Comparing Different Computers with N Queens Program

--D. Thiebaut (talk) 11:05, 24 October 2014 (EDT)

---

This is an example of how one can generate a quick comparison of the processing power of different systems using a CPU-bound (not using disk or network) application. This app in this case is a java program that finds the solution to the N-Queens problem while counting the number of probes it makes to the 2-D board. The source code is available here.

Here are some results gathered in Oct. 2014 on several systems I have access to. NxN is the dimension of the board. The different systems are:

1. Macbook Pro, 2009, 2.53 GHz, Intel Core 2 Duo, 8GB 1067MHz DDR3, 2 cores
2. Macbook Pro, 2014, 2.8 GHz Core i7, 16GB 1600MHz DDR3, 4 cores
3. MacPro, 2009, Intel(R) Xeon(R) CPU E5462, 4 cores
4. MacPro, 2014, Intel(R) Xeon(R) CPU E5-1650 v2 @ 3.50GHz, 6 cores
5. Beowulf2, (year?) Intel(R) Xeon(R) CPU E5-2650 0 @ 2.00GHz, 1 core
6. Laptop 1 = MSI GE62 Apache Pro 219, 2015, i7 5700hq @2.7GHz, 16.0 GB 1600MHz DDR3L, 4 cores
7. Linux 8 = AMD FX(tm)-8320 Eight-Core Processor, 3.5 GHz, 2GB cache
8. Arduino = Diecimilia, ATMega168 processor @ 16 MHz.

All times in the table below are in milliseconds (ms).

NxN	Macbook Pro 2009	Macbook Pro 2014	MacPro 2009	MacPro 2014	Linux Mint Beowulf2	Laptop 1	Linux 8	Arduino
8x8 9x9 10x10 11x11 12x12 13x13 14x14 15x15 16x16 17x17 18x18 19x19 20x20 21x21 22x22 23x23 24x24 25x25 26x26 27x27 28x28 29x29 30x30	0 0 0 0 1 0 26 17 37 28 87 19 144 36 750 81 242 92 248 284 1551 1518 33667	1 0 1 0 1 0 3 2 14 10 34 7 64 13 311 31 101 37 102 118 633 356 12586	0 1 0 0 1 0 8 5 27 19 67 13 119 26 662 63 199 72 201 232 1250 687 23558	1 0 0 0 0 0 5 2 18 13 41 9 70 17 384 39 119 44 121 142 734 413 14158	0 0 0 0 1 1 17 14 46 31 104 22 152 50 610 101 226 111 227 257 1213 684 22889	0 0 0 0 0 0 0 0 1 1 5 1 24 1 221 3 56 7 58 69 469 244 9041	0 0 0 0 0 1 9 6 33 23 60 16 99 29 463 58 157 64 161 184 938 522 17252	493 2,585 14,016 81,831 512,952 3,387,865 NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA |} </center> <br /> ==Comparing Java to C== <br /> This calls for a quick comparison of the Java to C implementation available [[N-Queens in C|here]], compiled by a C++ compiler with O3 optimization, and running on the fastest of the machines above, a MacBook Pro, 15" Retina Display, with a 2.8GHz 4-core processor. The computation uses only one core. All times are in milliseconds (ms). <br /> <center> {| class="wikitable" |- ! NxN ! Java<br />1.7.0_60 ! C<br />Compiled with g++<br />&amp; -O3 optimization |- | <source lang="text"> 8x8 9x9 10x10 11x11 12x12 13x13 14x14 15x15 16x16 17x17 18x18 19x19 20x20 21x21 22x22 23x23 24x24 25x25 26x26 27x27 28x28 29x29 30x30	1 0 1 0 1 0 3 2 14 10 34 7 64 13 311 31 101 37 102 118 633 356 12586	0 0 0 0 0 0 0 0 1 0 4 0 17 1 159 2 40 5 40 49 319 170 6320

For more information on the difference between Java, C, Python, and Assembly, go to this page.

N-Queens on a 16 MHz Arduino

The code for the N-Queens on the Arduino is available here.

NxN	Elapsed Time in ms
8x8 9x9 10x10 11x11 12x12 13x13	493 2,585 14,016 81,831 512,952 3,387,865

Note that 3,387,865 ms is 56 min, 28 seconds! The Arduino is spending almost an hour computing a result that took a Desktop machine less than a millisecond to compute (Java or C++).