New AMD decoder ring; 'Sea Islands' (Canary Islands) to come in 2013

[caveman-jim]

Now that the 2012 AMD Financial Anaylst day is over, you're probably struggling to keep up with the new codenames for unreleased products.

Here's AMD's roadmaps:

you’ll note the addition of the “Sea Islands” family to our graphics roadmap. As is typical with our graphics roadmap, we don’t disclose many details beyond the codename. You will have to wait until 2013 to get the rest of the story on that one.

http://blogs.amd.com/work/2012/02/02/your-new-amd-decoder-key/

Read here for why "Sea Islands" is really 'C-Islands' and in fact 'Canary Islands'

[noko]

The 2013 Steamroller and GCN graphics in the mobile roadmap looks enticing.

No client Steamroller in 2013 , looks like PileDriver will have to last for awhile before next major update. I take it SteamRoller will be new chipset/Tri-memory controller or quad, pcie 3 etc. or DDR4/5 maybe.

PileDriver, when? End of this year?

[caveman-jim]

Piledriver in Trinity in middle of year I think, Piledriver in FX is 2013 I think. Not sure.

[NoHero]

Maybe I am over looking here, but I don't see "Sea Islands" on any of the linked road maps.

[caveman-jim]

Hit the link for the originals/more; it's in their blog as the new codename, so Sea Islands is the official name right now; it's really 'C Islands' because it's 'Canary Islands'.

[NoHero]

Hit the link for the originals/more; it's in their blog as the new codename, so Sea Islands is the official name right now; it's really 'C Islands' because it's 'Canary Islands'.

aahhh that clarifies it. Thanks.
2013 makes it little obvious they are aligning Canary islands with 20nm availability. TSMC states that 20nm low power process will be ready by Q1 2013, two quarters ahead of schedule. I wonder if we will see another "RV740" testing the waters.

[caveman-jim]

Given the statements in the AMD Financial Analyst day (don't lead with your chin, don't adopt new process technology unless you can deliver) I wonder if AMD aren't going to do a second round on 28nm instead of 20nm in 2013. Course, they might just wait like NV do, intro later to get a better handle on the process tech.

I think we won't see 'pipecleaners' anymore, as AMD have made significant improvements to their back-end infrastructure; their worldwide cloud; for engineering modelling and simulation. With better simulation and modelling power and tools they need less physical runs to find out about the process node, and TSMC have learned from the 40nm/32nm problems, too. Also don't forget that AMD want to use as GloFlo wherever they can, will GF 22nm be a better bet? Interesting stuff.