v053qgzj@ubvmsd.cc.buffalo.edu (David M Snyderman) (11/16/90)
Summarized from the Buffalo News, 11/15/90: "Gilbert Hyatt, the engeer who was granted a pantent last summer for the computer microprocessor after a 20 year battle... announced Wednesday at the Comdex Computer show that he recently received a patent for a technology that triples the speed of... Dynamic Random Access Memory chips." Any confirmations out there? Does this mean 20 ns DRAMs? Will memory boards with these chips be able to work in a standard ISA bus? Any idea how soon these chips will be out on the market? Costs in comparison to standard DRAMs and also fast static RAMs? What are the long term implications? I see some, providing they are as cheap as standard DRAMs: 1) The death of cache memory. If you can have 30 ns DRAMs, these would be fast enough for even 33 MHz processors. What does this bode for the 486 (that has a *albeit small* 8K internal cache)? (I know, probably nothing. But will this affect INTEL's future desings?) 2) *ANY* advancement in RAM technology would seem to raise the "minimum" RAM required to run the "topline" applications. 3) Computer prices WILL NOT be very affected by this. A (now standard) configuration of 4 megs of memory is only costing (ballpark) $300 on a machine that costs 6 to 12 times that much. 4) Development of some "hard disk replacement" memory cards. Especially if these chips are capable of higher capacity. This might be possible-- the article says that Hyatt achieves his speed by eliminating "unnecessary operations in the chips." Less hardware support internal to the chip might imply less space required and less power consumption. Hope you found this interesting. David M. Snyderman V053QGZJ@UBVMS
daly@ecs.umass.edu (11/17/90)
In article <46539@eerie.acsu.Buffalo.EDU>, v053qgzj@ubvmsd.cc.buffalo.edu (David M Snyderman) writes: > Summarized from the Buffalo News, 11/15/90: > > "Gilbert Hyatt, the engeer who was granted a pantent last summer for the > computer microprocessor after a 20 year battle... announced Wednesday at the > Comdex Computer show that he recently received a patent for a technology that > triples the speed of... Dynamic Random Access Memory chips." > > Any confirmations out there? Does this mean 20 ns DRAMs? Will memory boards > with these chips be able to work in a standard ISA bus? Any idea how soon > these chips will be out on the market? Costs in comparison to standard > DRAMs and also fast static RAMs? What are the long term implications? I don't have any confirmations, just some commentary: Memory boards in a standard ISA bus would never need 20 ns DRAMS, and could never take advantage of them. The ISA bus runs at 8MHz (I think), no matter what the processor speed is (hence, some of the greater desire for a more advanced bus) and thus memory on the bus can not be accessed at faster speeds than that. So memory on the bus would only need to operate at 1/(8MHZ)= 125ns, since anything faster would not be useful. Actual memory on computers is usually implemented so that it can run at (or as near as possible to) the processor speed, and not at the bus speed. Some motherboards offer a special slot that a memory board (one specially manufactured for that motherboard) can plug into and run at the same speed as the motherboard's memory (ie > regular bus speed). A normal memory board (such as an add-on EMS memory board) is forced to run at the normal 8MHz bus speed, and 20 ns would do nothing to improve performance. You question probably should have been "Will these chips be able to replace standard DRAMs in PC's?" > > I see some, providing they are as cheap as standard DRAMs: > > 1) The death of cache memory. If you can have 30 ns DRAMs, these would be fast > enough for even 33 MHz processors. What does this bode for the 486 (that has > a *albeit small* 8K internal cache)? (I know, probably nothing. But will this > affect INTEL's future desings?) If fast enough DRAMs are cheap enough, cache memories will disappear, because who needs a cache if your main memory can run fast enough for zero waits? Intel's designs will probably still include caches, though, as they keep talking about 50, 80, 100 MHz chips in the future, and who knows when DRAMs for those speeds will be available. > > 2) *ANY* advancement in RAM technology would seem to raise the "minimum" RAM > required to run the "topline" applications. I don't know if RAM speeds would affect application memory requirements as much as RAM size would: i.e. 20ns DRAM probably won't affect RAM requirements as much as something like, say, cheap 16Mb DRAM chips would. > > 3) Computer prices WILL NOT be very affected by this. A (now standard) > configuration of 4 megs of memory is only costing (ballpark) $300 on > a machine that costs 6 to 12 times that much. Everything affects computer prices (especially new technology)...New high end systems featuring "*TRUE* Zero Wait State Performance" will appear, costing $$$ (until something else comes along to bring prices down); systems that now cost $300-$900 more because of their cache memories will become much cheaper, and non-cache systems at the cheaper end would gradually disappear the way of the 16 & 20 MHz 386's (squeezed out from between the low end and medium range systems). > > 4) Development of some "hard disk replacement" memory cards. Especially > if these chips are capable of higher capacity. This might be possible-- > the article says that Hyatt achieves his speed by eliminating "unnecessary > operations in the chips." Less hardware support internal to the chip might > imply less space required and less power consumption. "Hard disk replacement" implies mass permanent storage. DRAMs are not suitable for replacing mass storage, unless they could retain their memory on power-off. DRAMs need to be constantly refreshed (hence Dynamic RAM) and so would need some sort of battery backup or something to retain memory. Also, the bus bottleneck comes into play again, and ultra-fast DRAM's of 20ns probably wouldn't be necessary. > > Hope you found this interesting. I have. ;-) Let's hope that it's more than a misprint or distortion...(Although I tend to be skeptical). > > David M. Snyderman > V053QGZJ@UBVMS Bryon Daly, Graduate student, Electrical and Computer Engineering, Univ of Mass daly@ecs.umass.edu --- I tried to speak for my employers, but they told me to shut up! (Possibly with good reason)