Now i see what Intel is getting at. My Guess
Broadwell wont be coming on to Desktop. Haswell Refresh will be a tweaked stepping with better thermals. Higher Clock Speed etc.
Along with 9 series chipset that may be produced at 22nm. Since Desktop Sales is declining rapidly. When Mobile and Smartphone hops onto 14nm those 22nm capacity will be enough for Desktop and Chipset, as well as the eDRAM required for GT3e.
14nm looks like to be designed with Low Power in mind. At least initially. Hence the reason why there wont be a desktop Broadwell, which normally made with High Performance Node.
The Low Power 14nm Node will be given first priority to Mobile Atom. Where Intel is desperate to get some market share. Intel again will be using their manufacturing advantage to full length. My guess is that at 14nm and their new design and work accumulated over the pass few years, this looks like a sure win in terms of Power / Perf. The only thing left is price. At least both Apple and Samsung are very unlikely to buy x86 chips from Intel. And TSMC 16nm is only few months to half year behind Intel. Compared to what we are used to which is a whole node cycle.
So Broadwell will be Mobile only with a new GPU architecture. And it is very likely Intel is going to double the GPU Power in their top Model.
What worries me is that Intel has a habit of dropping older graphics driver development once a new GPU architecture arrives. And their drivers development as been slow already. Although for Mac users this dont matter so much since the drivers development are done by Apple ( I think )
In terms of die size, my guess is that like the rest of the industry 16/14nm will be an expensive node compare to previous node. So a smaller die size is required to balance out the cost. But i still think the die size and cost for Intel are minor compared to their die shrink. More likely they are miking their Desktop and Server Business where they have little to no competition.
Broadwell wont be coming on to Desktop. Haswell Refresh will be a tweaked stepping with better thermals. Higher Clock Speed etc.
Along with 9 series chipset that may be produced at 22nm. Since Desktop Sales is declining rapidly. When Mobile and Smartphone hops onto 14nm those 22nm capacity will be enough for Desktop and Chipset, as well as the eDRAM required for GT3e.
14nm looks like to be designed with Low Power in mind. At least initially. Hence the reason why there wont be a desktop Broadwell, which normally made with High Performance Node.
The Low Power 14nm Node will be given first priority to Mobile Atom. Where Intel is desperate to get some market share. Intel again will be using their manufacturing advantage to full length. My guess is that at 14nm and their new design and work accumulated over the pass few years, this looks like a sure win in terms of Power / Perf. The only thing left is price. At least both Apple and Samsung are very unlikely to buy x86 chips from Intel. And TSMC 16nm is only few months to half year behind Intel. Compared to what we are used to which is a whole node cycle.
So Broadwell will be Mobile only with a new GPU architecture. And it is very likely Intel is going to double the GPU Power in their top Model.
What worries me is that Intel has a habit of dropping older graphics driver development once a new GPU architecture arrives. And their drivers development as been slow already. Although for Mac users this dont matter so much since the drivers development are done by Apple ( I think )
In terms of die size, my guess is that like the rest of the industry 16/14nm will be an expensive node compare to previous node. So a smaller die size is required to balance out the cost. But i still think the die size and cost for Intel are minor compared to their die shrink. More likely they are miking their Desktop and Server Business where they have little to no competition.
Facebook
Twitter