Author: Robert Hyatt
Date: 07:17:15 09/06/01
Go up one level in this thread
On September 05, 2001 at 19:08:40, J. Wesley Cleveland wrote: >On September 05, 2001 at 15:29:31, Roy Eassa wrote: > >>On September 05, 2001 at 15:26:53, Uri Blass wrote: >> >>>On September 05, 2001 at 15:15:24, Dann Corbit wrote: >>> >>>>On September 05, 2001 at 13:07:50, Robert Hyatt wrote: >>>>[snip] >>>>>They will come, but it is not clear that the 18 month cycle will be preserved >>>>>for the forseeable future. It has already started to lengthen... >>>> >>>>I'm not sure about really recent trends, but for the past decade it had actually >>>>shortened from 18 months to 12. >>> >>>If you are right then it means that the latest computers are 1024 >>>times faster than the computers of 10 years ago. >>> >>>I think about machines that everybody can buy >>>and not about supercomputers. >>> >>>It seems to me to be wrong. >>> >>>I believe that the 386 is less than 1000 times slower than the latest >>>pentiums even if you give them 2 processors(computers with more than >>>2 prcessors are almost not used by the public so I do not count them) >>> >> >>In August of 1991, I bought a 486/33, which was about the fastest you could get >>at that time. I'd agree that a 1.4 GHz Athlon or a 2.0 GHz Pentium 4 is indeed >>LESS than 1000 times as fast as that 1991 system was. > >The 486 was not superscalar and had slower L1 cache. I would guess that >processor-intensive tasks would be about 1000x faster. Note that memory speeds, >particularly latency, have not increased anywhere 1000x. More accurately, note that memory speeds (random access speeds) have not increased _at all_. DRAM was 100ns (or slightly less) 10 years ago. It is _still_ that slow today. Electrical properties like resistance, capacitance, and inductance haven't been eliminated... Another issue is super-pipelines. Breaking up simple operations into multiple operations that are even simpler doesn't always make the machine run faster. The longer the pipeline, the more problems there are with branch prediction and speculation.
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.