GLOBALFOUNDRIES shows hope with 2.5GHz ARM SoC at 28nm

by Alistair Lowe on 15 December 2011, 09:43

Tags: GLOBALFOUNDRIES, ARM

Quick Link: HEXUS.net/qabagf

Add to My Vault: x

Following on from several stories hinting that GLOBALFOUNDRIES had been responsible for the poor yields and performance at AMD; the firm yesterday announced that it had successfully taped out a test chip, based on a dual-core ARM Cortex-A9 running at 2.5GHz, an industry first.

The chip test was conducted at the Dresden Fab 1 in Germany and based on the firm's 28nm High Performance technology; GLOBALFOUNDRIES hopes to translate this success to its High Performance Plus technology, which will allow for a low, 0.85V operating voltage.

The company has attributed this success to close collaboration with ARM in the development of a Technology Qualification Vehicle (TQV), which essentially simulates the design of a Typical Cortex-A series SoC, enabling standardised and close-to-reality testing, ensuring that the firm's fabrication process successes will translate well over to customer ARM designs.

Further to this announcement, the company also revealed that with its new TQV as a template, it has successfully taped-out silicon on its upcoming 20nm fabrication process. GLOBALFOUNDRIES claims that its 20nm TQV is designed to deliver up to 35 per cent greater performance at half the power consumption of an equivalent 28nm design. Production of any final silicon aimed for market on the 20nm node is a long way off, though this news shows that the firm is taking appropriate first steps to ensuring a smooth transition for clients.

We've all heard bad press coming from GLOBALFOUNDRIES lately, the firm is quite open in its intent to show to potential customers that, thanks to its continued close collaboration with ARM, those coming to it with ARM Cortex designs will not be steered down a path of low-yields and poor design performance. Time will reveal all.



HEXUS Forums :: 5 Comments

Login with Forum Account

Don't have an account? Register today!
Hexus
duel-core

….
ddy
….

fight!

:)
DDY
….

bah :Oops:
Creating one high-performance test chip on a new process isn't that big a deal - what we need to know is how many they *tried* to create in order to get a working one. One good chip out of an entire wafer isn't such a triumph…
scaryjim
Creating one high-performance test chip on a new process isn't that big a deal - what we need to know is how many they *tried* to create in order to get a working one. One good chip out of an entire wafer isn't such a triumph…

True, though to squeeze 2.5GHz out of the chip does indicate that they were able to push the design quite hard, indicating a high quality of silicon and thus high yield at lower performance, though that's just trend and certainly not a rule so perhaps indeed yields could also be low, but I think less likely to be the case.