AFAIK upgrading a 386 to pentium was not a simple drop in replacement of the CPU. Indeed all the 386 PC's I'd seen-mainly HP- had the CPU soldered to the MB. So I have doubts as to the cost of this upgrade being a cheap one as you imply. Pentium PC's were $3000 plus at the time. A lot of money, but PC users did pay it.
No I'm talking about the ever dropping price of PCs being a plus and a minus point. The plus point is you just sell your dodgy old 386SX 16 to some n00b and put it towards a 486-25 machine. And as my 4mb 486-25 only cost about £50 more than an A2000 68040 accelerator when you remove the price of the monitor it was actually cheaper anyway.
But people upgrading to a new faster PC could keep their old games and revisit them, A500 users buying an A1200 could play the same games at A500 speed with the same faults (like Lotus III for example) so where is the incentive to either accelerate your A2000/1500/500 or even buy a new A1200 until new cutting edge games to make your jaw drop appear? None

Lotus 3 ran fine on my Apollo 68040. Blizzard 040's were RC units-recycled CPU's and I don't recall them being 500 pounds. The 1260 boards were about 500 pounds . You wouldn't buy a 68060 to run Lotus or Streetfighter to get a better frame rate. You'd buy it to run a lot of serious apps.
The point is Lotus III ran dog slow on Amiga floppy machines (only CD32 Lotus III ran as fast as Lotus II) and was really frustrating to see. And the point there is that adding even a 68060 to your A2000 would not improve the speed of Lotus III BUT if you had a 286 and later bought a 486 your could dig out Lotus III and enjoy it with improved speed and hence playability. That was the point, you think Amiga mass market was people running serious software? I think not, it was to games players foremost.
And unlike when buying a new PC or an accelerator card for my ST the games wouldn't improve at all on Amiga except for stuff like Starglider/Flight Simulator II etc.
A 40/50 mhz 68030 is significantly quicker than a 68020, and AFAIR lets you use more RAM. And the MMU did come in handy for emulation, and virtual memory (gigamem).
Doesn't matter 8mb+2mb was enough even up until Windows 95 and beyond PC era (1997-98?) for games programmers. The point was mhz for mhz the 030 was a poor choice. A 28mhz 020 cost about £100-125 less in 94/95 than a similar speed 030 board. And as 020 does nothing an 030 can't do as far as games programming goes it shows the proposed Amiga 1400 with 28mhz 020 and Fast ram and CD-ROM for £600 in 1994 was a much better buy than the overspeced and priced 4000/030 that was too damn slow for serious work and zorro+£1000 price too much for gamers to buy into.
This led to Amiga 3D games being produced based on A1200 spec (ie Nintendo Star Fox for SNES level if you are lucky!) compared to texture mapping routines on PC 3D games being experimented with.
OK so a faster CPU wasn't for you.
A faster CPU WAS for me and every gamer but not via brown boxes from unrecorded sales of mail order companies that Ocean et al would never see and hence never develop for. IT HAD TO BE VIA SALES OF SPECIFIC MODEL OF AMIGAs like the A1400 prototype.
It had to be in an affordable machine too and quickly, Commodore messed up badly by going for a crippled CD32 with no fast ram possible (unless you bought something that turned it into an A1200 for more than the cost of a damned CD-ROM drive for an actual A1200) instead of the A1400/A1800 prototypes. Both 28mhz 020 (so same speed as 25mhz 030 accelerators costing about 150 bucks) with fast and chip ram to maximise CPU speed and would be sold for £400 without CD and £500-600 with CD all in an Amiga 3000 style slimline case.
This never happened so we got the same old crap and games like TFX which were finished were never even released as sales of accelerator cards is not necessarily to games players and difficult to prove so games companies ignored them.
Amiga hardware did not depreciate anywhere near as quickly as PC hardware. The hardware always had better re-sale than a PC. I upgraded to a Cobra 40 mhz 68030 for $299, used it for a 2 years, sold it for $250, then bought an Apollo 68040 for about $400. Later added a CDROM, multiscan monitor. And with each upgrade there was an immediate boost in performance and amount software that I could run.
Except unlike the 25% who were only interested in accelerating serious software I had no interest in owning an 040 based A1200 if Lotus III/Power Drift/SF2 were all going to be crap unlike our PC cousins who would see imrpovements in ALL game styles when they did upgrade.
And most people sold up around 1996ish or a year later, and my A4000/030 was worth just £175 back then so that's a load of crap too. Rare machines may be worth a lot now on ebay but that doesn't count.
What I do know is that the Amiga market was made up some of the biggest tight-arses I've ever met. Buying an A1200? Nah too expensive, rather run the old 1 meg A500 and complain why I can't run Doom. Hard drives? Too expensive, but I'll complain about why all the disk swapping. Monitor? Nah just use the TV. Workbench 3.1? Nah 1.3 is OK.
The reason most people didn't buy an A1200 is because
1. Commodore didn't have a clue what they were doing and left out some pretty vital things like decent parallax/better sound/no HD floppy support/CPU crippled until you invest another £150 in a RAM board (where were the bloody SIMM slots!?).
2. The games still looked way to similar despite all the '32 bit power' hype. 3D games were barely improved unlike when we went from wireframe 3D on C64 to solid 3D on Amiga. 2D games STILL looked inferior to 1989/90 Megadrive AAA titles let alone the £150 SNES and it's superb 256 colour SF2 ports.
I did buy an A1200 but this was for my love of animation and digitiser work.
THERE IS NO WAY THAT YOU WOULD GET THE SAME LONGEVITY FROM A PC FOR THE SAME MONEY. The software ( Windows, games and apps) would force you to upgrade the hardware to the tune of thousands, or you'd need to bin your PC. This concept never caught on in the same way with Amiga users, so all we got was games that were made to run in 512k off two floppy drives.
Well it was Commodore's job to upgrade the CPU. 7mhz 68000 in A1000, 12mhz in A500 and 16mhz in A500+/A600. Without a faster CPU the thing that demanded more power from PC games was 3D games like Falcon or F15 etc.
A500/A500+/A600/A1000 owners had limited CPU accelerator options and games companies would never write games for accelerators costing more than an A500 anyway. Like I said if Commodore had upgrade the machines in an evolutionary fashion for the base model the games would have improved.
Only an idiot ran the latest version of Windows if at all, and anyway Win 95 onwards is where this is an issue and by then Commodore had been dead 2-3 years.
(ESCOM's crappy A1200 for £400 scam in 1995 was doomed to fail, it was an iffy price/performance in 1992 when launched let alone 1995!)
The point is gamers got bugger all benefit for 2D games on Amiga if they did invest in an accelerator, a thing which was only popular of base model Amiga's AFTER A1200 launch anyway. A1200 was only sold by Commodore for 18 months, so regardless of how many accelerator cards were sold games companies did not commit to a dead platform (Amiga was dead the minute Commodore filled for chapter 11 etc in the eyes of the software houses).
And as ALL the big box Amigas were overpriced and underpowered accelerator card sales for those machines were of no concern and pimple on the ass of the games buying Amiga user base.
PS You can do Doom on a standard Amiga 500, it's just the game window would be 80x50 pixels
