TSMC test Cortex A9-chip with a clockspeed of 3,1GHz

[Bosbeetle]

I just saw this article in a dutch tech blog : http://tweakers.net/...komma-1ghz.html


I was just wondering how much they can squeeze out the cortex system, it seems that they are doing a good job.


Just to let you guys know, but probably most of you already read this news.


it also mentions the work TI is already doing on cortex A15, sounds yummy

 

[Exophase]

Goes to show that attainable clock speed varies a lot depending on process parameters. This is probably using an HP process that optimizes switching speed over leakage and operates at a voltage that's well past the efficiency knee for the design.


The other side of this is how fast the L2 cache is, if their test chip even has L2 cache. They probably wouldn't get away with trying to clock a Cortex-A8 like this because of its tightly coupled L2.

 

[Bosbeetle]

is it less coupled in the A9?

 

[Exophase]

is it less coupled in the A9?

Yes. L2 cache is external. It's on the other side of the AXI bus. An A9 doesn't need to have L2 cache at all. NEON is also optional. Would not surprise me at all if this chip lacked that.