In order to address concerns of a difficult transition to 28nm, AMD demonstrated its next-generation 28nm GPU at IDF last month. The demo system featured a mobile variant of the next-gen GPU running DiRT 3. At the time, AMD told us to expect the GPU's release later this year.
Since then there have been rumors of a delay until 2012, but no official update from AMD. I suspect we'll get something to that effect soon enough, but today AMD let us know that it ran another public demonstration of 28nm GPU silicon - this time in Taiwan. The demo was apparently pretty similar to last time: a mobile 28nm GPU running a game. In this case the game was Dragon Age 2, but obviously there are no performance or power details to accompany the announcement.
AMD didn't provide any more info nor an update on the GPU's release timeframe. As with any major process transition, scaling to be able to produce mass quantities of chips on a new process can be extremely challenging. The move to 28nm is particularly rough because unlike 40nm, AMD didn't have a pipecleaner part to begin its process learnings on. How this impacted overall development remains to be seen.
Source: AMD
1) Full Node shrinks reduce layout area by a factor of 2 - so dimensions reduce by sqrt(2).
2) Half nodes should shrink area by a factor of sqrt(2) so dimensions reduce by 2^1/4.
Full Nodes 45, 32, 22.5, 16, 11, 8
Half Nodes: 38, 27, 19, 13.5, 9.5, 7
In practice this is limited by what the technology allows and what the customers need, so often they are rounded up or down.
Copyright © 1997-2012 AnandTech, Inc. All rights reserved. Terms, Conditions and Privacy Information.
Click Here for Advertising Information
