bit-tech.net

Intel claims Moore's Law is alive and well

Comments 1 to 10 of 10

Reply
edzieba 30th March 2017, 11:02 Quote
Quote:
we wouldn’t be at all surprised if Intel shifts its stance on the desktop to meet the challenge from AMD
I don't think we'll see the result of any 'stance shifting' for many years. Architecture development takes a [n]LONG[/b] time, so the next few nodes and generations we'll see will be things that were in development long ago. Remember that it took AMD a decade to respond to Core 2's smash hit desktop line, and 6 years if you count from Bulldozer/15h's flop. Even if you assume AMD did nothing whatsoever on the development of Zen until after Keller returned, that's still a 5 year lag time.
Birdy 30th March 2017, 15:18 Quote
This is just Intel marketing, no? I think some critical though could be applied. If this is an Intel sponsored article that ought to be mentioned.

Abstract:
Intel managed to come up with a few tricks to make transistors smaller. Turns out making really small stuff is really complicated, such much so that that said tricks need a fancy, meaningless marketing name like Hyper-scaling. Intel says it is keeping up with Moor's law.
Intel marketing slides included.

Oh, and no-one mention Samsung.
( Debate topic: Is Samsung's 10nm node equivalent to Intel's? )
somidiot 31st March 2017, 20:33 Quote
I don't know that I agree, there's not that big a difference in performance (15% in most benchmarks) between the same price point of 5 years ago and today's processors, looking at the i5-2400 and the current i5-6400. So Moore's law is dead to me. If they've been getting faster results, etc. then they're keeping their hands on that performance to themselves, greedy *******s.
Maki role 31st March 2017, 21:23 Quote
Quote:
Originally Posted by Birdy
This is just Intel marketing, no? I think some critical though could be applied. If this is an Intel sponsored article that ought to be mentioned.

Abstract:
Intel managed to come up with a few tricks to make transistors smaller. Turns out making really small stuff is really complicated, such much so that that said tricks need a fancy, meaningless marketing name like Hyper-scaling. Intel says it is keeping up with Moor's law.
Intel marketing slides included.

Oh, and no-one mention Samsung.
( Debate topic: Is Samsung's 10nm node equivalent to Intel's? )

Well Antony did fly out there and back again to attend this press event, seems a bit complex for just a simple Intel paid job dontcha think?
rollo 1st April 2017, 16:45 Quote
Birdy no its not. The below post explains it easier between Intel and TSMC.

Understanding how foundries name their process nodes: Intel’s 10nm versus TSMC 10nm and beyond

The answer lies in the fact that Intel’s naming standards and TSMC’s naming standards are drastically different. Intel’s processes use the same backbone as the advertised node (a 14nm process will use a 14nm backbone) while as all pure play foundries use a mixture of process technologies. TSMC’s 16nm FinFET tech for example uses a 20nm backbone (BEOL). So it is almost a certainty that they will be using a 14nm BEOL for their ’10nm finfet’ node.

Intels Transitor gate pitch is 14nm smaller than Samsung at the same marketed nodes. Samsung is 84nm Intel is 70nm thats at 14nm node size. Intels old 16nm is still smaller technically than Samsungs 14nm stuff.
GreekUser 2nd April 2017, 12:24 Quote
Intel claims we can't read a gad-damn diagram.

1st picture
Between 2008-2010. Shows 1.75 years for an 2.3x increase. That's 1.31x increase per year.
Between 2014-2018. Shows 3.50 years for an 2.7x increase. That's 0.77x increase per year.
A good 40% slow down. Almost halved.

2nd picture
It shows about the same thing. At the begging gradient is steep. The angle is bigger than the average. Which means that it has to be smaller as we move to the right. And indeed, not only the angle decreases but the decrease continues all the way to the last section (2014-2018) where is the smallest.

Honestly, I am insulted.
Corky42 3rd April 2017, 08:35 Quote
Moore didn't say anything about performance increases, that was David House, Moore said that the number of transistors on a given chip could double every year, something he revised a decade later to two years.
Gareth Halfacree 3rd April 2017, 09:21 Quote
Quote:
Originally Posted by Corky42
Moore didn't say anything about performance increases, that was David House, Moore said that the number of transistors on a given chip could double every year, something he revised a decade later to two years.
To be specific, Moore's "Law" was this short extract from a longer editorial written for Electronics Magazine in 1965:
Quote:
Originally Posted by Gordon Moore
The complexity for minimum component costs has increased at a rate of roughly a factor of two per year. Certainly over the short term this rate can be expected to continue, if not to increase. Over the longer term, the rate of increase is a bit more uncertain, although there is no reason to believe it will not remain nearly constant for at least 10 years.
Corky42 3rd April 2017, 09:48 Quote
Yea i was going to quote that but worried i didn't understand the complexity cost bit. :o

One part of that article i found interesting though was when he wrote something like 65k transistors would be feasible, I've not looked up where we are today but I guess that 65k number is an understatement half a century later.

Things like that always make me smile as what must have seemed like an impossible task at the time turned out to be a massive understatement. :)
Gareth Halfacree 3rd April 2017, 10:01 Quote
Quote:
Originally Posted by Corky42
Yea i was going to quote that but worried i didn't understand the complexity cost bit. :o
"Complexity for cost" basically means how complex a chip you can have for a particular cost. Year one you pay $5, you get a five-thousand transistor chip; year two you pay the same $5 but get a ten-thousand transistor chip; year three you're still paying the same $5 but getting a twenty-thousand transistor chip.

Basically, anyone tracking Moore's Law shouldn't be tracking the absolute number of transistors on a chip, but the transistors-per-dollar - 'cos that's what he was originally talking about.
Quote:
Originally Posted by Corky42
One part of that article i found interesting though was when he wrote something like 65k transistors would be feasible, I've not looked up where we are today but I guess that 65k number is an understatement half a century later.
Well, Nvidia's latest graphics chips have 12 billion transistors on 'em...
Log in

You are not logged in, please login with your forum account below. If you don't already have an account please register to start contributing.



Discuss in the forums