Jump to content

Benchmarks: m68000 -> Apple M1


Recommended Posts

There's a benchmarking thread where we've started benchmarking CPUs all the way back to the m68000 in a Mac 512K. See here for history:

 

https://68kmla.org/forums/topic/62602-a-month-might-be-what-it-takes…/

 

This thread is to collect results that'll help us make good estimates about performance, such as how much overhead the video generation on early compact Macs causes. If you have the time, please download and run, then post your results with your machine's information :)

 

https://www.klos.com/~john/primesbench.sit

Link to post
Share on other sites

Have you heard of nbench?  It has been around a long time and tried on a wide variety of hardware/vintages:

 

https://www.math.utah.edu/~mayer/linux/bmark.html

 

That list of results has a 68030.  I wonder if it would compile and run on a 68000?

 

I don't have any 68k hardware, but if you post a .c file for your test I can try to compile and run it on my powerbook.

Link to post
Share on other sites
20 hours ago, herd said:

I don't have any 68k hardware, but if you post a .c file for your test I can try to compile and run it on my powerbook.

 

It's in the original thread:

 

Tweet sized benchmark

 

Other benchmarks are interesting, too, but this purely synthetic benchmark bears out very specific behavior.

 

nbench is a port of BYTEmark, and both exist in pkgsrc. However, I've only started learning how to compile stuff on Mac OS, so it may be a while before I can compile something with any complexity.

Link to post
Share on other sites
  • 68kMLA Supporter

Motorola StarMax 5000 (275 Mhz 603ev | no-SpeedDoubler)
==========================================
primeslong_000    562.400000
primeslong_020    599.250000
primes_000    11.700000
primes_020    12.416667

 

PowerComputing PowerTower (210 Mhz 604e | no-SpeedDoubler)
==============================================
primeslong_000    769.333333
primeslong_020    869.216667
primes_000    16.066667
primes_020    17.966667

 

PowerComputing PowerTower (210 Mhz 604e | SpeedDoubler enabled)
=================================================
primeslong_000    705.916667
primeslong_020    870.133333
primes_000    18.016667
primes_020    18.033333

 

Power Macintosh 4400 (200 Mhz 603ev | no-SpeedDoubler)
=========================================
primeslong_000    768.700000
primeslong_020    817.366667
primes_000    15.950000
primes_020    16.933333

 

Power Macintosh 4400 (200 Mhz 603ev | SpeedDoubler)
=======================================
primeslong_000    702.666667
primeslong_020    817.266667
primes_000    16.850000
primes_020    16.900000

 

 

Didn't know how much system info detail was needed, but I can revamp this if need be. Also, if there's a better method for formatting the scores, I can re-upload these.

Link to post
Share on other sites
  • 68kMLA Supporter

Oh, for tables...

 

SE, System 6.0.8, primes_000 ——— 2037.083333

 

SE + Radius SE25 (25MHz 020), primes_000 ——— 218.533333 (!)

SE + Radius SE25 (25MHz 020), primes_020 ——— 186.966667

 

Link to post
Share on other sites
On 4/19/2021 at 5:35 PM, cheesestraws said:

SE + Radius SE25 (25MHz 020), primes_020 ——— 186.966667

 

Neat! Thank you.

 

There's not much information about the Radius SE25 online that I can find. Is it the same as this accelerator?

 

https://youtu.be/Dniqwec7_ko?t=4656

 

If so, then is it safe to assume that this is using the SE's 16 bit memory?

Link to post
Share on other sites
  • 68kMLA Supporter
20 hours ago, johnklos said:

There's not much information about the Radius SE25 online that I can find. Is it the same as this accelerator?

 

Looks like it.  There were two versions, a 16 MHz and a 25 MHz, but as far as I know apart from the CPU speed they're identical (at least, the ROMs are identical and interchangeable).  There's no memory on the accelerator, so yes, it's talking to the SE's 16 bit memory.

Link to post
Share on other sites

I've updated the Stuffit to have nine binaries: three each for m68000, for m68020+, and for PowerPC. The three for each processor family are:

  • primes16bit, which counts the first 6,542 primes, which would be all the primes that can fit in 16 bits, unsigned
  • primesshort, which counts all the primes up to 1,000,000
  • primes, which counts the first million primes

primes16bit runs too quickly to be of much use on a fast m68k or on PowerPC, but I think it'll be interesting to compare 16 bit to 32 bit math on an original m68000, and it'll make a nice comparison for 8 bit and 8/16 bit processors.

 

Get them here: https://www.klos.com/~john/primesbench.sit

Link to post
Share on other sites
  • 68kMLA Supporter

If it's not too much of a pain, it might be educational to generate an equivalent to primes16bit, counting the first 6,542 primes, but using 32-bit maths to do so, and just carting around a load of extra 0s?  Intentionally pessimised, if you like.  That would give an apples-to-apples comparison of 16- and 32-bit maths, and just what the performance penalty involved is, without having to do mental arithmetic to scale times to match each other.  Mental arithmetic makes me sad ;-).

Link to post
Share on other sites
  • 3 weeks later...

Does anyone have the time to try 1,000,000 primes on an original m68000? I think it'd take about 26 to 32 hours or so. If so, we could compare that with the 1,000,000 primes on the ARM M1:

 

1000000, 15485863
./primes  2.42s user 0.01s system 96% cpu 2.528 total

 

I'll post a collection of these results soon.

Link to post
Share on other sites
  • 3 weeks later...
  • 68kMLA Supporter

Results for an SE/30 vs. an SE/30 + Carrera040 at 40MHz:

 

Quote

Time of run: 212.666667

 

vs

 

Quote

Time of run: 36.766667

 

Haven't had the /20 out to check your 16-bit version yet, sorry...

Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Loading...
×
×
  • Create New...