![]()
PCMark04 features a series of tests that have different functions within PCMark. Part of the tests stress mainly the CPU, while part stresses multiple components at the same time. The default run also contains three test pairs to be run simultaneously multithreaded. The System Tests are used to produce the PCMark score. In the Professional Edition of PCMark04 there are also tests for individual components, such as CPU, Memory, Graphics and Hard Disk Drive. You can get a component score for an individual component by running the test suite specifically designed for it.
System Tests - OverallThere are 13 system tests - each one is designed to represent a certain type of PC usage. By running these tests, PCMark04 stresses the components in a similar manner as they are stressed in normal home usage. More information on the selection of these tests can be found on the PCMark04 White Paper.
The system test suite contains tests measuring different components of a PC and the results of the tests as well as the total score depend on all components. The system tests stress CPU, system bus, Memory and Graphics performance.
If not stated otherwise, all the tests’ data is first loaded in the memory to minimize the effect of hard disk performance. In normal PC usage the hard disk performance often is the bottleneck on several operations, and implementing all the tests in PCMark04 to utilize also the hard disk drive, would in many occasions result in measuring mostly hard disk drive performance instead of CPU or Memory performance as intended. PCMark04 offers a separate hard disk drive test suite for HDD performance testing.
System tests 1 and 2 – File Compression and DecompressionFile compression and decompression are very popular tasks in today’s PC usage, especially files transferred over a network are almost without exception always compressed. These tests use the open source Zlib data compression library (http://www.gzip.org/zlib)
Technical detail:
This test measures mainly CPU performance on integer computing. Data files that are used include a 2MB executable, 2MB document, 7.4MB video file and a 2.7MB .dds file. The files are processed (compressed or decompressed) as many times as possible during the 20 second runs and Megabytes processed is calculated. Metrics for File Compression and Decompression tests is Megabytes processed per second.
The File Compression test is run simultaneously multithreaded with the File Encrypting test and the File Decompression test is run simultaneously multithreaded with the Image Processing test in the default run.
System Tests 3 and 4 - File Encryption and DecryptionEncrypting is becoming more and more popular task with file sharing. Many companies already have a policy where all emails for instance must be encrypted. This test involves encrypting of different kind of files using the open source Crypto++™ Library 5.0 (http://www.eskimo.com/~weidai/cryptlib.html) implementation of Blowfish algorithm (http://www.schneier.com/blowfish.html).
Technical detail:
The encrypting and decrypting tests measure CPU performance on integer operations. The files used for encrypting and decrypting include 2MB executable, 2MB document, 1.1MB picture and a 1.7MB audio file. The files are en/decrypted in 8 byte padded blocks and the encryption key size is 56 bytes. The amount of operations processed during the 20 second run time is calculated for both tests and the metrics is Megabytes processed per second.
System Test 5 - Virus ScanningVirus scanning has become a very common and crucial task for using PC in everyday life. New viruses are reported constantly and they spread rapidly through email and internet browsing. Virus scanner engines prevent these viruses to harm the user's system and they have become an essential part of a home user's PC as well. The component used in this test is the virus scanner from F-Secure Corporation (www.F-Secure.com).
Technical detail:
Virus Scanning test measures CPU and memory performance with integer operations. The files scanned by the virus scanner include files for approximately 23MB of total size. The virus scanner itself scans different kind of files in different ways, and this test includes different types of files (e.g. executables, documents, text files) to reflect the real behavior of a virus scanning engine. The files are scanned consecutively in a loop, and the number of loops processed in 20 seconds is measured. The result is further calculated to Megabytes scanned per second.
Because virus scanning is a kind of task that is often done in the background while doing something else, this test is run simultaneously with the Grammar Check test multithreaded in the default run.
System Test 6 – Grammar CheckThis test simulates a user checking the grammar in a document. As text editing is among the most common tasks in using PC at home, and checking grammar from text is very popular while editing text, such test should be included in determining the performance of a system in home environment. In the default run, this test will be run simultaneously with the Virus Scanning test multithreaded. This test uses the Link Grammar Parser v4.2 (http://www.link.cs.cmu.edu/link/).
Technical detail:
The Grammar Check test measures mainly CPU performance with integer operations. A set of sentences are processed through the parser for 20 seconds and the amount of processed sentences is calculated. The data used in this test is a text file sized approximately 130kB.
The Grammar Check test is run simultaneously with Virus Scanning test multithreaded in the default run.
System Test 7 - Audio ConversionA part of PCs evolution is the increase of entertainment usage. Listening to music and creating own playlists are common tasks to perform while operating a PC. Audio Conversion test uses the Ogg Vorbis audio compression. (http://www.gnu.org/directory/oggvorbis.html).
Technical detail:
The Audio Conversion test measures mainly CPU performance with floating point operations. Audio Conversion test compresses a 10 second wave format audio file sized apporximately 1.7MB to Ogg Vorbis format and the time for encoding is measured.
System Test 8 - Image processingIn today’s world, where a big part of PC usage is internet browsing and viewing images and digital photos, JPG decoding is one of the most common tasks of PC home usage. This test uses the Standard JPEG library from Independent JPEG Group (www.ijg.org), which appears to be the most widely used format.
Technical details:
The Image Processing test measures mainly CPU performance. The JPEG decoding pipeline uses fixed-point IDCT and RGB-24 output pixel format, so this test uses only integer operations. The image is decoded one scan line at the time to a one scan line length buffer to provide cache coherent memory usage. The data files to be decoded are of jpg format including a 130kB picture, two 900kB pictures and 1.1MB picture. The content of the pictures include actual photographs and PC screenshots. The files have been chosen to cover different image compression ratios which now vary between 3 and 18 in the data files. The result metrics is Mpixels/second and the test runs for 20 seconds.
The Image processing test is run simultaneously with the File Decompression test multithreaded in the default run.
System Test 9 - Web Page RenderingInternet browsing is a task that can not be bypassed when discussing today’s PC use. In this test the system performance is measured when opening web pages through the Internet Explorer browser (http://www.microsoft.com/windows/ie/default.asp). This test requires Internet Explorer 6 installed.
Technical detail:
Web Page Rendering test stresses both the CPU and the memory subsystem. Html files including text, pictures and tables are opened through the browser for 20 seconds time and the amount of pages opened per second is measured. The data files this test utilizes are three different types of internet pages. The pages represent different types of web pages including a company site, a page containing pictures, and a page containing text. To ensure valid results, certain checks are made and this test is run only when the right conditions meet. These conditions include allowing picture decoding and running scripts. To make sure your system can run the test, it is suggested to change the security settings of the browser to medium for the time of testing. The test does not have any Internet activity, all the content is run locally.
System Test 10 - WMV Video CompressionWith digital video cameras increasingly gaining popularity, video conversion and editing is becoming a very popular task in home PC usage. PCMark04 video compression tests measure two different types of video file conversion. In the WMV Video Compression test, a Windows Media format video file is encoded. This test uses the Windows Media Encoder 9 (http://www.microsoft.com/windows/windowsmedia/9series/encoder/default.aspx) which needs to be installed separately.
Technical detail:
This Video Compression test stresses mostly the CPU with floating point operations. The video source data file used for the encoding is approx. 1.8 MB sized mpg file with a resolution of 320*240. Video compression is done through Windows Media Encoder to same resolution WMV format with video bitrate of 1000kbps.
System Test 11 – DivX Video CompressionThe DivX Video compression test is similar to the Wmv format encoding test, but uses the DivX format. The two different video encoding tests are both included because they are both widely used and they shouldn’t be generalized under one test. The tests also use different types of data. This test uses the DivX Video for Windows codec v5.0.5 (http://www.divx.com) for encoding the file.
Technical detail:
The DivX Video Encoding test stresses mainly CPU with floating point operations. The video source data file used in this test is a DV format file with a resolution of 720*480. The DV file format is commonly used by digital camcorders. The file is converted to DivX format with the same resolution. DivX is a popular format for storing digital video clips, due to the higher compression ratio. The encoding to DivX format is done aside the decoding of DV in separate threads, as the real applications do. The time for the process is measured and metrics is frames processed per second. The DivX Video Compression test reads the input file is from the hard disk and writes the output to the hard disk, so hard disk performance may have some effect to the test result. Note that PCMark04 includes a version of the DivX codec that can be used only for running this test.System test 13 - Physics Calculation and 3DGaming is a very popular task when using a PC. Futuremark has a benchmark series specifically designed to measure 3D gaming performance of a system, 3DMark. 3DMark series is encouraged to be used instead of PCMark04 to measure 3D gaming performance more deeply. However, gaming cannot be totally ignored when measuring PC performance in home usage and that’s why PCMark04 also includes a test representing a game-like workload. This test includes real time physics calculation based on Havok physics engine v 2.1 (http://www.havok.com/) and 3D content. This test requires a DirectX 7 compliant graphics card and DirectX 9 runtimes installed.
Technical detail:
The physics calculation test measures CPU and graphics performance with floating point operations. Havok’s physics engine v2.1 is used for the physics simulation and the 3D in this test uses DirectX fixed function multitexturing and vertex processing. The physics in this test consist of rigid body calculations and collisions. Visual content includes boxes falling and colliding with each other and with the surroundings. The test is run frame-based, i.e. the test renders the same frames every time, and this makes the test length to depend on the system's performance.
System Test 14 - Graphics Memory TestThis test is built to stress the video memory sub-system in the way regular Windows desktop applications utilize it. The test pushes both the graphic card's internal memory bandwidth as well as AGP bus transfer speed. The common usage of desktop applications like scrolling of documents and moving/resizing windows stresses the video memory subsystem in the same manner as this test does.
Technical detail:
The video memory test measures somewhat the CPU and memory speeds, but mostly the AGP bus and graphics memory performance. The test starts by creating back buffered primary surface in 1024x768x16bit resolution in DirectX exclusive mode and one off-screen work surface (twice as high) in the same pixel format. During the test, the work surface is updated each frame by transforming data through the AGP bus, with scrolling speed of 64 scan lines / frame. The work surface is blitted every frame to primary surface to stress the internal memory bandwidth. The test is run for 10 seconds and the frame updates made during that time is given as result.
System Test SuiteThe default test suite is the System Test Suite including tests combining the use of different components. The tests is the System Test Suite include:
Multithreaded Test 1 - File Compression and Encrypting
This test runs the File Compression and File Encrypting tests at the same time multithreaded.
Multithreaded Test 2 - File Decompression and Jpeg decompression
This test runs the File Decompression and Image Processing tests at the same time multithreaded.
Mutlithreaded Test 3 - Virus Scanning and Grammar Check
This test runs the Virus Scanning and Grammar Check tests at the same time multithreaded.
Test 4 - File Decrypting
Test 5 - Audio Conversion
Test 6 - Web Page Rendering
Test 6 - WMV Video Compression
Test 7 - DivX Video Compression
Test 8 - Physics and 3D
Test 9 - Graphics Memory
CPU Test SuitePCMark04 also includes test suite for testing CPU performance. The tests selected in this suite are the most CPU intensive and lack activities in other components, such as memory. The tests in the CPU suite include:
Multithreaded Test 1 – File Compression and Encrypting
This test runs the File Compression and File Encrypting tests at the same time multithreaded.
Multithreaded Test 2 - File Decompression and Image Processing
This test runs the File Decompression and Jpeg Decoding tests at the same time multithreaded.
Test 3 - Grammar Check test
Test 4 - File Decrypting test
Test 5 - Audio Conversion test
Test 6 - Wmv Video Compression test
Test 7 - DivX Video Compression test
Memory Test SuiteAs separated memory tests PCMark04 has tests stressing the memory sub-system with different sizes of raw memory read, write and copy operations and browse through a linked list, with elements pseudo-randomly in memory.
Technical detail:
The Memory Test suite includes raw read, write and copy operations performed for four sizes of arrays: 8MB and 4MB used for system memory and 192kB and 4kB for caches. The result is balanced towards the system memory performance, and the cache test results are first averaged before calculating the geometric mean. Each size of block is tested for five seconds and the amount of accessed data is given as result. There are additional block sizes included for advanced testing and also a test for measuring the peak bandwidth of the memory subsystem in copying operations.
The random access test uses a large list of data items. The size of each data item in the list is 64 bytes. The list is unsorted (or sorted using a pseudo-random key). The list is iterated through as many times as possible in 5 seconds and Megabytes of items accessed per second is given as the result. The list in the Memory Test suite is 8MB and 4MB for the system memory and 192kB and 4kB for caches and additional sizes are available in the advanced test selection.
The measurement unit for read, write, copy and random access is Megabytes per second.
There are also latency tests in the advanced selection of the memory tests. These tests are synthetic and designed to isolate the performance of that particular aspect of the memory subsystem performance. They measure the time it takes to access data memory. These tests use data blocks of 16 Megabytes for main memory, 192 Kilobytes for the L2 cache and 4 Kilobytes for the L1 cache. The results are in nanoseconds spent to access the data. Note that latency test results are not included in calculating the overall memory score.
Graphics Tests SuiteGraphics tests - overall
Since graphics is an important part of using today's PCs with graphic operating systems and applications, PCMark04 includes also tests measuring the graphics sub-system. The graphics tests in PCMark04 include both 2D and 3D content. However, since PCMark04 is a benchmark measuring general home usage of a PC, and not only graphics, the tests include some basic operations and not necessarily the latest features of today's graphics hardware. To fully measure a system's 3D performance, Futuremark's 3DMark should be used instead.
2D TestsThe 2D tests in PCMark04 include tasks that are common for today's 2D graphics and playback of video files to determine the performance of the graphics hardware.
Transparent Windows
The graphics performance is measured with several windows moving across the screen with visibility changing during the test.
Technical detail:
The test creates 10 windows and uses a 1024*768 space for drawing. Each window has its own fading (alpha blending) speed as well as moving speed. Each window is created fully visible, and they are faded away, after what the loop starts again. The time it takes to draw the windows and the number of windows drawn is measured and a metric of number of windows drawn per second is given as a result.
Graphics Memory tests
The Graphics Memory tests measure the performance of the graphics subsystem by stressing the AGP bus and graphics card memory. Note that one of the graphics memory tests is run in the System test suite, as it represents a common use of the graphics subsystem.
Technical detail:
The video memory test measures CPU, memory, AGP bus and graphics memory performance. The test starts by creating back buffered primary surface in 1024x768x32bit resolution in DirectX exclusive mode and one off-screen work surface (twice as high) in the same pixel format. During the test, the work surface is updated each frame by transforming data through the AGP bus, with scrolling speeds of 32 and 16 scan lines / frame. The work surface is blitted every frame to primary surface to stress the internal memory bandwidth. The test is run for 10 seconds and the frame updates made during that time is given as result.Video Playback tests
As a lot of today’s graphic usage on PC is playing video files, PCMark04 includes tests for measuring video playback performance also. There are two video playback tests, one displaying a clip of WMV format and other a DivX clip. Note that the Video Playback tests don't have affect the PCMark04 Graphics score.
Technical detail:
The WMV video file is a Windows Media 9 format file with resolution of 1024*768.The DivX content is a DivX format file with resolution of 720*480 resolution video file.
The video files are not limited to framerate, i.e. they are played as fast as possible and the framerate is measured. As the display's vertical refresh rate is the maximum achievable framerate, the tests only scale to a certain point. That is why they are excluded from the graphics total score. If a system gets a result of these tests that equals the vertical refresh rate, it should be considered quite good enough for video playback.
3D Tests3D graphics is yet mostly used in games, but since a 3D implementations are every day becoming more common in other applications too, and since practically all the available graphics adapters support 3D acceleration, 3D tests are justified also in a benchmark like PCMark04. The 3D tests in PCMark04 include only basic 3D operations (fill rate and polygon throughput) and don't require the latest hardware accelerated features to run. To fully test the 3D performance and features of graphics hardware, we recommend using the 3DMark series of benchmarks.
Technical details on 3D tests:
There are two tests measuring the fill rate of the graphics hardware. Fill rate is a measurement of how fast the graphics card is capable of drawing textures onto 3D objects. The test draws a number of large surfaces perpendicular to the viewer that cover the whole screen and fill these objects with large textures. There is a single-texturing and a multi-texturing test.
Fill Rate Test 1: There are 64 surfaces with one texture each. This means that the graphics hardware fill each of these objects separately, no matter how many texture layers that card is capable of drawing in a single pass.
Fill Rate Test 2: 64 textures are drawn using as many layers per pass as possible. DirectX 9 has a limitation of max. 8 texture layers per pass in fixed function multitexturing, so no more than 8 textures can be drawn per pass, although hardware might be capable of more.
The metrics for the fill rate tests is Mtexels (milloin texture pixels) per second.
Polygon throughput test
These tests measure the polygon throughput of the graphics hardware. A scene with no textures but a large number of polygons is drawn, and the amount of polygons drawn per second is measured. The tests also use specular lighting. There are two tests measuring polygon throughput:
Polygon Throughput Test 1: A single directional light is rendered.
Polygon Throughput Test 2: The same scene with 8 lights in all.
The polygon throughput tests use the measurement unit Mtriangles (million triangles) per second.
Hard disk drive testsHDD Tests - Overall
The HDD (Hard Disk Drive) tests measure the performance of the HDD subsystem. Hard disk performance depends on several things, and may vary even within the same system quite a lot depending on the consistency of the drive. In PCMark04's Hard Disk Drive tests the variability is been minimized and the results are comparable between different systems and disks. An important thing to remember while testing hard disk drives is that when a hard disk is connected to and operated through a PC, the performance is always somewhat dependant also on other components of the PC, such as CPU and the hard disk controller in the motherboard chipset.
The HDD tests simulate the usage of a hard disk in real applications, such as starting Windows, opening applications, copying files, and using applications. At the beginning of program execution, all the logical disks will be examined and the drives available for testing will be listed in the HDD Test target list on the Tests - Advanced dialog.
The HDD tests in PCMark04 are an implementation of RankDisk, developed by Intel®
NOTE! In order to receive reliable HDD test results, you should always defragment the tested drive before running the HDD tests. The defragmentation run can be run on Windows XP through Start - All Programs - Accessories - System Tools - Disk Defragmenter.
Technical detail on Hard Disk Drive test:
This test uses the RankDisk tool, developed by Intel® and is included in the iPEAK storage toolkit. RankDisk uses a predefined trace for disk activity and runs the "recorded" activities of the disk during a test. RankDisk disk accesses are generated to the raw disk using the NT device driver, but bypassing the file system and the operating system’s buffer cache. When running these HDD tests, a “dummy” file is first generated, which is then used for all activities of the test. This file is created to make sure that the traces are run in the same way on all systems, and to make sure no data is accidentally damaged or deleted on the target disk. The “dummy” file is created on the same or closest possible physical location of the hard disk on all systems, offering repeatable and comparable results. However, due to the nature of hard disks, a defragment run should be executed before running the HDD tests. The defragmentation of a hard disk can be done in Windows XP by selecting Disk Defragmenter from the System Tools (Start menu - All Programs - Accessories - System Tools)
The traces are created from real usage and the disk idle times have been compressed to 50 milliseconds to speed up the playback time. Compressing the idle times to 50 ms doesn’t affect the performance and the actual idle times have been reduced from the run times when reporting the result. The result is reported as the total transferred data (written + read) per second.
The traces for HDD tests include the following:
HDD test 1: The Windows XP start trace which contains HDD activities occurring at start of WindowsXP. This trace contains no user activity, all the activity is caused by operating system.
HDD test 2: The Application Loading trace contains HDD activities from loading applications. This trace is collected from opening and closing several applications, such as Adobe Acrobat Reader, Windows Media Player etc. This test, is somewhat similar to test 1, but differs from it by not including any user activity with the applications. All the activities collected are caused by the applications.
HDD test 3: The File Copying trace contais HDD activities from copying files. This trace simulates approx. 400MB of files copied.
HDD test 4: The General Usage trace contains HDD access from several common applications, such as Microsoft Word, Winzip, Internet Explorer etc.
![]()
![]()