3GHz Cortex A9, in a tablet

krumme

Diamond Member
Oct 9, 2009
5,956
1,595
136
Well 3GHz on an a9 is damn impressive :), but does it really say something usefull about FD-SOI? - isnt this more like ppt slides?
 

NTMBK

Lifer
Nov 14, 2011
10,380
5,539
136
Well 3GHz on an a9 is damn impressive :), but does it really say something usefull about FD-SOI? - isnt this more like ppt slides?

Have you seen the A9 hit those speeds on any other process? ;)
 

Face2Face

Diamond Member
Jun 6, 2001
4,100
215
106
So the A9 is a 40nm chip? Not too bad, pretty much being equal to a 1.8Ghz A15 - And i am sure a lot cheaper to produce.
 
Last edited:

NTMBK

Lifer
Nov 14, 2011
10,380
5,539
136
So the A9 is a 40nm chip? Not too bad, pretty much being equal to a 1.8Ghz A15 - And i am sure a lot cheaper to produce.

Nope, ST-Ericsson's 28nm FD-SOI- the same process which is getting ported to Global Foundries, too.
 

Exophase

Diamond Member
Apr 19, 2012
4,439
9
81
Raw wafer cost is higher, but number of steps involved in producing a chip is reduced, giving a roughly equivalent net cost. http://semiaccurate.com/2013/01/17/st-ericksson-puts-out-a-2-5ghz-phone-chip/#.USzYlKUqysY

Thanks, hadn't seen it. I'm taking it with some salt though (as I do all company figures/promotional materials), I think there's a reason that not a lot of the industry has heavily embraced FD-SOI.. David Kanter seems to think it's not that economical for high volume..

Hmmmm.. Then why not just go A15? I guess they wanted to prove the can design chips too?

This was probably in development for a long time.. ST-E aren't exactly quick in time to market. They've had an A15 SoC announced since forever ago; don't know if it'll use FD-SOI or not.

We really don't have a good idea of what the perf/W curve looks like for this vs Cortex-A15 on the same process and Cortex-A9s and A15s on other 28/32nm processes. This may use less power at the lower end of the curve, at the very least, and of course it uses less area. So it might be a better fit for some applications. It's all about compromise.

If it's true that it delivers the same power consumption at 2.5GHz as the previous one did at 1.85GHz on 40nm bulk that's pretty impressive.
 
Last edited:

Khato

Golden Member
Jul 15, 2001
1,251
321
136
If it's true that it delivers the same power consumption at 2.5GHz as the previous one did at 1.85GHz on 40nm bulk that's pretty impressive.

It's quite possible, sadly the one good STE presentation of FD-SOI only compares dynamic power versus frequency against a bulk 28nm LP process. Guess I'll go ahead and post the link to that presentation one more time for reference - http://www.soiconsortium.org/fully-... below for extremely power-efficient SoCs.pdf Page 27 has the aforementioned chart which shows that the 28nm FD-SOI part has a dynamic power of approximately 2.3 watts when run at 2.5 GHz. Extrapolating the curve out to 3 GHz would have it somewhere north of 4.5 watts. Yes, there's a good reason why they don't mention power usage with these overclocking experiments.