|
wtf happened? computers are supposed to get twice as good every year but i was in walmart yesterday and all the computers are just the same as last year, this is bullshit
|
# ? Jan 27, 2015 15:35 |
|
|
# ? Mar 19, 2024 04:09 |
|
i dont think you're going to find the twice as good computers in a walmart
|
# ? Jan 27, 2015 15:37 |
|
It broke about 10 years ago. Another 10 years or so and computers won't ever get faster again until someone finds a new way of making them.
|
# ? Jan 27, 2015 15:38 |
|
Serious answer, Moores law is about doubling transistor density every 18 monthes, which is still happening. The problem is for various reasons it is difficult to get extra performance out of typical CPUs by adding transistors. GPUs and mobile SOCs are still seeing rapid performance gains.
|
# ? Jan 27, 2015 15:39 |
|
Germstore posted:Serious answer, Moores law is about doubling transistor density every 18 monthes, which is still happening. But going to run into a serious wall into a few years.
|
# ? Jan 27, 2015 15:42 |
|
Submerging Computers In Hyperspace So Their Calculation Speeds Aren't Limited By The Speed Of Light
|
# ? Jan 27, 2015 15:42 |
|
Germstore posted:Serious answer, Moores law is about doubling transistor density every 18 monthes, which is still happening. The problem is for various reasons it is difficult to get extra performance out of typical CPUs by adding transistors. GPUs and mobile SOCs are still seeing rapid performance gains. so the robot apocalypse/nerd rapture is still on schedule? ok good I was worried
|
# ? Jan 27, 2015 15:44 |
|
Rutibex posted:so the robot apocalypse/nerd rapture is still on schedule? ok good I was worried I'm building a drone that will kill me live over YouTube. With any luck, I'll be the first of many.
|
# ? Jan 27, 2015 15:48 |
|
Germstore posted:Serious answer, Moores law is about doubling transistor density every 18 monthes, which is still happening. The problem is for various reasons it is difficult to get extra performance out of typical CPUs by adding transistors. GPUs and mobile SOCs are still seeing rapid performance gains. It's been a while since I followed this. Isn't the problem that conventional means of manufacture are getting limited, in that we can't use our normal tools and processes to make smaller and smaller circuits? Instead the manufacture needs to switch over to a nano-scale level, building circuits atom by atom from the ground up instead. I remember IBMs advanced storage division was in the preliminary stages of this a good 18 years ago, demonstrating proof of concept stuff, by being able to lay out a dozen (hydrogen? carbon?) atoms to spell the word "HI" and having them hold their position. Anyone know if there's been any advances in nano-scale manufacture in the last couple decades?
|
# ? Jan 27, 2015 15:53 |
|
Tsinava posted:Submerging Computers In Hyperspace So Their Calculation Speeds Aren't Limited By The Speed Of Light Gets me hard. Hyperspace ocean similes are the sickest.
|
# ? Jan 27, 2015 15:56 |
|
consoles killed moores law. theres no point doubling ur ram if all youre gonna play is COD modern warfart (satire)
|
# ? Jan 27, 2015 15:57 |
|
Internetjack posted:It's been a while since I followed this. Isn't the problem that conventional means of manufacture are getting limited, in that we can't use our normal tools and processes to make smaller and smaller circuits? Intel seems to be pretty confident that they'll get down to 5nm. Current state of the art is 14. So we can probably expect another 10x increase in transistor density, and then possibly nothing after that. There are several more important problems. When transistor density doubles power usage only drops by the square root of 2 (or something like that), so we can't power all of those transistors and keep the clock frequency the same. GPUs have an easier time because they can work well with lower clock frequencies. That's why even intel is starting to use a large portion of its die space on integrated graphics. It's too difficult to use up all that space for CPU cores without making a chip that would melt itself.
|
# ? Jan 27, 2015 16:14 |
|
most of the serious investment for stuffs is going in to capital management compustuffs like gpus for financial simulations http://www.moneyscience.com/pg/events/Admin/read/651826/gpus-monte-carlo-simulation-and-kooderive-with-professor-mark-joshi theres a lotta this crap floatin aboot
|
# ? Jan 27, 2015 16:15 |
|
Moore's law continues on, but an increase in transistor count only goes so far. Eventually we'll have application specific ASICs for everything, but we'll probably have to change the way computers work to see increases beyond that.
|
# ? Jan 27, 2015 16:31 |
|
quantum computers are already a thing they're just not yet practical or ready for market yet a moot point after the gamma ray burst in august destroys all life on earth though, oh well
|
# ? Jan 27, 2015 16:55 |
|
it's not really a law, it's more of an observation op. nothing happened to it, it's just not valid anymore since we've reached the physical limits of the materials we're making transistors out of at the moment.
|
# ? Jan 27, 2015 16:58 |
|
it's not a law op laws are to keep them queermos from marryin
|
# ? Jan 27, 2015 17:05 |
|
In my limited understanding the problem with creating smaller transistors is that we're getting to the scale where classical mechanics gives way to quantum mechanics where things are less intuitive and harder to control. Electrons can tunnel out of potentials they would not normally be able to and poo poo. And there's nothing we can do to fix it, it's just the way it works. So some clever people are making quantum computers which manipulates quantum scale things in a way we can control, allowing a huge increase in calculation speed so you can watch Kill La Kill well into the future.
|
# ? Jan 27, 2015 17:20 |
|
rip andy grove, he died
|
# ? Jan 27, 2015 17:22 |
|
Beige posted:In my limited understanding the problem with creating smaller transistors is that we're getting to the scale where classical mechanics gives way to quantum mechanics where things are less intuitive and harder to control. Electrons can tunnel out of potentials they would not normally be able to and poo poo. And there's nothing we can do to fix it, it's just the way it works. So some clever people are making quantum computers which manipulates quantum scale things in a way we can control, allowing a huge increase in calculation speed so you can watch Kill La Kill well into the future. didn't that one company make a working quantum computer and iirc it was only really efficient for some calculations and traditional computers were better at most stuff we use computers for
|
# ? Jan 27, 2015 17:22 |
|
The main thing holding back performance isn't cpus. It is memory. A couple decades ago it took a handful of cpu cycles for a main memory access. It is hundreds of cycles now. An algorithm with good cache behavior can have an order of magnitude better performance than an algorithm with bad cache behavior when both perform the exact same number of calculations. Main memory is the new disk.
|
# ? Jan 27, 2015 17:26 |
|
Can we grow a brain in a jar that acts like a computer?
|
# ? Jan 27, 2015 17:48 |
|
|
# ? Mar 19, 2024 04:09 |
|
Beige posted:In my limited understanding the problem with creating smaller transistors is that we're getting to the scale where classical mechanics gives way to quantum mechanics where things are less intuitive and harder to control. Electrons can tunnel out of potentials they would not normally be able to and poo poo. And there's nothing we can do to fix it, it's just the way it works. So some clever people are making quantum computers which manipulates quantum scale things in a way we can control, allowing a huge increase in calculation speed so you can watch Kill La Kill well into the future. a transistor needs to consist of multiple atoms and we're down to the point of poo poo being the size of single atoms the quantum stuff is making single atoms do multiple tasks
|
# ? Jan 27, 2015 20:44 |