Report: IBM researcher says Moore's Law at end
Moore's Law is maxing out. This is an oft-made prediction in the computer industry. The latest to chime in is an IBM fellow, according to a report.
Intel co-founder Gordon Moore predicted in 1965 that the number of transistors on a microprocessor would double approximately every two years--a prediction that has proved to be remarkably resilient. But IBM Fellow Carl Anderson, who researches server computer design at IBM, claims the end of the era of Moore's Law is nigh, according to a report in EE Times.
Exponential growth in every industry eventually has to come to an end, according Anderson, who cited railroads and speed increases in the aircraft industry, the report said.
"A generation or two of continued exponential growth will likely continue only for leading-edge chips such as multicore microprocessors, but more designers are finding that everyday applications do not require the latest physical designs," Anderson said in the EE Times' report. Anderson also cited the staggering costs of research and fabs (factories) as a formidable barrier for continued advancement. Few companies can afford chip plants that typically cost billions of dollars to build and maintain.
So, what does the future hold? Anderson cited three technologies: optical interconnects, 3D chips--which have circuits and components stacked on top of each other--and accelerator-based processing as seeing significant advancements, the report said. The latter technology, accelerators, is hot right now.
In addition to IBM, companies such as Nvidia and Advanced Micro Devices' ATI unit supply graphics-processor-based computers to accelerate scientific, engineering, and animation applications. Intel is also expected to bring out its Larrabee chip later this year or early next year that can be used as an accelerator.
Brooke Crothers has served as an editor at large at CNET News, an editor at Dow Jones' Asian Wall Street Journal Weekly, and a senior editor at InfoWorld. His CNET blog covers chip technology and computer systems, and how they define the computing experience. He also contributes to The New York Times' Bits and Technology sections. He is a member of the CNET Blog Network and is not an employee of CNET. Disclosure. Follow Brooke on Twitter @mbrookec. 
accelerator-based processing, will all come into play with the price
of chip fabs increasing. I would also suggest that these are near term
issues (ten-twenty years) and that the future of circuit production has
a very bright future with Moore's law still playing a major guidance roll.
The difference I see in Moore's law is that it does not state that future
chips must run with electrons, or be flat. On the high end, I would suggest
that "Free-Space Photonics" (light intersecting light) will rule the roost.
On the smaller, mobile market, I would suggest that genetically engineered
photonic crystals will replace near term 3D silicon. I also see circuit
fabrication plants of the future costing one tenth the cost we see today.
I can't wait to see what's next. What happened to the experiments in quantum computing?
If "3D computing" is the next phase, and if this is stacking circuitry within a microprocessor, then I would suggest Moor's law will live on for a while yet. Moor never put a limitation on the physical size of an IC.
Also, the line "designers are finding that everyday applications do not require the latest physical designs" has been true ever since computers came into the main stream. This is where the phrase "good enough computing" came from. The computers we have at work are years and years old, heck one of them is a Duron! But it still does its job perfectly well. It's used for viewing service manuals, using office applications, spread sheets and databases (nothing too large). There are no applications run on it now, or at any time, that require better hardware. You'd have to be a pretty "power user" to require better hardware. Give it a few more years before new "every day" apps require faster hardware.
I think More's law is safe as long as IC packages can be made larger,
They question is, is this a pause or the end?
...and I think IBM was one of the companies that solved that issue.
I also agree with an earlier poster that there was no mention of quantum computing, which may still be years or even decades off, but it will be a game changing.
BTW, the argument that people are finding they don't need all this processing power is pure bunk. I guess they never heard 'Matt's Law"...
"Software Developers abhor a vacuum and will fill it with useless code at the earliest convenience."
If you have any doubt, think about it next time you install an application skin, or run more than one way launch an app (Docks, Toolbars, etc.), or have fancy 'glassy' UI effects turned on.
In any case, we may have reached a limit with this technology, and there is something new on the horizon - just as 60 years ago they had reached a limit with vacuum tube technology.
Still, 70 million users of all computers that are known to be on the Internet are Windows 98. Whether that's Windows 98 or 98 SE is unknown. Also, I don't know how many companies I go into and see the office still using Windows 95, 98, Me, or 2000 on a clunky, bad old Pentium, Pentium Pro, or Pentium II/Celeron/AMD K Series computer. When I suggest buying a cheap, new eMachines, Gateway, HP, Compaq, or Dell that still has Windows XP on it, then give me this crap about how they paid "X" thousands of dollars for it and that it doesn't need to be replaced. Most of my service calls are for the these computers not running YouTube and iTunes and such. I understand there is iTunes for Windows 2000, but that's if it's on the correct computer, not these antiques. Moore's Law died with businesses and consumers that, since around 1997, don't have to "state of the art", whatever that means. They get along just fine with their old hardware.
Then, for the gamer, Moore's Law died in 1999, with the release of the only 128-bit consumer computer, the Dreamcast. It used a Hitachi SH-4 CPU. The SH-5 and up are 64-bit CPUs, taking a step back. The PlayStation 2's Emotion Engine (***?) is composed of two 64-bit CPUs slapped onto one socket and given a 128-bit bandwidth. True, the system is 128-bit, but the actual computer is 64-bit. In essence, this could be considered a multi-core CPU. Since then, the original Xbox was 32-bit, the GameCube was a G4, ergo, 32-bit system as well, the DS and PSP are both 32-bit, running ARM CPUs, as is the iPod since 2005 (sans Shuffle models), the Xbox 360, nearly running at a whopping 3 GHz, is 64-bit with it's solo-core G5 CPU, the Wii is also G5, but not near the clock speed of the 360, and the PS3's Cell CPU is actually a bastardized, prostituted version of the PowerPC G series CPU line. In essence, it could be considered a G6, if it wasn't so poorly designed. It, too, is only 64-bit. And recall, that four 64-bit consoles existed in the 1990s, and two were poorly, and I mean poorly, designed. The first on market was Atari's Jaguar, which used, oh, I don't know, a bunch of off-the-shelf 32-bit CPUs to reach 64-bit bit, and Atari just had the bandwidth go up to 64-bit. If it's true 64-bit, I doubt it. The second were both based on the Hitachi SH-2 CPU: Genesis 32X and Saturn, both from Sega. The 32X was just a cartridge that plugged into the Genesis slot, and was still at the Genesis video resolution, sound capabilities, and 16-bit bandwidth. Unless somebody knew some fancy, smancy programming tricks (nobody did), this was poor excuse to slap two 32-bit SH-2 CPUs into a machine. The Saturn was designed almost the same, but had a ****-poor polygon processor inside of it. I don't recall correctly, but I think the Saturn had this thing about how complex the two SH-2 CPUs had to be programmed, yadda yadda yadda wash rinse repeat. With the N64, we've basically have gotten to 64-bit and stopped. Otherwise, we'd be at, according to Moore's Law, for computers, somewhere near 4096-bit by now, or even out of bits an near a 1-byte CPU, running at a Hz of unknown prefix. As for video games, we'd be near 2048-bit, also running at a Hz of unknown prefix.
Moore's Law really died, to simply put it, when the CPU speeds around 3.5 GHz started melting the sockets, ends, boards, and wires. Why? Well, somebody at Intel, Moore, forgot CPUs get hot when in use. If he thought it out, a healthy, say, 4096-bit CPU would be about two feet by two feet. And is there anything wrong with that if it works in a tower unseen? Think about it.
Also, what I just wrote I ripped off from Robert X. Cringely/Mark Stephens.
Moore's Law has nothing to do with bits, speed or Windows. Doubling the number of transistors.
Why is this so hard for 99% of the press and 95% of the "Techies" out there.
Doubling the number of transistors.
Doubling the number of transistors.
Doubling the number of transistors.
Doubling the number of transistors.
Doubling the number of transistors.
Doubling the number of transistors.
Doubling the number of transistors.
Doubling the number of transistors.
Doubling the number of transistors.
All that aside, there are physical limitations that prevents the continuation of the "Moore's Law". However, engineers usually want to curb expectations because the solutions have not been found (yet).
Just like the auto industry, its been plateauing for a while, wait till a new fuel efficient or super efficient design is adopted.
However, microprocessors are commodities, are manufactured using an mass production method (lithography on wafers), and must be manufactured efficiently. In hindsight, wise people should have seen this as the limit years ago. DEC never made money on its Alpha chip, as the fab costs were too high to support a single line. The other big RISC players (SPARC and MIPS) were all fabless. That was over a decade ago. The economics of microprocessor manufacturing were already an issue.
Today, Intel, IBM, and Fujitsu have organic foundries. AMD spun its off into GlobalFoundries. Texas Instruments plans to go fabless for its DSPs sub 45nm. IBM and Fujitsu probably exit at some point as well. But eventually, even mighty Intel will hit the wall.
Processors with heterogeneous cores (including accelerators) will be here soon. They tackle the heat barrier.
Chip stacking and wafer scale integration will become the next big thing once the financial costs of migrating to the next fabrication process becomes the limit.
There really is, from a Physics point, where things can be made only so small before they no longer work well. As for Quantum Mechanics, which goes hand-in-hand with Quantum Physics, it gets harder and harder for Chipsets, Graphics Cards and so on to get small enough to be sub-atomic.
I'm not decrying Gordon Moore. He said it first. Things have gotten to the point that making them smaller is nearly pointless. I see 22nm, but smaller is just that. Smaller. As it is, most 64-bit chipsets are in reality two 32-bit chipsets jammed together. True 64-bit chipsets are hard to make, very expensive to fab, and require more than double the cost of a fab that makes a 32-bit chipset.
Gordon Moore, in saying his Law was near it's end, I seem to recall he predicted somewhere in the neighborhood of five or so years, is right. The technology may be there, but the expense will no longer allow that technology. His Law may live on, but the bucks are stopping and not in the hands of greedy bankers. Instead, it would be too expensive to build or retrofit fabs for makers of the Chipsets to justify the cost. Not to mention the costs to even the most 'nerdy' of computer geeks.
I've been around computers for 40+ years now and have seen punch cards become punch tape become magnetic tape, hard discs that were 20 inches, 5 1/4" floppies to 3 1/2" floppies, to the miniscule-memory of the first commercially available (very expensive) production hard drives and on until now we have NAND memory. I can see the day when there are 128-gig SD cards that will replace all hard drives, with people just having to transfer whatevfer they wish to save from that card to another of the same size or larger.
Gordon Moore is right, the 'Law' is nearing it's end, at least it's usefull end, but a new 'Law' perhaps postulated by Mr. Moore or even in this posting, will be the next thing. NAND and Solid State memory to something like the SD Card, with true 22 nm chipsets, true 128-bit chips the last of the monetarily feasible chipsets. And then Personal Computers without all the latest bells and whistles, just most of them if even then, will run in the $10-$15,000 range.
Take what Gordon Moore said about his 'Law'. 'Postulation', whatever you want to call it to the bank. And realize that money is everything, from the maker of a device to the end user. If the Law isn't dead as stated above. then the end is near.
No, Gordon Moore did not PREDICT anything. He OBSERVED a trend and remarked upon it, which later became known as his law.
I recommend a modicum of research by you and your editors before publishing. I am not sure how much to trust anything else stated in this article. In case you are having trouble researching, might I suggest a simple google search:
http://www.google.com/search?q=moores+law
The first hit (Wikipedia, unsurprisingly) states: "Moore's law describes a long-term trend..." Note how the word "PREDICTION" is conspicuously absent. It further goes on to state that, "The trend was first observed by Intel co-founder Gordon E. Moore in a 1965 paper".
Now, you might not consider Wikipedia to be the answer to all of your questions, and if that is so, I would applaud you. But if Wikipedia is contradicting your statements, don't you think you should at least dig further to find out the truth?
Moore's law is about the rate at which the size of transistors (a semiconductor technology, usually silicon based), will decrease over time as the fabrication technologies advance to allow each transistor to be smaller.
This process of reducing the size of the transistors allows more transistors to be placed on a chip of a given size, and since the smaller transistors are closer together, increases the speed of the device by reducing the distance the electrons must travel between transistors. This has been the main driving force behind the increase in computing power over the last 40 or so years.
Moore's law makes no comment about other technologies such as biotechnology or quantum computing, nor does it address other ways of increasing computing power such as 3D chips with mulitple layers of circuitry.
Moore's Law, as stated, in relation to chips based on transistors, will not go on forever, As mentioned in other posts, there are limits at the atomic level and whether we're five years or twenty years away from those limits is irrelevant. We are getting very close to those limits.
However none of this says that we won't continue to find ways of hugely increasing computing power in the future. It just says it won't be with transistors, and it won't necessarily follow the pretty constant rate described by Moore's law. There may be periods of more gradual increase, interspersed with sudden leaps brought about by radically new ways of performing computation with new technologies.
Moore's Law is indeed getting close to it's limits, but great increases in computing power are none the less coming.
- by mikeburek April 12, 2009 8:53 PM PDT
- Moore's Law is not some clairvoyant prediction about the future, it was Moore noticing a trend in the past of what has happened already, and assuming is would keep happening. Moore was a smart man, but he did not have insight into the far future of microchips.
- Like this Reply to this comment
-
Showing 1 of 2 pages (32 Comments)When a project goes according to plan, I'm surprised there aren't press releases declaring Murphy's Law dead.