GlobalFoundries preps 22-, 20-nm processes for 2013

Backed by a steady stream of cash from Abu Dhabi, GlobalFoundries is quickly establishing itself as a leader in the independent foundry business. Yesterday, the company revealed its roadmap for 22- and 20-nm fabrication processes, and it also announced plans for a “High Performance Plus” version of its 28-nm process.

GlobalFoundries says it intends to run “test chip shuttles” at 22 and 20 nm in the second half of next year, with “risk production” to follow in the second half of 2012. Actual product production is expected to begin some time in 2013. The 20-nm process will be available in high-performance and super-low-power variants; GlobalFoundries also intends to offer a 22-nm super-high-performance process for “devices requiring the utmost in performance.” (Sadly, the announcement doesn’t get into more detail.) All three processes will be based on next-gen high-k metal gate technology.

Speaking of super-high performance, this announcement also foretells the arrival of a 28-nm “High Performance Plus” technology, which will purportedly enable up to 10% greater performance than the regular 28-nm high-performance process. GlobalFoundries will begin risk production for the 28-nm HPP tech in the fourth quarter of 2011. This time, the firm does name concrete applications: “smart mobile devices and high-performance processors requiring more than 2GHz of processing power.”

Comments closed
    • JumpingJack
    • 12 years ago

    @sschaem

    The glo-flo 28 nm process is on bulk, and is lower performance in terms of Fmax, but optimized for low power. Perfectly suitable for ARM, but it is not suitable for high performance CPUs that AMD needs, you will never see an AMD CPU on that 28 nm process — at least not what you will find in notbooks, desktops, and servers.

    If the 28 nm is slated for end of next year, Intel is also slated for 22 nm in the same timeframe, therefore Intel remains in the lead.

    Intel launched 32 nm products in the first week of January of this year, AMD will launch 32 nm products sometime next year.

    Between 90 nm, 65 nm, and 45 nm AMD has historically been 10-12 months behind, they will likely fall further behind at 32 nm.

    • blastdoor
    • 12 years ago

    All excellent points. If I had mega-bucks I’d be investing in real-world physical things, too. The whole thing of trading one piece of paper with a face/building on it for another piece of paper with a face/building on it isn’t really paying off anymore. I think trading paper for real world things is a good move.

    Heck, I’m even doing that to a certain extent with my mini-bucks. I recently paid for an SSD — that has a real, perceivable pay-off that I experience every time that I use my computer (so the return is guaranteed — no risk!). If I put that money in an equally safe investment (say a CD or treasury bond) with a similar lifespan, I’d be making less than 2%.

    • NeelyCam
    • 12 years ago

    You’re confused if you say Gulftown and Clarkdale were not “proper” CPUs. And they both spank pretty much everyone in performance per watt.

    And Clarkdale desktop platforms idle at far lower power levels than Core2 or any Nehalem/Lynnfield.

    • NeelyCam
    • 12 years ago

    Would you care to please enlighten me; what does this “schedule” you are referring to look like?

    New process every two years, consistently. How is this “way, way off schedule”?

    • sweatshopking
    • 12 years ago

    the end of next year on the 28nm

    also in teh ARM vs intel camp
    §[<http://www.brightsideofnews.com/news/2010/9/2/arm-and-globalfoundries-attack-intel-atom-252bghz-cortex-a9.aspx<]§ 2.5ghz cortex a9 is otw

    • sschaem
    • 12 years ago

    Didn’t Intel release 32nm earlier this year? and GF will have 28nm production by the end of the year.

    So it will only take ~10 month for GF to BEAT Intel, not catch up.

    And didn’t Intel stated 22nm for late 2011 ?
    So it look like Intel will be a full year (possibly more) behind GF.

    How do you get that Intel beat GF by 24month ?!

    And from this article, GF will then again match or beat Intel 6-10 month later.

    • sweatshopking
    • 12 years ago

    nice man. real nice. wait. lame and dumb post. leave such stuff to the religion and politics forum if you need to post your crap

    • UberGerbil
    • 12 years ago

    You’re not the server market, and the server market matters.

    • UberGerbil
    • 12 years ago

    They have more than half a trillion dollars, and they have to put the money /[

    • OneArmedScissor
    • 12 years ago

    The only way I look at it is that none of this matters, at all, until it means something beneficial to me.

    I’m still waiting for Intel to catch up to themselves in that department. Their 45nm platforms spank their 32nm platforms in battery life and affordability. If it takes them until the end of 2011 for the “full” Sandy Bridge, which, if you ask me, will be their first “proper” 32nm CPU, then Bulldozer may come out first.

    Is AMD really behind, then?

    • OneArmedScissor
    • 12 years ago

    And yet, they cancelled quite a few CPUs over the past two years, were late with 32nm, and are still mixing 45nm off-die GPU/northbridges with tiny, tiny Core 2-esque dual-cores.

    I’m sure their yields are fine in their eyes – just enough to keep the market for $1,000 CPUs fed, which is about all they’re using it for a year after it was supposed to arrive.

    That doesn’t mean they have a huge manufacturing capacity. There aren’t even 32nm Atoms, for crying out loud.

    By the time they have anything close to beneficial 32nm mainstream CPUs, it will have been two years past the target “tick” point. They are way, way off schedule.

    • bcronce
    • 12 years ago

    AMD said 32nm Q4 of 2009 and is now pushed to 2h 2011

    Intel said 32nm in Q1 2010 and had purchasable products in March 2010

    Q1 ’10 to Q1 ’11 is 1 year. But 2h usually means late Q3 or Q4, so closer to 2 years. 2 years is “more” correct.. ma’b 1.5 is best. Even then, by the time they start selling 32nm, Intel will had 22nm around th corner in 3-6 months.

    • NeelyCam
    • 12 years ago

    Not quite true – Intel is simply manufacturing 45nm CPUs at the same time with 32nm CPUs just to extract more profits from 45nm fab investments… why? Because the competition is… well, not very competitive.

    This in no way implies that Intel’s 32nm is somehow not ready. Intel press releases have consistently stated that yields are great or better than expected.

    Meanwhile, IvyBridge (i.e., non-atom 22nm CPU) is scheduled to Q1/2011 – are you betting on them not meeting this schedule…?

    • NeelyCam
    • 12 years ago

    Um… let’s think about this:

    l[

    • TaBoVilla
    • 12 years ago

    absolute mega lol

    • OneArmedScissor
    • 12 years ago

    Q1 2012 is probably being optimistic, but the kicker is that Intel has been increasingly stretching what constitutes the introduction of a new manufacturing process. It used to mean cheaper and better CPUs across the board. Now it means something gimmicky or something psychotically expensive just so they can say they did it.

    I’d bet money they do something like 22nm Atoms for phones and then just push everything else way back.

    • moritzgedig
    • 12 years ago

    certainly better than terrorism.
    but I still don’t get what they want.
    10% on the investment annually? Never!

    • mesyn191
    • 12 years ago

    And? You know they can’t do single atom lithography or traditional circuitry either right? The leakage, heat, and electromigration kill the chips within days at room temp if they get anywhere near that level, yes even at 1nm.

    They have to develop a new way of fundamentally building the chips and circuits.

    • blastdoor
    • 12 years ago

    this caused me to laugh 🙂

    • ClickClick5
    • 12 years ago

    1nm = 5 silicon atoms side-by-side.

    • sweatshopking
    • 12 years ago

    hey hey hey!!! how’s it going boys!

    • MadManOriginal
    • 12 years ago

    When was it scheduled for early 2010? The original ticktock announcement way back at the Core 2 launch timeframe? All I can remember about it is that it was Q4 2010 for some parts and Q1 2011 for others as far as ‘official’ announcements. I do recall Intel slipping a bit off the original ticktock schedule but not by more than a quarter and partly because launches are messy with different lines like 1366 and 1156 – Nehalem launched with 1366 but went mainstream with 1156.

    • bcronce
    • 12 years ago

    what do you mean “now two years”?

    Current industrial estimates is that the nearest fab plant is 12-18 months behind Intel already. Close enough to 2 years for me.

    • can-a-tuna
    • 12 years ago

    Well TSMC can’t even handle 40nm.

    • jdaven
    • 12 years ago

    Be careful. If you say sweaty three times, the sweatshopking will appear.

    • jdaven
    • 12 years ago
    • SoulSlave
    • 12 years ago

    Well, not so much, intel’s Sandy Bridge is already a bit late, some say it’ll come in early 2011. Considering it was scheduled for early 2010, it’s not that much of a stretch to consider the possibility of intel’s 22 nm process coming only in 2012.

    Also, it’s really weird that there are currently only a few parts in 32 nm…

    • khands
    • 12 years ago

    Little harsh there, I do wish they’d planned to get the processes out by the end of 2012 (which I’m sure Intel will be at by then) but I don’t think they’ll be the year behind that they are now at that time.

    • Triskaine
    • 12 years ago

    That is incorrect. Intel will introduce Ivy Bridge in a 22nm process in
    Q1 2012 which puts Globalfoundries ~ 1 year behind Intel, just like it always has been in the last 10 years.

    • tejas84
    • 12 years ago

    Lots of talk from GF. Not much action though. Would like to be able to hold an actual working 32nm CPU from AMD.

    At this rate AMD are now two years behind Intel on process.

    Bulldozer should have it’s name changed to Fairy….

    • mesyn191
    • 12 years ago

    Yea they’re getting close to practical limits with current tech.

    There is stuff in the works to get them down to around 10nm or so but after that they can’t really get any smaller with lithography and traditional circuits.

    They better have spintronics or some shit ready to go by then, but “by then” is probably 10 years or less away so I kind’ve doubt it.

    • MadManOriginal
    • 12 years ago

    It won’t be long until process tech stalls, at least for traditional or modified traditional techniques. That might give GF some time to catch up. Going sub-20nm is going to be tough for everyone.

    • Ardrid
    • 12 years ago

    So that puts AMD a good 2 years behind Intel on manufacturing assuming schedules hold. It seems like the gap gets larger with each year. I can’t see this being a good thing for them.

    • Anonymous Coward
    • 12 years ago

    Sounds like an excellent use for oil money.

    • GodsMadClown
    • 12 years ago

    Which means that the investment money is all sweaty?

    • TaBoVilla
    • 12 years ago

    pfff, Abu Dhabi, hottest place I’ve ever been

Pin It on Pinterest

Share This

Share this post with your friends!