[Bytesmiths]

$this->bbcode_second_pass_quote('in Scientific American (Nov 2004, pp 96-101), W. Wayt Gibbs', 'I')t was never a question of whether, but only of when and why. When would microprocessor manufacturers be forced to slow the primary propulsive force of their industry... And would it be fundamental physics or simple economics that raised the barrier to further scaling? The answers are: in 2004, and for both reasons.

[johnmarkos]

Peak computing by what measurement? The industry tends to measure in terms of ghz/chip, which does seem to be flattening out. However, in terms of ghz/US$ or ghz/cubic foot, we're still advancing. What about ghz/kilowatt?

In terms of P.O., we would be most concerned with the last one.

Even when processor speed was advancing happily along with Moore's Law, things like disk I/O, RAM, and hard drive space limited the speed of a home or office computer. By far the greatest limiter of speed was software bloat. It took less time in seconds to boot up my Terak 8510/a from a floppy than it takes to boot my laptop running GNU/Linux. Yes, GNU/Linux is guilty of the bloat too -- not just Microsoft. But to their credit, free/open source operating systems allow programmers to write smaller, more efficient versions.

[Bytesmiths]

$this->bbcode_second_pass_quote('johnmarkos', 'P')eak computing by what measurement? The industry tends to measure in terms of ghz/chip, which does seem to be flattening out. However, in terms of ghz/US$ or ghz/cubic foot, we're still advancing. What about ghz/kilowatt?

[johnmarkos]

$this->bbcode_second_pass_quote('Bytesmiths', '
')As to having all the computing power you'll ever need, I guess that leaves out stuff like real-time holographic video feeds and such. I'm sure my great-grandparents had all the computing power they ever needed as well!

[Bytesmiths]

$this->bbcode_second_pass_quote('johnmarkos', 'T')he computing measurement I'm most concerned with is interesting problems/time. By that measurement, computing topped out at a completely satisfactory level (no improvement needed) in 1981, when I started teaching myself Pascal on the aforementioned Terak 8510/a.

[johnmarkos]

$this->bbcode_second_pass_quote('Bytesmiths', '
')I would submit that you have benefitted from the growth of computing power to a similar extent that you have benefitted from the growth of cheap energy from fossil fuel -- and that you are just as much in denial of that fact as the person who drives their SUV four blocks to the 7-11 to buy one loaf of bread is in denial of Peak Oil.

[johnmarkos]

$this->bbcode_second_pass_quote('Bytesmiths', '
')In 1981, there was no Internet. (Although I used the predecessor, called the Arpanet, at that time.) Networking of any kind was limited to 56kb/s synchronous connections, available only on minicomputers (or larger) costing hundreds of thousands of dollars and more.

So by your metric, you shouldn't even be experiencing this forum!

[Jack]

There's a semi-recent paper to the same effect at http://www.sea.uni-linz.ac.at/conferenc ... sion_e.pdf

I suspect that we need to make advances in parallel processing first; transforming applications to perform in parallel takes some careful coding and will require some substantial work to convert from the present uniprocessor model. Of course, the software exists - MPI, for example - but converting uniprocessor applications will take some serious work.

There's also the question of what applications really need that additional power. Other than the gaming sector, most of us do a little surfing (with bandwidth being the limiting factor), some word processing, and some things with spreadsheets. I realize that Bytesmith has more demanding applications; graphics have always been resource intensive. And yet, I'm none too sure that the current processors are the real bottleneck, even there. Most of us use chips with relatively small cache sizes - certainly nothing like the Itanium 2 with 9MB of L3 cache!

Our motherboards tend to be slow relative to the speed of the chip, we tend to have pathetic amounts of RAM (anything less than a gigabyte), and hard drives are a factor of a million slower than RAM.

So...before I'd worry too much about chip speed, I think I'd look at having RAIDs with striping, a faster board configured for a pair of Itanium 2's, and 10 gigs or so of RAM.

[Bytesmiths]

$this->bbcode_second_pass_quote('johnmarkos', 'i')f the development of computing power (in terms of processor speed) had ground to a halt in 1981, would the Internet have been impossible?

[bart]

The effect of too much computer power is similar to the effect of too much energy.

Designs become bloated and complex. Bells and whistles turn a compact program into a multi-Megabyte bug-filled monster. Instead of thinking, people turn to expensive high-tech solutions.

During the first part of my years in high tech, the additional computing power was a boon. It enabled more functionality, better interfaces (my specialty), and more modular programming (didn't have to count every cycle). At a certain point, though, the increases just led to waste. Management seemed to lose about 30 IQ points. It was a dramatic example of Tainter's theories about societies inevitably becoming complex and dysfunctional.

I think good design requires constraints, whether the design is for a web interface, a house, or a civilization.

[0mar]

A good example here is the design of the chip not the megahertz.

AMD processors have long been "slower" in GHZ than Intel chips. However, the chips themselves show no indication of this. In fact, AMD's new chips only run at 2.0 - 2.4 ghz, yet they routinely outperform 3.4-3.6ghz.

It's not how many GHZ you have, but how you use them . The "barrier" of GHZ was really only felt by Intel, because other companies saw that such a venture was untenable. AMD has long been championing dual-core chips and new sorts of logic. Added to that is that the x86 system is just about out of steam and a new set of instructions should be made.

[johnmarkos]

$this->bbcode_second_pass_quote('Bytesmiths', '"')Impossible" is a strong word. Do you ever download music or movies? Do you like websites with pictures?

[Dvanharn]

My 2-year old AMD Athlon 700mHz has three fans - one for the HUGE processor assembly, one for the PS & one for the case. I changed from a very noisy 10Gb IBM SCSI hard disk to a silent Seagate 40Gb Barracuda single platter HD. But the system is still noisy as hell - it takes 3 fans to carry away the waste heat of the newer machines. I hope to get a "silent PC" next year, but having been out of the LAN/WAN world for over three years, I have lost touch with the details of the technology.

What I really want is a super-quiet, energy efficient PC with a LCD monitor, moderately fast (1.3-1.7mHz) AMD processor and a reasonably large cache.

A currently active network guy stopped in the store to look at solar toys, and we talked computers a bit. He suggested that I get a used IBM laptop with a 15" screen. A lot of 4-5 year old IBM's like the one's we used at Chevron's HQ are still being used and actively bought and sold.

I wonder what percentage of energy in developed countries is used for computers, & how much will be saved as lead-laden CRT monitors are hauled to landfills and replaced iwth LCD's? (Certainly not enough to stave off peak oil!)

Dave

[pilferage]

It's hitting the wall because Intel just realized that when you go to about
90nm you have trouble with electrons tunneling where they shouldn't, etc...
http://www.abxzone.com/forums/showthrea ... ge=2&pp=15
AMD is realizing the benefits of having more operations per cycle since their newer processors operating at ~1ghz less are beating intel's best offerings
http://www.tomshardware.com/cpu/2004031 ... 53-31.html
I've heard ppc architecture can complete more instructions per cycle than AMD's stuff... so it's not just about speed, it's speed and operations per cycle.

[Carmiac]

It sounds like what you really want is one of the new flatpanel iMacs. Check them out.

[holmes]

house in two years will be 100% independent. solar and wind. It will be able to run a computer. Computers are one of the most important tools of the future without them we will hit the dark ages. poeple need to build knowledge bases.

[savethehumans]

I sure will miss posting on these boards. Also the wonderful access to the world's information. ALSO, e-mail--it's soooooooo quick and efficient!

That's why we all need to make the best use of them while we still have them!

[khebab]

The current technology used to make chips is actually peaking because the size of the transistor patterns are becoming too small relatively to the wavelength used in the lithographic process. Much hopes were placed on the extreme ultraviolet litography but it was recently abandonned because too expensive and too complex. Other ways were explored such as using water in order to obtain smaller patterns (because of the change in the reflective index). The size of the transistors is not the only issue:
- heat is becoming tremendous ans is increasing with the ghz (an Intel P4 radiates as much heat as a iron on the cotton position!);
- circuitry is becoming the bottleneck (copper is nout enough);
- quantum effect (ex: tunneling effect) are becoming predominant while transistors are shrinking.

The current approach is to improve the architecture (Hyperthreading, multi-processor) but multi-tasking/parallelism is theoratically a nice feature but practically very difficult to implement and very application dependant. Other improvements are possible in a near future: spintronic, optotronic and so on.

Another aspect is that the CPU speed is not as a selling factor as it was a decade ago. Most of the companies are actually satisfied with their computers and do not want necessarily to upgrade all their infrastructure for a less than certain profit.

[tmazanec1]

Well, my Windows XP SP2 (my current system) seems a little more stable than the earlier ones I used.
I agree, though...MIPS/milliwatt are the way to go, especially with laptops (and battery life) becoming more prominent. And when your electric bills start going up...

[Trab]

Most home users of PC's have way more horsepower than they need for most tasks. Other than gaming or high-end audio/video editing, there are few tasks the average person would need the latest and greatest equipment for.

One thing I have thought about off and on over the last several months is the concept of low power computing. Most newer PC's have a 400W or higher power supply in them. This is overkill for the most common tasks (email, web, IM, etc). Finding the right mix of components necessary to give the average user of PC's a good computing experience while cutting unnecessary power consumption is a worthy task.

I'm not an expert in the area, but by using low-power chips and reducing the number of moving parts in a PC, it should be possible to reduce the amount of power needed by at least 50% if not more.

Monitors are another story.