- Saturday, 21 November 2009
- INQ Mobile
- RSS
You call it moving the goal posts -- I call it dynamic e-business
GlobalFoundries goes wafer waggling
GLOBALFOUNDRIES was showing off some shiny new 32nm and 28nm wafers today in Taipei, pre Computex, having apparently having to secretly smuggle them into the country in a plastic bag.
Jon Carvill, head of communications at GloFo also showed the INQ a 45nm six-core Istanbul wafer which he claimed was the company's "smoothest ramp and conversion yet," boasting a holy grail of high performance and low leakage thanks to technology like immersion lithography and APM (automated precision management).
The 45nm Istanbul wafers are already being churned out by GloFo, which is claiming the offering to be "AMD's most complex and largest microprocessor to date."
The backslapping doesn't stop there either, with the AMD spin-off saying it has delivered a full quarter ahead of schedule and that no other foundry in the world is making processors this complex today, in high volume, something TSMC and UMC would vehemently contest, but seeing is believing.
Carvill told the INQ the 45nm process was already pumping out "processors in volume production at mature yields in record time." he added, "We look to build on this with 32nm and 28nm to the benefit of AMD and prospective third party customers."
As for 32nm, the firm reckons development is still on track and that 32nm pilot lines are already running in Dresden with test chips. "We'll be able to accept bulk designs in 2H09 with the ability to ramp production in 1H10," Carvill told us.
As we oohed and ahhed, Carvill procured a shiny 28nm wafer deftly from behind his ear, possibly Glofo's most strategically important node to date.
The main benefits of 28nm will purportedly be felt in the area of power efficiency and will be particularly valuable in terms of the graphics and wireless markets.
Carvill was confident the 32nm "experience" would "ensure a very smooth transition to 28nm with minimal risk for our customers," adding "We're already running development SRAM chips already in Upstate, NY with IBM and plan to accept customer designs and begin ramping production in 2010"
"With 28nm, we'll be on second generation high-k and third generation immersion, which we believe will provide an optimal balance of performance and power efficiency for our customers," he concluded. µ
Share this:
ShareBusiness Intelligence Software
Customer Relationship Management (CRM)
Enterprise Accounting Software
Enterprise Resource Planning (ERP)
Enterprise Systems Management
If GloFo can deliver 32nm in 1H10, does this mean that they are catching up with Intel?
(with Intel's 32nm due in late '09)
They are catching up with implementing though in reality they were never behind in terms of technology. Intel with, what is it, 364 fab's can just incredibly easily shut one down to change over to a new process with ease as they don't need to run any of their fabs at max capacity.
Amd have the problem of limited fabs and downtime hurts them far more. I think the dell deal really hurt AMD as much as it helped in so much as, they suddenly had a massive contract where they HAD to provide a certain number of chips, the time involved in shutting a fab down and switching to a new process is incredibly strained and due to their commitments they had to delay and delay till the yields were so good they could cope with the downtime. Thats seemingly the main reason for AMD's slow transition over to 65nm originally and 45nm though that was somewhat quicker.
Now they honestly have the money to push through new process's faster and accept the loss in production, deep pockets go a LONG way in these situations. LIkewise when AMD have their NY fab up and running it will get easier and quicker to switch to new processes.
Ok 3+1- 4770 cards into 4 FULL 16X pci-e, say 4.6 GHz/s nehli, & your get 64 2.0-16X LANES Potential for NewMain or PENTIUM II true W/ Hypertheading & Boost. This IS Good Vision.
Yet After Grape Jelly Wears off, Its Still Old Kitchen Table & yes,yes, Fart My Asce OFF, Dual GPU Models to Follow.
Alleaged multi Core Marketier, In colorful Diamond & Silk. Trade Your Ivory TUSKS For BenchMark.
Balance Hasn't BEEN Taught In School of HARD Knocks, Assuming People Want Much LESS Than Paid for, Yet, Some Ultee', HIGH I.Q. & que one, why is it so HARD for oriental Mind to UnderStand?
Then: Kiddnap BoToxEta,Too.
These GOALS Are MUCH More Important As Nm Scale reachs Absolute Limit. Remember Each Cor of Ultee' Pat PEND, WorkStation Main superProcessor multi core is 2.0 Ghz/s, Y2K, limit often fabled. Seldom, if ever, exceeded. So Thats it Folks!
From FlaiLIng ARMS of BUD.
Good to hear that GloFlo's issues were more with capital access than with talent.
And congrats to them on progress at 32nm and 28nm. What's next? 15nm?
Cut me a slice of that diecious New York Style sweet-n-sour Asian pizza and secretly snuggle it out of old Taipei. Just keep it haut.
They still talk about going beneath 10 nm, apparently the distance between silicon atoms is 0.3 nm so at what point it not go smaller?