by [TC]²
A monthly column of technology rambling, rumination and reality
By: Jud Early, Corporate Vice President, Research, [TC]²
March 2005
Hello, faithful readers,
This month I'd like to examine some items that, as a gearhead, I find interesting. I hope you will too.
The Law and Commodities
Last week, Intel sponsored another of its annual developer conferences. These conferences are for developers of hardware and software that will incorporate one or another of Intel's products. For those who may have been in a Rip Van Winkle sleep for the past forty years, the “Intel Inside” logo refers to the microprocessor that is the heart of today's modern personal computer. Or is it so modern? Before we get to that, the conference revealed several new areas of technology that will be interesting to follow, and one that I'll describe later in this column.
In one of my columns in 2003, I spoke of my first programming training for Intel microprocessors using machine language. The year was 1974, and the 4004 microprocessor had just become available for general use. At that time there were no high level languages for microprocessors, so machine language was the only way to tell these unique control devices what to do. For anyone who has had the same experience, the architecture of three 12-bit registers and single accumulator now seems positively primitive. Through a succession of new architectures, the 4004 spawned the 4040, that led to the eight-bit 8008 and the 8080, and the 8086, made popular by its inclusion in the IBM personal computer. Each new architecture added more registers, more addressable memory, and sparked the development of higher level languages and compiled code. It was the venerable 8086 that “froze” the x86 instruction set, and was the foundation architecture for chips that are still in use today. We'll come back to the x86 later.
In 1965, Gordon Moore, who at the time was R&D Director for Fairchild Camera & Instrument, authored a paper titled “Cramming More Components Onto Integrated Circuits” In this paper, he graphed the growth of the number of components that could be feasibly included in a given unit of area, and predicted the growth potential to be at least “two times, if not more” for each year of the next ten. As this prophesy became fulfilled, the press picked up on it, and over time, the interpretation of “Moore's Law” has been said to relate to the doubling of transistors in a circuit, to the doubling of computing power each year, or to other variants, each involving the doubling, or halving of some factor related to some other fixed factor. If you would care to read for yourself the original paper, it is on the web at ftp://download.intel.com/research/silicon/moorespaper.pdf .
Whatever your interpretation of Moore's Law, it has stood the test of time for almost forty years. My personal version of it is that for every computer that I purchase, the price for a similar machine will immediately be less than one half of what I paid. Moore's Law, reflecting the relentless advance of technology has enabled the personal computer to become commoditized, and the processor chip to be the focus of a silicon “arms race” between Intel and several other chip developers, some of which are no longer in existence, or have exited the processor market.
AMD, the competitor that causes Intel management heartburn, was initially regarded as a copier, but not an innovator. In 1989 Intel released the 80486 chip. Soon after AMD began to produce a clone of it, but because it was of a different design, did not infringe on the Intel patent. Pipelining, an innovation at AMD, provided parallel execution of instructions, as well as partial execution, in order to have pre-computed results available for use in subsequent cycles, or discarded if not needed. This anticipatory pipelining brought performance to a level near that of the Intel version. The Pentium (80586) was Intel's answer to AMD's steady advance. With the race mounted, the succession of Intel's Pentium II, III and 4 were being pushed along by The AMD K6, Athlon and Athlon 64 from AMD. The Athlon 64 bit chip will soon be met in the market place by Intel's catch-up 64 bit chip. All of these chips use the basic x86 instruction set, although over the years, several additions to it have been added to incorporate new features such as sound and graphics. Remember, the first PCs were text only! Threading, multi-threading and the latest from AMD, Hyper-transport, all have been added to the original machine instructions.
During the mid-1990's I had the opportunity to make a number of visits to the United States Department of Energy National Laboratories. During one of the visits, we were discussing with lab scientists the “next evolution” of chip fabrication, the drive to achieve 200 nanometers in the spacing of the conductive and non-conductive elements in silicon chips. It was to have been a spectacular breakthrough. That barrier was broken, as well as the 130 nm, the currently adopted 65 nm, with plans to achieve 23 nm in the not-too-distant future. It is the ability to more tightly pack transistors into a given unit of silicon that allows Moore 's Law to continue to be proven. The expected limit of 23 nm technology is not the laser litography, but that the number of atoms between elements that are more tightly packed would not provide the operational characteristics of transistor switches.
Are we at the end of the development road? Is chip technology doomed to never be any better that that which is envisioned for the near term?
One of the items revealed at the developer conference last week is a new technique for connecting multiple chips within a package. Current technology packages the silicon chip on a carrier, with wire leads electrically bonded to the chip and leading to the contact pins on the carrier. With more than nine hundred pins to connect, the technology to do so has evolved to provide a reliable, robust package for a single processor. Dual-core processors are only now emerging, and along with them the problem of how to connect more than one chip. Imagine a stack of waffles, carefully aligned so as to have each cell in the waffle directly above or below the matching cell of the next layer. Cells in a waffle are depressions, so would not make contact. If the cells were to be protrusions instead of depressions, the mating protrusions would touch. How, then do the chipmakers produce a chip with a protrusion from a flat sawn and polished wafer of silicon?
On an unusually small scale, when chips are to be stacked, Intel has developed a process where small pits are formed from one side of the silicon wafer. Microscopic in size, these “pits” are internally plated with an electrically conductive material, then the top side of the wafer is ground away until the conductive nubs that are made of the plating deposited in the “pits” is exposed. These conductive nubs are analogous to the protruding waffle cells described above. It is hard to imagine the small scale of these conductive nubs, but manufactured in this way, the stacking of multiple layers of chips is possible, with connections through and through the stack. Imagine having a stack of sixteen or thirty-two processors in a package that is no larger in real estate than one of the current single or dual core processors. Eight way or sixteen way processor clusters will then live in one package. Heat is a problem. The conduction of heat away from the chip must be done to avoid damaging thermal rise. In stacking the chips, heat will be concentrated in a smaller area, but is also expected to be less heat that that generated by current designs because of the tighter spacing. A laptop computer processor of today may dissipate as much as 67 watts. A sixteen layer stack might produce as much as a small space heater. This, rather than small scale manufacture may be the stopper for Moore's Law.
Last year I told you about Firefox, the open-source browser. I am still pleased with its utility, and user convenience items. Now that more than twenty-five million downloads have been done, Firefox has become a larger target.
As might be expected, some security holes were discovered. Crooks found it possible to create false web sites that displayed URLs that appeared to be the real thing, but were actually used to harvest credit card and personal information to aid identity theft. It was done by substituting a Cyrillic character in place of a Latin character, giving the appearance of an all-English URL, but actually pointing to a rogue site. The Firefox team has come up with a solution. Firefox 1.0.1, released last week, shows Web addresses with foreign scripts in code, preceded by the letters "xn." So "paypal.com" with a Cyrillic "a" becomes "xn--pypal-4ve.com." The Opera browser has also adopted the scheme. Microsoft's Internet Explorer does not display foreign characters in the same way, so presents less a threat for this exploit.
A more formidable problem for Firefox is Microsoft. For the first time, Internet Explorer slipped below 90% use by web surfers. As a result, an announcement that a new, improved IE will be forthcoming, and will not be tied to the release of a new operating system. As we know, Longhorn, the new O/S, release date has slipped back into 2006, so Microsoft cannot afford to wait for that release to unleash the new IE. It is reported to have tabbed browsing like Firefox, as well as improved security measures. Even when released with Longhorn, it will not be necessary to install the IE browser. For those of us that have made the switch, going back, may be an option, but not a desire.
At my house, just as it is here in the office, data protection is not an option. I have written about some of the miscreants out there, and solutions to protect against unwanted intrusion. My ongoing efforts are not paranoia, but an attempt to prevent theft and the loss of time and money in recovering from an unplanned event. Anyone who follows communication technology should be aware of these dangers. For those with newer cell phones, spam is already being seen, and virus infections are now possible with certain cell phone operating systems. The data that is contained in your cell phone may be just as valuable as the data in your notebook or desktop system. The companies that develop anti-virus software are at work on preventive products, but for now, you are on your own.
During the Christmas holidays in 2004, I took the time to research the alternatives to the use of Norton Internet Security Suite. Not that it is a bad product, but the boot-up time is long, and despite it's auto-pilot mode of operation, getting new virus definitions and installing them automatically, I just felt there was something better and which would not require the long boot time, even when warm booting.
After much research and reviewing forum discussions, I decided on PC-cillin from Trend Micro. I installed it on my personal notebook computer, and have, over the last two months, monitored its automatic updating and ability to stop virus infection. The last part was hard, since I have had none. I also researched various software firewalls. PC-cillin has a built-in firewall, but I wanted one that would monitor outbound attempts to connect to the internet, not just block unwanted intrusion. The product that I selected is Zone Alarm Pro, by Zone Labs.
Yesterday I set about removing the Norton Internet Security Suite and installing the two new security applications. Despite much negative press and forum discussion around the topic of incomplete removal of Norton, I used the un-install process within Windows, and after doing so, went completely through the registry and scanned all drives looking for remnants of Norton. None were found anywhere on the machine. With a clean machine, I installed PC-cillin, and configured all settings that I wanted to set personally. I disabled the firewall, then installed Zone Alarm Pro. After reviewing the default settings, and making a tweak or two, I was ready for a test.
For those who may wish to test their protection against intrusion, Steve Gibson has a site that has a great tool to test your security configuration. Visit the secure site https://grc.com/x/ne.dll?bh0bkyd2 to learn a lot more than you need to know about security and test your own system. My security tests were all great. I feel that I now have security as good, if not better than was provided by Norton, and I promise to let you know of any bad outcome that may result from this combination. The best feature that I have found so far is that Zone Alarm does not leave open ports to allow permitted traffic through. In sending a shipping report to UPS, it politely asked if I wanted to permit the transmission, and also provided a check box to allow the UPS application to access the internet in the future. It will open the port, allow the transmission, and then will close the port, blocking intrusion from the web. Neat! With Norton, before uploading shipping info to UPS, I had to disable Norton, do the upload, and re-enable Norton. I preferred this to the alternative of leaving an open port for UPS to connect through, although there was the risk that I might accidentally leave the protection disabled for an extended period.
Miami Beach Convention Center is where you will find me next week. I'll be in the [TC]² booth # 1345. Be sure to stop by and say hello. On Wednesday March 16, Dr. David Bruner of [TC]² will be presenting “ 3D Comes of Age ” in the Technology Solutions area.
If this column is oriented too much to the gearhead, or if you would like to see other subjects, please send an e-mail with your ideas for future columns. Just click on http://www.techexchange.com/askjud.html and send your comments. I'll be looking for your input.
Until next time, be safe, compute wisely.
Jud