The Inquirer-Home

Only 10,000 GeForce 6800 series to be produced?

IBM Microelectronics yields could be the problem
Mon May 03 2004, 07:27
PERHAPS FILE THIS in the rumour mill, but the INQUIRER has been given the heads-up on some tasty internal correspondence which has just flowed within ATI Technologies.

This correspondence comments on rumours indicating that 'all Nvidia NV40 (GeForce 6800) 3D graphics business with IBM has been concluded'.

ATI appears to believe that the entire run of Nvidia NV40 GPU's from IBM's multi-billion dollar East Fishkill fab (semiconductor fabrication facility) is a relatively small 10,000 units.

Furthermore, it's believed that IBM's fabrication rival in Taiwan, TSMC, has not yet taped out NV40 for Nvidia.

It seems that the next project for Nvidia and TSMC is to be Nvidia NV41. This is described as a 175 million transistor, 8x or 12x pipeline part set for announcement at E3 very soon .

The production future of Nvidia's previous generation mid-range 3D accelerator GeForce FX 5700 Ultra (NV36) might also be in some doubt, as this was one of the first chips produced by IBM at its East FishKill fab.

If what ATI believes is true, what's behind the cessation of NV40 or any other Nvidia production at IBM?

Various suggestions have been put forward by exceptionally well informed individuals, but more than one has basically suggested that 'yields suck at IBM' and that 'maybe other (IBM) customers are having the same issues'. We've reported on yields problems at IBM in previous weeks.

Following the launch of the ill-fated GeForce FX 5800 Ultra (NV30) we seem to recall some public comments by Nvidia, and definitely from one of its partners which, referred to process problems at TSMC, and which essentially seemed to lay significant blame at TSMC's feet for the failure of NV30.

Subsequently when Nvidia started to canoodle with IBM, we heard lots of loving noises from Nvidia PR regarding IBM. TSMC might not be too happy with Nvidia - certainly ATI is now one of its very valued customers. µ


Share this:

blog comments powered by Disqus
Subscribe to INQ newsletters

Sign up for INQbot – a weekly roundup of the best from the INQ

INQ Poll

Heartbleed bug discovered in OpenSSL

Have you reacted to Heartbleed?