This page shows a list of benchmark programs I use to test my hardware. In case you want to see results click here for VGA and here for CPU results.


3DMark2000 is a synthetic DirectX 7 benchmark and doesn't always represent real-life performance. It's very CPU-limited so you'll need a fast CPU to max out your graphic card. This means a GeForce4 Ti4200 will probably get the same results as a Radeon 9800 Pro if you benchmark it with a slow Pentium... > Read more

3DMark2001 Game4: Nature

This is a popular benchmark because it is not much CPU-intensitive (especially for it's time). Some benchmarks are limited by the CPU because the CPU cannot feed enough data to the graphic card. When this happens the graphic card cannot show it's full potential. The nature demo from 3DMark2001 doesn... > Read more

3DMark2003 Game4: Nature

Futuremark's first DirectX9-benchmark. I only use the fourth demo; Mother Nature to test SM1.0 and SM2.0 pixel and vertex shaders.... > Read more


3DMark05 uses DirectX9.0c (with all tests requiring Shader Model 2.0) and puts more than 2 million polygons on the screen.

I run 3DMark05 using the normal/default settings on 1280x1024. After running the benchmark I click on the 'View your results online' button (why does it have to be online!?) ... > Read more

3DMark99 MAX

3DMark99 MAX is an updated version of 3DMark99. It runs a number of tests using DirectX 6.1. I use this benchmark and run all the tests on 640x480 at 16-bit colors (800x600 is default) to test older 3D-cards.... > Read more


An open-source compression tool that can handle all sorts of filetypes (.ZIP, .RAR, .GZ etc) but also uses it's own .7z file-format. It also incorporates a little benchmark to measure how fast a CPU would be with 7-Zip algorithms.

I use this benchmark for CPU's with a minimum of two cores. Becaus... > Read more

Aquamark 3 GFX

Aquamark is based on the game Aquanox and uses a default resolution of 1024x768 with 4xAF (Anistropic Filtering, it will enhance image quality). As I benchmark with default settings the 4xAF setting is enabled.

Aquamark 3 is pretty straight forward. After installation click on the first option in... > Read more

AutoCAD R10

A non synthetic benchmark that utilises a floating point unit (FPU, co-processor). For ease of handling I've stripped down AutoCAD R10 to the bare minimum files needed to load the sample 'STAIR.DWG' file.

This application works only with an FPU installed. The range of systems I'd use with this b... > Read more

ChipTest RAM

Memory bandwith benchmark for DOS.... > Read more

Cinebench 2003 (Multi CPU rendering)

This benchmark is suitable for CPU's from 2000 and upwards to modern dual and quad-core CPU's. It will run quite slow on an Athlon 500 or similar but it can give a good picture of the performance between a single CPU of dual CPU (or either dual-core) system. Cinebench 2003 is great of testing perfor... > Read more


Codecreatures is a DirectX 8.1 benchmark written by Codecult in 2002 that involves vertex- and pixelshaders. I use it to benchmark cards starting from the GeForce 4 Ti era. Codecult doesn't seem to write new benchmarks anymore so this will be the only program I will use for older generation of graph... > Read more

CPU Free Bench

For me this was an unknown benchmark. I downloaded a few benchmarks because I needed various programs to test performance. After benchmarking for a while I noticed this benchmark doesn't stress the FPU much but rather goes to integer performance. Based on my feeling the score seems to represent the ... > Read more


This benchmark runs a series of tests and seems to indicate general performance. It doesn't stress a specific part of the processor so the results aren't representative for 3D-games like Quake or Half-Life or any other specific task/application. I will use this benchmark for CPU's between Pentium 10... > Read more

CPUMark '99

A single-threaded benchmark that is around for quite a while. Many people used it in the past and it seems like a good benchmark to test integer performance. Note that CPUMark '99 likes L2-cache while a large L2-cache doesn't always improve performance in, for example, office applications.

I have... > Read more

Crystalmark ALU

This ALU benchmark is part of CrystalMark 2004 and uses four (Fibonacci, Napierian, Eratosthenes, QuickSort) ALU benchmarks. ALU stands for Arithmetic Logic Unit and is a part in a CPU with is responsible for i.a. AND, NOT, OR and XOR calculations. All x86 (starting from 8086) have at least one ALU.... > Read more

Crystalmark FPU

This FPU benchmark is part of CrystalMark 2004 and uses four (MikoFPU, RandMeanSS, FFT, Mandelbrot) FPU benchmarks. FPU stands for Floating Point Unit. Not all x86 processors featured an FPU which meant that it had to be emulated, which is much slower. Modern (actually starting from the 80486DX) CPU... > Read more


This game is practically forgotten but was 'the' Unreal of the arcade racing games back in the old days. It run's fine on 3dfx cards so I used it primarily as a 3dfx-benchmark. According to the readme.txt it can also works good on the ATI Rage Pro (but with far lower image quality compared to 3dfx),... > Read more

dhry1od (Dhrystone)

This is a synthetic benchmark that runs on DOS to benchmark integer performance (ALU). Dhrystone was very popular but today people use 'SPECint' instead. Dhrystone is the opposite of Whetstone which stresses the floating point unit (FPU).

I use this benchmark on 25MHz-233MHz systems.... > Read more

dhry2ont (Dhrystone)

More or less like dhry1od for MS-DOS, but for Windows.... > Read more


Doom is a classic DOS game. Back in 1993 this caused a revolution in the 3D-shooter world! Because it's an old benchmark it's suitable to measure the performance of 486 CPU's. Luckily Doom scales up quite well so you can easily use it to benchmark up to fast Pentium, Cyrix 6x86 or AMD K6 CPU's. Be s... > Read more

Doom 3

Doom 3, based on the ID Tech 4 engine, is an OpenGL rendered game. When Doom 3 was new the graphical effects were considered high quality and needed the fastest graphics cards around. This makes it ideal to put the OpenGL qualities of card from 2004 to the test! The Doom 3 engine has been used in va... > Read more


A little benchmark program I wrote by myself in Turbo Pascal 7. I wrote it to create a benchmark that would run on 8086/8088 and 286 systems. The algorithm is quite simple as it only does a few mathematical calculations. Because there are some floating point calculations systems with a floating poin... > Read more


Create a shortcut of 'go.exe' in the game-folder and use -timedemo as parameter. When the demo is finished there will be a new file called timedemo.txt in the game-folder with the details.

In order to change settings you have to start the Expendable executable and choose setup. I use basic... > Read more

Far Cry

The first version of Far Cry was released on March 31, 2004. It was known for it's graphics and out-door freedom (see screenshots with islands, water, green trees). Far Cry uses the CryEngine (developed by Crytek) which is very suitable for benchmarking graphic cards starting from the 2004-era.

I... > Read more


FFMpeg is a widely used tool to encode video, but often it's driven by a (third-party) GUI. This allows users to easily select their settings like codecs, resolution and what not more. It's also very suitable as a video-encoding test, also because it can utilize up to eight cores. I will use this be... > Read more

FPUMark '99

The brother of CPUMark '99. This benchmark tests the FPU (floating point unit) rather then integer performance. Because games like Quake use the FPU a lot the scores of FPUMark '99 should be pretty much the same.

FPU WinMark is part of WinBench 99. I use [url=]Crys... > Read more


According to the website on (Geeks3D): FurMark is a very intensive OpenGL benchmark that uses fur rendering algorithms to measure the performance of the graphics card. Fur rendering is especially adapted to overheat the GPU and that's why FurMark is also a per... > Read more

Half-life (timerefresh)

Does this game need an introduction? Half-life is released in 1998 and a popular 3D-shooter. It runs on a modified Quake engine and allows people to customize the game by using 'mods'. These mods (modifications) often feature new models, maps, sounds or code for enhanced (on-line) gameplay. Due to t... > Read more

Hexus PiFast!

Hexus PiFast! calculates Pi and is quite a popular program to measure performance. Websites or forums about overclocking often use PiFast! to indicate the performance of CPU (and RAM) setups. The application is CPU and memory intensive and lower scores are better.

My benchmark method

Unpa... > Read more

Lame 3.98.2 MP3 encoding (6:15 WAV)

A real-life single-threaded benchmark. This benchmark isn't L2-cache intensive so a Duron or Celeron can be as fast as their more expensive Athlon or Pentium counterparts if other factors like the front side bus (FSB) are the same.

This benchmark is suitable for anything between 500MHz sixth gene... > Read more

Landmark 6

Landmark System Speed Test Version 6.00 made by Landmark Research International Corporation. It benchmarks your system and tries to relate the speed of your system to an AT system. A somewhat weird method of indicating performance because a PC has many factors that 'make' performance and especially ... > Read more


A DOS game that needs a Pentium-class (fifth generation) CPU to run at all. You can run it on the Cyrix 5x86, Pentium-class or newer CPU's.

The demo version, though being 3D-rendered, doesn't use 3D accelerators. In case you use a decent PCI graphics card this game is excellent for benchmarking C... > Read more


Released in the year 2000 and sequel to MDK from 1997. In MDK2 you play Kurt, Max and Doctor Hawkins. They all face action and puzzles in order to complete the game.

MDK2 requires DirectX5.0 and runs on a Pentium 200 with a 4MB 3D-accelerator, but a Pentium II 300 with 8MB 3D-accelerator is recom... > Read more


NSSI stands for Navratil Software System Information and is an utility that provides general information about your PC and it includes a benchmarking feature to test either CPU or FPU performance. I use this benchmark on 286, 386 and low-end 486 systems.

My benchmark method

Start NSSI.E... > Read more


PC-Tools CPU benchmark index. PC-Tools 9 requires an 286 or higher which makes it suitable for benchmarking 286 to 486 systems.... > Read more

PoV RAY 3.7

I use the PoV RAY benchmark for CPU's with at least 4 real (non-hyper threaded) cores available.... > Read more


Another classic DOS game that uses the floating point unit (FPU) intensively. This is why you need a FPU to run Quake at all.

The benchmark results will be quite predictable in the long run. The Intel Pentium is known to have quite a decent FPU while the non-Intel parts like Cyrix 6x86, IDT Winch... > Read more

Quake 2 640x480 software (demo1)

I used Quake II to test performance of CPU's under Windows '98. Because I used Quake at 320x200 for 386, 486 and fifth-generation CPU's I thought Quake II at a higher resolution would be great to indicate performance of lower-end fifth-generation CPU's up to faster 1000MHz CPU's.

To benchmark I i... > Read more

Quake 3

A classic benchmark! It has been used a million times in the past to benchmark cards like the Voodoo5, GeForce 2,3 and early Radeon cards. It's suitable for faster cards as well but then you have to crank up the resolution (and perhaps detail or Anti-Aliasing settings) to prevent the CPU from being ... > Read more

Soldier of Fortune

Based on a modified Quake II engine. Soldier of Fortune was criticized for being violent. Especially regarding blood and the GHOUL damage model which allowed the player to dismember body parts from killed enemies.

The game also featured a difficulty setting in which enemies appeared 'on the fly'.... > Read more


Chris' 3D bench. The benchmark-suite contains both a VGA and SVGA version. I use the SVGA-version and note the frames per second (FPS).... > Read more


A system information tool from the past. Their website is still operational and can be found here: This application can also perform a small benchmark for hard drives and graphics cards. I used it to benchmark textual speed of old VGA cards. Note that I took ... > Read more

Turbo Pascal 7.0

The command-line compiler built-in Turbo Pascal 7.0 for DOS shows how long it took to compile a certain source code. To benchmark early systems with a real-life benchmark I use Turbo Pascal 7.0 with the Bloxinies II source code.

I've found that the compil... > Read more


A game from 1998 with lot's of graphical features. Because it runs great with 3dfx Glide and OpenGL I selected it to benchmark graphic cards. The cards that won't run in OpenGL are tested in Direct3D mode. Often OpenGL looks better than Direct3D with Unreal on older cards so I prefer to test with th... > Read more

Unreal Tournament 2003 FlyBy

The sequel to the original Unreal Tournament from 1999. According to Wikipedia this game set a record in number of downloads when the demo-version was released. The Unreal Tournament 2003 engine has been used for several games such as Rainbow Six, Splinter Cell and America's Army.

Unreal Tourname... > Read more


Vidspeed is a benchmark that tests graphical performance under DOS. I Vidspeed with the L parameter in order to benchmark the standard 320x200 pixels VGA resolution.

Benchmarks are in '16-bit writes'.... > Read more

whetcod (Whetstone)

Whetstone is the opposite of Dhrystone and instead of integer performance it tests floating point performance. This is why CPU's without a FPU or a decent FPU score very low. For example and as seen in the benchmarks: a 386DX25 with FPU installed can easily beat a 486SX2/50 with a big margin.... > Read more

Wolfenstein 3D VGA

I use a patched version of Wolfenstein 3D. It goes by the name WOLF_286.EXE and you can download it here. You can use the executable with Wolfenstein 3D 1.4 full-version. To run the benchmark start 'WOLF_286 TIMEDEMO'. The TIMEDEMO suffix tells the executable to perform a benchmark :).

Unfortunat... > Read more

Wolfenstein: Enemy Territory

Wolfenstein: Enemy Territory is a free multiplayer game from ID software. It's available for both Windows and Linux platforms and includes a timedemo function.

My benchmark method

I used a demo I found on the internet which is suitable for version 2.60. To run the demo I open up the con... > Read more


I use wPrime only for CPU's with at least 4 available threads (i.e. 2-core with HT-enabled or true 4-cores). As benchmark setting I use 1024M setting with as many threads as a CPU has available.

To run the benchmark start it as 'Admin'. In order to speed up the starting process of wPrime press an... > Read more

X-Vesa reads

X-Vesa shows information about your graphics card and is able to measure VRAM speeds using different kinds of graphic modes. You need a VESA-compliant graphics card for this benchmark.

I use this DOS benchmark for ISA, VLB and older PCI cards to get insight in their VRAM performance. Note that th... > Read more

X-Vesa writes

X-Vesa shows information about your graphics card and is able to measure VRAM speeds using different kinds of graphic modes. You need a VESA-compliant graphics card for this benchmark.

I use this DOS benchmark for ISA, VLB and older PCI cards to get insight in their VRAM performance. Note that th... > Read more

X3 Reunion Rolling Demo

EGOSOFT's X3 Reunion Rolling Demo is a specific version that runs a demo of X3 while recording the FPS.

I run X3 at 1280x1024 in order to benchmark the 2002 - 2007 era video cards and I use the first part of the demo for the measurement. This means I press [Escape] a few times during the loading ... > Read more