Nvidia Boosts Production of DirectX 10 Graphics Chips

Nvidia Places Additional Orders on “80nm GeForce 8000” Chip Production

by Anton Shilov
01/16/2007 | 07:02 AM

Nvidia Corp., a leading developer of discrete graphics processors and core-logic sets, has reportedly placed “urgent” orders on production of its flagship GeForce 8-series lineup. The reasons behind the move are unclear, but the most likely explanation would be expectations for tangible increase in demand for DirectX 10-compatible graphics cards.


DigiTimes web-site reports that shipments of the “urgent orders” of Nvidia’s “GeForce 8000 GTX and GTS” chips will begin in March and will be equivalent to 3 to 4 thousand 300mm wafers a month.

It is claimed that the wafers will use the “80nm process technology”. Currently Nvidia produces its GeForce 8800 GTX and GTS chips using 90nm process technology at Taiwan Semiconductor Manufacturing Corp. (TSMC), hence, it is uncertain whether the web-site actually meant new 80nm flavour of the chip code-named G80 or additional orders of the 90nm chips. The likely scenario is that Nvidia has placed orders to manufacture code-named G84 and G86 products.

Given that die size of the G80 chip is about 420 square millimeters, it is possible to obtain 140 – 150 of such chips of a single 300mm wafer, which will give 420 – 600 thousand of additional high-end graphics processor candidates per month (or 1.26 – 1.8 million a quarter). The number of graphics chips that are functional is lower than the amount of candidates, though, the actual yield of the G80 is unknown.

The market of graphics cards that cost from $250 to $700 accounted for 4% of revenue – or about $200 million – of the leading add-in card suppliers in Q3 2006, according to Jon Peddie Research. Average sales price of high-end enthusiast graphics cards may be as high as $475, which means that around 420 thousand of such boards are sold quarterly. Given that the yield of chips in mass production is much higher than 50%, it makes no sense for Nvidia to produce that many high-end chips.

Nvidia and TSMC did not comment on the news-story.