Number cruncher charts rise of the personal computer...

[7im]

... and the fall?



Discuss...

[jrweiss]

Until something revolutionary comes out that can REALLY replace the PC (which apparently includes laptops), I think the top of the curve is going to be relatively broad.

[MtM]

From the linked article in OP:

The key conclusion: smartphones and tablets are well on their way to becoming the defining personal computing platforms - in terms of units shipped - for the early 21st Century.

First of all, consumers consume by definition, and most consumers consume media ( be it in the form of television or internet sites, or social media like facebook or the likes ). So it's not surprising most consumers 'settle' for something focuses on enabling this and decide they don't need more.

Would that mean lesser growth in absolute donor numbers for f@h unless f@h refocuses not on consumers but rather entities like foundations, schools and corporations? That would still mean lesser growth in computing ability when looking only at the number of donors, leaving out the exponential increases in compute abilities harnessed in modern and future compute focused devices. Because, even while consumers might settle for consumer devices, those other entities rely more and more on devices with high compute ability if not for their focus area ( think of the webservers for aforementioned Facebook ) then for enabling their core focus like how an accountancy office needs pc's to work efficiently ( it was possible with pen, paper and abacus but I don't think that would be as efficient as having a pc do most of the repetitive work ). So in the end, computing power will keep increasing, and the availability of said compute power will also keep increasing, but where the compute power will be located will shift towards bigger entities then lone consumers.

That's my impression of it anyway, I think it's a correct one. Not sure about implications, but I do know it feels good to have heard PG was also putting some attention towards the cluster client sometimes requested on this forum by people with access to these compute powerhouses. There might come a time when a client such as that would contribute more cycles then all us common/simple donors combined.

Until something revolutionary comes out that can REALLY replace the PC (which apparently includes laptops), I think the top of the curve is going to be relatively broad.

Those sliding tablets with win8 seem to be able to do that... and those are not that far away according to some coverage on CES I read the last week. And yeah, win8 will run on x86(-64) but it will also run on ARM and I'm not sure which one will be more popular if and when the software base for ARM get's the same software packages needed for professional use like a proper Office suit with synchronising abilities. ARM is better suited for low power computing then x86, at least it has been so far. And even if it comes down to ARM levels of power consumption, it wouldn't be capable of offering higher compute abilities with that power envelope.

[Jesse_V]

That is a fascinating graph. Very interesting and informative; thank you for posting it!

But I think we've hit or have almost hit the peak speed a single computing core can do. Over the next decade, I expect to see processors with more and more cores become available. The thing that holds this back is getting applications (a large portion being games) to scale to all these additional cores. I mean F@h has no problem doing this now, but I think in general scalability is going to be the one thing that holds core counts back. I think this is evident from the graphs. The runtime of applications can double if the speed of the single-core processor doubles, which makes getting faster computers a really attractive idea. But unless there are programs that can use dual-cores efficiently, adding another core doesn't necessarily make the single application twice as fast. I mean on a dual-core you can now run two applications at the same time at normal speed. Currently using dual-cores is not a big deal in a lot of applications (except Minecraft ) but how about quad/hex/oct-cores? I think there comes a point that most people either don't require the extra speed, or there is simply less and less of a gain from upgrading their processors since their applications can't take advantage of the full processor. F@h users don't have either of these problems.

[bruce]

FIRST: That graph shows a definite migration from devices that use a wired source of power and a rechargeable source of power. Though I'm exaggerating slightly, I'm willing to state that FAH is designed to take advantage of hardware that runs on power from the mains, not rechargeable power, simply because the "wasted resources" that it was designed to take advantage of were originally the idle state of hardware that consumed power 24x7.

Would that bias be corrected if we removed the tablet-type devices and replaced them with something else that FAH might be able to use?

Where are the statistics for big-screen TVs and other forms of Entertainment Centers? Their computational power is still growing but I don't see a graph in units-per-year.

What's happening in the game-console market? Was the PS3 client just a one-time event or is it rational for folding AT HOME to tap that source of computing power?

What else is missing and if it were included, how would that change this discussion?

SECOND: We still have Moore's Law. A flat PC curve in units per year still increases processing power at 2x every 1.5 years. Sure, it's nice to beat that rate, but FAH is still growing with that as it's base rate.

[7im]

The author of the article pointed out the potential/eventual decline of the PC. However, it won't decline like some of it's predecessors, IMO, more of a leveling off before an eventual, but slow decline.

Also note that Macintosh line is still rising well. And note how there are several smart phones, all with sharp increases and high volumes. And yet, we do *not* see a sharp decline in PC numbers as a result. Hand held devices are supplementing our personal computing power in vast numbers, but not yet replacing it. Everyone has an iPhone AND an iBook. They haven't tossed the iBook yet...

Even if the "universal" hand held device finally arrives in a few years and we all toss our laptops and personal computers, we'll all still be folding on our powerful Media Centric HTPC / DVR / Streaming / Gaming Console / OLED Edutainment Wall Screens / Personal Cloud Server. Intel, AMD, Sony, Vizio and the rest will make sure of that.

[Hisuichan]

The graph simply shows that what has become a rich market in IT slowly merged into two main branches: PC and Apple, since all the rest couldn't sustain itself over the years. Will that curve decline? I don't think so, at least for a long long time. At worst, it will indeed be as 7im said, there will be a steady leveling off since there will be a saturation of the market, everyone and their dog will have one or two computers. I imagine future purchases will still be made, either to upgrade individual parts or the entire computer, as technology progresses and becomes faster as per Moore's Law. The fact that it was possible to add some variety with the advent of smartphones and tablets the last half dozen of years has not had a negative impact on PC sales, if anything, these new tools were complimentary. I doubt most people with a smartphone don't have a computer.

However I don't see tossing away PCs (laptops) happen at all, because PC will always have the edge on mobile computing and it would be counter-productive to act otherwise (unless we hit the absolute performance ceiling... but sky is the limit). Rather, we'll just keep accumulating new technologies and increase our folding power x-fold, as long as we can sustain our electricity hunger. Either way, good times ahead for science I say.

[beer]

If "personal computer" is a normal Desktop then I dont think it will disappear.
There is people like me who work in front of a computer every day, and therefor needs comfort. I have not seen anything that can replace any setup that can replace a desktop if one needs comport. If one dont need 2+ screen the dekstop can be replaced by a laptop/taplet with mouse,keyboard and monitor but I need 2 screen to work ( I have a lot of stuff that I need to look at)
Another aspect is price/performance. If one just need a computer you can get the cheapest as a desktop.
a third area is performance segment. Some people need decenet amount of power at the desk. It might be latest video games, 3d modeling, videoedeting , picture edeting or math modeling. Personal I am a gamer so I would not change to laptop since they cost to much compared to what I get in games. I have tried to buy gamelaptop but it went to quickly to old. At work I am in mat modeling (full scale farm model http://fasset.dk/) . Even with 60% speed improvements I made last somme it can stil put a i7 2600 down for 1 to 30 min depend on the size of the information it need to process. There will go many years before I can conisering buying anything small

But I also see a trend for people who want to pay a bid extra to have small device that they can take with them. With other words: I see that the market for Desktop computers will at some point fall but never vanish. That part will be taken over by laptop, tablets or mobile phones. Or maybe not replaced but make a good supplement to the desktop.

Sorry for my bad english. I am not native english speaking

[k1wi]

It looks like PC is laptops and desktops, but it would be rather interesting to see the two differentiated in the chart. I've found that lately that I've been rolling out an increasing amount of desktops or laptops that are used basically as desktops, and this is primarily to Iphone users.

My suspicion is that tablets and smart phones are complementary to immobile computing, rather than direct substitutes. People like having the mobility of a phone, but also really like the fixed immobility of a desktop/stationary laptop. At a philosophical level I feel it appeals to humans. So at least at the moment, my clients want a fixed computer as well as a mobile internet device - even if the fixed computer is just to view photos and check email (which they can do on their phone). Try and dissuade them from that and they will look at you odd - Desktop PCs are still 'the computer' in the minds of many. That said, I do think that there will be some important evolutions in the area, NVidia's plug in portable computer is an interesting concept of what will happen when or if smartphones 'catch up' to a level of computational power that will rival desktop/laptop usability while the rise of all-in-ones is going to ease the way for T.V/PC convergence - once the 12ft interface is resolved.

[7im]

Still have that last mile issue in the US, and many countries, in addition to that 12 ft. thing.

IR2 that complimentary thing, look at the latest Mac commercial for iCloud. THAT is the "universal computing" and "convergence" people have been waiting for... download a song on one of the devices, and play it on any of the others. And you take along the computing device that suits your needs for the next hour or next day. Phone for on the go, laptop for work or school, home system for powerful computing and multimedia. Plus, no one wants to do their home finances or run tax software on a mobile device that is all too easily stolen, lost, or dropped off a table. Home server, home network, home whatever will always be there as the computing center/anchor for all of this new supplimental computing technology that's coming down the road, IMO.

[k1wi]

This discussion makes me think about the implications of the above discussion to the 'potential and actual supply market for FAH'. By 'potential and actual supply market for FAH' I mean the amount of FLOPS donated by the total FAH community. To my knowledge, these factors have been tackled in the forums, but officially, have only been touched upon in the section five of paper 67 (Folding@home: lessons from eight years of distributed computing). In my opinion they do have a significant impact on the future of FAH/DC...

As a starting point, holding all other thing constant, FAH increases because of Moore's Law + increasing number of global (internet connected) computers. If the number of global (internet connected) computers becomes constant, FAH increases because of Moore's Law.

Folding@home: lessons from eight years of distributed computing adds additional variables by considering the effect of publicity and paper releases, participant psychology (i.e. the effect of competition and altruism) and per-volunteer costs (which it sees as insignificant unless power prices increase rapidly in a geographical location).

However, the above discussion raises additional concepts, such as what computers they will use in the future and how users will use them.

What computers, such as a Desktop or a Laptop or a 'handheld device' (tablet/smartphone) is important as each has a different utilisation profile - Desktop = best candidate, Laptop = good candidate, handheld = poor candidate (at least in the current - immediate future). Changes in the total number of devices and in the proportion of each type will have impacts on the growth of folding.

In between What and How is the specialisation of computers and the increasing stratification of these categories, eg in the desktop PC category has evolved to range from low powered 'good enough' systems such as ATOM or Brazos for web browsing and word processing to high powered gaming rigs. No longer do you see a P4 or AMD equivalent in virtually every PC sold. Basically what we've already been discussing, but in effect, as people specialise their computers for a given task, there is relatively less excess for FAH to mop up.

How is just as complex. One could argue that Per-volunteer costs are increasingly less insignificant. This is not a criticism of the above paper, or a suggestion that F@H does not value the investment by donors okay? Let us be clear that I for one know F@H values its users, I see it every time a staff member responds to me. One area in which per-volunteer costs are increasing is because offset against increasing power and increasing power efficiency/FLOP is the gains made in idle efficiency. Idle efficiency is the downclocking of cores, power gating, hurry up and get idle and faster start up/restore/wake times etc, and is arguably one of the main ways manufacturers are improving power efficiency at the moment. As computers become more efficient at being idle, the delta in power usage between idle and full utilisation increases. This delta is amplified by real electricity prices increasing in most markets (in mine real power prices have increased by 42% since 1992, that is 42% higher having adjusted for inflation). The result is the marginal cost of utilising idle processing power increases and this absolutely puts pressure on DC. As per-volunteer costs increases, DC becomes less unattractive relative to local computation. This would be manifested in it being more efficient to donate money to F@H and leave idle processing power idle than to simply fold for them. In this specific case, per-volunteer costs will always have to be less than the cost of local computation minus any tax incentives regarding donation to a tax deductable charitable organisation (Stanford). Perhaps this can be seen, or measured, in terms of the shift to the Cloud. Or one might say more specifically, the re-centralisation of computing through PaaS, SaaS and IaaS...

Increasing numbers of computer devices per user (i.e. from One desktop or laptop each to one desktop or laptop and one mobile device) may increase the idle computational time on DC friendly devices such as the desktop, as tasks such as browsing are shifted from the PC to the smart phone, but conversely, may reduce the amount of time that DC friendly devices are powered on or 'awake', particularly if start up/restore/wake times decrease.

Quite how decreasing hardware costs interacts with that I'm not too sure. Perhaps it increases marginal per-user cost relative to the fixed cost of purchasing a computer, making utilising idle processing power less attractive. Perhaps it increases people's specialisation of devices. Perhaps decreasing hardware costs shortens the purchase cycle, meaning people future-proof their computers and have less idle capacity at the beginning of the products life-cycle. Perhaps is opposite to the three previous and it allows people to purchase computers over and above what is used!

Finally, intimately related to hardware costs is the second to last area I want to acknowledge in this post: software demands. Positively through innovation and negatively through bloat, increasing faster than hardware can supply. As 'active and user' applications directly influences the amount of idle processor power this too influences DC. I think software is still pushing hardware and pushing pretty hard, but I am not sure whether this is positive thing for FAH or a negative thing.

I am interested in hearing about what other areas people think will have an impact on DC.


...You did say discuss in the OP 7im