Intel is said
to be planning for 15 more years of Moore's Law progress
An article has just appeared on The Economist website which mentions that Intel has promised 15 more years of Moore's Law progress at the present rate (doubling of transistor densities every 2 years), with some experts predicting 50 more years of such progress through the harnessing of new techniques such as spintronics.
"Intel is confident that it will be able to maintain this pace of improvement in silicon for another 15 years. Recent breakthroughs by researchers at IBM and Hewlett Packard in molecular electronics lead many experts to believe that Moore's law will continue to apply for perhaps another 50 years. Similarly dramatic advances in storage and transmission technologies are also in prospect."
Table 1 - Moore's
In 1971 a memory chip held 1,024 transistors.
In 2000, "workhorse" RAM chips employed 64,000,000 transistors to store 64 megabits of data.
The above chart predicts a 100-fold increase in RAM capacities between 2000 and 2010, with a 25-fold further increase between 2010 and 1/1/2018.
Table 2 - RAM
To be consistent with other tables, these prices need to be multiplied by8 to convert them to costs per megabyte.
What they show is a price decline of about 10,000 to 1 over a 24-year period, from $50,000 a megabyte in 1974 to $5 a megabyte in 1998. (My records show RAM prices between $2 and $1.25 a megabyte in 1998.)
Fifteen years from now would be the beginning of 2018, and as
our computer technology forecast table indicates, we would be
conjuring with 8-nanometer circuit feature sizes, estimated RAM
costs of $35 a terabyte*, and hard disk prices of, e. g., $2,750
per petabyte ($275 for 100 terabytes of disk storage capacity).
These prices should support human-level artificial intelligence at a consumer-level price..
One of the consequences of such a development could be the replacement of tapes, CD's, and DVD's by memory cards... a sea-change which is already underway.
* - derived by dividing the $84 a gigabyte cost of RAM on 1/1/2003 by the above factor of 2,500.
50 more years of Moore's Law progress?
Fifty more years of Moore's Law progress boggles my imagination. As our computer technology table suggests, we would hit atomic dimensions by 1/1/2035. I'm not sure how circuit refinements would be expected to proceed between 2035 and 2053, but I'm all in favor of them.
In any case, these quotations are encouraging.
With 250 gigabyte disks available for desktop computers, 100 terabytes doesn't seem beyond the pale, especially with a forecast of 3.5", 30 to 50 terabyte drives, Conventional drives can reach a terabyte per square inch (30 to 50 terabytes in toto).
It's hard to imagine how Moore's Law improvement could continue for 50 more years. If individual atoms could store bits or bytes of information, how would they be addressed? Optical schemes would require x-rays, with wavelengths of a few Angstroms. A solid crystal might work. A crystal one centimeter on a side would have 1018 cells that were 10 nanometers by 10 nanometers by 10 nanometers on a side. These cells might be written and read by an extreme-ultraviolet laser. (The material would have to be transparent to extreme ultraviolet, and most materials aren't.)
In the meantime, we still have along way to go in display technology. In 1991, there was quite a bit of talk about large-screen displays with 1,024 by 1,920-pixel resolutions for HDTV. The plans at that time called for large LCD displays. In the meantime, mirror and plasma displays were also runners in the race.
Plasma display panels are finally arriving, but I haven't seen any large-screen displays of any description that offer more than 1,280, 1,024, or 1,366 by 768 resolution. Plasma display panels and mirror displays are still very expensive... e. g., $4,000 for a 42" plasma display. Nineteen-inch CRT's were readily available in 1991 that offered 1,600 X 1,200 resolutions, with a $40,000, 20" by 20" air traffic control display available that gave 2,000 by 2,000 resolution. Today, you can buy a 22" display for $679 that affords 2,560 by 1,920 resolution. But we're still a long way from what an IMax movie can provide: wall-sized 3-D screens that make it look as though you could step through them. Maybe organic LCDs will eventually enable them at affordable prices.