Geekbench result for upcoming iPhone

By: Maynard Handley (, July 2, 2018 12:49 pm
Room: Moderated Discussions
Andrei F ( on July 2, 2018 12:06 pm wrote:
> Maynard Handley ( on July 2, 2018 11:21 am wrote:
> > Niels Jørgen Kruse ( on July 2, 2018 10:30 am wrote:
> > > A pretty modest performance increase, so it might be legit.
> > >
> > > linky
> > >
> > > Oddly, it has 128I + 128D for the A12, when the A11 had just dropped to 32I + 32D.
> >
> > Oh god. This really is Lucy and the football.
> > EVERY year at this time we get these faked results. And EVERY year people take them seriously.
> > Come on guys, you're better than that!
> >
> > At the very least the real leaks
> > - will happen a LOT closer to release time
> > - will involve the ENTIRE GB4 set of results, so that we can at
> > least look over them and see whether they make sense as a body.
> >
> > As for this, a faked screenshot is even easier and lamer than a faked submission to the GB website.
> >
> is it that hard to just look up the result in the DB before crying fake?
> Plus the caches are just whatever GB4 likes to read out - they've never been reliable.

Well, that's what happens, isn't it, when you use a BS click-bait link rather than the actual link?

More convincing is that we also have submitted a GPU Compute benchmark

If it is legit, it's clearly not the grand leaps forward of previous designs, just a pathetic ~6% frequency boost and (maybe?) larger caches, or a few other slightly enlarged structures.
Does this mean Apple have hit the end of the line WRT performance improvements?

Perhaps, but I'd suggest another possibility -- this is basically EXACTLY the same CPU as an A11, only "hardened". They took the A11 design, added whatever was believed necessary to protect against every speculation attack they could imagine (presumably things like tagging various structure field by PID or protection level) and the "real" A12 was delayed for a year to get the same treatment.
Then whatever could EASILY be added given the new process (perhaps, legitimately, large L1's were feasible given the clocks [designed for 10nm, not optimized for 7nm] so why not?) were added.

If I'm correct, I expect we will hear about this at the Sept announcement which will go something like "[[Intro about Spectre, TLBleed, etc.]] Apple takes security seriously. Today we're announcing the A12 chip, codenamed Fortified. It's the first chip designed from the ground up to protect against the various issues we've just described, and it does so with zero performance cost, and without requiring any special compiler or OS intervention."

Seem plausible? That's my best interpretation of everything we're seeing, assuming it's legit.
< Previous Post in ThreadNext Post in Thread >
TopicPosted ByDate
Geekbench result for upcoming iPhoneNiels Jørgen Kruse2018/07/02 10:30 AM
  Geekbench result for upcoming iPhoneDoug S2018/07/02 10:44 AM
  Geekbench result for upcoming iPhoneMaynard Handley2018/07/02 11:21 AM
    Geekbench result for upcoming iPhonesomeone2018/07/02 11:54 AM
    Geekbench result for upcoming iPhoneAndrei F2018/07/02 12:06 PM
      Geekbench result for upcoming iPhoneMaynard Handley2018/07/02 12:49 PM
        Geekbench result for upcoming iPhoneanon2018/07/02 06:13 PM
      Geekbench result for upcoming iPhoneDoug S2018/07/02 01:16 PM
    Geekbench result for upcoming iPhoneDoug S2018/07/02 01:15 PM
Reply to this Topic
Body: No Text
How do you spell purple?