<xentrac>I think the cost savings from going from a 16-bit ALU to an 8-bit ALU stopped mattering around 1980 for most applications; the remaining 8-bit applications were embedded microcontrollers that often really cared about timing <xentrac>well, maybe 1990. there was a long period of time when computers were expensive and buying a CPU from a number of generations back was a reasonable financial choice <xentrac>I don't agree that there is no point in having a less-than-16-bit address space; as I mentioned, the PDP-8 got by with a 12-bit address space <oriansj>xentrac: but one does not see new pdp-8s being made or sold <xentrac>the 16-bit PDP-11 (and Data General machines!) replaced them almost completely <oriansj>also the reason for cheap processors from 2+ years ago was because Moore's law was alive and kicking <oriansj>once it ended (approximately after 32nm); the performance gains have been so small from one generation to the next that the prices have adjusted accordingly <oriansj>aka a 10 year old PC in 1995 was a complete non-starter but today a laptop from 2008 is entirely acceptable in terms of performance <oriansj>2^5 enhancements are hard to ignore (1985 to 1995) <xentrac>well, what I mean is that in 1990 you could buy a PS/2 Model 25 with an 8086 (or 8088?) in it <xentrac>and people did buy them, because the machines were much cheaper <xentrac>the difference in performance from a 386 was enormous, but the difference in price was quite significant too <xentrac>in the 1990s that kind of thing sort of stopped happening <xentrac>and I don't know why, but I don't think we can attribute it to the end of Moore's Law, since as you say, that happened later <oriansj>xentrac: because people realized you can't run a gui on an 8086 <xentrac>in 1995 or 2005 it would have been rather extreme to buy a new personal computer with a CPU from 1983 or 1993, but in 1990 the PS/2 Model 25 and Model 30 were considered reasonable new buys <xentrac>so I feel like there was a time in the middle there, roughly 1985 to 2005, where obsolescence was faster than before or after <xentrac>the weirdest part is that the DRAM pricefixing bubble was right in the middle of htat time <oriansj>xentrac: well 1995 was the release of the Pentium Pro and the period of x86 picking up all of the performance enhancing tricks of RISC processors (Literally beating DEC Alpha in Integer benchmarks) <oriansj>Windows 3.0 could get away with 2MB of RAM and an 8086 but NT 3.1 and Windows 95 required a 386 or better <oriansj>by 1996 Windows required 486 or better <oriansj>By 1998 Pentium was the oldest one could use <oriansj>Windows ME just wouldn't really run on anything slower than a Pentium 2 @300Mhz <oriansj>and When Windows XP came out in 2001; you needed a Pentium 3 (1999 chip) or better. <oriansj>So xentrac I think what happened between 1993 and 2005; is Microsoft and Intel actively worked together to drive each other's monopolies forward. But a split occurred because of Linux's growth in the Dot com boom in the late 1990s combined with Intel's Netburst blunder; which never managed to restart because of the physical realities of Moore's Law ultimately ended in 2008. <oriansj>Now we are in a period of just more tightly packed transistors that haven't gotten smaller in nearly a decade; processor performance gains only from architectual improvements and a massive rash of hardware security vulnerabilities hiding that fact from the general public. <oriansj>I'd even suggest the RAM price fixing even gave Microsoft a solid excuse for pumping up the System requirements in terms of processing power (which ment new computer with new Windows License); It makes me ponder what the implications of the curent Hard drive price fixing will result in. <oriansj>now the Price doubling in 2011 is entirely related to the Thailand floods; but after the price of a 2TB only dropped from $109 in 2013 to $60 in 2019 (6TB is still $180) <oriansj>I'd think it was the end of Kryder's law (Which in 2009 estimated that if hard drives were to continue to progress at their then-current pace of about 40% per year, then in 2020 a two-platter, 2.5-inch disk drive would store approximately 40 terabytes (TB) and cost about $40.) if we were not still seeing a 15% annual improvement rate <oriansj>but price decay is slower than the projected 15% rate <oriansj>Either because of price fixing or the current market demand for disk storage exploded to the highest growth rate in history 3 years prior to Edward Snowden's revelations. (With the average person purchasing 28TB of storage a year) <oriansj>Or the current scope of Surveillance capitalism has reached a level of detail which makes 1984 look quaint and privacy preserving. ***CORDIC is now known as DKordic
***pgreco_ is now known as pgreco
<oriansj>xentrac: it honestly raises a simple related question for me; do programming languages have a cognative preload requirement, which ultimately determines how successful the language (or its family) becomes <mid-kid>Hi, I'm trying to follow guix' commencement.scm manually, and I'm running into a problem with make-3.80 not wanting to run scripts with shebangs like ./move-if-change. <mid-kid>Is this a problem that's been encountered before or am I on my own? <mid-kid>I suspect it might be something with the mes libc not liking something about my setup. <mid-kid>But execve is just a straight syscall, so it's kinda odd.