13

There was a time where computers often just stayed on all the time. Computer monitors did not have power save functions... they were either on or off. Processors such as the 6502, 68000, 8086, 286, 386, 486 and the Pentium typically ran at a fixed speed all the time and did not attempt to conserve energy.

But apparently there was a point sometime in the 1990s where a move to trying to save energy started to occur. Is there a specific cause or reason for this?

For Intel processors, it seems like the turning point may have started with mobile processors and speedstep, to try to get more battery life.

  • Think about the total amount of computers consuming energy in 1980, then end of the 90ies. You'll find the share of energy consumed by computers is a "bit" bigger... – tofro May 27 '17 at 19:45
  • 4
    In the PC world, the 386SL was introduced in 1990 with energy-saving features, and Energy Star was founded in 1992. Pentiums, SL chips and late 486DX4 would reduce their energy consumption when halted. But I suspect the real answer lies in super-computing territory, or at the other extreme in battery-powered micro-controllers (space probes anyone?). – Stephen Kitt May 27 '17 at 21:55
  • 11
    I suspect it really was driven by laptops. Space probes, etc. are a small market. Laptops are a HUGE market and that was pretty clear early on even though they started out expensive. Once it became practical to run a laptop on battery for a couple of hours, minimizing power usage to maximize battery life became very important. Many of the innovations - e.g., sleep/hibernate of entire systems, monitors (really video controllers) going to sleep instead of just to a screen saver, even hard drives powering down, started with laptops and moved to desktops since the cost to do so was minimal. – manassehkatz May 28 '17 at 2:05
  • 1
    @alephzero I wasn’t thinking of the need to save money, just the need to run very fast CPUs as usefully as possible without frying them; lots of the issues there are relevant for energy saving (power distribution and clock gating over large areas of silicon, powering down parts of CPUs...). I suppose the question is more about mass market than niches though, and more about actually reducing energy consumption than optimising it. – Stephen Kitt May 29 '17 at 17:54
  • 1
    I would say it's the 2002 era when Pentium 4 and PowerPC 900 proved unsuitable for laptop, which forces Intel to develop Pentium M. Desktop continued for a few years before Intel realized 4GHz is unfeasible and another couple years before they get core 2 ready. CPU cooling fan industry took quite a hit at that moment. – user3528438 Jun 12 '17 at 14:37
9

While the earliest date can be disputable, big shift can be pinpointed as the Transmeta Crusoe CPU release in 2000.

Transmeta was a secretive company working since 1995 on a "mystery product", employing some celebrities from the IT world (like Linus Torvalds.) Its "mystery product" was a new CPU with focus of power consumption. It offered unprecedented battery life to laptops.

The fame of the company was short-lived. Crusoe, while revolutionary, didn't entirely perform to promised specs. It was fairly expensive for what it offered performance-wise and it had minor compatibility problems with x86 platform. Still, the idea caught on, and both Intel and AMD released their own "low-power" CPUs soon after (putting the last nail in Transmeta's coffin.)

For a time the development ran in parallel, high-performance, high-power CPUs for desktop, and "energy-saving" portable CPUs. The manufacturers ran against a wall when trying to maintain Moore's Law concerning CPU speed though - power consumption for anything exceeding 4GHZ was excessive, and heat dissipation was difficult.

That's where the race for higher clock speed ended, and the race for more cores in one chip began. Simultaneously, lower clock speeds allowed, and heat dissipation problems necessitated reduction of power consumption and tout that as a marketing point, "energy efficient".

Another high point in the race for energy efficiency was the One Laptop Per Child project.

OLPC, starting around 2005, was a non-profit developing special child-friendly laptops with many completely novel solutions - including a system-on-chip, Geode processor, with power demand low enough that the laptop could be charged by an attached crank.

One of points of the program was "Give one, get one" - the system would be unavailable in normal trade, except this program: buy two, one goes to a child in need in a 3rd world country, you keep the other.

Asus, tasked with development of the device, did something rather mean - using the experiences gained developing OLPC XO-1, created Asus EEE PC - the first netbook. The price, while higher than of a single XO-1, was still lower than that of two, the new competition murdering the OLPC G1G1 program, and the new fad for netbooks started - with EEE 901 ('2008) boasting over 9 hours of battery life using Intel Atom. Along with iPhone (2007), Google G1 (2008) and iPad (2010) the race for "super-low-power CPU" began for real, and ARM became a major player.

  • 1
    Pedantically: Psion launched a product literally called the netBook in the late '90s that was a real-keyboard sub-notebook with 10 hours of battery life, Wifi and a web browser; they just happened to pick Symbian as their OS-that-isn't-Windows rather than Linux. On ARM, naturally. Asus created the market, but not the category. – Tommy May 31 '17 at 15:21
  • 1
    @Tommy: Asus eee PC was an actual PC, fully featured and fully x86 compatible, only small and low specs. Psion was a weird form-factor palmtop with a lot of features. Despite superficial similarities, they were fundamentally different devices. And palmtops, while popular in business, never gained enough foothold to really change the market (especially Psion with its weirdo inventions...). I got myself Psion Revo where they started costing as much as a common lunch, and while it seems like a superb device, it's more trouble than it's worth. Got eee900 and it's just a PC. – SF. May 31 '17 at 17:54
  • 1
    I disagree: there's no distinction in form factor as both are clamshells with have full-size keys and screens within a small delta of each other — the Psion's is larger at 7.7" but the original EEE is close enough at 7". I also disagree that processor architecture along amounts to "fundamentally different devices". The designer's intention with both was that they were net-focussed appliances. You're focussing on aftermarket usage as a differentiator, but the market's attempt to turn them into "actual PCs", including crippled versions of Windows, is what killed the segment. – Tommy May 31 '17 at 18:16
  • 1
    20 minutes later, a picture of the two side-to-side: farm3.static.flickr.com/2379/2278546566_5199527de3.jpg?v=0 . The Psion has a touchscreen, so no trackpad. But I really don't see how you could describe one as a palmtop but not the other. Also, apologies for hitting that point so hard as I know it was hardly the one you considered most compelling, it's just the easiest to find third-party evidence for. – Tommy May 31 '17 at 18:36
  • @Tommy: 20 minutes of using the two later you'd withdraw your statement. Yes, they look similar. But this is like comparing VeloX4 to Zerotracer. You can list all the similarities and differences in chassis, wheels, windows, shape, but that doesn't change the fact that one is a bicycle and the other is a motorcycle. – SF. May 31 '17 at 19:44
5

Power savings has always been critical for microprocessors and other integrated circuits because of the physics of heat transfer. The more heat generated per unit of area on chip, the more aggressive heat dissipation needed to protect the chip from over-heating. This effects system cost, chip reliability, and maximum attainable speeds. Since heat dissipation and power consumption are inseparable, it's always been a critical consideration to minimize both.

Better power usage and easier heat dissipation is the main reason that, since 1976, CMOS has dominated integrated circuit construction. This has supported greater densities and speed while keeping the heat dissipation manageable. Without CMOS, not only would we have shorter battery life in our devices, but also less powerful processors owing to the limits higher power consumption imposes on the chip's transistor density and speeds.

An excellent example of how CMOS and battery life intersect is the original Apple PowerBook 100. It was released in 1991 and was notable for being an early battery-powered laptop built around the 68HC000 processor, which was a CMOS variant of the Motorola 68000. This processor was chosen in order to improve battery life and heat characteristics when compared to the HMOS 68000 used in most desktop computers. To complement the low-power microprocessor, the PowerBook 100 also included a power management IC that would power-down various I/O subsystems when they weren't in use. The combination of low-power IC's and power management over various subsystems is the basic model used to optimize battery life for all portable computing devices.

Modern power-saving processors can shutdown or "sleep" internal subsystems, such as the GPU, when they aren't strictly needed.

3

I would say many factors influenced:

  • The performance race drove to an unsustainable consumption, not only because the consumed energy but for the heat and noise.

  • The common-user market is tending to mobility that require less heat and more battery life.

  • The number of datacenters has grown exponentially. Saving energy means saving millions of dollars in electricity bill.

  • The processors were performant enough to afford sacrificing some performance to reducing consumption/heat and increasing battery life.

3

According to Wikipedia, the ability to put CRT monitors on standby (VESA DPMS) arose from the US Energy Star initiative of 1992. That matches my recollection, of computers in the 1993-6 timeframe booting with Energy Star logos.

2

It began for real a few years after the "Heatsink and fan required" started appearing on 486-class CPUs - to be specific, about very late 1996 or 1997. At this point, heat was starting to be a major issue starting with Pentium II processors, and got worse in 1998 when the first Celeron CPU was released. At this point, mobile computers started to get cheaper, and more mainstream; Many of them had either Pentium or Pentium II processors, which weren't very power efficient and created a lot of heat (to make things worse, laptops of the era were mostly passive-cooled ones); In 1999, with release of Intel Pentium III, Intel also released Pentium III Mobile that was a slightly slower and less powered CPU, but it also was a lot more power-efficient and produced less heat.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.