Is M2's Speed Designed on Purpose?

[zhaoxin]

According to MacTracker, the multi-core point of M2 on a MacBook Pro 13 is 8755. And the same score of M1 is 7369.

We could easily calculate base on this when the score will be doubled. That is ln2/ln(8755/7369) = 4.02. It means for Apple's M chip, it is designed to double its speed in every 4 years.

 

[Gudi]

Multicore speed only depends on how large and expensive you want the chip to be. M1, M1 Pro, M1 Max, M1 Ultra. There you have all your doubling and quadrupling in the same year.

 

[maflynn]

We could easily calculate base on this when the score will be doubled. That is ln2/ln(8755/7369) = 4.02. It means for Apple's M chip, it is designed to double its speed in every 4 years.

I don't think we can extrapolate a linear performance progression as the data doesn't exist yet to support such a hypothesis

 

[Toutou]

We could easily calculate base on this when the score will be doubled. That is ln2/ln(8755/7369) = 4.02. It means for Apple's M chip, it is designed to double its speed in every 4 years.

What.

 

[theorist9]

You can't project a doubling rate from only two data points! More broadly, you can't even determine the functional form of the speed increase from only two data points, since everything fits two data points—linear, exponential (as you're assuming), power law, and any other real-valued functional form.

More broadly, I assume Apple is going for the best balance between performance, efficiency, practicality, and cost it can achieve, rather than trying to target a certain rate of performance increase.

 

[TSE]

Do not ever insult the M2. It is not marginally faster. It is a supercomputer processor from the year 3000 that exists today.

 

[1096bimu]

Yea of course they knew roughly what speed it would be at before they produce the first chip

According to MacTracker, the multi-core point of M2 on a MacBook Pro 13 is 8755. And the same score of M1 is 7369.

We could easily calculate base on this when the score will be doubled. That is ln2/ln(8755/7369) = 4.02. It means for Apple's M chip, it is designed to double its speed in every 4 years.

But ABSOLUTELY NO, you don't plan to increase performance at specific rate. You plan to increase your performance by as much as possible, while maintaining your profit margin.

BTW while you might have grown up in times where processor speed can just double every X years. I wouldn't be surprised if we're already close on the FINAL doubling of single core performance, yes, final, meaning forever, 10 year, 100 years, 1000 years, it will never double again.

And no, quantum computers aren't going to change it, quantum computers are specifically much faster on certain tasks, they don't run geekbench any faster.

 

[Sydde]

Do not ever insult the M2. It is not marginally faster. It is a supercomputer processor from the year 3000 that exists today.

No, the Яблоко supercomputer of 3000 will be the size of a testicle and, using its ΨUI, it will know what you are going to do before you have considered thinking about doing it. The M2 is very good, but it still needs some tweaking before it can fit in your shorts.