There are sites that attempt that, but they rarely go back further than the Pentium 4 era.
https://www.cpubenchmark.net/compare/1146vs62 … D-Ryzen-9-9950X
And even then, the accuracy is questionable.

You could build a database on catalogue data, theoretical computational performance, transistor count. For real world applications I think it's way more different. A coworker of mine made a little program that did fractal calculations to stress test old CPUs, but you should have to make something that takes into account if the CPU doesn't have a floating point unit, and scales well with CPU (core) count.

I can't give you actuall numbers, but I've owned and used a lot of systems from
1970's on up (See "Daves Old Computers").

I've written emulators for several of my favorite systems that I had over the
years. These were written and run under DOS, and include:

1974 MIL Mod-8 (8008)
1975 Mits Altair 8800A (8080)
1976 NorthStar Horizon / Vector 1+ (Z80)
1977 Heathkit H8 (8080)
1983 D6809 (homebuilt 6809 portable)
(I've just been discussing these in another forum on "non x-86 systems" which
is why I have this information handy)

I can tell you this:

At that time I wrote them my main desktop was probably a Pentium 2 or 3,
running DOS. Even on that system, the emulators ran faster than the original
machines ... but that was only 1 level of CPU emulation.

Currently my lab Desktop is an I7 running Win7-64. Since the emulators are
16-bit DOS programs, then don't run naively on Win-64.

So I run them under DosBox, which means that the emulated systems are being
run through 2 levels of CPU emulation (for this example I'll use D6809 which
emulates my homebuilt 6809 portable system):

X86-64 -> X86-16 -> 6809

And the CUBIX O.S. that I wrote for the 6809 portable, includes SIM80 which
is an 8080 emulator that I created to be able to run much of the code I'd
previously written. So if for example, I write a BASIC program in the BASIC
interpreter that I had written for the Altair and want to run it in the D6809
emulator, the CPUs being emulated are:

X86-64 -> X86-16 -> 6809 -> 8080

(This means that each instruction "executed" on the 8080 is actually performed
by 100s-1000s of actual CPU instructions)
... and said BASIC programs STILL run faster than they did on the Altair!

Most people don't truly realize how fast modern CPUs are! This is mainly
because software has gotten bigger and slower (taking advantage of the faster
hardware) - so the overall system speed seems a lot faster, but not that many
orders of magnitude faster than it is.

Back in the 70s, writing code for my Altair, I spent many hours carefully hand
crafting code in assembly language, both to make it fit in the available RAM,
and to make it run fast enough.

I think that if someone were writing a simple BASIC interpreter like I did for the
Altair on a modern system today, they'd probably do so in C++ or something like
that, and the executable could be anywhere from 50 to 100+k of code (depending
on what libraries got included etc. - and lets not even consider DLLs etc.)
My BASIC for the Altair was written in .ASM and produced a 3k executable.

Dave ::: https://dunfield.themindfactory.com ::: "Daves Old Computers"->Personal

Good overview I think in this article...
https://en.wikipedia.org/wiki/Instructions_per_second
Then a kind of benchmarking through the ages site, comparing methods and other issues relating to figuring out how fast CPUs are..
http://www.roylongbottom.org.uk/

Basically we've gone from half a MIP at home on your desk to pushing half a TIP (Terra or Trillion Instruction Per second)

edit: Ah whoops, MIPS isn't plural, the S is for seconds.

BitWrangler wrote on 2024-11-25, 18:37:

...
Basically we've gone from half a MIP at home on your desk to pushing half a TIP (Terra or Trillion Instruction Per second)
...

That's funny - the page you linked to shows "TIPS" as "Thousands of Instructions Per Second" 😀
(like anything is that slow these days! - but of course it's historical)

Dave ::: https://dunfield.themindfactory.com ::: "Daves Old Computers"->Personal

DaveDDS wrote on 2024-11-25, 18:11:

So if for example, I write a BASIC program in the BASIC interpreter that I had written for the Altair and want to run it in the […]
Show full quote

So if for example, I write a BASIC program in the BASIC
interpreter that I had written for the Altair and want to run it in the D6809
emulator, the CPUs being emulated are:

X86-64 -> X86-16 -> 6809 -> 8080

(This means that each instruction "executed" on the 8080 is actually performed
by 100s-1000s of actual CPU instructions)
... and said BASIC programs STILL run faster than they did on the Altair!

Most people don't truly realize how fast modern CPUs are! This is mainly
because software has gotten bigger and slower (taking advantage of the faster
hardware) - so the overall system speed seems a lot faster, but not that many
orders of magnitude faster than it is.

this is a great practical example of code speed on modern cpus, that a modern cpu is faster than an emulation target even when 2-3 times 'removed' by additional emulation layers

that computers often don't seem faster is almost entirely based on user interface factors and secondarily upon other factors like scripting rather than compilation, layers of libraries and apis and the like

I've noticed for years that the cheapest PC is advertised as "good for basic browsing, documents and so forth" when the computer of 2020 must have been so much faster than the computer of 2010 advertised on the same basis (admitting that 'just browsing' has become more expensive in cpu and ram over the years...)