How much faster is an iMac than an A500?

[Rotareneg]

And for a slightly more productive post:

The only way to get the emulated software running anywhere near the speed you're looking for is by turning on JIT mode set to fastest possible. That recompiles the 680x0 instructions into x86 instructions and runs them directly on the real CPU, which is far faster than the normal emulation.

[BastyCDGS]

To be fair, one should compare a modern CPU having all caches disabled, since 68000 did not have any caches by itself.

The performance drop when disabling all caches on modern CPUs is amazingly high!

[Starglider 2]

Let me rephrase this question:

As an average figure between the high and low possibilities, what would you guess the multiple is?

[Pixelfill]

Originally Posted by Starglider 2

Let me rephrase this question:

As an average figure between the high and low possibilities, what would you guess the multiple is?

42

[th4t1guy]

You could probably just reference this table:
https://en.wikipedia.org/wiki/Instru...ons_per_second

I assume this question will not have a practical application, so a guesstimate should be good enough.

[rare_j]

Eleventy billion.

[gulliver]

You have hit a wasp nest: they are not easily comparable, as others
have pointed out, and benchmarks always suck, no matter which one you
choose.

But lets get wild:

A stock Amiga 3000 is aproximately 6 Linux BogoMips.

An Amiga 500 is aproximately ten times slower in performance, so lets
assume 0.6 Linux BogoMips aproximately for an A500.

Then lets go to the computer to compare it with:

iMac 3.6GHz 8-core 9th-generation Intel Core i9 processor
This one seems to report a little over 7000 BogoMips.

So 7000 / 0.6 = madness

That iMac is more than 11,666 times faster than a stock A500 when
using the Linux BogoMips benchmark.

So again. benchmarks are stupid, and unreliable, and this one is no
exception.

I hope it helps to at least get some sort of unfair comparison.

[Starglider 2]

Okay cool, so it's somewhere between 2000, 11666, and eleventy billion!

[Starglider 2]

Originally Posted by th4t1guy

You could probably just reference this table:
https://en.wikipedia.org/wiki/Instru...ons_per_second

I assume this question will not have a practical application, so a guesstimate should be good enough.

Okay, that works out around 450x.

So we're somewhere between 450, 2000, 11666, (and eleventy billion).

This gives a nice round figure of: 4705

[gulliver]

And that is why when you see an Amiga demo coder you should vow with utmost respect:

They could probably rebuild the universe and seven parallel words with an Intel Core i9.

;-)

[th4t1guy]

Originally Posted by Starglider 2

Okay, that works out around 450x.

Shouldn't it be much higher? 68000 @ 8mhz (a500 example) is listed as 1.4 mips. While the i9 is not on there, the i7 ranges anywhere from 50k to 100k mips.

[Starglider 2]

Originally Posted by th4t1guy

Shouldn't it be much higher? 68000 @ 8mhz (a500 example) is listed as 1.4 mips. While the i9 is not on there, the i7 ranges anywhere from 50k to 100k mips.

I was looking at per clock cycle

[th4t1guy]

Originally Posted by Starglider 2

I was looking at per clock cycle ¯\_(?)_/¯

Ah, then that would be 450x faster than a 68000 clocked at 3 ghz

[Hewitson]

All this discussion is meaningless without figures. What would be a simple but effective way to benchmark the two systems? Something like calculating pi to 10,000 decimal places?

[meynaf]

Originally Posted by Rotareneg

The only way to get the emulated software running anywhere near the speed you're looking for is by turning on JIT mode set to fastest possible. That recompiles the 680x0 instructions into x86 instructions and runs them directly on the real CPU, which is far faster than the normal emulation.

While a lot faster than normal emulation, it's still slower than native code.
It does however allow to execute the exact same benchmark program on both machines.

Originally Posted by gulliver

That iMac is more than 11,666 times faster than a stock A500 when
using the Linux BogoMips benchmark.

So again. benchmarks are stupid, and unreliable, and this one is no
exception.

This kind of benchmark is biased because it also tests compilers - and 68k compilers are known to produce poor code.
But if you badly want numbers, you can estimate the compiler made the code 4 times too slow on the 68k...

Originally Posted by gulliver

And that is why when you see an Amiga demo coder you should vow with utmost respect:

They could probably rebuild the universe and seven parallel words with an Intel Core i9.

;-)

Nothing special a demo coder would do with an intel cpu. These are an horror to code in asm.
But a multi-ghz 68k derivative would have been fun, yeah.

Originally Posted by th4t1guy

Ah, then that would be 450x faster than a 68000 clocked at 3 ghz

Let's do an estimate for the ipc (instructions per cycle).
The 68000 has peak ipc of 0.25.
Modern (cough) x86 cpus have peak ipc probably of 4.
That's not more than 16 times faster.
Yet this is only academic - 68000 can't reach 3ghz.

Originally Posted by Hewitson

All this discussion is meaningless without figures. What would be a simple but effective way to benchmark the two systems? Something like calculating pi to 10,000 decimal places?

I'm afraid 10,000 pi decimals would be too fast on intel core cpu to be just measurable ;-)
And if running the same program, it would again be sensitive to compiler differences.

[robinsonb5]

Originally Posted by Hewitson

All this discussion is meaningless without figures. What would be a simple but effective way to benchmark the two systems? Something like calculating pi to 10,000 decimal places?

Better yet, stick with the chess theme and use the Queens problem? (Or a few hundred / thousand iterations of it on the Intel CPU!)

(It's a simple enough problem that anyone who objects to compiler differences giving the 68k an unfair penalty could fairly easily write an asm version!)

[Glen M]

Can you not just settle on "a lot faster".

What is the context of your future video?

According to wikipedia an intel i7 6950x achieves 317,900 MIPS at 3ghz compared to the 68k getting 1.4mips at 8mhz. So simple calculator time assuming MIPS scale linear against cpu clock (which they don't)

The 68k at 3ghz would get 525mips and the Intel at 8mhz would get 847.73mips.

I think its safe to say the Intel is a faster processor.

[roondar]

It's almost impossible to answer this question accurately. There are far too many variables to consider. It's already quite difficult to accurately measure the performance difference between two computers of roughly equal power that are using the same CPU family. Even statements such as "an A3000 is 10x the speed of an A500" are not as "simply true" as you might think. For some things, the A3000 is only marginally faster. For others it might be 20x+ the speed.

Plus, in the real world it doesn't actually matter as much what any synthetic benchmark tells you the performance difference is. What actually matters is how it performs for your specific uses. As a small example: I have a PC at home and at work. The one at work is six years newer and has a much faster processor. Yet, my PC at home is much faster in real world use because it does have an SSD and has plenty of memory, while my work PC doesn't have either.

Sure, the work one can calculate PI to 10,000 decimals a lot quicker. But I'm still far more effective on the one at home because it doesn't take a minute to start the applications I'm using and it doesn't need to swap out memory nearly as often.

[Jope]

Please everyone: do not use bogomips as any kind of comparison between different CPUs. It is a totally meaningless value.

[meynaf]

And all of this is for the cpu only. But the rest of the hardware also has to be taken into account.
As an example, Atari ST has 8Mhz 68000, Amiga 500 has 7.09Mhz 68000. So in theory the ST is faster...