hwh wrote on 2020-07-04, 08:23:
See, this is why I thought the whole dual core thing was stupid. Practically all software mainstream users had was single threaded. So we got a bunch of benchmarks and hype over what, initially, had no effect.
Wrong, as soon as you were multitasking, even with purely single-threaded stuff, you reaped instant and big benefits. If all you do is use the PC as a game console, running one single-threaded game only, then indeed, no benefit. But if you were running a word processor, an MP3 player (in P3-days, Winamp was a non-negligible load) or - perhaps most telling - wanted play a game while you waited for a CD to burn, the benefit was immediate. With single CPU, that CD burn failed if the game ever pushed the CPU near 100%. With dual CPU, the CD burnt correctly.
For retro gaming the multitasking use case is less relevant, but in the day it was the reason why these things sold at all outside of the 3DStudioMax/Autocad community. But the time the first dualcores came along a few years later, things like Photoshop has also gone multithreaded and the utility increased - but don't overestimate how much stuff is still single-threaded even today. The vast majority of programs, even games, still only uses one thread. Thing is you're running hundreds of them concurrently. More cores mean faster servicing of processes waiting for CPU. If you don't think that matters I challenge you to limit yourself to single-threaded programs and disable all but the first core on whatever CPU you most modern machine uses. You will have a measureable but more importantly a noticeable slowdown.
Of course I felt that way about a lot of things. Consider first generation SATA drives, which in some cases were slower than their ATA counterparts due to controller woes. Or the first PCIE cards - more bandwidth, but why, there was no card that could use it. In time it all made sense. But the dual core thing, damn, it was years and years before that had a material impact on what you were doing.
I clearly remember upgrading from an So939 Athlon64 to an Athlon64 X2 (the first mainstream dualcore) somewhere in late 2005 and being amazed by the increased usability of the system, despite clock speed & cache remaining the same (3200+ > X2 3800+)
Take a look at the multitasking benchmarks here. The very first dualcore in-depth review shows massive improvements in performance running two (single-threaded) applications at the same time.
Same thing with 64 bit. I got a 64 bit processor in 2003. Absolutely no use for it, XP-64 was renowned for its incompatibilities and driver woes, that's just what AMD was selling at the time. So it sounded cool but it was a completely academic capability.
It was all about memory, or rather address space. In 2003, 1GB was a lot of RAM and a normal user wasn't going to hit any limits, but 2003 was when the first mainstream memory controllers were able to support 4GB RAM. It therefore made sense to have a CPU that could also address all of it natively, and to have an OS that could support that. Of course it was a niche application back then, unless you were going to hit your 32b 4GB barrier you were better off with 32b, particularly given the bad state of OS support (Linux 64b libraries and 32b compatibility were also a mess at the time), but it added a lot of value to those users who actually needed the >3GB RAM.
By 2008 4GB of RAM was mainstream and OS support had improved to a generally usable level (as far as Vista could ever really be considered usable).
But back to retro stuff: the argument that SMP makes no sense if you only play games is more than valid, but that it made no sense for all-round systems back in the day: no way.