Roy Longbottom

UK Government · Central Computer Agency

Benchmarks and stress tests were run on existing devices, confirming that they ran successfully via upgrades to Android 12 and 13. A new phone was included with one Cortex-X2 cpu core and seven others of two varieties producing highest speeds but with questionable performance characteristics.

Folders contain source code with compile commands, compiled programs, example results and script files to select run time parameters. A preprint of the report is included.

Analysing results of a large number of my benchmarks and stress tests run on the new Raspberry Pi 5. Comparing performance with a Pi 4, PCs and Supercomputers. Evaluating cooling arrangements and power supplies with extreme stress testing.

The compressed file contains C source code and compiled program of a stress test involving integer SIMD calculations.

This is a list of files with download links and summaries, where reports for various projects include results on different platforms, running the same benchmarks and stress tests or alternative categories of historic performance.

Celebrating 50 years of computer benchmarking and stress testing Background to support claims made in other publications.

Benchmarks run on the original Raspberry Pi Pico were repeated to identify performance differences, with additional ones covering the new WiFi facilities.

This report covers performance from running my standard data storage unit test procedures on Windows and Linux based PCs, Android devices and Raspberry Pi systems, comprising writing and reading large files, numerous small files and random access, covering main. USB and network drives. Common, but variable, problems are identified associated with f...

The main comparisons are based on benchmark results used to verify performance of the first Cray 1, with variations from two similar vintage benchmarks. Maximum MFLOPS performance, 100% vectorisation and multiprocessor effects are also considered. Samples of historic benchmark results are included along with others for the latest compilations. Bott...

I ran my Android benchmarks using the Operating Systems versions 10 and 11 to verify that they continued to operate successfully. The later version was installed on a new phone using more advanced technology. Both phones employed big.LITTLE CPU Architecture, the old one having four fast and four not so fast cores and the new one two faster and six...

I ran my Android benchmarks using the Operating Systems versions 10 and 11 to verify that they continued to operate successfully. The later version was installed on a new phone using more advanced technology. Both phones employed big.LITTLE CPU Architecture, the old one having four fast and four not so fast cores and the new one two faster and six...

The Raspberry Pi Pico is a microcontroller with a number of advanced options, but this report concentrates on relative performance of Pi 4, Pi 400 and Pico, running simple bit banging operations, using the same procedures in programs written in C and Python. They comprise six tests with sleep delays aimed at executing between 5 and 500000 cycles pe...

Benchmarks and stress tests were run on a Raspberry PI 400 desktop computer, using the 32 bit and 64 bit Operating Systems. The PC comprises an upgraded version of the Raspberry Pi 4B CPU, fitted and fanless within a Raspberry Pi keyboard, running at a MHz of 1800 instead of 1500. Benchmark results were compared with those run on the original Pi 4B...

I have compiled and run my benchmarks to run under the 64 bit Beta Raspberry Pi OS, with tests including variations to exercise the Pi 4B with 8 GB RAM. Full results, with comparisons to 32 bit working, are included, also a link to download all the benchmarks and source codes. Some results show significant 64 bit performance gain and others no gain...

I have a Raspberry Pi 4 with a PoE HAT that incorporates a cooling fan and can convert power supplied by an Ethernet cable to use instead of the normal power supply. The Pi can then be used, at a distance, with a one wire connection. A unit to add the power is required, with pass through Ethernet, at the other end. Benchmarks and stress tests were...

My Android benchmarks required a minor change in order to run via Android 8. This report covers the latest benchmarking exercise, intended to verify that they run using Android 9. This was successfully achieved. Two 2019 inexpensive devices were used that include the same 8 core ARM CPU technology, one running at 32 bits and the other at 64 bits. S...

On running stress tests on a Pi 4, without a cooling fan attached, CPU temperature can increase, leading to clock speed throttling in stages, normally between 1500, 1000, 750 and 600 MHz. In turn, this leads to slower performance, proportional to to clock speed reduction, for the processor speed limited programs. This report covers a number of diff...

Previously, I have run my 32 bit and 64 bit benchmarks and stress tests on the appropriate range of Raspberry Pi computers, up to model 3B+. Details of the benchmarks, results and download links are available from ResearchGate in a PDF file. I have also run the 32 bit versions on the Raspberry Pi 4, with results in Raspberry Pi 4 Benchmarks PDF fil...

This report contains details and results of the programs used for stress testing Raspberry Pi 4B. They cover multi core CPU integer and floating point tests with data covering caches and RAM, Input/Output exercisers for the main, USB and network connected drives and graphics activity via OpenGL. The programs used, or new test functions, are ones th...

This report contains results of my numerous benchmarks run on the Raspberry Pi 4. In this case, they were 32 bit versions using Raspbian Buster Operating System. Existing benchmarks were used to provide comparisons with the old 3B+ model. The benchmarks were also recompiled using gcc 8, that came with Buster, to provide further comparisons. The ear...

This report includes results of repeated runs of High Performance Linpack Benchmark on a Raspberry Pi 3B, using an original precompiled version that was subject to producing the wrong numeric calculations and system crashes. These were repeated and also run via a newer version of Raspbian and all on a Raspberry Pi 3B+. A newer HPL program was produ...

This report contains results of my numerous benchmarks, running on a Raspberry Pi 3B+, providing comparisons with the older model 3B and 32 bit versus 64 bit working. One major objective is to identify strengths and weaknesses, rather than an overall rating. Single core Benchmarks generally show 3B+ performance improvements proportional to the fast...

Updated OpenMP MFLOPS Benchmark using 64 bit gcc 7 to use more efficient vector instructions.

Contains source code and benchmarks for MP-MFLOPS compiled by gcc 7 for highest speed.

Contains source code and benchmarks for MP-MFLOPS compiled by gcc 7 for highest speed.

My original benchmarks generally ran successfully on devices controlled by up to Android 7. They could be installed, using Android 8, but failed to run. This was found to be associated with an unimportant function that obtained the version of Linux used for Android. The earlier applications were produced using Eclipse, running via Linux Ubuntu. o...

The report provides details of a series of stress tests that cover processor integer and floating point calculations, cache and memory speeds, disk and other drives, including local network traffic, and graphics hardware. The programs can either be run standalone or multiple copies of any at the same time. Some detailed results are provided plus re...

My first graphics benchmarks were written in the days of dial-up Internet access, to provide a range of performance measurements without the need to download large volumes of image data, starting with DOS in 1997 and progressing via Windows and Linux to present times, with added 64 bit varieties. The benchmarks run a series of different tests, wi...

CUDA, from nVidia, provides programming functions to use GeForce graphics processors for general purpose computing. These functions are easy to use in executing arithmetic instructions on numerous processing elements simultaneously. The benchmarks measure floating point speeds in Millions of Floating Point Operations Per Second (MFLOPS), with versi...

These benchmarks execute the same code as the original ones, designed to exercise a single CPU, but implementing multithreading to use up to all available cores. Some employ a single method of manual procedures, where there might be more suitable options, with others using OpenMP and QPAR to automatically generate parallelism. In most cases, 32 bit...

During the 1980s, relatively inexpensive UNIX based systems were being purchased for UK Government Departments, for use by limited user populations. Some of the application software was not very efficient, often generating 100% CPU utilisation. Some departments were happy with performance, except for certain transactions that could normally be show...

These benchmarks provide performance measurements over a wide range of data sizes, covering all caches and RAM using different processing scenarios. In many cases, the programs have been compiled for both 32 bit and 64 bit systems. They emphasise the danger of comparing computer system performance by using a single number. For example, performance...

Separate calculations are provided, of performance, using data from CPU only or L1 cache, L2 cache and RAM, in terms of %MIPS/MHz and %MFLOPS/MHz. based on 29 measurements from 10 of my benchmarks. The actual count of floating point operations and most integer assembly level instructions were known. Intel CPU types covered are from 80486 to Core i7...

This program was initially produced to identify the CPU under test, using such as CPUID functions. Later assembly code procedures were included that could potentially demonstrate better performance than compiled programs, using long series of add instructions to 1, 2, 3 and 4 registers, in turn. Speeds are measured in Integer MIPS (real assembly le...

The report contains descriptions and results of my Android Benchmarks, with links to download the apps and source code projects. The latest automatically install code to execute on either ARM or Intel CPUs and to run via 64 bit Android, when available.

The report contains details, links to execution files and source code, and results from Raspberry Pi, Pi 2 and Pi 3. Benchmarks and stress tests cover CPUs, memory, I/O, graphics and multiprocessing in more than 30 different categories.

Contains results of benchmarks via DOS, OS/2, Windows, PC Linux, Android and Raspberry Pi Linux

Contains results of benchmarks via DOS, OS/2, Windows, PC Linux, Android and Raspberry Pi Linux

Contains results of benchmarks via DOS, OS/2, Windows, PC Linux, Android and Raspberry Pi Linux

The Livermore Loops, Whetstone, Dhrystone and Linpack benchmarks are are the first programs that set standards of performance for computers. This report provides background information, sample standard output and lists of files included in this project.

To complete the benchmark set, this provides links to technical reports in my Whetstone Benchmark Project

In 1980 I created a database to contain information about mainframes, minicomputers, supercomputers and workstations. The information came from manufacturers’ publications and press reports. Performance numbers are claims, based on benchmarks or manual calculations. They are not likely to be particularly accurate but are probably suitable for histo...

Before standard benchmarks were available, average speed rating of computers was based on calculations for a mix of instructions with the result given in Kilo Instructions Per Second (KIPS). This document contains results produced by the UK Government Central Computer & Telecommunications Agency (CCTA). KIPS calculations are provided for 190 system...

The later Whetstone Benchmark provides performance measurements of each of the eight test loops, besides the overall rating in MWIPS (Millions of Whetstone Instructions Per Second). More than 1000 sets of these results are reported here. For the main C/C++ compilations, single and multiple core benchmark results are included, with program versions...

Several hundred results, from more than 50 computer manufacturers, for systems produced between the 1960s and 1990s, plus PCs to 2014.

Whetstone Benchmark, Excel Visual Basic version

"A Wiley-Interscience publication" Book mainly based on analysis of reliability statistics and acceptance trial results of numerous UK Government and University computers - 12 chapters, 321 pages.

Equipment failures on computer systems give rise to more interruptions to service and excessive periods of repair time than a user would expect. This paper examines theoretical and actual reliability and repair statistics, showing how the shortcomings are influenced by test techniques, maintenance engineers, application of the system and spares ava...