8 ms·

Why Moore's Law, not mobility, is killing the PC

45 points by Libertatea 14y ago

ghshephard 14y ago
Excellent article - I've been waiting for someone to capture the fact, that as a reasonably productive Network Engineer/Technical Account manager, I'm able to work, every day, completely proficiently with my Dell Precision 650 -Circa January 2004. Visio, Power Point, Excel, VMWare Workstation (with 5-6 OpenBSD + Linux VM instances running) , Cisco Dynamips Simulations, Outlook+Lookout, PowerPoint, and lots more Visio. Prior to 2003, I really felt the need to upgrade every 2-3 years, and by year 3, my PC was really, really starting to drag. I've just never felt the need to lean on IT to upgrade/replace my current Desktop though. Runs just fine.
- DeepDuh 14y ago
  While I share your sentiment I'm honestly kind of impressed that you run 5-6 linux VMs on a Pentium 4/AMD64 generation system. Really? Those things didn't virtualization support on the CPU yet, did they? Also, Ram probably maxes out at around 4GB which is quite low for an updated winXP with so many instances. Does that really run smoothly? I recommend you get an i5, 16GB, 500GB SSD and then come back and tell me the upgrade wasn't worth it ;-).
  - drucken 14y ago
    Vmware Workstation was first released in 1999. It never used to require hardware virtualization support until recently. Also, 4GB is plenty for Windows XP. Perhaps you have not used it for so long now that you've forgotten, but it only ever required 64MB and ran perfectly fine with 128MB-256MB even with high performance apps. All that has happened since is a factor increase in memory requirement for a limited set of applications. With Linux guest instances, it is easy to cap and keep very low the maximum memory usage. 6x VM with 256MB each is still only 1.5GB. Without a change in software requirements, I doubt many consumers would feel the need to upgrade anything. CPUs/PCs really have been fast enough for many years now.
    - DeepDuh 14y ago
      Did you also update your XP to the latest service packs and patches? My updated XP in a VM currently runs at 780MB after one morning of very light usage. It has only one application open at 100MB. Also, yes, virtualization wasn't always needed, but without it anything slightly CPU intensive will be awfully slow. I can see it working for CLI-only linux/bsd systems, but not much else. Do you ever compile something bigger on your guest systems as an example?
  - ghshephard 14y ago
    Always two (2) Linux instances (One CentOS and One Ubuntu), and almost always 4 OpenBSD images (for networking simulations). The VMware instances don't really seem to cause any foreground processing pauses (unlike the Craptastic VMware Workstation on my MacBook Pro + WIndows XP which leaves me in constant jerk-spinning beachball state. Grr...) - Runs super smoothly. I do all my interactive work on a (Circa 2010) Macbook Air - the Windows XP system is for Visio, Dynamips, Vmware Workstation, PowerPoint, Visio, Microsoft Word, and Outlook+Lookout (I have 10 years of Exchange PST's cached on that box) - I don't think a faster system would give me much more than 10% improvement in performance - because I'm never really CPU bound on that box. And, honestly, I spend maybe an hour a day on it - the other seven I spend on my MacBook Air. I just don't see any justification to upgrade it. It gets the job done, and it's fast enough. With all that said - the upgrade from my 2010 MacBook Pro to my SSD based 2010 MacBook Air was an awesome performance jump. Totally worth it. CPUs may not make much of a difference but WOW, do SSDs in a MacBook rock your world!
kiba 14y ago
I think mobile phones will eventually move up and displace traditional PC with a docking system that connect to your big monitors, your mouse, and your keyboard, and maybe even a big fat hard drive. Perhaps we will see a hybrid model in which your phone gets additional CPU and RAM to work with when plugged in? What about google glasses and other devices? The PC should find a way to interface with that too.
- VLM 14y ago
  "Perhaps we will see a hybrid model in which your phone gets additional CPU and RAM to work with when plugged in?" Yes, but indirectly over the network. The hybrid model I've been using for about 20 years now is connect a wimpy machine to a big one using SSH/NFS and a couple other technologies. For a decade or so the end users at wife's employer have been using rdesktop/vnc for their hybrid solution. Its actually "easier" for enduser devices to connect to a (private?) cloud than to try to do stuff themselves. So that's the more likely hybrid model. I don't run exchange server at home, or even run exchange client at home. Exchange webmail, yes.
  - kiba 14y ago
    Its actually "easier" for enduser devices to connect to a (private?) cloud than to try to do stuff themselves. So that's the more likely hybrid model. It depends on the latency tolerance that a use case might have. Why not have all of these above?
antonios 14y ago
Yes, I remember when my aunt was to decide between an iPad and a new PC. Once I explained Moore's law to her, the iPad won her over. It's definitely Moore's law.
UnoriginalGuy 14y ago
The article goes on and on about CPUs. They're right of course, anything after a core 2 duo is likely "fast enough." Tons of people have been buying the i5s instead of the i7s (myself included) for that very reason. But they're wrong in general. Memory capacity has continued to climb (from 2 GB with Vista, to 4 GB with Win7, to 8 GB today). I wouldn't build a PC today without 16 GB of RAM. Then we get to talk about the MASSIVE amount of improvement in IO in the last few years. The 7200 RPM HDD is dead. SSDs are now affordable. You can get a 256 GB SSD for £100. When consumers start to experience an SSD system compared to their old "slow" HDD system, they will be at the stores in droves wanting to upgrade. SSD's speed improvement is very noticeable indeed, like night and day.
- Spooky23 14y ago
  You're missing te point -- computers for the mass market are no longer an enthusiasts market. They are like TVs. Where I work, we're stuffing users into 2GB/1core VDI instances. They work fine. The only new client purchases are replacements of dead stuff, devices that improve mobility, and some dev/kiosk machines.
  - ghshephard 14y ago
    So, your users aren't working on traditional "Desktop PC" instances anymore, but inexpensive terminal clients that connect to a backend Hypervisor running on a server? I've been waiting to see when companies would start doing that "en masse" - back in 1999/2000, I thought it was going to take off when Sun was promoting the Thin Client Concept. Then VMWare took it up - but, for some reason, I never saw it get any traction, and the vast majority (Well, in my experience, all) companies seemed to go the other direction (countering where my instincts thought we were going to go) - and started issuing everyone laptops. The exact opposite of thin clients connected to servers. I had to eat a lot of claim chowder. If I'm understanding your situation correctly, how do your thin clients work out in practice, and, how do people deal with scenarios in which they would normally use laptops? (Business Trips, Airplane Flights, etc...)
    - Spooky23 14y ago
disturbedfood 14y ago
I think what needs to be focused on is mostly software, to make it less clunky and slow, and teach developers to write efficient sofware rather than just write bloaty software since their super-powerful computers can handle it just fine. Not everyone can afford to get the best computer every couple of years.
JonnieCache 14y ago
Forget your grandmother. There are loads of hobbyists out there doing high quality video, audio and software development work on boxes that are 5+ years old, quite happily. The set of things for which you need a superfast machine is really very small (compared to the whole,) especially if you are happy working within the limitations of the tools you have to hand, which covers a huge quantity of creative people, and most of the really good ones.
- yareally 14y ago
  One reason to need a fast CPU and lots of RAM is building Android from the source (primarily within the Android modding/hacking community). I mostly point it out because it has a wider appeal than most other large projects that are built from the source by typical users regularly these days and also for its demanding build requirements. Rebuilding the entire source can be lengthy without a fairly new CPU, SSD and lots of RAM. Granted one does not have to rebuild the entire source most of the time, but rebuilding some parts like the framework base (the java userland area libraries), takes a while. Following specs came from Google[1] the other year when ICS was released: -6GB of download. -25GB disk space to do a single build. -80GB disk space to build all AOSP configs at the same time. -16GB RAM recommended, more preferred, anything less will measurably benefit from using an SSD. -5+ hours of CPU time for a single build, 25+ minutes of wall time, as measured on my workstation (dual-E5620 i.e. 2x quad-core 2.4GHz HT, with 24GB of RAM, no SSD). Lots of RAM allows you to switch the compile output over to a ramdrive (or if not, move it to a separate SSD if you have two). Moving the entire source to a ram drive or fast ssd also makes editing much faster (especially the initial load into an IDE like Intellij or Eclipse) or grepping through the codebase. Doing it on anything less becomes somewhat tedious from my experience. [1] https://groups.google.com/forum/?hl=en&fromgroups=#!topic/android-building/N1exifTpPfA https://groups.google.com/forum/?hl=en&fromgroups=#!topi...
  - tsunamifury 14y ago
    I don't understand why anyone would do this locally when you can compile on an elastic cloud instance at many times the speed for pennies. Why keep upgrading your hardware when you should just be submitting these jobs to an ultra-cheap server that will be many times faster anyways?
DGrutt 14y ago
"Meanwhile, Intel's push towards ubiquitous computing--gesture controls, speech recognition, and so on--not only advances traditional interface models, but the technologies involved also require strong computing heft. Sneaky, sneaky." Lol ya this must be the reason. I bet that they have a hand in the self-driving cars too. Its all an excuse to require more computing power.
- coldtea 14y ago
  No, Intel is doing it out of the goodness of its heart. It might be a business but they never cared about money. The push these technologies to people out of pure love of new technology.
  - fthssht 14y ago
    You make money by selling what people want.
    - coldtea 14y ago
      You'd be surprised.
DanBC 14y ago
The article doesn't mention the bring your own hardware stuff that's increasingly happening at work. Docking stations have been tried before, and they didn't work particularly well, but the time's right now for a great tablet and a great docking station. The dock needs to have keyboard & mouse inputs, network connectivity, and be able to output to a nice monitor. (Allowing user to have two monitors?). Someone needs to sell that as a kit, with some kind of Office suite installed. MS could, if they put a lot of effort into trimming down MS Office, make a lot of money here. (As could a Lenovo / LibreOffice collaboration. Because although Excel is still much better than Libre's spreadsheet it's good enough for many people.)
- learc83 14y ago
  Speaking of docking stations, my thinkpad dock is one of the best purchases I've ever made. It makes it my laptop essentially much more portable since I don't have to worry about disconnecting reconnecting a monitor, speakers keyboard, mouse external hdd, and power.
- jiggy2011 14y ago
  I don't know, if you accept the premise of the article that old/cheap PC hardware is good enough for most applications then what do you gain by having a docking station vs a desktop PC? There is only a small cost saving since the expensive thing is going to be the peripherals and screen and you now have to deal with problems like employees taking the tablets home and losing/breaking them , getting malware on them or filling them up with personal stuff. The cost of having a few computer components nailed to the desk is probably less than the loss of productivity when somebody can't work because their tablet is broken. Are there many companies who are doing full on bring your own hardware apart from maybe startups etc who can't afford to provide extra computers to staff?
zurn 14y ago
Nobody is mentioning the multicore problem? Wide out-of-order CPU designs have hit diminishing returns a while ago in adding more instruction-level parallelism to the single thread of execution. Processor speedups haven't been coming dominated by faster transisitors for a while, it's been smarter OOO. You can see this now in ARM-land as they start climbing the OOO ladder. Multicore has been the best "next step" they've come up with, but the required difficult software rewrites prevented consumers from seeing speedups. Total on-chip exceution resources have essentially kept on the Moore's law path as CPU vendors punted to multicore designs. 99% of PC software is serial since parallel software is too hard to make. Runing the same code faster was the reason that software used to become more hungry each CPU generation and forced people to upgrade their machines. New CPUs used to make the programmer more productive, not less.
- mark-r 14y ago
  The problem is that in a fight between Moore's Law and Amdahl's Law, Amdahl will win. Extra cores reach diminishing returns very quickly.
qwerta 14y ago
It is not just that 'old is good enough', in last years PCs are actually getting worse. I would love to spend money, but there is no stuff to buy. My 7 years old laptop has higher display resolutions than anything on the market (except two new laptops). It is also impossible to buy laptop with 4:3 screen ratio. My 2 years old desktop CPU overclocked at 5GHz is faster than anything on the market, last two CPU generations can not reach such performance. And some newest SSD generation actually have worse performance characteristics than previous generation.
- fierarul 14y ago
  >It is not just that 'old is good enough', in last years PCs are actually getting worse. I would love to spend money, but there is no stuff to buy. So true, and it applies to Macs too! Long ago, I was looking forward to a 17" MBP but Apple discontinued it. My 2008 15" MBP has an Intel SSD plus a 2nd large spinning drive (optibay), mate display and a removable battery. Sometimes I wish I could just keep buying spare parts for it forever. Too bad you can't upgrade the CPU too.
- jiggy2011 14y ago
  My 2 years old desktop CPU overclocked at 5GHz is faster than anything on the market, last two CPU generations can not reach such performance. That's a fairly extreme overclock! I imagine you must be able to get some impressive results from a modern i7 with the right cooling though? I remember when overclocking was a thing that anybody into PC gaming did for economical reasons, these days it hardly seems worth the effort of fiddling around in the bios menu.
  - qwerta 14y ago
    Cooling is regular Antec H2O 920, it fits into case without any problem. It took about 3 hours to assemble and overclock, I just followed manual. It runs 24/7, I do number crunching and Scala developing (slow compiler). Intel does not bother soldering heatsink into chip anymore (saves 50 cents). So newer CPU can not be overclocked without disassembling and replacing some parts. Results are not bad, but in reality it is just 30% speedup.
brisance 14y ago
What's the point of all that computing power if there's no "killer app" to make use of it? These days the devices that interest me are those that consume little power, have perpetual internet connectivity (WiFi or cell tower), and are instant on/off. It has to be portable and durable. Everything else is icing.
- amalag 14y ago
  As of the core duo generation, people don't need to upgrade their computers in order to browse the web or use Microsoft Office. Office is the eternal killer app for the masses IMHO (excluding the browser), and it doesn't need the latest generation intel chip. As we know so many apps have moved to the browser.
- noahl 14y ago
  Just at a guess, really really good speech recognition might be a "killer app" for big, vectorized chips. I haven't tried to implement it, though, so I don't know how much power it would really take. Apple and Google seem to do all right with a little phone CPU.
  - api 14y ago
    As far as I know, Siri off-loads something to Apple's servers. But maybe I'm mistaken.
- jiggy2011 14y ago
  You need something requires low latency as well as power, otherwise it usually makes more sense to offload the extra work onto a server (with the added bonus that it's difficult to pirate something when there is code isn't running on a client machine, e.g the new Sim City). So the most common example would be games, with the new console refresh we might see a bump in technical requirements for PC games in order to keep up.
saturdaysaint 14y ago
Hyper-competition to make a faster browser and faster websites has to be a factor. In the days of packaged software, you might be happy to get a handful of performance updates before a flashy (and more CPU-intensive) new version came out, but any web service worth its salt is built to load almost instantly; snappy usability is important when competition is a click away. And just think of how much horsepower has been saved by abandoning Flash. The end result is that the average software user (Facebook, Google, Gmail/Hotmail, etc.) has probably seen the software they use actually get faster in the last 10 years.
api 14y ago
It's economics. I can attest to the fact that most five-year-old chips are good enough for 95% of software I want to use. So there's no real demand outside of big data and data center applications. The only way to keep pushing performance would be for software writers to invent new and compelling things to do that genuinely require more power. It can be something as simple as a game-- I was a teenager when Quake came out, and I remember scraping together as much money as I could to get the absolute fastest machine I could (A P120) to play it. Before Quake came out, my old 486 DX2/66 was fast enough for the Linux hacking, adventure games, and BBS/Internet stuff I was doing at the time. We need another Quake... or something equally cool. We need killer apps. Otherwise computers are going to go over like the aerospace industry did. We stopped pushing the envelope there in the 70s, again for economic reasons.
- EthanHeilman 14y ago
  Technologies like the oculus rift will force a resolution and latency arms race (the resolution arms race is already in full swing thanks to attempts to surpass apple's "retina display"). Of course this arms race will take place mostly on the GPU and display side of things. I see the CPU becoming more of a coordinator between custom processors (think the video decoders on iphones) than the main show. * What are the computational bottlenecks for driverless cars? Vector operations? * If every car needs a high performance computer to drive how will this change the semi-conductor landscape?
- qznc 14y ago
  The point is, it must be about something noticeable other than the graphics. For example, a good AI should easily be able to exhaust all the cores you can throw at it. However, it is much harder to sell. In those days they could show you some pictures of Quake and you know you wanted that. It is harder to make "pictures" of AI behavior. Alternatively, exchange AI for physics simulation. Maybe the sandbox genre (Minecraft, Dwarf Fortress) will grow. Simulating huge worlds requires lots of processing power.
  - api 14y ago
    I did a lot of work on evolutionary computation a while back. Now there's something that can soak up some cycles. It's also embarrassingly parallel. (For those who want to trot out the "GAs don't work well," "evolution is slow," etc. mantras, please delve into the literature first. EC has some interesting and unique results, the most important of which is that it's the only "AI" technique I'm aware of that can demonstrate true creativity. EC can invent novel structures ex-nihilo. Everything else can just optimize, find maxima, etc. Another way of saying it is that EC can find maxima involving unknown, novel parameters that it invents along the way, while everything else requires that you know all the parameters to be optimized ahead of time. So EC can work with unknown unknowns, not just unknown values of knowns.)
api 14y ago
Starting in the 90s up until today, the CPU surged ahead of everything else: disk, memory, and network. Today we have very, very fast chips that are bound by the speed of their connection to their memory, disk, and each other (the Internet). Memory has improved with DDR technology. (What are we up to now... DDR4?) Disk is improving with SSDs, which offer a huge speed improvement. If you've never used a box with a SSD in it, it must be seen to be believed. But the real bottleneck I think is the Internet. Imagine what we could do with gigabit peer-to-peer IPv6 networks in areas like games, collaborative development tools, teleconferencing, etc. Things like virtual reality "metaverses" could really arrive, etc. These in turn would require more CPU power and parallelism. Super-fast gigabit+ WANs would deliver the killer apps that we need to keep pushing the speed barrier. Once again, it is the Internet carriers that are the barrier to all progress. Breaking the telco oligopolies is the thing that must happen to unlock the future. Everything-- maybe even the continuation of Moore's Law-- depends on that. The telcos, cable companies, and wireless carriers are the trolls guarding the bridge to the future.
- stevenameyer 14y ago
  The internet has become a necessity in a lot of peoples lives and with such a high barrier to entry and only a few telcos controlling peoples access there is too much room for abuse of power by the telcos. We need to recognize that access to the internet is turning from a tech issue to a social issue and start policing it as such.
- TillE 14y ago
  Having a speedy home connection is well and good, but what really needs to improve is mobile data. Think of stuff like Google Glass, and all the possible advantages of having a constant connection everywhere you go for a two-way exchange of data. Now imagine your data bill after just a couple hours of live streaming HD video. Something's gotta change. The bandwidth keeps increasing as technology improves, but the price per bit is actually getting higher.
  - api 14y ago
    Mobile data is the worst, by far, but if we're talking about pushing the performance envelope then we're talking about PCs. Mobile is constrained by power consumption as well as network bandwidth, so you're not going to see blazing fast continuations of Moore's law there. I really think Google Fiber nationwide would lead to killer apps that would revitalize the PC and the high-end chip market. I'm thinking of rich constant collaboration, virtual reality, next generations of MMORPGs where you can really richly interact with the world's physics (e.g. shaping materials, in-world chemistry, etc.), tele-presence, digital telepathy with brain-computer interfaces, distributed supercomputing, ... lots of stuff. All those things are things that are CPU-hungry. Many are embarrassingly parallel, so they can utilize many-core chips efficiently. If Intel, AMD, Dell, Apple's PC/laptop division, IBM, etc. knew what was good for them they'd get together and invest in gigabit+ broadband nationwide. It's also got to be network-neutral, since all the things I mentioned above are at least somewhat P2P. The PC industry is spiraling down because of a lack of killer apps for high-end boxes. Most of these are waiting on fast networks.
MilesTeg 14y ago
If you want to see Moore's Law in action today you really need to look at GPUs rather than CPUs. Increased performance in recent years has mostly come from increased parallelization which has limiting returns on the CPU( quad cores anyone?) but does wonders for computer graphics.
bhauer 14y ago
I agree with the author, although I don't buy the literary ploy that this realization just hit him recently. This has been a development building within the PC sphere for years. Not only that, many professional and countless more no-good hobbyist opinion-makers (like myself) have made this point over and over. I'm nobody, but I have opined about this and the insipid taint of Good Enough several times [1]. Like "api" at the top of this comment thread, I feel there are bottlenecks that keep the desktop PC, and the laptop PC to a degree, constrained to the roles that we have presently envisioned for those devices. My opinion is that desktop display technology is ultimately the root of desktop stagnation and, speaking positively, could be the root of a desktop renaissance [2]. I agree that bandwidth progress has slowed and storage capacity progress has slowed [3], but point out that only display technology has actually outright stopped and regressed. In 2001, the IBM T220/221 monitor achieved high-DPI on the desktop (3840x2400 at 22 inches). Since then, we banished high DPI from the desktop and obsessed about a marketing term named "HD" because we allowed the rules that govern our living rooms' displays to take over on our desktop. That we accept 24" desktop monitors with 1080 vertical lines in 2013 is heartbreaking. I agree with api that bandwidth is to blame. But bandwidth hasn't stopped progressing and regressed (well, with one exception: mobile data caps). More importantly, I feel that behind the bandwidth complacency is display complacency. Put a 40" high-DPI desktop monitor on your desktop and I think the demand for ultra-high-definition content, high-clarity user interfaces, bandwidth, computation power, and storage capacity will all increase. Intel sells more processors, HP/Dell/Lenovo sell more PCs, and oh-boy would I be so happy to finally get a real monitor upgrade. [1] http://tiamat.tsotech.com/technology-sucks http://tiamat.tsotech.com/technology-sucks [2] http://tiamat.tsotech.com/displays-are-the-key http://tiamat.tsotech.com/displays-are-the-key [3] http://tiamat.tsotech.com/boring-technology http://tiamat.tsotech.com/boring-technology

45 points by Libertatea 14y ago

Excellent article - I've been waiting for someone to capture the fact, that as a reasonably productive Network Engineer/Technical Account manager, I'm able to work, every day, completely proficiently with my Dell Precision 650 -Circa January 2004. Visio, Power Point, Excel, VMWare Workstation (with 5-6 OpenBSD + Linux VM instances running) , Cisco Dynamips Simulations, Outlook+Lookout, PowerPoint, and lots more Visio. Prior to 2003, I really felt the need to upgrade every 2-3 years, and by year 3, my PC was really, really starting to drag. I've just never felt the need to lean on IT to upgrade/replace my current Desktop though. Runs just fine.

DeepDuh 14y ago
While I share your sentiment I'm honestly kind of impressed that you run 5-6 linux VMs on a Pentium 4/AMD64 generation system. Really? Those things didn't virtualization support on the CPU yet, did they? Also, Ram probably maxes out at around 4GB which is quite low for an updated winXP with so many instances. Does that really run smoothly? I recommend you get an i5, 16GB, 500GB SSD and then come back and tell me the upgrade wasn't worth it ;-).
- drucken 14y ago
  Vmware Workstation was first released in 1999. It never used to require hardware virtualization support until recently. Also, 4GB is plenty for Windows XP. Perhaps you have not used it for so long now that you've forgotten, but it only ever required 64MB and ran perfectly fine with 128MB-256MB even with high performance apps. All that has happened since is a factor increase in memory requirement for a limited set of applications. With Linux guest instances, it is easy to cap and keep very low the maximum memory usage. 6x VM with 256MB each is still only 1.5GB. Without a change in software requirements, I doubt many consumers would feel the need to upgrade anything. CPUs/PCs really have been fast enough for many years now.
  - DeepDuh 14y ago
    Did you also update your XP to the latest service packs and patches? My updated XP in a VM currently runs at 780MB after one morning of very light usage. It has only one application open at 100MB. Also, yes, virtualization wasn't always needed, but without it anything slightly CPU intensive will be awfully slow. I can see it working for CLI-only linux/bsd systems, but not much else. Do you ever compile something bigger on your guest systems as an example?
- ghshephard 14y ago
  Always two (2) Linux instances (One CentOS and One Ubuntu), and almost always 4 OpenBSD images (for networking simulations). The VMware instances don't really seem to cause any foreground processing pauses (unlike the Craptastic VMware Workstation on my MacBook Pro + WIndows XP which leaves me in constant jerk-spinning beachball state. Grr...) - Runs super smoothly. I do all my interactive work on a (Circa 2010) Macbook Air - the Windows XP system is for Visio, Dynamips, Vmware Workstation, PowerPoint, Visio, Microsoft Word, and Outlook+Lookout (I have 10 years of Exchange PST's cached on that box) - I don't think a faster system would give me much more than 10% improvement in performance - because I'm never really CPU bound on that box. And, honestly, I spend maybe an hour a day on it - the other seven I spend on my MacBook Air. I just don't see any justification to upgrade it. It gets the job done, and it's fast enough. With all that said - the upgrade from my 2010 MacBook Pro to my SSD based 2010 MacBook Air was an awesome performance jump. Totally worth it. CPUs may not make much of a difference but WOW, do SSDs in a MacBook rock your world!

kiba 14y ago

I think mobile phones will eventually move up and displace traditional PC with a docking system that connect to your big monitors, your mouse, and your keyboard, and maybe even a big fat hard drive. Perhaps we will see a hybrid model in which your phone gets additional CPU and RAM to work with when plugged in? What about google glasses and other devices? The PC should find a way to interface with that too.

VLM 14y ago
"Perhaps we will see a hybrid model in which your phone gets additional CPU and RAM to work with when plugged in?" Yes, but indirectly over the network. The hybrid model I've been using for about 20 years now is connect a wimpy machine to a big one using SSH/NFS and a couple other technologies. For a decade or so the end users at wife's employer have been using rdesktop/vnc for their hybrid solution. Its actually "easier" for enduser devices to connect to a (private?) cloud than to try to do stuff themselves. So that's the more likely hybrid model. I don't run exchange server at home, or even run exchange client at home. Exchange webmail, yes.
- kiba 14y ago
  Its actually "easier" for enduser devices to connect to a (private?) cloud than to try to do stuff themselves. So that's the more likely hybrid model. It depends on the latency tolerance that a use case might have. Why not have all of these above?

antonios 14y ago

Yes, I remember when my aunt was to decide between an iPad and a new PC. Once I explained Moore's law to her, the iPad won her over. It's definitely Moore's law.

UnoriginalGuy 14y ago

The article goes on and on about CPUs. They're right of course, anything after a core 2 duo is likely "fast enough." Tons of people have been buying the i5s instead of the i7s (myself included) for that very reason. But they're wrong in general. Memory capacity has continued to climb (from 2 GB with Vista, to 4 GB with Win7, to 8 GB today). I wouldn't build a PC today without 16 GB of RAM. Then we get to talk about the MASSIVE amount of improvement in IO in the last few years. The 7200 RPM HDD is dead. SSDs are now affordable. You can get a 256 GB SSD for £100. When consumers start to experience an SSD system compared to their old "slow" HDD system, they will be at the stores in droves wanting to upgrade. SSD's speed improvement is very noticeable indeed, like night and day.

Spooky23 14y ago
You're missing te point -- computers for the mass market are no longer an enthusiasts market. They are like TVs. Where I work, we're stuffing users into 2GB/1core VDI instances. They work fine. The only new client purchases are replacements of dead stuff, devices that improve mobility, and some dev/kiosk machines.
- ghshephard 14y ago
  So, your users aren't working on traditional "Desktop PC" instances anymore, but inexpensive terminal clients that connect to a backend Hypervisor running on a server? I've been waiting to see when companies would start doing that "en masse" - back in 1999/2000, I thought it was going to take off when Sun was promoting the Thin Client Concept. Then VMWare took it up - but, for some reason, I never saw it get any traction, and the vast majority (Well, in my experience, all) companies seemed to go the other direction (countering where my instincts thought we were going to go) - and started issuing everyone laptops. The exact opposite of thin clients connected to servers. I had to eat a lot of claim chowder. If I'm understanding your situation correctly, how do your thin clients work out in practice, and, how do people deal with scenarios in which they would normally use laptops? (Business Trips, Airplane Flights, etc...)
  - Spooky23 14y ago

disturbedfood 14y ago

I think what needs to be focused on is mostly software, to make it less clunky and slow, and teach developers to write efficient sofware rather than just write bloaty software since their super-powerful computers can handle it just fine. Not everyone can afford to get the best computer every couple of years.

JonnieCache 14y ago

Forget your grandmother. There are loads of hobbyists out there doing high quality video, audio and software development work on boxes that are 5+ years old, quite happily. The set of things for which you need a superfast machine is really very small (compared to the whole,) especially if you are happy working within the limitations of the tools you have to hand, which covers a huge quantity of creative people, and most of the really good ones.

yareally 14y ago
One reason to need a fast CPU and lots of RAM is building Android from the source (primarily within the Android modding/hacking community). I mostly point it out because it has a wider appeal than most other large projects that are built from the source by typical users regularly these days and also for its demanding build requirements. Rebuilding the entire source can be lengthy without a fairly new CPU, SSD and lots of RAM. Granted one does not have to rebuild the entire source most of the time, but rebuilding some parts like the framework base (the java userland area libraries), takes a while. Following specs came from Google[1] the other year when ICS was released: -6GB of download. -25GB disk space to do a single build. -80GB disk space to build all AOSP configs at the same time. -16GB RAM recommended, more preferred, anything less will measurably benefit from using an SSD. -5+ hours of CPU time for a single build, 25+ minutes of wall time, as measured on my workstation (dual-E5620 i.e. 2x quad-core 2.4GHz HT, with 24GB of RAM, no SSD). Lots of RAM allows you to switch the compile output over to a ramdrive (or if not, move it to a separate SSD if you have two). Moving the entire source to a ram drive or fast ssd also makes editing much faster (especially the initial load into an IDE like Intellij or Eclipse) or grepping through the codebase. Doing it on anything less becomes somewhat tedious from my experience. [1] https://groups.google.com/forum/?hl=en&fromgroups=#!topic/android-building/N1exifTpPfA https://groups.google.com/forum/?hl=en&fromgroups=#!topi...
- tsunamifury 14y ago
  I don't understand why anyone would do this locally when you can compile on an elastic cloud instance at many times the speed for pennies. Why keep upgrading your hardware when you should just be submitting these jobs to an ultra-cheap server that will be many times faster anyways?

DGrutt 14y ago

"Meanwhile, Intel's push towards ubiquitous computing--gesture controls, speech recognition, and so on--not only advances traditional interface models, but the technologies involved also require strong computing heft. Sneaky, sneaky." Lol ya this must be the reason. I bet that they have a hand in the self-driving cars too. Its all an excuse to require more computing power.

coldtea 14y ago
No, Intel is doing it out of the goodness of its heart. It might be a business but they never cared about money. The push these technologies to people out of pure love of new technology.
- fthssht 14y ago
  You make money by selling what people want.
  - coldtea 14y ago
    You'd be surprised.

DanBC 14y ago

The article doesn't mention the bring your own hardware stuff that's increasingly happening at work. Docking stations have been tried before, and they didn't work particularly well, but the time's right now for a great tablet and a great docking station. The dock needs to have keyboard & mouse inputs, network connectivity, and be able to output to a nice monitor. (Allowing user to have two monitors?). Someone needs to sell that as a kit, with some kind of Office suite installed. MS could, if they put a lot of effort into trimming down MS Office, make a lot of money here. (As could a Lenovo / LibreOffice collaboration. Because although Excel is still much better than Libre's spreadsheet it's good enough for many people.)

learc83 14y ago
Speaking of docking stations, my thinkpad dock is one of the best purchases I've ever made. It makes it my laptop essentially much more portable since I don't have to worry about disconnecting reconnecting a monitor, speakers keyboard, mouse external hdd, and power.
jiggy2011 14y ago
I don't know, if you accept the premise of the article that old/cheap PC hardware is good enough for most applications then what do you gain by having a docking station vs a desktop PC? There is only a small cost saving since the expensive thing is going to be the peripherals and screen and you now have to deal with problems like employees taking the tablets home and losing/breaking them , getting malware on them or filling them up with personal stuff. The cost of having a few computer components nailed to the desk is probably less than the loss of productivity when somebody can't work because their tablet is broken. Are there many companies who are doing full on bring your own hardware apart from maybe startups etc who can't afford to provide extra computers to staff?

zurn 14y ago

Nobody is mentioning the multicore problem? Wide out-of-order CPU designs have hit diminishing returns a while ago in adding more instruction-level parallelism to the single thread of execution. Processor speedups haven't been coming dominated by faster transisitors for a while, it's been smarter OOO. You can see this now in ARM-land as they start climbing the OOO ladder. Multicore has been the best "next step" they've come up with, but the required difficult software rewrites prevented consumers from seeing speedups. Total on-chip exceution resources have essentially kept on the Moore's law path as CPU vendors punted to multicore designs. 99% of PC software is serial since parallel software is too hard to make. Runing the same code faster was the reason that software used to become more hungry each CPU generation and forced people to upgrade their machines. New CPUs used to make the programmer more productive, not less.

mark-r 14y ago
The problem is that in a fight between Moore's Law and Amdahl's Law, Amdahl will win. Extra cores reach diminishing returns very quickly.

qwerta 14y ago

It is not just that 'old is good enough', in last years PCs are actually getting worse. I would love to spend money, but there is no stuff to buy. My 7 years old laptop has higher display resolutions than anything on the market (except two new laptops). It is also impossible to buy laptop with 4:3 screen ratio. My 2 years old desktop CPU overclocked at 5GHz is faster than anything on the market, last two CPU generations can not reach such performance. And some newest SSD generation actually have worse performance characteristics than previous generation.

fierarul 14y ago
>It is not just that 'old is good enough', in last years PCs are actually getting worse. I would love to spend money, but there is no stuff to buy. So true, and it applies to Macs too! Long ago, I was looking forward to a 17" MBP but Apple discontinued it. My 2008 15" MBP has an Intel SSD plus a 2nd large spinning drive (optibay), mate display and a removable battery. Sometimes I wish I could just keep buying spare parts for it forever. Too bad you can't upgrade the CPU too.
jiggy2011 14y ago
My 2 years old desktop CPU overclocked at 5GHz is faster than anything on the market, last two CPU generations can not reach such performance. That's a fairly extreme overclock! I imagine you must be able to get some impressive results from a modern i7 with the right cooling though? I remember when overclocking was a thing that anybody into PC gaming did for economical reasons, these days it hardly seems worth the effort of fiddling around in the bios menu.
- qwerta 14y ago
  Cooling is regular Antec H2O 920, it fits into case without any problem. It took about 3 hours to assemble and overclock, I just followed manual. It runs 24/7, I do number crunching and Scala developing (slow compiler). Intel does not bother soldering heatsink into chip anymore (saves 50 cents). So newer CPU can not be overclocked without disassembling and replacing some parts. Results are not bad, but in reality it is just 30% speedup.

brisance 14y ago

What's the point of all that computing power if there's no "killer app" to make use of it? These days the devices that interest me are those that consume little power, have perpetual internet connectivity (WiFi or cell tower), and are instant on/off. It has to be portable and durable. Everything else is icing.

amalag 14y ago
As of the core duo generation, people don't need to upgrade their computers in order to browse the web or use Microsoft Office. Office is the eternal killer app for the masses IMHO (excluding the browser), and it doesn't need the latest generation intel chip. As we know so many apps have moved to the browser.
noahl 14y ago
Just at a guess, really really good speech recognition might be a "killer app" for big, vectorized chips. I haven't tried to implement it, though, so I don't know how much power it would really take. Apple and Google seem to do all right with a little phone CPU.
- api 14y ago
  As far as I know, Siri off-loads something to Apple's servers. But maybe I'm mistaken.
jiggy2011 14y ago
You need something requires low latency as well as power, otherwise it usually makes more sense to offload the extra work onto a server (with the added bonus that it's difficult to pirate something when there is code isn't running on a client machine, e.g the new Sim City). So the most common example would be games, with the new console refresh we might see a bump in technical requirements for PC games in order to keep up.

saturdaysaint 14y ago

Hyper-competition to make a faster browser and faster websites has to be a factor. In the days of packaged software, you might be happy to get a handful of performance updates before a flashy (and more CPU-intensive) new version came out, but any web service worth its salt is built to load almost instantly; snappy usability is important when competition is a click away. And just think of how much horsepower has been saved by abandoning Flash. The end result is that the average software user (Facebook, Google, Gmail/Hotmail, etc.) has probably seen the software they use actually get faster in the last 10 years.

api 14y ago

It's economics. I can attest to the fact that most five-year-old chips are good enough for 95% of software I want to use. So there's no real demand outside of big data and data center applications. The only way to keep pushing performance would be for software writers to invent new and compelling things to do that genuinely require more power. It can be something as simple as a game-- I was a teenager when Quake came out, and I remember scraping together as much money as I could to get the absolute fastest machine I could (A P120) to play it. Before Quake came out, my old 486 DX2/66 was fast enough for the Linux hacking, adventure games, and BBS/Internet stuff I was doing at the time. We need another Quake... or something equally cool. We need killer apps. Otherwise computers are going to go over like the aerospace industry did. We stopped pushing the envelope there in the 70s, again for economic reasons.

EthanHeilman 14y ago
Technologies like the oculus rift will force a resolution and latency arms race (the resolution arms race is already in full swing thanks to attempts to surpass apple's "retina display"). Of course this arms race will take place mostly on the GPU and display side of things. I see the CPU becoming more of a coordinator between custom processors (think the video decoders on iphones) than the main show. * What are the computational bottlenecks for driverless cars? Vector operations? * If every car needs a high performance computer to drive how will this change the semi-conductor landscape?
qznc 14y ago
The point is, it must be about something noticeable other than the graphics. For example, a good AI should easily be able to exhaust all the cores you can throw at it. However, it is much harder to sell. In those days they could show you some pictures of Quake and you know you wanted that. It is harder to make "pictures" of AI behavior. Alternatively, exchange AI for physics simulation. Maybe the sandbox genre (Minecraft, Dwarf Fortress) will grow. Simulating huge worlds requires lots of processing power.
- api 14y ago
  I did a lot of work on evolutionary computation a while back. Now there's something that can soak up some cycles. It's also embarrassingly parallel. (For those who want to trot out the "GAs don't work well," "evolution is slow," etc. mantras, please delve into the literature first. EC has some interesting and unique results, the most important of which is that it's the only "AI" technique I'm aware of that can demonstrate true creativity. EC can invent novel structures ex-nihilo. Everything else can just optimize, find maxima, etc. Another way of saying it is that EC can find maxima involving unknown, novel parameters that it invents along the way, while everything else requires that you know all the parameters to be optimized ahead of time. So EC can work with unknown unknowns, not just unknown values of knowns.)

api 14y ago

Starting in the 90s up until today, the CPU surged ahead of everything else: disk, memory, and network. Today we have very, very fast chips that are bound by the speed of their connection to their memory, disk, and each other (the Internet). Memory has improved with DDR technology. (What are we up to now... DDR4?) Disk is improving with SSDs, which offer a huge speed improvement. If you've never used a box with a SSD in it, it must be seen to be believed. But the real bottleneck I think is the Internet. Imagine what we could do with gigabit peer-to-peer IPv6 networks in areas like games, collaborative development tools, teleconferencing, etc. Things like virtual reality "metaverses" could really arrive, etc. These in turn would require more CPU power and parallelism. Super-fast gigabit+ WANs would deliver the killer apps that we need to keep pushing the speed barrier. Once again, it is the Internet carriers that are the barrier to all progress. Breaking the telco oligopolies is the thing that must happen to unlock the future. Everything-- maybe even the continuation of Moore's Law-- depends on that. The telcos, cable companies, and wireless carriers are the trolls guarding the bridge to the future.

stevenameyer 14y ago
The internet has become a necessity in a lot of peoples lives and with such a high barrier to entry and only a few telcos controlling peoples access there is too much room for abuse of power by the telcos. We need to recognize that access to the internet is turning from a tech issue to a social issue and start policing it as such.
TillE 14y ago
Having a speedy home connection is well and good, but what really needs to improve is mobile data. Think of stuff like Google Glass, and all the possible advantages of having a constant connection everywhere you go for a two-way exchange of data. Now imagine your data bill after just a couple hours of live streaming HD video. Something's gotta change. The bandwidth keeps increasing as technology improves, but the price per bit is actually getting higher.
- api 14y ago
  Mobile data is the worst, by far, but if we're talking about pushing the performance envelope then we're talking about PCs. Mobile is constrained by power consumption as well as network bandwidth, so you're not going to see blazing fast continuations of Moore's law there. I really think Google Fiber nationwide would lead to killer apps that would revitalize the PC and the high-end chip market. I'm thinking of rich constant collaboration, virtual reality, next generations of MMORPGs where you can really richly interact with the world's physics (e.g. shaping materials, in-world chemistry, etc.), tele-presence, digital telepathy with brain-computer interfaces, distributed supercomputing, ... lots of stuff. All those things are things that are CPU-hungry. Many are embarrassingly parallel, so they can utilize many-core chips efficiently. If Intel, AMD, Dell, Apple's PC/laptop division, IBM, etc. knew what was good for them they'd get together and invest in gigabit+ broadband nationwide. It's also got to be network-neutral, since all the things I mentioned above are at least somewhat P2P. The PC industry is spiraling down because of a lack of killer apps for high-end boxes. Most of these are waiting on fast networks.

MilesTeg 14y ago

If you want to see Moore's Law in action today you really need to look at GPUs rather than CPUs. Increased performance in recent years has mostly come from increased parallelization which has limiting returns on the CPU( quad cores anyone?) but does wonders for computer graphics.

bhauer 14y ago

I agree with the author, although I don't buy the literary ploy that this realization just hit him recently. This has been a development building within the PC sphere for years. Not only that, many professional and countless more no-good hobbyist opinion-makers (like myself) have made this point over and over. I'm nobody, but I have opined about this and the insipid taint of Good Enough several times [1]. Like "api" at the top of this comment thread, I feel there are bottlenecks that keep the desktop PC, and the laptop PC to a degree, constrained to the roles that we have presently envisioned for those devices. My opinion is that desktop display technology is ultimately the root of desktop stagnation and, speaking positively, could be the root of a desktop renaissance [2]. I agree that bandwidth progress has slowed and storage capacity progress has slowed [3], but point out that only display technology has actually outright stopped and regressed. In 2001, the IBM T220/221 monitor achieved high-DPI on the desktop (3840x2400 at 22 inches). Since then, we banished high DPI from the desktop and obsessed about a marketing term named "HD" because we allowed the rules that govern our living rooms' displays to take over on our desktop. That we accept 24" desktop monitors with 1080 vertical lines in 2013 is heartbreaking. I agree with api that bandwidth is to blame. But bandwidth hasn't stopped progressing and regressed (well, with one exception: mobile data caps). More importantly, I feel that behind the bandwidth complacency is display complacency. Put a 40" high-DPI desktop monitor on your desktop and I think the demand for ultra-high-definition content, high-clarity user interfaces, bandwidth, computation power, and storage capacity will all increase. Intel sells more processors, HP/Dell/Lenovo sell more PCs, and oh-boy would I be so happy to finally get a real monitor upgrade. [1] http://tiamat.tsotech.com/technology-sucks http://tiamat.tsotech.com/technology-sucks [2] http://tiamat.tsotech.com/displays-are-the-key http://tiamat.tsotech.com/displays-are-the-key [3] http://tiamat.tsotech.com/boring-technology http://tiamat.tsotech.com/boring-technology