Nvidia — the six-month cadence

Out-ship every graphics competitor on the pace of new chips

Graphics rivals leapfrogged each other on an 18-month cycle, so whoever shipped a faster architecture first took the market, since a single late chip could sink the company.

Three engineering teams working in parallel, each on a different architecture, so a new generation ships even while the next two are already in development.

Release a new architecture roughly every six months instead of every eighteen, built to open industry standards rather than proprietary lock-in.

1. 01Run three GPU design teams on staggered, overlapping schedules
2. 02Ship a new architecture every six months, not every eighteen
3. 03Build to DirectX standards so games run on Nvidia out of the box
4. 04Emulate new features in software when the silicon isn't ready yet
5. 05Sample chips to game developers months before launch