will anyway be very limited.
For the same reasons NVIDIA will be unable to offer a better value version of the new NV30 targeted for a more mainstream market at first. This is exactly why they scheduled the NV31 chip only for Q2 2003.
Summing up I would like to state that firstly, NVIDIA has every chance to break the "6-month-cycle" schedule, and secondly, the results in Q3 may turn out even worse than those in Q2, which we have recently discussed in our news.