Some mandelbrot benchmarks for the PCs I own.
I have a collection of weird and wonderful computers, and I thought it would be interesting to benchmark them. The Mandelbrot Set seemed like a suitable non-trivial but relatively easy to port problem. So here in this repo, there is a program which can general an image of the Mandelbrot set, along with the results of running it on some of my computers.
The code code came from https://rosettacode.org/wiki/Mandelbrot_set#C. I've modified it to compile on a wide range of C compilers and platforms.
I used Visual Studio 2022 Community Edition, which came with Microsoft (R) C/C++ Optimizing Compiler Version 19.42.34433 for x86.
C:\Mandel> CL /O2 mandel.c
Microsoft (R) C/C++ Optimizing Compiler Version 19.42.34433 for x64
Copyright (C) Microsoft Corporation. All rights reserved.
mandel.c
Microsoft (R) Incremental Linker Version 14.42.34433.0
Copyright (C) Microsoft Corporation. All rights reserved.
/out:mandel.exe
mandel.obj
C:\Mandel> mandel
Ran at 5693 kpixels per second
C:\Mandel> magick output.ppm -normalize output.pngI used gcc version 12.2.0 (Raspbian 12.2.0-14+rpi1).
user@host:~/mandelbrot $ gcc -O3 mandel.c -o mandel
user@host:~/mandelbrot $ ./mandelI used Apple clang version 16.0.0 (clang-1600.0.26.4).
user@host:~/mandelbrot $ cc -O3 mandel.c -o mandel
user@host:~/mandelbrot $ ./mandelI used MIPSpro ANSI C 6.2 on IRIX 6.2.
% cc -O3 -o mandel mandel.c
% ./mandelBecause it was compiled on an R8000 machine, the C compiler produced a 64-bit executable using MIPS IV instructions.
I used gcc 3.4.3 on Solaris 7 for SPARC.
$ gcc -O3 -o mandel mandel.c -mcpu=v9
$ ./mandelThis particular C compiler produced 32-bit SPARC ELF files. I wasn't able to get it to produce 64-bit SPARC ELF code.
I used gcc 3.2 on HP-UX 11.00 for PA-RISC.
$ gcc -O3 -o mandel mandel.c
$ ./mandelThis particular C compiler produced 64-bit PA-RISC 2.0 ELF files.
The benchmark is relatively short on fast machines, and there's a lot of noise. They are really just to give you an order-of-magnitude difference between systems.
| Machine | CPU | OS | Compilation | kPixels Per Second | Cycles/pixel |
|---|---|---|---|---|---|
| MacBook M1 Pro | Apple M1 Pro @ 3.2 GHz | macOS 15.1 | 3 | 6230 | 513 |
| HP Z1 Entry Tower G5 | Intel Core i9-9900 CPU @ 3.10GHz | Windows 11 x64 | 1 | 5693 | 544 |
| Raspberry Pi 5 | Arm Cortex-A76 @ 2.4 GHz | Debian Linux 12.8 | 2 | 5300 | 452 |
| Raspberry Pi 4 | Arm Cortex-A72 @ 1.8 GHz | Debian Linux 12.8 | 2 | 1783 | 1009 |
| HP Visualize C3000 | HP PA-RISC 8500 @ 400 MHz | HP-UX 11.00 | 6 | 534 | 749 |
| Sun Ultra 80 | UltraSPARC II @ 450 MHz | Solaris 7 | 5 | 421 | 1068 |
| SGI POWER Indigo 2 | MIPS R8000 @ 75 MHz | IRIX 6.2 | 4 | 57.2 | 1311 |
Notes:
- On the MacBook M1 Pro you have to run it a few times in a row to get the CPU to ramp up to maximum performance.
- Some of these systems have multiple processors, or multiple cores within a processor, or multiple hardware threads within a core. This benchmark is strictly single-threaded though.
- Some of these systems have variable clock frequencies, and so for 'Cycles/pixel' we've assumed the system is running at its nominal clock speed, which might be wrong.