Questions by Sander Olson. Answers by John Cholewa. John Cholewa is the creator, author, and administrator of JC's Home Page, a website devoted to microprocessors and computer technologies. JC's Home Page is one of the most popular computer sites on the web, and is repeatedly cited by numerous other technology websites.
Question 1: Tell us about yourself. How long have you been interested in microprocessors? How long have you been doing JC-news?
My interest in microprocessors
goes back far, far back, perhaps as far as I can remember. Well, roughly
three or four years, actually. I was always peripherally into computers,
ever since getting to play with the Commodore PETs at school and futzing
about with my father's TRS-80 and Tandy 1000. But my interest in microprocessors
themselves probably started with the
transition from 80486 microarchitecture
to the various designed put together by Intel, Cyrix, AMD, and others.
That was when I really started paying attention to the differences between
microprocessors, and that was when I realized that some of this stuff had
the potential to be quite fun.
My website is a bit of a
different story. You see, prior to 1998, my only outlet for technology
education were mainstream computing magazines. These were good for grabbing
a very basic idea of how technology works, but they were comparatively
expensive, and magazines had this tendency to overgeneralize and miss important
points. For example, a relative of mine
who works at a major technology
company must rely on a combination of hearsay and magazines to keep his
knowledge up. Said combination leads him to currently believe that
the 1.00GHz Pentium III is a higher performance processor than the Pentium
4, regardless of clock speed. He's a sharp fellow, so the problem obviously
isn't his assimilation of the data -- he's
merely not getting the right
input.
In very late 1997, the company for which I worked (and still do at the moment) tried to create a business division for reselling PCs. Although this was pretty much fated to fail from the start, it did allow me the opportunity to do some website design for this potential service. As a part of this endeavor, we signed up with a service called "LinkExchange", which is a banner swapping system where each website in a particular category has a banner on their page, and this banner randomly presents linked banners going to another website in that category. At some point in January of '98, I happened to notice a banner from LinkExchange pointing to something called the "System Optimization Page". This was an incredible experience! Imagine: A place on the WWW set aside for people to talk about computers and technology, put together by individuals and not some huge mainstream media company! Before this, I hadn't even the concept in my mind of such a thing existing. It was really cool.
On the spot, I decided to make one of these websites. My first attempt was... strange. The format was pretty, but there wasn't much on content (and, looking back, I notice that I made a surprising amount of factual errors about the technology in question, despite there only being two or three pages there!). I shelved this and just sat around for a while, hunting around for other websites. One of the things that really impressed me about these websites was the community aspect. Each website has a section set aside for "Links" to other websites of similar nature. This, to me, was the most important part of the whole "enthusiast website" phenomena -- the "glue" that stuck it all together. If there were no links, there would be no way for newcomers to find all the websites. After all, your average Joe Webmaster is not a huge company, and therefore cannot afford to advertise all over the place. I love websites that go out of their way to tell you of the existence of other sites, and I inherently distrust websites that go out of their way to *avoid* linking to others (for one thing, it's a sign that the webmasters in question are more interested in running a "business" than they are in furthering the propogation of knowledge and happiness).
April of 1998 was when my website started. I admit that I was a bit immoral there, as I announced my site's existence on all the message boards that I could find (this is, after a fashion, spam). Still, these actions brought me to an immediate hit rate of one hundred page views per day. I remember sitting in the HP UX lab at my university (SUNY Stony Brook, for anybody over there), catting and recatting the text file that contained the number of hits that my web page accrued. It was really groovy and allowed me to further delay the useless class-oriented tasks that my teachers wanted me to perform.
May ... I think it was May 2nd that I gave my website its first name. My nick at the time was "JC" (it was actually forced upon me by others, but I have learned to live with it), and my website was to specialize in "PC News" and "Links" (the "Links" part was very important, because I wanted this to be as prominent on my website as the actual news), so I called the site "JC's PC News'n'Links", a name which actually lasted until earlier this year when I decided to simply refer to the site as my "Home Page". :)
Question 2: How long do you think that the current PC slump will last? What do you see as the primary cause of the slump?
Feh. This is all psychological. The only real reason why the semiconductor market is down is because people stopped buying, and the only reason why people stopped buying was because they were pressured by fear into saving their money instead of spending it, and the only reason why they did that is because all the analysts cried wolf in the first half of last year until people believed them. Er, additions to the psychological damage was caused by a famous fellow named Alan Greenspan, who could have perhaps been a bit more forgiving on rate cuts. But the basic idea is that a self fulfilling prophecy was caused, and it was a needless loss, to say the least.
The annoying side effect
of this slump is that computers are becoming more commoditized. Even
though new microprocessors have *always* been met with claims of "we won't
need anything faster", people are suddenly starting to believe it.
Are there applications that run slow with today's processors? Yup, anything
made by Microsoft runs at a crawl on the highest performance workstation,
and many other programs could use some serious boosting. Of course,
hard drive and memory speeds are a factor here, more than they were before,
so the cpu increases alone aren't as cool (this is why we're seeing
accelerated pushes for DDR333
north bridges and ATA133 south bridges) as they used to be.
Question 3: Many scientists and engineers are stating that Moore's Law cannot continue much longer. How much longer do you think it will continue?
I think that this may just
depend on what you define Moore's Law to be. If I recall correctly,
Moore's Law was initially just an offhanded comment made by our friend
Gordon suggesting that transistor densities will double every eighteen
to twenty-four months. Since then, most people have shifted the "Law" to
simply every eighteen months and many have attributed the law to
operating frequencies of
microprocessors. Basically, it's just a process thing and nothing more.
The idea is that process sizes will decrease an average of 30% every year
and a half to two years (a 30% decrease in both width and length is the
same as a 50% decrease in area). The real kicker is that eventually these
feature widths (of the process technologies) are going
to become so small that
each feature will be composed of only a handful of molecules. As
we get closer to this, it becomes harder and harder to draw these features
with precision. These silicon etchings can't really get smaller than molecules,
so you're going to hit a hard limit there where we'll have to rewrite the
laws to compensate (quantum computing is one anticipated workaround to
this as a limiting factor to potential chip performance), but the limits
start to crop up at much, much larger sizes
than that.
For the details, I must defer to the experts, who say that we have most of a decade left of these semiregular process shrinks.
Question 4: What do you think about the prospects for molecular computing? Which post-lithography technology do you think is most promising?
Whoosh! I fear that this
one goes over my head, as I know little about molecular computing.
Well, other than, that is, that computers are composed mostly of molecules,
at least if you measure by mass. But the only future computing paradigm
that may even resemble this of which I am familiar would be quantum computing,
and I'm a bit rusty on that as of the moment, so I'll
stop right here for this
question. ;)
Question 5: It has become commonplace to find prognosticators claiming that the PC is dead. Do you find any merit to these conclusions?
People who say that tread
dangerously into the realm of blanket statements. The PC is still very
much alive, and it will continue to be so. There are certain applications
which require a non-PC -- I mean, handheld computers are obviously needed
for those silly people who like to walk around while they update their
websites (I'm one of those silly people). But a centralized desktop
computer will remain the means by which people perform their data communications
and productivity tasks for quite some time
forward. I think that the
PC will gradually morph until it becomes something different -- the desktop
computer as a concept suddely and spontaneously dying and being completely
replaced by something altogether different is a bit hard to swallow, given
the nature of momentum.
Everybody I know has a desktop
computer. NCs are definitely not popular for the common home user.
PDAs don't do everything needed, and notebooks simply are chopped down
in what they offer while sporting much higher prices. The
momentum is with the companies,
and they will resist sudden changes. I mean, we've been waiting since well
before I was born for electric cars and the like, so I fully expect them
to slow down any similar alteration in the computing market.
Question 6: Many industry analysts argue that the PC industry will stay depressed until a new "killer app" is found. Do you agree? What do you think the next "killer app" will be for the PC?
I have never found a computer
fast enough to satisfy my needs. I think that people are simply becoming
accustomed to slow loading and slowly reacting programs. And we're at the
point where the most common types of performance increase will not impact
these particular performance problems. The real bottlenecks are probably
more along the lines of internet connection
bandwidth, hard drive bandwidth/latency,
and (to a lesser extent) memory latency. The "killer app" today is
the broadband internet access device, which brings in all those mpegs and
mp3 files that are more popular than ever (despite rather interesting actions
on the part of the MPAA). And we're going to need bigger, faster storage
mediums to hold this extra data. We're going to need faster systems to
encode and decode all these files, and....
Bah, who am I kidding?
There's no such thing as a "killer app", really. I mean, what was
the killer app for the 486? There was no big huge thing which required
that everybody upgrade immediately (save for bloatware OS upgrades, but
we get those now, too). Computer games have been gradually increasing their
requirements from the days of Wing Commander to today's
Serious Sam and Black&White.
Like the PC slump of today, I think that the killer apps are more of a
psychological thing. It's more a combination of marketing and general public
willingness to spend money.
Question 7: Some hobbyists have doubled the speed of their systems by cryogenic cooling techniques. Have you ever done this? Do you think that the mainstream computer industry will go over to cryogenic cooling?
If you are merely talking
about methods of cooling below room temperature (what you find in somewhat
common peltiers), then you are talking about something that could find
some feasability in the mainstream. However, I am fairly certain that the
more aggressive type, the form which generally involves something like
liquid nitrogen or helium, is quite a ways away. Cooling methods like that
are inherently dangerous to objects around it. Currently, a broken system
could at worst end you up with some burnt out components (well, a fire
if you're not careful, but that's a matter of personal safety). If
we used these exotic cooling techniques, then a leak in the system would
be have really, really nasty effects on the entirety of the computer as
well as anything else laying around. Can you imagine what your wife would
say if she saw the carpet stain caused by a centiliter of 75
Kelvin liquid? It wouldn't
be pretty.
Question 8: There are a number of bottlenecks that currently constrain microprocessor performance. These bottlenecks include the Interconnect, and cpu-memory speed discrepancies. As the speed of CPUs increase, these discrepancies will become ever more problematic. How do you think the industry will handle these problems?
See my above answers.
Question 9: The Intel Pentium 4 is arguably the most criticized and controversial chip that Intel has ever introduced. Some even go so far as to claim that it was designed by Intel's marketing department. What is your opinion of the Pentium 4?
Most controversial? Hrmph, I suspect that you forgot about the names "Covington", "i860", "i432", "Merced". The Pentium 4 gets a bad rap, but it has its strong points. It does get a marketing edge that it does not deserve (the whole frequency thing), but it is at the moment at least a competitive offering compared to other microprocessors out there.
Question 10: The concept of distributed computing has become popular recently, with SETI and other project. How much potential do you think that distributed computing has? Do you think that companies will ever be able to make money from it?
Money? Pfft!
I vastly prefer the other route that this is taking. Distributed computing
allows people to make significant contributions to the needs of others
on a voluntary basis without having to exert any actual effort. This
is the perfect form of charity, as it merely bleeds off excess, unused
resources that otherwise would not be noticed by said user. To exploit
it for commercial value is to ignore why it can be a benefit. I'm sure
that the general idea of distributed computing has its place.
After§éoK, that's
basically what a cluster is, and there are businesses that use clusters.
But talking about capitalizing on this as a standalong thing has the feeling
of, say, trying to patent a Grand Unified Theory (of Physics). It
sounds ... icky!
Question 11: Some writers, such as Ray Kurzweil and Hans Moravec, are arguing that computers will become sentient within the next several decades. Do you agree? What is your opinion of the concept of Artificial Intelligence?
Sentience is a term that has never been solidly defined. We have little in the way of reliably gauging it. We used to have the Turing Test (I think that's what it was called -- having a conversation with an unseen entity and seeing if you cannot tell whether the entity is human or machine), but that particular methodology seems to have failed along the way. A recent and rather humourous suggestion was made recently stating that the ultimate proof that a species is sentient is that it can come up with the idea of making a Turing Test to detemine sentience of other beings. Wacky, huh? :)
Without actually pretending to be an expert or even an informed novice on this topic, I will suggest that artificial intelligence is too constrained to produce sentience in the way that you are thinking. I would point towards A-Life, a related field but something a bit more open-ended.
Question 12: What is your opinion of molecular nanotechnology?
I know little of the details, but I should plug Brian L. Halperin (SF writer, has something to do with the question, I believe) while I have the opportunity.
Obviously, molecular nanotechnology will allow for smaller devices to be constructed, with finer detail that will allow them to consume less power, performing more efficiently than they would otherwise.
Hmmm ... it would be kind of neat to get out of the rigid, two dimensional approach to cpu fabrication. I mean, molecular nanotechnology could perhaps allow us to create such systems from solid blocks instead of thin (though somewhat multilayered) wafers.
But there are those out there that know more of this stuff than I. :)
|
|||||||||||||||||
