Moore's Law ended in 2012

[Balog]

http://lionoftheblogosphere.wordpress.com/2014/12/01/moores-law-has-ended-but-no-one-noticed/

Interesting read. I guess skynet may have to wait.

[AJ Dual]

IMO, the slowdown in Moore's Law, it's almost needed. There's a LOT of back-end computing problems and optimization that have gotten swept under the rug because you could always just "Throw more flops at the problem".

Moore's Law slowing to a crawl is going to force a number of technology industries to dig deeper into improvements in architecture, code, and more elegant formulas and how problems and data sets and inputs are represented. We'll also see improvements in compression technology.

Also, even if the doubling of chip performance stops, the chips will continue to get cheaper, and spreading the intelligence around, making for a de-facto distributed computing model even within individual devices will continue to have an impact. 

It could even be a boon for AI, since some of the most promising results seem to be on this distributed model. In one example, if you have a self-driving car, or a legged robot, there's a processor or subsystem for the legs, one for balance, one for navigation, a processor for the vision system, then they all communicate via shorthand and make decisions collectively, or in a predetermined hierarchical basis. Which is a closer parallel to the way actual human intelligence, or even just the self-directed activities of lower animals is carried out.

We also have a lot of technologies that are nearing perfection, self-driving cars, speech recognition, facial recognition etc. that were once thought impossible without AI that are now everyday items, or about to be everyday items.

I don't think we'll see a huge decrease in the rate of progress of technology because Moore's Law has stalled. We've gotten this far with massive vertical growth in CPU performance. Now we can see what's possible with a broader more parallel and distributed approach.

[Ben]

Seems like it's been developing from even before 2012. At the old job, in the late 90's to early 2000's, if I didn't get a new workstation every couple of years, I couldn't keep up with the software I was using and the amounts of data being created. I could also see significant differences in computing speed - something that took two days on a SPARC station to two hours on a P4.

For the last few years of the job, I got along just fine with a ~$2K Dell Mobile Workstation. I could have gotten things done faster with some $6K quad Xenon, but we were talking a difference of computing time in tens of minutes versus hours. It was basically take a coffee break (or check in at APS), and come back and things were done.  So the value of the bleeding edge hardware became less important.

[RoadKingLarry]

http://lionoftheblogosphere.wordpress.com/2014/12/01/moores-law-has-ended-but-no-one-noticed/

Interesting read. I guess skynet may have to wait.

That's what Skynet wants you to believe.

[Perd Hapley]

Maybe it's because I quit playing around with computers, but I haven't noticed PC tech exploding in the past couple of years. I got 8 G of memory, when I put this box together a couple of years ago, and I figured that would be a pitifully small amount by now. But I still see new computers selling with 2 G. The chip speeds and number of cores seems to have stayed put, as well.

Then again, all the innovation seems to be going toward mobile devices and software.

[Hawkmoon]

Four years ago I bought an Acer Aspire One netbook. I came with one gig of RAM and Windows 7 Starter Edition. As long as I didn't go nuts trying to run everything simultaneously, it was more than adequate for my normal tasks. It ran the entire Microsoft Office 2003 suite, and it ran AutoCAD LT 2000.

Truth be told, after the trip for which I needed it was done I gave it to a grandson ... and now I wish I hadn't done so.

[Zardozimo Oprah Bannedalas]

But now that transistors have shrunk down to 28nm, it’s proving difficult to make chips with smaller transistors that are as inexpensive per transistor as chips with 28nm transistors.

It's more of an issue of changing behavior - when you get below a certain size, power consumption and signal transmission issues crop up.

[AJ Dual]

It's more of an issue of changing behavior - when you get below a certain size, power consumption and signal transmission issues crop up.

And at some point when the size is small enough quantum tunneling becomes an issue.. the electrons just jump and say "I feel like being over there now..." 

Although there's the possibility someone may figure out how to use that as a feature, and not a bug, and figure out making transistors and interconnects even smaller, or none at all.

[BryanP]

I think it's more of a matter of what we're looking for.  We have plenty of processing power.  As the link says, most of the focus is now on on mobile devices and balancing enough power to do the job with how long you can make a battery last while making the device as light as possible.

And they've done it.  My wife gave me an iPad mini back in spring of 2013.  It has become my most commonly used computer at home.  I don't even bring a laptop on vacation anymore unless I'm on call.  I use it to browse the web, read books (currently reading Marko Kloos' "Lines of Departure"), play a few games (Baldur's Gate Enhanced Edition is currently stealing all my game time), watch shows online, and stream MKV files from my desktop PC by way of nplayer (a not-free app, but the best one I've found for playing videos over my network, especially ones in formats the ipad does not natively support).

It could stand to be a little snappier, but it's fast enough.  Worst thing I can say about it now is the battery doesn't last as well as it did brand new.  I just need to get around to replacing it.

I use an Android phone for a lot of the same stuff (not the game) and from what I've seen, any Android tablet that cost more than about $175-$200 should be plenty for most purposes.  The iPad is (of course) more expensive, but I have to admit that it works very well. 

[Balog]

Wasn't there some promising research on new super batteries that are orders of magnitude better than what we have now? That'd be the real game changer.

[AJ Dual]

When a few thousand dollars can build you a gaming PC capable of 4k resolution at over 60fps, and render games at that resolution dynamically in realtime, unless you're trying to some really hard-corps nuclear weapon simulations or fluid dynamic studies, and just a few years earlier it would have taken a multi-million dollar rendering farm months to deliver that video output for a Pixar movie... I don't think we'll be hurting for a "lack of flops" for awhile.

http://kotaku.com/i-built-a-4k-ultra-hd-gaming-pc-and-i-love-it-1564135136

[BryanP]

Wasn't there some promising research on new super batteries that are orders of magnitude better than what we have now? That'd be the real game changer.

We hear things like that every few years.  I keep hoping it will pan out one day.

[Balog]

We hear things like that every few years.  I keep hoping it will pan out one day.

Cold fusion is ten years away! 

[Headless Thompson Gunner]

I think it's more of a matter of what we're looking for.  We have plenty of processing power.  As the link says, most of the focus is now on on mobile devices and balancing enough power to do the job with how long you can make a battery last while making the device as light as possible.

This.  We've reached the point where we have enough processing power for everyday stuff.  What we want is more battery power and more portability, things like that.

[230RN]

Truth be told, after the trip for which I needed it was done I gave it to a grandson ... and now I wish I hadn't done.

You can probably retrieve it from the back of his closet by now.

[Firethorn]

What we want is more battery power and more portability, things like that.

Sadly, thinner seems to be winning over battery power.

[MicroBalrog]

Bear in mind that while processor density is not improving, computing power continues getting cheaper. Computing power cost $1.80 per GFLOPs in 2011, and by late 2013 the cost fell to $0.12, an order of magnitude decline.

http://en.wikipedia.org/wiki/FLOPS#Hardware_costs

[KD5NRH]

Moore's Law slowing to a crawl is going to force a number of technology industries to dig deeper into improvements in architecture, code, and more elegant formulas and how problems and data sets and inputs are represented. We'll also see improvements in compression technology.

The problem is that even common user applications like Office have always kept their bloat expanded to barely run smoothly on the predicted top of the line for their expected release date.  They overshot the mark, and there's still a lot of optimization to be done.

I'm somewhat curious whether, even if someone managed to build a minimalist OS emulating just enough to handle the binary without tons of OS overhead, the current Windows versions of things like Notepad and Calculator could run effectively on a 12MHz system, because I sure remember the old versions working well on a 286.

[AJ Dual]

The problem is that even common user applications like Office have always kept their bloat expanded to barely run smoothly on the predicted top of the line for their expected release date.  They overshot the mark, and there's still a lot of optimization to be done.

This is exactly what I'm talking about. Bring the pain. 

Most large institutional applications are modular, with the legacy code often just carried forward indefinitely and slowly picked apart, or added to for features, and they're now too large and too complex for even a large organization to do functional testing on every variable in the test matrix, and essentially, with the internet for updates and the cloud to quickly distribute fixes, most products now exist in a state of perpetual beta testing anyway.

If Office really bogs down, people/organizations will stop buying Office, or will stick with 2010 past MS' planned EOL, and give Redmond the finger. And Microsoft will be forced to dig under the hood for optimizations they'd not otherwise have done.

Popular end-user applications and enterprise software has been almost analogous to the financial crisis, where the profit is privatized, but with .gov "bailouts" the risks have been socialized. In a similar fashion, software has been "bailed out" by the ever increasing performance of PC's and servers.

Just as if the bailouts don't happen, and the market learns the hard way "Don't do that!" through actual financial pain and loss, the software market and makers getting burned by unusable applications will force them to improve their practices.

[KD5NRH]

If Office really bogs down, people/organizations will stop buying Office, or will stick with 2010 past MS' planned EOL, and give Redmond the finger. And Microsoft will be forced to dig under the hood for optimizations they'd not otherwise have done.

I do wonder how ugly it's going to get at the end of Win7 support if they haven't done something better than 8.

[lee n. field]

I do wonder how ugly it's going to get at the end of Win7 support if they haven't done something better than 8.

I think 7 is going to last a looong time.

[Regolith]

I do wonder how ugly it's going to get at the end of Win7 support if they haven't done something better than 8.

I'm cautiously optimistic about Windows 10. My brother has the beta, so I'll have to ask him about it this weekend, but they seem to have un-*expletive deleted*ed everything wrong with Windows 8 while keeping the good things.

If true, the Microsoft publishing cycle will remain unbroken:

[RoadKingLarry]

This is exactly what I'm talking about. Bring the pain. 

Most large institutional applications are modular, with the legacy code often just carried forward indefinitely and slowly picked apart, or added to for features, and they're now too large and too complex for even a large organization to do functional testing on every variable in the test matrix, and essentially, with the internet for updates and the cloud to quickly distribute fixes, most products now exist in a state of perpetual beta testing anyway.

If Office really bogs down, people/organizations will stop buying Office, or will stick with 2010 past MS' planned EOL, and give Redmond the finger. And Microsoft will be forced to dig under the hood for optimizations they'd not otherwise have done.

Popular end-user applications and enterprise software has been almost analogous to the financial crisis, where the profit is privatized, but with .gov "bailouts" the risks have been socialized. In a similar fashion, software has been "bailed out" by the ever increasing performance of PC's and servers.

Just as if the bailouts don't happen, and the market learns the hard way "Don't do that!" through actual financial pain and loss, the software market and makers getting burned by unusable applications will force them to improve their practices.

We've had make serious mods to our field level work practices to adjust new craptastic corporate mandated software. We've all come to the conclusion that the software writers that were too incompetent to write for Obamacare have found long term employment developing software for us. What used to take 2 -3 clicks and a time entry now takes now requires 8 -10 clicks and putting "something" in 4 or 5 empty drop down boxes, adding a user alias and a time entry. To add even more fun the corporate mandated version of IE is not compatible with the bug report website and throws you out with an error when you try to report a problem.
The response to complaints is that we just aren't tech savey enough to get with the program.

[zahc]

There are people running Photoshop I for DOS on smartphones with a DOS emulator. Even with the emulation it's reportedly usable.

I use ancient Sun and DOS machines in the factory installed right next to "upgraded" tools with Windows 7. The Windows machines are slower and the operators prefer the 90's era machines.

[lee n. field]

I use ancient Sun and DOS machines in the factory installed right next to "upgraded" tools with Windows 7. The Windows machines are slower and the operators prefer the 90's era machines.

I had a customer ask me a while ago if I knew PDP-11.    He has a machine tool with that embedded.