The z80 speed vs today’s modern processors

ページ 1/5
| 2 | 3 | 4 | 5

By Bengalack

Paladin (747)

Bengalack さんの画像

16-01-2023, 16:44

Sometimes I feel it could be ok to try to explain to (young) people how slow the z80 really is. It is of course very hard to compare, with todays multithreading and additional GPUs. They even have FLOPs which does not make sense on the z80.

Still, we could possibly compare based on integers and bit-shuffling. Are there any numbers known on comparisons?

I found this… which potentially could be used?
https://www.grauw.nl/projects/tiger/

I’m not sure how this Alpha compares to a modern high end Ryzen, but from the numbers, it seems to be 11 000+ times faster than a normal z80.

ログイン/登録して投稿

By erpirao

Paragon (1315)

erpirao さんの画像

16-01-2023, 17:02

you can use this link

By ducasp

Paladin (680)

ducasp さんの画像

16-01-2023, 17:48

Bengalack wrote:

Sometimes I feel it could be ok to try to explain to (young) people how slow the z80 really is. It is of course very hard to compare, with todays multithreading and additional GPUs. They even have FLOPs which does not make sense on the z80.

Still, we could possibly compare based on integers and bit-shuffling. Are there any numbers known on comparisons?

I found this… which potentially could be used?
https://www.grauw.nl/projects/tiger/

I’m not sure how this Alpha compares to a modern high end Ryzen, but from the numbers, it seems to be 11 000+ times faster than a normal z80.

I have no idea how to make such comparison, even because there are so many metrics (i.e.: integer performance, branching, floating point operations, cryptography, execution of a given script language, etc) and many of those wouldn't even apply for a z80...

The easiest way to make people understand how difficult things are for a z80 nowadays is to tell that most webpages simple logo images won't fit the address space of 64KB that it can directly address, that it has no provision for simple tasks (nowadays) like multiplication, division, floating point operations... Just telling it is 10000 (or whatever number of times) times slower won't even scratch the surface of how impossible it is to even make a mundane task like render a google search webpage because SSL would take 3 minutes to make initial handshake and then you have all those scripted content...

People has no idea of how much resources are used to render "mundane" tasks nowadays, they just take it for granted, probably they will say: ok, it is that slow, but still like an old phone it will take 20seconds to open a google page but it will do it... :P

By Bengalack

Paladin (747)

Bengalack さんの画像

16-01-2023, 20:10

erpirao wrote:

you can use this link

Tricky. But it seems like the Drystone possibly can be used... in a way.

By Prodatron

Paragon (1843)

Prodatron さんの画像

16-01-2023, 20:22

I remember that 10 years ago my notebook had about 300.000 times the power of my Z80 machine.
I had to explain this to some people at this time:
When you build a line of 300.000 Z80 machines (in my case an Amstrad CPC, but for an MSX it will be similiar) it will have a length of 160km, which is about the distance from Amsterdam to Roermond.

By Bengalack

Paladin (747)

Bengalack さんの画像

16-01-2023, 20:36

ducasp wrote:
Bengalack wrote:

Sometimes I feel it could be ok to try to explain to (young) people how slow the z80 really is. It is of course very hard to compare, with todays multithreading and additional GPUs. They even have FLOPs which does not make sense on the z80.

Still, we could possibly compare based on integers and bit-shuffling. Are there any numbers known on comparisons?

I found this… which potentially could be used?
https://www.grauw.nl/projects/tiger/

I’m not sure how this Alpha compares to a modern high end Ryzen, but from the numbers, it seems to be 11 000+ times faster than a normal z80.

I have no idea how to make such comparison, even because there are so many metrics (i.e.: integer performance, branching, floating point operations, cryptography, execution of a given script language, etc) and many of those wouldn't even apply for a z80...

The easiest way to make people understand how difficult things are for a z80 nowadays is to tell that most webpages simple logo images won't fit the address space of 64KB that it can directly address, that it has no provision for simple tasks (nowadays) like multiplication, division, floating point operations... Just telling it is 10000 (or whatever number of times) times slower won't even scratch the surface of how impossible it is to even make a mundane task like render a google search webpage because SSL would take 3 minutes to make initial handshake and then you have all those scripted content...

People has no idea of how much resources are used to render "mundane" tasks nowadays, they just take it for granted, probably they will say: ok, it is that slow, but still like an old phone it will take 20seconds to open a google page but it will do it... :P

Yes, young people today would possibly use the web as a baseline for comparison, but I think I'd have to try to get measures that does not include network latency. Of course it is hard, the full ram of MSX fits into the cache on the CPUs today, haha. But still, I thought it could be possible to get some numbers. Some kind of standard test written in C (and maybe where each C-compiler could have some kind of "effectiveness-factor" to adjust the results for poor compilers). Then one could do single-core comparisons. Maybe Drystone-MIPS and Grauw's tiger-comparisons is sufficient for me.

By santiontanon

Paragon (1805)

santiontanon さんの画像

16-01-2023, 20:41

Salutte created a benchmark suite for different languages (assembler, C, Fortran, etc.) for MSX: https://github.com/MartinezTorres/z80_babel

Since some are in high-level languages with compilers for both z80 and modern CPUs, that seems to me like a very simple way to get an approximation. Of course, getting to run one of the classic standard benchmarks like Dhrystone, but the advantage is that is should be pretty low effort to get numbers (as Salutte already calculated the numbers for MSX).

By smx

Expert (72)

smx さんの画像

17-01-2023, 11:14

I wonder why even though the performance of the processors is hundreds or thousands of times higher than the Z80, a correct emulation is still not possible.

I add, even if I go off topic, why not use for example multiple raspberries to do what they do in practice with the FPGA. One raspberry PI Zero dedicated to Z80 emulation only, another for PSG and another for VDP and so on.

By BadWolf359

Resident (33)

BadWolf359 さんの画像

17-01-2023, 11:11

I think it all depends on what you want to test. There are several aspects that you can test in terms of speed. How fast is each instruction, but also if things are missing in the CPU compared to another CPU, this must be done on a slower work around.
A modern CPU has several layers of cache that can help a lot in performance. CPUs also have hardware multipliers and dividers for integers as well as for floating point calculations. And lots more. The Z80 does not have all this. So a fair comparison should include this.

If you just wanted to compare the CPU, a prime numbers calculation would probably be a good one. It uses a lot of these functions including multiple cores. But with this you don't include the video interaction or moving large blocks of memory internally. But I think that you're not looking for an exact here, so this should suffice.

By santiontanon

Paragon (1805)

santiontanon さんの画像

17-01-2023, 12:13

@smx what part of the Z80 is not currently correctly emulated?

By smx

Expert (72)

smx さんの画像

17-01-2023, 14:07

Mine is just an observation that the emulation despite faster processors and still behind. It is no coincidence that we talk about FPGAs as the future for this sector

ページ 1/5
| 2 | 3 | 4 | 5