6 ms·

"Most of us don’t even need all the computing power in our pocket. We’ve gotten to a place of sufficiency, when just about any computer or smartphone is good en

"Most of us don’t even need all the computing power in our pocket. We’ve gotten to a place of sufficiency, when just about any computer or smartphone is good enough for what we want them to do." This is a fairly timeless sentiment. Even more ubiquitous is looking back and laughing at just how primitive formerly-sufficient technology was. You'd think we'd know by now to just not have this thought anymore!

sanderjd 11y ago
While there is always power in this sort of "yeah, we've always said that and always been wrong!" argument, I still feel that the benefits of computing "power" have slowed to a crawl recently. Which is not to say that computing devices don't improve, they just do so differently; they are better designed, lighter, more durable, have better battery life, etc. rather than running noticeably faster (in fact, we are usually happy to trade "power" away for improvement to all those other things, especially battery life).
- hunter2_ 11y ago
  This only holds up in the absence of things like Wirth's law and its variants. Not enough people are incentivized to be as efficient (in terms of CPU, etc.) as possible. Device makers are, because it affects their battery life (marketing) numbers, but only in closed systems (like iOS) could we even begin to assume that device makers have much influence over whether the downstream software development is sufficiently resource-efficient.
gizmo686 11y ago
My laptop has a 2.7GHz possessor. When I need to conserve power, I lock in down to 0.8GHz, sometimes I forget to unlock it. Most of the time I don't even notice the difference. Maybe, at some point, we will come up with a use-case that needs more compute that what we currently have, but for most people's current usecase, compute is not the limiting factor.
- agumonkey 11y ago
  I used to do that too. The most limiting factor for "mainstream computing" is crapware / ads / shifting UX. That's always the reason why average joe neighbors ends up buying a new machine. Random popups slowing down and frustration over 4-5 years adds up to enough for someone to believe that shelling 500$ out again on new shiny will finally bring them the land of click abundance they dream of. Very few hard requirements are valuable today for most users: battery lifetime, hd video for skype/facetime. And maybe SSD, which are now large and cheap enough for people to switch, and get back the bliss of instantaneous reaction again. Other than that no important task will require more than a core 2 duo. They won't saturate Excel (which was not long ago still single core only). In the case they really do, that spreadsheet is probably worth enough to buy a full fledge MBP.
- TheOtherHobbes 11y ago
  Cycles are very much the limiting factor for any kind of creative computing. I could literally eat a 1000X speed increase without blinking. I think everyone else could too. The difference isn't about speed, but about quality of interaction. Today's minimal UIs are a good impedance match for the limited cycles available. iOS was particularly good at leading the way on this, creating a UI that was relatively undemanding of raw cycles but still created an unexpectedly immersive experience. A big part of that was the move from type-and-point to drag-and-tap. The tactile mode, the skeuomorphism, and the portability and instant access compensated for having to squeeze the UI into a tiny screen. It felt like a more responsive experience even though it was quite a bit slower and more rationed than equivalent interaction on a full-sized desktop. That's still true today. You can get stuff done on a tablet, but the most efficient and productive UIs pre-ration the affordances. A desktop is much more flexible/general and gives you much more visual bandwidth. If you apply this idea to hypothetical 1000X hardware, you won't get more of the same but faster - you'll get completely new modes of interaction, which will likely combine the immediacy of touch with the high bandwidth of desktop interaction, and take them much further than anything available today. It's possible to imagine all kinds of science fiction experiences. The reality will probably still be a surprise. Even so - the point is really that today's users feel like they're doing everything they want, because UIs are designed without affordances that would be slow and frustrating given today's hardware. That will change with faster hardware and new designs, and phone UIs will seem as quaint as a 720 x 350 monochrome desktop PC screen running DOS.
marssaxman 11y ago
Perhaps I'm misunderstanding you, but what you're saying doesn't match up with my experience. Pretty much any computer built in the last decade has been fast enough to do everything I need without making me wait. Screen resolution has improved somewhat, lately, but new machines feel just like old ones when it comes to performance, and that definitely was not the case from 1985-2005.
- hunter2_ 11y ago
  A machine from 10 years ago will run software developed 10 years ago with perfectly adequate performance -- probably true even prior to 2005 as well as any reasonable point in the near future. But the bit I quoted is breaking out of that completely, basically claiming that we're at some sort of hockey stick curve where that software-hardware vintage parity won't be required anymore.
  - marssaxman 11y ago
    Well, yes, that's exactly what I was saying: we reached that point on the curve somewhere close to a decade ago, and old machines are perfectly capable of doing everything I need them to do now, today, with current software. The changes which have actually made a difference to my computing experience over the last decade were the arrival of affordable SSDs and a jump upward in LCD pixel density.
    - hunter2_ 11y ago
      I certainly agree that we can do most of the same things if our needs don't change, but usually our needs do change. Even if the software isn't simply more bloated than it used to be, we usually have new needs: say, creating or consuming something in a crazy new format. I guess we've pretty much maximized the number of pixels that fit on a hand/pocket-sized screen, but if filling our entire living room wall with 16K resolution video is the new norm someday, and we're showing off a video we just took using our pocket/wrist device, we need serious horsepower. This example is hyperbole but the point is, for the most part, I find it hard to predict that we won't find a way to use serious increases in processing whenever it's available.
traek 11y ago
Fair point, but there are some trends in technology that it doesn't take into account. As technology becomes more centralized, the computing power of your phone or laptop becomes much less important than the power of Google's and Amazon's data centers. Your devices essentially become clients for computation done in the cloud. Technological advancements generally don't require increases in the client's computing power anymore.
- hunter2_ 11y ago
  It's the case now, but the trend toward the client+server model (terminal+mainframe, app+cloud, whatever it's named any given decade) versus the local model (personal computers, etc.) will probably be somewhat cyclical as fads are across industries. The contributing factors are wildly unpredictable... networks (or the internet) existing, privacy tolerance, battery breakthroughs, wealth distribution, who knows. We're cool with thin clients now, but it could go out of fashion the moment some confidentiality breach, mainstream, hits home a little too hard (for example).
busterarm 11y ago
For everyone commenting to counter your point: Try to encode high-resolution 10bit RGB 444 video and stream it to the internet. While playing a AAA game on the same machine. Try to get the latency as low as possible and make sure the video is smooth, without artifacts on the stream. This is the kind of problem a lot of folks are trying to solve right now and the current hardware is barely good enough to keep up...and that's with adding highly specialized add-on cards. Try emulating Sega ST-V games on your machine -- and realize quickly that accurate emulation is a matter of clock speed over most other factors...and which has mostly slowed down since the Intel C2Q era. When you provide compute power/bandwidth/whatever, someone will find a need for it.
- gizmo686 11y ago
  I don't think anyone here is arguing that no one has a need for more power, but all of what you are describing is a niche use-case. We seeming to be passed the time where we need more power for the general use case.
zkhalique 11y ago
I would say that a lot of the time software just keeps getting slower as developers care less about efficiency. I'm not talking about stuff you can't get to run faster. If people had the same kind of care they had in the 80s and 90s, the software would be incredibly fast. But these days they can be more careless. Sometimes it's a good thing, as things ship faster and frameworks are more expressive. Other times, though, it's a PITA. Like when half the time we upgrade just to run the ton of bloatware from different vendors on the web / multitasking OSes at a decent speed. Chrome is usually the biggest hog on my machine. Or like when Apple's iOS keeps making the iPhone slower with every version, until you're practically forced to upgrade. Is it really that hard to make an OS which hardly does any major multitasking from getting slower and slower with every release? One almost wonders if they do it on purpose ;-)

sanderjd 11y ago

While there is always power in this sort of "yeah, we've always said that and always been wrong!" argument, I still feel that the benefits of computing "power" have slowed to a crawl recently. Which is not to say that computing devices don't improve, they just do so differently; they are better designed, lighter, more durable, have better battery life, etc. rather than running noticeably faster (in fact, we are usually happy to trade "power" away for improvement to all those other things, especially battery life).

hunter2_ 11y ago
This only holds up in the absence of things like Wirth's law and its variants. Not enough people are incentivized to be as efficient (in terms of CPU, etc.) as possible. Device makers are, because it affects their battery life (marketing) numbers, but only in closed systems (like iOS) could we even begin to assume that device makers have much influence over whether the downstream software development is sufficiently resource-efficient.

gizmo686 11y ago

My laptop has a 2.7GHz possessor. When I need to conserve power, I lock in down to 0.8GHz, sometimes I forget to unlock it. Most of the time I don't even notice the difference. Maybe, at some point, we will come up with a use-case that needs more compute that what we currently have, but for most people's current usecase, compute is not the limiting factor.

agumonkey 11y ago
I used to do that too. The most limiting factor for "mainstream computing" is crapware / ads / shifting UX. That's always the reason why average joe neighbors ends up buying a new machine. Random popups slowing down and frustration over 4-5 years adds up to enough for someone to believe that shelling 500$ out again on new shiny will finally bring them the land of click abundance they dream of. Very few hard requirements are valuable today for most users: battery lifetime, hd video for skype/facetime. And maybe SSD, which are now large and cheap enough for people to switch, and get back the bliss of instantaneous reaction again. Other than that no important task will require more than a core 2 duo. They won't saturate Excel (which was not long ago still single core only). In the case they really do, that spreadsheet is probably worth enough to buy a full fledge MBP.
TheOtherHobbes 11y ago
Cycles are very much the limiting factor for any kind of creative computing. I could literally eat a 1000X speed increase without blinking. I think everyone else could too. The difference isn't about speed, but about quality of interaction. Today's minimal UIs are a good impedance match for the limited cycles available. iOS was particularly good at leading the way on this, creating a UI that was relatively undemanding of raw cycles but still created an unexpectedly immersive experience. A big part of that was the move from type-and-point to drag-and-tap. The tactile mode, the skeuomorphism, and the portability and instant access compensated for having to squeeze the UI into a tiny screen. It felt like a more responsive experience even though it was quite a bit slower and more rationed than equivalent interaction on a full-sized desktop. That's still true today. You can get stuff done on a tablet, but the most efficient and productive UIs pre-ration the affordances. A desktop is much more flexible/general and gives you much more visual bandwidth. If you apply this idea to hypothetical 1000X hardware, you won't get more of the same but faster - you'll get completely new modes of interaction, which will likely combine the immediacy of touch with the high bandwidth of desktop interaction, and take them much further than anything available today. It's possible to imagine all kinds of science fiction experiences. The reality will probably still be a surprise. Even so - the point is really that today's users feel like they're doing everything they want, because UIs are designed without affordances that would be slow and frustrating given today's hardware. That will change with faster hardware and new designs, and phone UIs will seem as quaint as a 720 x 350 monochrome desktop PC screen running DOS.

marssaxman 11y ago

Perhaps I'm misunderstanding you, but what you're saying doesn't match up with my experience. Pretty much any computer built in the last decade has been fast enough to do everything I need without making me wait. Screen resolution has improved somewhat, lately, but new machines feel just like old ones when it comes to performance, and that definitely was not the case from 1985-2005.

hunter2_ 11y ago
A machine from 10 years ago will run software developed 10 years ago with perfectly adequate performance -- probably true even prior to 2005 as well as any reasonable point in the near future. But the bit I quoted is breaking out of that completely, basically claiming that we're at some sort of hockey stick curve where that software-hardware vintage parity won't be required anymore.
- marssaxman 11y ago
  Well, yes, that's exactly what I was saying: we reached that point on the curve somewhere close to a decade ago, and old machines are perfectly capable of doing everything I need them to do now, today, with current software. The changes which have actually made a difference to my computing experience over the last decade were the arrival of affordable SSDs and a jump upward in LCD pixel density.
  - hunter2_ 11y ago
    I certainly agree that we can do most of the same things if our needs don't change, but usually our needs do change. Even if the software isn't simply more bloated than it used to be, we usually have new needs: say, creating or consuming something in a crazy new format. I guess we've pretty much maximized the number of pixels that fit on a hand/pocket-sized screen, but if filling our entire living room wall with 16K resolution video is the new norm someday, and we're showing off a video we just took using our pocket/wrist device, we need serious horsepower. This example is hyperbole but the point is, for the most part, I find it hard to predict that we won't find a way to use serious increases in processing whenever it's available.

traek 11y ago

Fair point, but there are some trends in technology that it doesn't take into account. As technology becomes more centralized, the computing power of your phone or laptop becomes much less important than the power of Google's and Amazon's data centers. Your devices essentially become clients for computation done in the cloud. Technological advancements generally don't require increases in the client's computing power anymore.

hunter2_ 11y ago
It's the case now, but the trend toward the client+server model (terminal+mainframe, app+cloud, whatever it's named any given decade) versus the local model (personal computers, etc.) will probably be somewhat cyclical as fads are across industries. The contributing factors are wildly unpredictable... networks (or the internet) existing, privacy tolerance, battery breakthroughs, wealth distribution, who knows. We're cool with thin clients now, but it could go out of fashion the moment some confidentiality breach, mainstream, hits home a little too hard (for example).

busterarm 11y ago

For everyone commenting to counter your point: Try to encode high-resolution 10bit RGB 444 video and stream it to the internet. While playing a AAA game on the same machine. Try to get the latency as low as possible and make sure the video is smooth, without artifacts on the stream. This is the kind of problem a lot of folks are trying to solve right now and the current hardware is barely good enough to keep up...and that's with adding highly specialized add-on cards. Try emulating Sega ST-V games on your machine -- and realize quickly that accurate emulation is a matter of clock speed over most other factors...and which has mostly slowed down since the Intel C2Q era. When you provide compute power/bandwidth/whatever, someone will find a need for it.

gizmo686 11y ago
I don't think anyone here is arguing that no one has a need for more power, but all of what you are describing is a niche use-case. We seeming to be passed the time where we need more power for the general use case.

zkhalique 11y ago

I would say that a lot of the time software just keeps getting slower as developers care less about efficiency. I'm not talking about stuff you can't get to run faster. If people had the same kind of care they had in the 80s and 90s, the software would be incredibly fast. But these days they can be more careless. Sometimes it's a good thing, as things ship faster and frameworks are more expressive. Other times, though, it's a PITA. Like when half the time we upgrade just to run the ton of bloatware from different vendors on the web / multitasking OSes at a decent speed. Chrome is usually the biggest hog on my machine. Or like when Apple's iOS keeps making the iPhone slower with every version, until you're practically forced to upgrade. Is it really that hard to make an OS which hardly does any major multitasking from getting slower and slower with every release? One almost wonders if they do it on purpose ;-)