iirc we didn't have much info on jaguar or the gpu families either. We might get hard numbers some time soon though. If steamroller is another 15%(including clocks) there is no need for AMD to go about advertising it as such. Also that pic seems like the design goals of the big cores have been changed after bd release.
Times are a-changin'. AMD's market share is borderline nonexistent on the server side and their server strategy has consequently been heavily reorganized.
They are now committed to targeting microservers with 64-bit ARM and smaller more efficient cores and leveraging GPGPU rather than big cores for traditional servers. Their <5% market share is lower than ARM, and that's ARM on 32-bit. There's no way in hell AMD is going to claw back there unless they get 14nm-XM access in a relatively quick time frame, but that process is being tuned primarily for mobile SoCs; I've seen no mention at all of a high power compatibility on the 14nm-20nm BEOL node at GloFo.
Intel makes three types of chips: Atom for low end, Haswell/laptops, and Xeons for server.
AMD, too, makes three types of chips: Jaguar for low end, APUs for laptops, and Opteron for server. AMD's Opteron chips are essentially on the AM3+ platform just as SB-E/2011 is the enthusiast/workstation platform for Intel.
As soon as AMD stops advancing the server roadmap then consequently they've stopped advancing on their enthusiast desktop platform as well.
Ever since AMD's Brazos platform, it's been their APUs that have sold better than their desktop and server chips. The margins aren't quite as high, but their margins on the desktop side haven't been stellar since the introduction of the X2.
Bulldozer was awful, but some of the sacrifices the chip made, at least with respect to having a single FPU shared between two integer cores, makes a fair amount of sense and has the potential to tie in well with their GPGPU agenda via their APUs. So I'm not sure big cores will disappear from AMD, but the lack of anything Steamroller-derived on their Opteron roadmap is a telling sign that it might be only APUs from this point forward.
And they aren't even using soi now, it isn't bleeding-edge.
Common platform committed to using bulk for 28nm so AMD had no choice. Going forward, FD-SOI is certainly a possibility.
Globalfoundries, the world’s second largest contract maker of semiconductors, will start to produce test wafers using its hybrid 14nm eXtreme mobility (XM) process technology as soon as in the first or second quarter of 2014. Early adopters of the technology will be able to produce their test circuits to find out the pros and cons of the process. Unfortunately, the main customer of GlobalFoundries, will not be able to benefit from the 14nm XM FinFET technology.
The 14nm-XM offering is based on a modular technology architecture that uses a 14nm FinFET devices combined with 20nm-LPM process back-end-of-line (BEOL) interconnect flow. Leveraging the maturity of the 20nm-LPM technology will enable a rapid time-to-market as well as a smooth transition for customers looking to tap the benefits of FinFET system-on-chips as soon as possible. According to Globalfoundries, 14-nm FinFETs have a 48-nm fin pitch, which is identical to what Globalfoundries expects about Intel's FinFET fabrication process.
Mike Noonen, executive vice president of worldwide marketing and sales, told EETimes web-site that 14XM mutiproject wafer (MPW) runs could take place in the first or second quarters of 2013 and would be used by early adopters of the process to run test circuits.
Globalfoundries seems to have a clear roadmap for both SOI as well as fully-depleted SOI at 28nm and 20nm. But those technologies use traditional planar transistors, not innovative FinFET transistors, which means that AMD will not be able to match Intel’s performance and power saving capabilities (Intel claims that tri-gate transistors enable 37% higher speed, or 50% lower power consumption compared to planar transistors) in the foreseeable future; at least, till Globalfoundries and its partners develop ultra high-performance 14nm process technology.
I believe risk production for 20nm is beginning later this year for both TSMC and GloFo while 14nm-XM PDKs have been sent out several months ago.
http://www.xbitlabs.com/news/other/...s_to_Start_14nm_Tests_in_Q1_2013_Company.html
AMD's biggest issue is that these two nodes (or 1 and 3/4??) is that they're primarily being introduced for mobile.
As the market moves away from PCs, so do the foundries. As a result, that puts a lot of pressure on those on wrong side of these trends. While we might get a future Jaguar-ish SoC on GloFo's 14nm-XM, it's extremely unlikely we see anything at all for the big core chips without a significant delay.