Moore's Law

Cool visualization.

Previously, previously.

Tags: ,

8 Responses:

  1. jwz says:

    Didn't expect to see the TI Explorer Lisp Chip hold the lead for almost 3 years!

  2. DrizztVD says:

    It's fake. The law was 2X every 12 months between 1965 and 1975, and the price difference between the 1960s chips and the 2010s chips is 15x more expensive for the latter.

    • Jonny says:

      15x more expensive for something that is literally tens of millions of times faster is a pretty good deal. That said, I am almost 100% positive that your claim that it is 15x more expensive is flatly wrong unless you ignore inflation. Computing prices have dropped like a rock. In fact, part of the reason why "Moore's Law" got a little bit disrupted is because the chip makers stopped focusing so hard on raw power, and started trying to make CPUs more energy efficient and dropping build costs in an effort to reduce server costs.

      It's true that "Moore's Law" isn't real, it's just a guideline. The modern day "Moore's Law" isn't that far off though. Yeah, it isn't doubling transistors density every 12 months. Instead, it's a bit more of a holistic thing that hits power usage and efficiency as often as it hits raw speed and it takes closer to 18 to 24 months as the technology gets more complex and tricky to get a little lower, while at the same time physics becomes more of an asshole making everything harder. They happily cheat to get more computing power and lower power usage through methods other than doubling transistor density. Your current computer chips are functionally 2D skins on the surface of silicon a few tens or hundred of microns thick, so the most obvious place they are going is to simply (it's not simple) go 3D, which is already happening; it's just really hard.

      Getting pedantic about Moore's Law though misses the point. The point is that computing power is on an exponential growth curve. It has been for a very long time, and it continues to be on one. Further, it is likely to stay on one for the foreseeable future (20+ years). It might slow down a little bit here, and go faster there, but it has profound consequences and is most certainly not "fake". Hell, you are living through the consequences right now.

  3. thielges says:

    Some of those later “chips” are actually stacks of dice in a single package. So if they’re counting the transistors in the stacked dram die then it kind of cheats the Moore’s law numbers.

    • Jonny says:

      You are not "cheating" Moore's law by going 3D. You are just continuing the trend by means other than shrinking transistor size so that you can fit more of them onto a nearly flat 2D plane. Going 3D was the next obvious step once making transistors even smaller became increasingly hard and complex. Going 3D is no free lunch though. It is very hard to build those package due to the stress on them, it's very hard to get rid of the heat, and if you screw something up that isn't found until final test, you are throwing away a lot of expensive computing power.

      3D is one of routs that people are taking to push the the "Moore's Law" trend further. Going from a flat 2D skin of computing power a few tens or hundreds of microns thick but 10s of centimeters in the XY, to something that is more 3D, is a continued exponential expanding of computing power.

  4. jwz says:

    Ingrid Burrington:

    The tech industry uses Moore's Law in much the same way that I personally use astrology – as a framework dogmatically applied to justify bad decisions and shitty priorities

  5. Bob says:

    As my mum is constantly reminding me, Moore’s Law is not an equation that happens to describe innovation, it’s a business plan. It’s meant to describe the growing needs of the market so business can focus their efforts.

  • Previously