blaine (myname.delete@this.acm.org) on November 19, 2020 5:18 pm wrote:
> Doug S (foo.delete@this.bar.bar) on November 17, 2020 12:21 pm wrote:
> > Dummond D. Slow (mental.delete@this.protozoa.us) on November 17, 2020 11:18 am wrote:
> > > Which also tells you where the biggest threat from Apple is. It pretty much caught
> > > up with state of the art x86's single core performance AND has process advantage.
> > > It could shoot ahead in performance in two areas if it chose to:
> > > 1) SMT as discussed. Not having SMT leaves massive multithread performance
> > > gains (end energy efficiency gains, more importantly) on the table.
> >
> >
> > Does that really matter all that much for desktop use? Maybe it matters for
> > the Mac Pro, but that's just not a big enough piece of the overall Mac market
> > (let alone the overall 'Apple SoC' market) for them to add SMT, IMHO.
> >
> > Now if Apple was going build their own servers (for iCloud, Siri, maybe Search if antitrust issues
> > force Google to quit paying Apple for being the default search in Safari) then adding SMT would
> > make more sense since they can amortize the extra design/verification work across more units.
> >
> >
> > > 2) AMD and to a bit less degree Intel squeeze the single core
> > > frequency of the core during single-thread boosting,
> > > with very high voltage and advanced power management so that
> > > they can pretty much run it as fast as the silicon
> > > allows and as high or even higher than manual overclocking can reach. This is why the power consumption in
> > > their single core turbo boosts is so high (it does dial way lower during all-core load clocks).
> > > Apple seems to only have simple turbo that drops clocks a bit
> > > on multicore load, but it is only a small difference.
> > > That implies Apple could extract a lot of frequency if it
> > > went as advanced on power management and aggressive
> > > on turbo as AMD does. I don't know how high it could go - the low power it exhibits suggest there is a lot
> > > of headroom, but perhaps the wide engine just couldin't handle much more due to timing even if it doesn't
> > > have high power output. But some potential Apple has not tapped yet is likely there.
> >
> >
> > I agree that Apple clearly has some room to increase performance with binning/turbo, but again how much
> > sense does that effort make for their market? It would be nice for those of us watching on the sidelines
> > and wanting to see a real performance war, but in reality Apple is competing with the last version of
> > their own products not with what you can buy from Dell, and the M1 is already faster for ST than any
> > x86 Mac so they don't NEED to do this. Maybe they will explore it in the M2 or M3, I hope they do, but
> > I wouldn't be surprised if they feel their design effort is better spent elsewhere.
> >
> > Heck, Apple could get faster by simply not designing for low power. According to TSMC using HPC cells
> > gets you a 10% boost, and other tricks you can do when power draw is less important can get 10% more.
> > An M1+ clocked at 3.8 GHz would be one hell of a beast, though if it used 2x more power per core it
> > would really only be appropriate for a Mac Pro - every other Mac is quite cooling constrained due to
> > the form factor. Sure, they COULD pay for a separate tapeout when the Mac Pro starts at $5K, but they'd
> > only be able to fit half as many cores at the same power budget. So they probably will use low power
> > transistors there despite the potential for faster ST performance, EVEN if they do a separate design
> > for the Mac Pro (i.e. based on being able to leverage it for internal server use also)
>
> Doug is right. The value of multi-threading is workload dependent.

Designing for higher-perf libraries and other such things are not exclusive with trying SMT out, btw. They could benefit from both at the same time.