Wow that is one more reason why your system should be open.
All the closed systems (read windows) have had precedents of working closely with their partners (read OEMs) to achieve this hardware software harmony (which Steve was a big fan of BTW). But if your system is open anybody who wishes can be your partner. Amazing.
I think opening their system is one thing Google did right with Android over Windows.
Cortex A7 is more proof that Intel can't "enter" the smartphone/tablet markets. I like that ARM isn't going to pursue ever more performance just to disrupt Intel in the low-end of their market (and higher later) while maintaining their low power consumption at the same level (Cortex A9, Cortex A15, etc), but they also keep making chips that use even less power (Cortex A7) for people that want as much battery life as possible, rather than try to approach PC level performance with their smartphone.
> As important as the launch of a new core design is, ARM’s heterogenous multiprocessing plans are perhaps the biggest news to come out of Wednesday’s event.
I find it interesting mostly in its validation of NVidia's Kal-El strategy (the article also quotes the PS3, but I don't think it applies: PS3's heterogenous CPU also uses very disparate architectures and behaviors, it's not a similar core with a lower load envelope)
The difference is that you can't run the same executables on all of the PS3's cores, the SPUs don't even have the same instruction set, much less stuff like memory management units.
The big difference between this and Kal-El is that Kal-El uses a fifth core on the same architecture but a different (high Vt) process for the handoff, with less potential for power savings. Kak-El also takes 2mS to make a transition, rather than the 20uS that ARM is advertising for A7/A15 systems with structures in their caches built specifically to enable this.
On the other hand, Kal-El will be coming out long before we see A7/A15 hybrid systems.
Well, sort of. The ordering depends on the frequency each of the cores is run at and that depends on the process node as well as the architecture. The A7 is only expected to get 1.9 DMIPS/MHz while the A8 gets 2.0, but the A7 is designed for much faster clock speeds so it ought to perform better than even most A9s.
EDIT: I just submitted an Anandtech article that explains all this much better.
[+] [-] sharmajai|14 years ago|reply
All the closed systems (read windows) have had precedents of working closely with their partners (read OEMs) to achieve this hardware software harmony (which Steve was a big fan of BTW). But if your system is open anybody who wishes can be your partner. Amazing.
I think opening their system is one thing Google did right with Android over Windows.
[+] [-] nextparadigms|14 years ago|reply
[+] [-] masklinn|14 years ago|reply
I find it interesting mostly in its validation of NVidia's Kal-El strategy (the article also quotes the PS3, but I don't think it applies: PS3's heterogenous CPU also uses very disparate architectures and behaviors, it's not a similar core with a lower load envelope)
[+] [-] Symmetry|14 years ago|reply
The big difference between this and Kal-El is that Kal-El uses a fifth core on the same architecture but a different (high Vt) process for the handoff, with less potential for power savings. Kak-El also takes 2mS to make a transition, rather than the 20uS that ARM is advertising for A7/A15 systems with structures in their caches built specifically to enable this.
On the other hand, Kal-El will be coming out long before we see A7/A15 hybrid systems.
[+] [-] aphexairlines|14 years ago|reply
ARMv7 Cortex A15 > ARMv7 Cortex A7 > ARMv7 Cortex A9 > ARMv7 Cortex A8 > ARMv6 ARM11xx > ARMv3 ARM7xx
[+] [-] Symmetry|14 years ago|reply
EDIT: I just submitted an Anandtech article that explains all this much better.
EDIT2: Oops, someone else had already submitted it and now its off the top of the "new" page. How this article beat it out is beyond me. Anyways, you can find it here: http://www.anandtech.com/show/4991/arms-cortex-a7-bringing-c...
[+] [-] lgeek|14 years ago|reply
[+] [-] wmf|14 years ago|reply
[+] [-] masklinn|14 years ago|reply