Originally Posted by 8800GT
Well AMD's 290x is 20% more inefficient than the 780 ti whilst having about a 20% smaller die area. I'd say they could do better in the power department.
Indeed a much denser die and higher clocks and a horrible cooler to make consumption even worse. We had reports of people clocking down their cards to 700/800MHz and using only ~180W so I don't think they are behind in power.
Now with Maxwell's low/mid end 750 and 750ti they are making great advancements by what is discribed as a new layout and more cache but I don't think AMD is sleeping either with their GCN being in the tiny core based APUs the large core apus and their graphics and soon be in their ARM segment. GCN should allow AMD to exploit it's IP across all segments which is really good for them.
Originally Posted by firagabird
20nm with FinFETs.
You're right just re-read it it says "finfet migratable" my bad
Clearly that picture shows that compared to the previous super low power node 20nm will be smashing but I have no idea how effective it will be scaled up at low clocks. As for the increased cost they claim the cheapest 20nm node.
20LPM (Low Power Mobility)
Leading Edge Planar Platform for Mobile & Consumer Applications
FEATURES / BENEFITS
Industry's most 20nm cost-effective, Fin-FETmigratable platform
Single unified platform configurable for optimum cost/performance/power
2x gate density, up to 50% performance advantages vs. 28nm
Gate-last HKMG, single technology platform
Uses patented High Density Constructs to facilitate IP migration to 14XM
When we move over to 14nm XM we see that this 3d transistor fails to deliver on higher voltage something we will see more and more often. For what AMD intends to use 14nm XM namely ultra mobile chips it is a win but it really has no place in the high end for now.
The 14nm Samsung collaboration is interesting and leaves no doubt that Globalfoundries will become a player as big as TSMC.
The work that GLOBALFOUNDRIES and Samsung are doing together will help AMD deliver our next generation of groundbreaking products with new levels of processing and graphics capabilities to devices ranging from low-power mobile devices, to next-generation dense servers to high-performance embedded solutions.
Still doesn't change the fact that 28nm TSMC destroys Intel's 32nm and that 20nm TSMC even if it is not a finfet node it will probably match or destroy Intel's 22nm. Intel's node size however is actually said to be smaller than is really is compared to how the rest of the fabs names it as far as I heard.
There are plenty of cool things coming 2014 even if we don't see a node switch we may still finaly see HDL make it's way in the Carrizo APU late 2014 or early 2015 which is really the only good thing about them staying on 28 besides the savings it brings for AMD.