Michael S (already5chosen.delete@this.yahoo.com) on May 15, 2013 12:35 pm wrote:
> Exophase (exophase.delete@this.gmail.com) on May 15, 2013 6:43 am wrote:
> > RichardC (tich.delete@this.pobox.com) on May 15, 2013 3:26 am wrote:
> > > The conclusion at that link "Hyperthreading neither helps nor hurts when gaming".
> > > It wins a few, it loses a few.
> >
> > The idea isn't to take an average over several games from several years, but to show that there are
> > games that do in fact benefit even from HT at 4C (I guarantee that the effect would be far more pronounced
> > at 2C). Or by using a processor at stock clock speed instead of overclocked to 4.4GHz.
> >
> > The conclusion is that newer games tend to benefit more. Even where HT hurts older games it tends to
> > be mostly in maximum framerate, having little impact on average and even positive impact on minimum,
> > which would tend to be what you'd prefer. Since older games usually have lower minimum requirements
> > to begin with their performance deltas should matter less even if that's what you prefer to play.
>
> Or, may be, on newer engines less time was spent so far optimizing inner loops. So more performance
> is left on table. Part of what left could be picked up by SMT, but only part.
> Other explanations are possible as well.

Other possible explanations for those results:

a) The engine tries to use all logical CPUs, but 8T is simply beyond the engine's sweet spot.

b) The engine only uses 4T even when it has 8 logical CPUs but the OS doesn't do an effective job at distributing them across the 4 cores.