By: Alberto (git.delete@this.git.it), March 6, 2015 2:30 pm
Room: Moderated Discussions
Exophase (exophase.delete@this.gmail.com) on March 6, 2015 11:52 am wrote:
> Alberto (git.delete@this.git.it) on March 6, 2015 10:58 am wrote:
> > Or "less" performance vs. A57 clock to clock??? :-).
>
> If you're going to insist that Peter Greenhalgh is flat out lying (and to quite a degree) you'd
> better find some substantial evidence to back it up. Your reasoning so far is very flimsy.
>
> > Qualcomm considers A72 a lower end core than A57, if i have read well on their web site.
>
> Nonsense. You can't compare across different part generations like that. And you're overlooking that Qualcomm's
> 620 is manufactured on TSMC 28HPm while 810 was on 20nm and 820 will be on one of the FinFET processes.
> Not to mention, these different SKUs are artificially segmented by clock speed, meaning it says nothing
> about the full potential of the CPU. All it means is that for this part Cortex-A72 was a more attractive
> choice than Cortex-A57 (which should have been obvious) with no implication that the CPU trades peak performance
> or perf/MHz for perf/W - a step that we've never seen ARM take in succeeding CPUs.
>
Calm down, i see what i can see. I see a medium level SOC equipped with a 1.8GHz A72, only 200Mhz slower than A57 on the brand new powerful and expensive 810, point. No matter the manufacturing process, we are speaking of pure performance and clock speeds.
This is has nothing to do with this engineer and its words.
We not even know if "this" A72 will be the same A72 on track for the 16nm node.
> Alberto (git.delete@this.git.it) on March 6, 2015 10:58 am wrote:
> > Or "less" performance vs. A57 clock to clock??? :-).
>
> If you're going to insist that Peter Greenhalgh is flat out lying (and to quite a degree) you'd
> better find some substantial evidence to back it up. Your reasoning so far is very flimsy.
>
> > Qualcomm considers A72 a lower end core than A57, if i have read well on their web site.
>
> Nonsense. You can't compare across different part generations like that. And you're overlooking that Qualcomm's
> 620 is manufactured on TSMC 28HPm while 810 was on 20nm and 820 will be on one of the FinFET processes.
> Not to mention, these different SKUs are artificially segmented by clock speed, meaning it says nothing
> about the full potential of the CPU. All it means is that for this part Cortex-A72 was a more attractive
> choice than Cortex-A57 (which should have been obvious) with no implication that the CPU trades peak performance
> or perf/MHz for perf/W - a step that we've never seen ARM take in succeeding CPUs.
>
Calm down, i see what i can see. I see a medium level SOC equipped with a 1.8GHz A72, only 200Mhz slower than A57 on the brand new powerful and expensive 810, point. No matter the manufacturing process, we are speaking of pure performance and clock speeds.
This is has nothing to do with this engineer and its words.
We not even know if "this" A72 will be the same A72 on track for the 16nm node.