The technology changes that have occurred are listed in Table 4.1. The types of transistors have changed from grown junction to planar, from bipolar to metal oxide semiconductor (MOS), but the transistor effect has been the principal element in all silicon integrated circuits for the past 40 years and probably will continue to be for at least another quarter of a century. The silicon transistor has been the principal element in this technology since 1957. The productivity gains are made possible by continued technology improvements in the manufacture of silicon integrated circuits. It is this continual improvement in productivity in silicon integrated circuits, and the related lower cost of memory, computing power, and communications bandwidth, that is leading the world into the Information Age. This dramatic reduction in cost or increase in computing power over a period of several decades is unparalleled in any other industry. Fortune, the impeccable source of reliable information, recently showed that the cost of MIPS has declined by a factor of 200,000 over 30 years. The cost of computing has also decreased dramatically during this time frame. The cost of memory will have fallen by roughly five orders of magnitude from 1972, when the first 1,000-bit static random access memory chip (SRAM) was available, to the year 2010, when the 64-gigabit dynamic random access memory chip (DRAM) will be available. The productivity of silicon technology is usually measured in the cost reduction of memory or the increase in processing power. Ie: you're going to see an XX improvement in power consumption and speed when moving from a "45nm" process to the "28nm" process.Everyone knows the adage (variously attributed to baseball managers and ancient Chinese philosophers): ''It is difficult to predict, especially the future.'' With that in mind, this paper focuses on how productive the semiconductor industry has been over the past 50 years, its status in 1995, the road map for the industry for the next 15 years, some of the challenges that the industry faces, a few predictions, and a final caveat. Use it as a gauge of what to expect in terms of power, performance, and cost. So once you get a 'feel' for the scale of what you're dealing with, don't sweat the real nature of the precise size of things. Even right there it's claimed that the actual smallest 'feature' could be as small as 25nm. I mean shoot, go look at the wiki page for the definition of a 65nm process. I used to think that free_electron's description is / was the most accurate. There is likely to be some fudging in there. So in general, I don't think you can necessarily take a lowest level, hard and fast technical definition of a given process technology to be accurate. Well, guess what? Try finding a 22nm process with the likes of TSMC or other fab.
I see on sites like wiki and such they go from 45 to 22.
And yes, I said "28nm".that's NOT a typo.
But as of the last few years in working on projects that were done in 65nm, 45nm, and 28nm seems the lowest level of technical definition hasn't remained consistent over the years. I'm just a digital designer that doesn't get involved with the backend of things (have always had a vendor that handles that) so I haven't really had a need to be intimately familiar with how a given process node was actually defined. I've been doing IC design for quite a few years, and I used to think I understood what it meant to define a given process node.
0 kommentar(er)
