In the April edition of "News of the Ultranet", I reported that Intel Announces Plans for 10 GHz Microchip in 2011, Now Intel has announced, at the Year-2000 IEEE International Electron Devices Meeting, that it will reach this 10 GHz goal by 2005, and that it sees its way clear to manufacturing at least three more semiconductor generations between now and then, switching to 0.13-micron design rules in 2001, 0,10-micron rules in 2003, and 0.07-micron features in 2005.
    "Intel will show that it can build transistors with elements as narrow as eight angstroms, or three atoms, and Big Blue plans to make its chips 10 times faster. With a research paper and images from an electron microscope, Intel plans to show that Moore's Law will hold for another decade." (IBM, Intel chips: Smaller and faster )
    The above statement is somewhat misleading. The 8t-angstrom element is just the gate insulator.All the other features will be at least 700 angstorms wide. Intel is only saying that it can reach the 0.07 micron level that Moore's Law compliance will require in 2005.
     "'The achievement is significant in that it shows there are no known physical barriers to slow the pace of chip performance growth for the next five to 10 years', said Gerald Marcyk, director of Intel's (Nasdaq: INTC) Components Research Lab."
    In addition,
    "IBM Research recently also announced a breakthrough design technique, called V-Groove. Should the technique be used in manufacturing, IBM says it could help create chips with parts as small as 10 nanometers or less by reducing the size of the channels between transistors. With luck, the technique could be put into production in the coming 10 to 15 years, IBM executives said. "
    Ten nanometer features would permit a continuation of a Moore's Law rate of improvement  through 2013, giving us 400 gigops desktop computers at least 400 times faster than today's home computers, with typical 2013 RAM capacities in the 64-to-128 gigabyte (64,000-to-128,000 megabyte) range.
    The article linked in "2005 time frame" below explains that Intel is focussing on the mass market and upon minimum manufacturing costs, whereas IBM is targeting the high performance marketplace. Nevertheless, this is certainly a better prognosis than gloomy prior forecasts that had us hitting a wall in 2000.

    Intel closed out the year 2000 with 1.4 GHz Pentium 4's. The company has announced plans for a 2 MHz Pentium 4 in the second quarter of 2001, with higher speeds expected later in the year. The transition to 0.13-micron circuitry will begin in the third quarter of 2001. In the meantime,

2005:
    The semiconductor revolution appears to be fully on track. The 10-Ghz, 400,000,000-transistor chips mentioned in the article may be with us in the 2005 time frame rather than in 2011. An excellent survey of these prospects is given in PC Magazine's  5-year computer technology forecast in the Fall issue of "Ubiquity".  PC magazine is forecasting 12 GHz computers in 2005, using 0.07-micron design rules, with a gigabyte of RAM, 800 to 1,000 gigabytes of hard disk storage, rewritable DVD's, a high resolution video camera, and an array of microphones. High-end game systems may afford photo-realistic rendering. Three-D displays may be a possibility by that time, together with wider use of flat-screen displays. 2,000 X 1,500 or even higher-resolution graphics displays may be common by 2005. Communication will be wide-band, and possibly, in favored locations, second-generation wideband. (In 1995, a 15" 1,024 X 1,280 resolution monitor cost about $380. In 1995, a 17" monitor cost then about $700. In 1995, a 20" monitor might have been purchased for as little as $1,100.)

2010:
    Looking to 2010, we might extrapolate computer speeds to the 50 GHz to 100 GHz range, assuming that there are no hidden barriers to the attainment of such speeds. RAM should cost, perhaps, $6 to $10 a gigabyte, and we'll probably sport 8 to 16 gigabytes in a low-end computer. Our hard drives will provide storage in the multi-terabyte (1 terabyte = 1,000 GB) range, perhaps using multiple, low-cost drives.
    Disk drives require some discussion. In the early 1990's, IBM observed that magnetic flux leakage sets an upper bound on hard drive storage densities of about 62.5 billion bits, or 7 gigabytes per square inch of disk surface, or about 10 billion bits or 1.2 billion bytes per square centimeter of disk surface. This would permit the storage of, perhaps, 600 GB on a standard 3.5 hard drive, or 1.2 terabytes (TB) on a 5.25" hard drive. Note that we're already able to store 100 GB on a 3.5" drive (by adding a fifth sheave to Maxtor's 4-sheave, 80-GB drive), or 200 GB on a 5.25" drive--a factor of only 6 away from the theoretical limit.
    "According to Currie Munce, director for storage systems and technology at IBM's Almaden Research Group, disk areal density should double each year for the next two years and then grow at a somewhat slower rate in subsequent years. In IBM's case, that means moving from 17Gb per square inch in its densest disks to 34Gb per square inch, then to 68Gb per square inch. Munce says that beyond 50Gb per square inch, the physics get far tougher. He adds that IBM's Giant Magnetoresistive (GMR) head technology should carry the technology through to 100Gb per square inch with incremental improvements. That means we may see high-end PCs with 300GB disk drives in two years, and possibly 800GB to 1TB (terabyte) drives by 2005.
    It may be that disk technology will continue to improve beyond 2005. However, its future is particularly uncertain, and I'm loath to forecast beyond the  3.5", 1-TB hard drives or the 5.25", 2-TB disks that Mr. Munce is predicting. However, even if there were to be no further improvements in hard drive technology, a Moore's Law projection would indicate the year 2020 or beyond before RAM would reach a price-point ($60-to-$100 a terabyte in 2020) comparable to that of a 1-TB hard drive.
    Wideband connections might support full HDTV, and perhaps higher bandwidths. Monitors are hard to predict, but large, flat, high-resolution monitors may be economically with us by then. Three-D displays may be popular. (Our superb ink-jet color printers weren't even on the horizon in 1990, nor were CD's.)

2015:
    If we continue to chug along the Silicon Trail through the year 2015, we might extrapolate clock speeds to the 300-to-1,000 GHz range. RAM would run $600-to-$1000 a terabyte (TB), and we might be using 128 to 256 GB. Hard drive, or other non-volatile storage capacity is something I'm afraid to forecast. Similarly, wideband channel capacities and monitor capacities are hard for me to estimate, depending as they do upon deployment issues and manufacturing costs as well as upon technological advances.

2020:
    Simple extrapolation would lead us to clock speeds in the 3 to 10 terahertz regime. RAM would cost $6 to $10 a terabyte, and we might be using 16 to 32 terabytes. Non-volatile storage is hard to predict. We could use a petabyte (1,000,000 GB) hard drive if it were available. We should be within striking distance of the 100 terops computing power that Hans Moravec  has estimated to be required for human-level artificial intelligence. My crude  estimate of the information storage potential on the synapses of the human brain is about a petabyte (10^15 synapses at 1 byte per synapse), but the brain depends upon redundancy to establish reliability. Perhaps a hundred terabytes or so of storage would suffice to emulate the brain. Our extrapolations would lead us to estimate a cost of $600 to $1,000 for 100 terabytes of RAM or of non-volatile flash memory by 2020. And of course, a petabyte of random access storage would be available for $6,000 to $10,000.

    It will be a thaumaturgical tour de force if we can remain on a Moore's Law curve through 2020. However, if we do, there might still be some further improvements in manufacturing costs and other ploys that would yield improvements through 2025.

2025:
    This must be considered highly fanciful, but if semiconductor miniaturization reached its limits in 2020, manufacturing cost reductions might continue to cut costs for a few more years, culminating in a 2025 price of, perhaps, $1.00 to $1.25 a terabyte for RAM, and in low-priced  multiple processors delivering 100 terops. At that rate, 100 terabytes of RAM would cost $100 to $125, and one petabyte of RAM would run $1,000 to $1,250.  Digital signal processors might offer the same computational power at lower cost, or greater computational capabilities at the same cost.

    One other point that's worth noting: humanity has (we hope) a long future stretching out before it. For example, robotics and automation may lower manufacturing costs. And marginal improvements, or conceptually different approaches may carry these burgees farther.

    There are further discussions of the future of computer technology in the April, May, June, and Fall issues of "Ubiquity".

Late 2000:
    The 1-to-1.2 GHz computer for $600 to $700 forecast for April, 2001, has come in a little ahead of schedule. Its price tag is $760, which is above the $600 to $700 that I had forecast. However, this is a premium-quality computer that includes a 1 GHz Athlon Thunderbird processor, 128 MB of 133 MHz SDRAM, a 20 GB, 7,200-rpm hard drive, a DVD, a 56k modem, and an Ethernet card.
    (Two months ago, I bought a 900 MHz Athlon Thunderbird that's faster than a greased goose.)

April, 2002:
    I might expect to see a ~2 GHz microprocessor with 256 MB of 266 MHz Double-Data-Rate SDRAM, a 60-to-100 GB, 7,200-rpm hard drive with a UDMA100 (or 150) interface, a DVD, a 56k modem, an Ethernet card, and perhaps a rewriteable CD drive.

April, 2003:
    We might expect to see a 3-to-4 GHz microprocessor  with 256 MB of 266 to 400 MHz DDR SDRAM or RAMBUS, a 200 GB hard drive with a UDMA 150 interface, perhaps a rewritable DVD, a 56k modem, possibly a cable or DSL codec, and an Ethernet card.

April, 2004:
    I look for a 5-to-7 GHz processor with 512MB of 400/600 MHz SDRAM or RAMBUS, a 400+ GB hard drive, a rewritable DVD, perhaps a 56k modem, and, perhaps, a cable or DSL codec.

April, 2005:
    I'll spring for a 9+ GHz processor, 1 GB of RAM, a 600+ GB hard drive, a RW DVD, and, perhaps, a codec. (We might see 12 GHz speeds by 2005.)

    You may notice that I've upped my projections for computer speeds based upon the developments of the past eight months.
 
 

    Computer prices haven't dropped one iota in the past year, and may even have risen a bit. The cheapest price I could find for a 450 MHz K6-2 with 32 MB of 100 MHz SDRAM, a 4.3 GB hard drive, a 52X CD, a network interface card, and a V.90 56k modem was $335, plus shipping and handling, up from $300 for a 400 MHz version in December, 1999.

    This is certainly yesterday's news. Maxtor has upstaged IBM by offering an 81.4 GB drive for as low as $280. Maxtor has achieved this capacity level by adding a fourth platter to their 60 GB drive. Over the past year, prices have dropped from $7-to-$8 a GB to $3.50-to-$4.00 a GB. It's worth noting that no one has announced plans to sell drives with greater capacities than these. However, as mentioned above, under 2010, Currie Munce, director for storage systems and technology at IBM's Almaden Research Group, is projecting 300-GB disk drives in two years (Dec., 2002), and 800-1,000 GB disk drives in five years (Dec., 2005).

    Sun Microsystems is offering, free of charge, a very capable and advanced office suite called StarOffice 5.2. I. I haven't plumbed the full extent of its capabilities, but it includes a word processor, a web page editor, a spead-sheet, a drawing package, a charts-and-graphs package, an image editor, a slide-presentation generator, an equation editor, an equation editor, an integrated database, an e-mail client, a web browser, and a business card generator. It's an 80-megabyte download, but can be purchased on a CD from Sun Microsystems for $9.95. It seems to be a well-designed system, and well worth the price. (I haven't had time to use it much yet, but I'm impressed with what I've seen.)

    Viideo games are rapidly becoming interactive over the Internet, and general-purpose Internet appliances. Sony's Playstation 2 can be used as an Internet terminal and a DVD player, as well as a game console. Sony's Playstation 3, when it appears in a few more years, will reputedly employ 500,000,000 transistors, up from 42,000,000 in the newest Pentium 4 today. Three-D graphics, photo-realistic graphics, and tactile feedback might be among the enhancements that future video game controllers might possess.
 
 

    The ability to effectively take voice dictation is probably still years away. Voice dictation will have to become at least an order-of-magnitude more reliable before it begins to crowd out conventional typing. Progress in this area is glacial. I wouldn't recommend that anyone waste money on IBM Viavoice or Dragon Dictate unless they're typing-challenged. (Dragon has sold out, and I believe that Lernout & Hauspie has filed for bankruptcy.) Perhaps by 2005, when computers possess ten times their current speeds and stoage capacities, or 2010, when they may offer 100 times their current speeds and storage volumes, voice-writers will come into their own.

    One of our goals in February, 2000, was to try videoconferencing. I still haven't tried it. Have you? With wideband service spreading across the countryside. it may be a timely move.

    Voice chats are meeting with modest success. Probably the best application is in conjunction with online meetings.

    Voice mail hasn't really caught on, possibly because it requires somewhat more storage capacity and bandwidth than conventional email. Also, we may be more accustomed to organizing our thoughts when we write than when we talk. Then, too, it's my impression that those who use AOL can't receive anything but straight text.
 
 

    In my opinion, this service is as much needed now as it was a year ago.

    This still hasn't been done.

    Mutual support via the Ultranet has certainly been a success. We still haven't secured any grants, but we're working on them.

    The "wireless world" is making steady progress.

    This is happening rapidly. Cable modems, DSL, and satellite service are rapidly ramping up. My cable modem service typically supports download speeds of 200,000 bytes per second, whereas my Bellsouth ADSL service tended to run in the 50,000 bytes per second range.