Article: Power Delivery in a Modern Processor
By: Anon3 (non.delete@this.anon3.invalid), May 14, 2020 7:06 am
Room: Moderated Discussions
Mark Roulo (nothanks.delete@this.xxx.com) on May 13, 2020 8:22 am wrote:
> Short version: Validation and debugging are harder and the win is not as obvious as one would like.
Eta Compute recently had a async Cortex M3 based on a technology they call DIAL. However I think their latest designs have given up on this in favour of a more conventional design with much more agressive frequency scaling than normal something they call CVFS.
Another issue with async over and above merits of the technology itself is predicting the overall performance and binning is hard or doesn't really work at all. So if your business model is based on binning then you aren't even going to look at it.
> Short version: Validation and debugging are harder and the win is not as obvious as one would like.
Eta Compute recently had a async Cortex M3 based on a technology they call DIAL. However I think their latest designs have given up on this in favour of a more conventional design with much more agressive frequency scaling than normal something they call CVFS.
Another issue with async over and above merits of the technology itself is predicting the overall performance and binning is hard or doesn't really work at all. So if your business model is based on binning then you aren't even going to look at it.
