Michael S (already5chosen.delete@this.yahoo.com) on May 4, 2017 7:56 am wrote:
>
> But probably more important factor is that, if I am not mistaken, A9 is a 1st attempt
> of Sophia team at superscalar design. On the other hand, A53 is a product (or should
> we call it a masterpiece?) of much more experienced (Cambridge ?) team.

A9 was very successful, but it was a horribly flawed core. The L2 interface was laughably bad (not just the small lines, the L2 was behind the slow bus interface, iirc), and quite frankly, the fact that a later and much better designed in-order core can just barely beat it is very much not an argument for in-order and against OoO.

A12 (A17? I have no idea what happened there with the crazy naming) would probably be a much better example of an A9-like core that fixed some of the big faults, but I think the core was largely a commercial failure, simply because of timing (by the time it came out, people wanted 64-bit cores).

The crazy GB hardware crypto insistence also effectively killed anything without the AES hardware, so there was a double whammy for that line.

Although it's also possible that I just didn't know what to look for.

Linus