Jump to content

Welcome to FutureTimeline.forum
Register now to gain access to all of our features. Once registered and logged in, you will be able to create topics, post replies to existing threads, give reputation to your fellow members, get your own private messenger, post status updates, manage your profile and so much more. If you already have an account, login here - otherwise create an account for free today!
Photo

Charts, Graphs, and Statistics Of The Future

charts graphs statistics infographics future computing power solar Future Timeline information technology

  • Please log in to reply
490 replies to this topic

#481
Kynareth

Kynareth

    Member

  • Members
  • PipPipPipPipPip
  • 252 posts

Disappointing progress in supercomputing power.

 

I wonder what caused the slowdown. Hopefully things get back on track from 2025-30.

 

 

exascale-supercomputer-2021-future-timel

Cost per transistor is no longer falling like it's supposed to. So, to build faster supercomputers you need more money and more engineers.

 

Look, for example Zen 3 5600X CPU is 20% faster than Zen 2 3600X CPU and costs 20% more in MSRP. DRAM was supposed to drop in price by 2x every year and a half, but it doesn't. PS5 has only 2x more GDDR than PS4.

 

Haswell has 2.25x more transistors per core than Sandy Bridge but is not 2.25x faster per core.

 

And there is a question how far can you scale in parallel?



#482
Kynareth

Kynareth

    Member

  • Members
  • PipPipPipPipPip
  • 252 posts

Historical-Growth.png

 

slowdown.jpg

 

As you can see from these graphs, there seems to be a slowdown in acceleration of economic growth and in the number of electronic components in a computer at constant price (and in single core CPU performance).

 

This doesn't look like The Singularity can happen in 2045.

 

I hope this trend is only temporary and we'll get back on track in the 2030s, because it doesn't look like the 2020s will be any different.



#483
Kynareth

Kynareth

    Member

  • Members
  • PipPipPipPipPip
  • 252 posts

On a more positive note : how standard theoretical GPU performance improves. There is now info on specs of the next flagship AD102 product, so I made a graph which extends to 2030. Around 500x in 20 years. GTX 580 was on 40nm, RTX 3090 is on 8nm and RTX 4090 will be on 5nm. I have my doubts on whether they will be able to continue this trend, because things become hazy and complicated after 5nm. Performance/$ became an issue after 16nm. Their margins are higher and costs of production also higher. RTX 3090 costs 3x as much as GTX 580 1.5 GB version (there was also a 3 GB variant).

 

CHANGES-IN-PERFORMANCE-OF-TOP-NVIDIA-CAR



#484
Yuli Ban

Yuli Ban

    Born Again Singularitarian

  • Moderators
  • PipPipPipPipPipPipPipPipPipPipPip
  • 22,687 posts
  • LocationNew Orleans, LA

Historical-Growth.png

 

slowdown.jpg

 

As you can see from these graphs, there seems to be a slowdown in acceleration of economic growth and in the number of electronic components in a computer at constant price (and in single core CPU performance).

 

This doesn't look like The Singularity can happen in 2045.

 

I hope this trend is only temporary and we'll get back on track in the 2030s, because it doesn't look like the 2020s will be any different.

I've been starting to stress to people to not expect "the Singularity" to be a thing, just because it's impossible for computers to exponentially increase in power perpetually. Computing power still has a very long way to go before we top out, but we'll definitely hit limits. 

 

The Singularity as perpetually increasing computing power will almost certainly never be a thing.

 

The Singularity as a point where human intelligence is surpassed by artificial intelligence by insurmountable levels, however, will definitely be a thing with a level of confidence approaching certainty. And yes, I do think it's possible, even probable, we'll reach this by or even before 2045.

 

Probably the biggest misconception of Kurzweil's predictions I've seen on the Internet is the idea that he's banking everything on Moore's Law holding indefinitely, hence why Singularitarians are so desperate to prove that it is in fact not faltering or even entirely dead. For a lot of people, it's not "Moore's Law" but "Moore's Law of Accelerating Returns." Therefore Moore's Law's failings and complete breakdown in the past decade means that Kurzweil must be wrong, that he looked at a single chart of progress and promised the universe out of it, and that the final demise of Moore's Law means the Singularity is forever canceled.

 

In fact, Kurzweil himself outright stated that Moore's Law was going to give out, and even predicted it'd die completely as a trend in the 2020s. The real argument is "What takes over from Moore's Law?"  Kurzweil seems to believe it's 3D computing, and it very well may be. It may be something else.

 

Like I said, there's still quite a bit of room until we reach the top. Theoretically, we could have zettaflop mechanical computers the size of a sugar cube: 

 

It really is all going to depend on how we progress from here.


And remember my friend, future events such as these will affect you in the future.


#485
waitingforthe2030s

waitingforthe2030s

    Member

  • Members
  • PipPipPip
  • 51 posts
  • LocationOrion Arm, Milky Way Galaxy.

 

Historical-Growth.png

 

slowdown.jpg

 

As you can see from these graphs, there seems to be a slowdown in acceleration of economic growth and in the number of electronic components in a computer at constant price (and in single core CPU performance).

 

This doesn't look like The Singularity can happen in 2045.

 

I hope this trend is only temporary and we'll get back on track in the 2030s, because it doesn't look like the 2020s will be any different.

I've been starting to stress to people to not expect "the Singularity" to be a thing, just because it's impossible for computers to exponentially increase in power perpetually. Computing power still has a very long way to go before we top out, but we'll definitely hit limits. 

 

The Singularity as perpetually increasing computing power will almost certainly never be a thing.

 

The Singularity as a point where human intelligence is surpassed by artificial intelligence by insurmountable levels, however, will definitely be a thing with a level of confidence approaching certainty. And yes, I do think it's possible, even probable, we'll reach this by or even before 2045.

 

Probably the biggest misconception of Kurzweil's predictions I've seen on the Internet is the idea that he's banking everything on Moore's Law holding indefinitely, hence why Singularitarians are so desperate to prove that it is in fact not faltering or even entirely dead. For a lot of people, it's not "Moore's Law" but "Moore's Law of Accelerating Returns." Therefore Moore's Law's failings and complete breakdown in the past decade means that Kurzweil must be wrong, that he looked at a single chart of progress and promised the universe out of it, and that the final demise of Moore's Law means the Singularity is forever canceled.

 

In fact, Kurzweil himself outright stated that Moore's Law was going to give out, and even predicted it'd die completely as a trend in the 2020s. The real argument is "What takes over from Moore's Law?"  Kurzweil seems to believe it's 3D computing, and it very well may be. It may be something else.

 

Like I said, there's still quite a bit of room until we reach the top. Theoretically, we could have zettaflop mechanical computers the size of a sugar cube: 

 

It really is all going to depend on how we progress from here.

 

The problem with the concept of the Singularity is that there's no one definition of what is it. Is it when AI keeps improving itself until it becomes super intelligent? Is it when computers just keep getting better, forever? Is it when AI invents so much stuff we can't keep up? I've heard different definitions and no one can seem to clarify which one they're talking about.


This is Scatman's world, and we're living in it.


#486
Yuli Ban

Yuli Ban

    Born Again Singularitarian

  • Moderators
  • PipPipPipPipPipPipPipPipPipPipPip
  • 22,687 posts
  • LocationNew Orleans, LA

The "Singularity" actually has a fairly concrete definition: it means humans are unable to keep up with the rate of change in technology (and the technological change thus has similar osmotic effects throughout society), making it impossible for humans to fully comprehend in any minute detail the changes that are happening, even if we can guess using things like science fiction or fantasy some broadstrokes of what may be possible.

 

The ambiguity lies in why and how this is happening.


And remember my friend, future events such as these will affect you in the future.


#487
Kynareth

Kynareth

    Member

  • Members
  • PipPipPipPipPip
  • 252 posts

I have my serious doubts whether supercomputers will ever reach 1 zettaflops in double precision on current paradigm of computing. If I had to bet, I would choose that they won't.

 

1 exaflops is not only very achievable, but may be reached this year (Folding@Home doesn't count). 10 exaflops should also be possible on current silicon integrated circuits paradigm, which started in the 1970s (using 2nm chips?). 100 exaflops will however pose a huuge challenge.

 

I don't know how much exactly is necessary for whole real-time human brain simulation, but for every kind of simulations the faster the better.

 

In all books about the Singularity, there is talk about economic developments, growing GDP/capita and computing performance, memory etc. There's no Singularity with economic and hardware stagnation, that's for certain.



#488
Yuli Ban

Yuli Ban

    Born Again Singularitarian

  • Moderators
  • PipPipPipPipPipPipPipPipPipPipPip
  • 22,687 posts
  • LocationNew Orleans, LA

mx5e41sfjp961.png


And remember my friend, future events such as these will affect you in the future.


#489
Yuli Ban

Yuli Ban

    Born Again Singularitarian

  • Moderators
  • PipPipPipPipPipPipPipPipPipPipPip
  • 22,687 posts
  • LocationNew Orleans, LA

v23w0zlc87t31.jpg

 

It's disturbing how accurate this is


And remember my friend, future events such as these will affect you in the future.


#490
Kynareth

Kynareth

    Member

  • Members
  • PipPipPipPipPip
  • 252 posts

The blue line is the real performance and the orange line is how people like Kurzweil, Moravec or Vinge thought things would go. Michio Kaku thought that CPUs will continue to double in speed every 18 months until 2030s, when the trend would slow down to maybe doubling every 36 months.

 

In the mid 2030s we will very much need a completely new paradigm of computing. Like self-assembling 3D molecular lattices operating on photons.

 

1970-2030-FLOPS-in-PCs.png



#491
Yuli Ban

Yuli Ban

    Born Again Singularitarian

  • Moderators
  • PipPipPipPipPipPipPipPipPipPipPip
  • 22,687 posts
  • LocationNew Orleans, LA

RNA-vaccines-and-how-they-work.png?ssl=1


And remember my friend, future events such as these will affect you in the future.






Also tagged with one or more of these keywords: charts, graphs, statistics, infographics, future, computing power, solar, Future Timeline, information, technology

0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users