3DMark graphic explanation-Video Card Testing Software

Source: Internet
Author: User

To measure the performance between different video cards, a unified test standard must be set. This standard is software released by a third party. In the OpenGL field, the dominant positions of Quake3 and Doom3 are hard to shake, while in the DX field, there are various competitions. The 3D Mark series are naturally an indispensable force. With the release of 3D Mark 05, this test package has once again attracted worldwide attention.

3D Mark history review

● 3D Mark99 Max
3D Mark99 Max is Madonion.com's first video card testing software, but its influence is quite limited. On the one hand, the video card was used only as the assistant for CPU processing, and its impact on the overall system performance was not as great as it is now, in addition, the testing overlord was the 3D winmtr and Quake series. 3D Mark99 Max mainly investigates fixed function vertex conversion and illumination capabilities and the multi-material Filling Capability of graphics cards.

● 3D Mark2000

The advent of 3D Mark2000 made users feel refreshed. Before that, 3D graphics card testing was a laborious task, and the available software only stayed in some old-fashioned software: for example, the 3D winmtr series or Quake2. 3D Mark2000 is in the DX7 era. Compared with DX6, The DX7 video card integrates"HardwareThe light and mobile engine greatly improves the image quality and reduces the CPU burden. From DX7, the graphics card has gradually gained independent computing power and began to develop towards the next day's GPU.

3D Mark2000 has two games, each of which is low, medium, and high.ResolutionThere are also two CPU tests. 3D Mark2000 brings unprecedented visual impact and convenience to the video card evaluation. With this DX7 test software, Madonion.com is well known.

● 3D Mark2001/SE

3D Mark2001/SE is the most typical test software in the history of Madonion.com. Why? So far, Futuremark still provides the Online Result Browser Service (ORB) for 3D Mark2001/SE scores ). At the same time, the scores of 3D Mark2001/SE are still being refreshed.

3D Mark2001/SE 3D Mark scores are obtained from the scores of four games in seven scenarios. The first three games are based on DX7 in six scenarios, and the last Mother Nature scenario is a complete DX8 scenario. Hardware-based Pixel Shader and Vertex Shader. Only DX8-compliant graphics cards can complete all tests.

Madonion.com was originally designed to create a video card testing tool for 3D Mark2001/SE. However, it may be unintentional. In actual tests, the CPU has a significant impact on the final score. As a result, the CPU and video card jointly affect the 3D Mark2001SE score. This is related to the low starting point of 3D Mark2001/SE: Most scenarios are based on DX7, and the cost of DX8 or even DX9 graphics cards is very low. The real speed bottleneck lies in the modeling speed of the CPU. 3D Mark2001/SE more reflects the overall performance of a machine's CPU, video card, memory and other systems, not just the video card.

● 3D Mark 03-great fortune

The success of 3D Mark2001/SE makes Madonion.com and 3D Mark very famous. 3D Mark is almost the only Mark for classification of graphics card grades. Before the advent of 3D Mark03, Madonion.com was renamed as FutureMark, and the Commercial transformation was completed. With the infinite expectation of the industry, 3D Mark 03 is available. But what we did not expect was that 3D Mark 03 suffered a lot of criticism at the beginning of its release, and the strongest criticism came from NVIDIA: in practical applications, GeForce 4 Ti 4600 and ATi RADEON 9700Pro differ significantly in 3D Mark03. In addition, the confusing preference of FutureMark makes it possible to doubt the intricate interests between FutureMark and ATi and NVIDIA.

Another problem plaguing 3D Mark 03 is the vendor's hands and feet on the driver. NVIDIA and ATi are driving a greater or less optimized 3D Mark03 to improve the video card score. This leads to the de-authority of 3D Mark 03. 3D Mark 03 gradually lost its aura and was left cold by the media of the evaluation agency.

The 3D Mark 03 score is still the result of the game scenario. 3D Mark 03 is a software completely written by DX9 API, without using the MAX-PAYNE engine used before. The first scenario is the DX7 scenario, and the third scenario is the DX8.1 game, using the SM1.1/1.3/1.4 technology. The fourth scenario is Mother Nature 2, which is the DX9 scenario of CG image quality edition and uses SM2.0. Most DX9 video cards cannot run smoothly.

3D Mark 05-coming on schedule

● System requirements for 3D Mark 05
What kind of configuration can we run 3D Mark 05?
This is the lowest configuration recommended by FutureMark:
★X86 processors at least 2 GHz
★MB system memory
★MB hard disk space
★Meets DX9 graphics card, at least SM2.0 or higher
★Windows 2000 or Windows XPOperating System
★DX 9.0c or higher

3D Mark 05 has high requirements on the system, especially the video card. Because it rejects all DX8-level video cards. According to the current information, the starting video card that NVIDIA can run is GeForce FX5200, And the ATi is RADEON 9600SE (RADEON 9500 can also run, however, the ATi naming method does not reflect the gap between R9500 and R9600SE .)

● 3D Mark 05 main interface:

The main interface is similar to 3D Mark 03. If it is a registered version, you can select a specific test item. In "Setting", you can select resolution, anti-aliasing, filtering mode, and SM version.


Select Test Project

3D Mark 05 consists of four parts: Game test, CPU Test, 3D Feature Test, and Batch Size Test. The Game Test score is processed to generate a 3D Mark.
3D Mark 05 score = (Game Test 1 × Game Test 2 × Game Test 3) ^ 0.33 × 250
The weights of the three scenes are the same in 3D Mark 05, but they do not account for a large proportion of the last three scenes as 3D Mark 03 does.

● Shader Mode2
All using Shader Mode2 is the biggest feature of 3D Mark 05. This means that only DX9 graphics cards can run 3D Mark 05, and the old king Ti series will be completely rejected by 3D Mark 05.

In Direct X, there are two types of Shader: Vertex Shader and Pixel Shader. After DX9 is started, the Shader version has been upgraded to 2.0 (known as SM2.0 ). SM2.0 greatly extends the Shader's use range and flexibility. SM2.0 can execute longer and more complex commands. SM2.0 is added with the ability to use floating point values for computation. This is very important for improving image quality: in real-time computing and image-based light computing, floating-point data occupies a dominant position. SM no longer relies on approximate calculation. SM2.0 introduces a more advanced Shader Language and is compatible with the original Shader Language as much as possible.

For different GPU architectures, DX9 supports different Shader versions. The latest DX9.0c supports SM2.0, 2.0a, 2.0b, and 3.0. Currently, only GeForce 6800 and 6600 support SM3.0. 3D Mark 05 is widely used in SM2.0 and can determine the mode used for rendering based on the maximum version that the hardware can run. Of course, you can also manually select the Shader version.

● 3D Engine
Before 3D Mark 03, Madonion.com had been using the MAX-FX engine for Demo production. This has changed since 3D Mark 03. Due to the adoption of Shader technology, 3D Mark 03 directly submits most of the work to the API and graphics card, and the engine's job is to load the Shader Language into the Shader. 3D Mark 05 further reinforces this approach, so the 3D Mark 05 engine makes the system load heavier than 3D Mark 03.

● Game Testing
Game testing has always been the first place in 3D Mark testing. The images of every generation of 3D Mark games are being improved. From this point, we can see the development of 3D technology.

Game1: Return to Proxycon

Game scenario test

Return to Proxycon is a continuation of Battle of Proxycon, the second game in 3D Mark 03. The background of the story is still in space. A ship full of valuables encounters a Space pirate, and a fierce battle is inevitable. Return to Proxycon in the fierce battle scene is no less than Doom3. Return to Proxycon shows an unprecedented screen. This scenario is only part of the game. In the Demo of 3D Mark 05, you can see the complete process. In this scenario, eight light sources are used, including two pointing light sources and two static light sources, with a texture size of 2048x2048.


Test Results

In a game similar to Doom3 in this scenario, the scores are arranged similarly to doom3. NVIDIA graphics cards have obvious advantages, and even a 6800GT can beat X800XT. However, due to the inherent disadvantage of the pipeline, the FX series were defeated in the previous generation of graphics cards.

Game2: Firefly Forest

Game scenario test

On the night of the full moon, the moonlight is pouring in the forest, and the fireflies freely travel between them, as if they were the owners of the forest. This is the Firefly Forest scenario.

In the scenario, the processing of the ground material is similar to that of Game1, but the diffuse reflection effect is increased. The rock is made up of metallic glossy textures. The scene uses a geometric model with high details, as well as atmospheric scattering and atomization effects. In this scenario, moonlight points to the light source, and the texture size of the dynamic shadow is also 2048 × 2048. Firefly uses a texture of 512x512.


Test Results

The performance of the 6800 series and the X800 series is similar. 6800 U beat X800XT, but the X800 Pro with 12 pipelines beat the 6800 Standard Edition. 5950Ultra is less than 9600Pro's performance, reflecting NV35's architectural design mistakes.

Game3: Canyon Flight

Game scenario test

The third scenario describes the fierce battle between a ship flying in the canyon and a sea monster guarding the canyon. The complete story can be seen in the Demo of 3D Mark 05. The wide range of outdoor scenes are very complex because of the need to make the Effect of Water Reflection Canyon. The atomization effect is greatly applied to the treatment of water and the sky. In this way, you can see the sea monsters in the water.

The rock processing in this scenario is one of the most important parts of the system in 3D Mark 05. PS2.0 must always participate until dynamic shadow rendering is completed, and the surface effect must be rendered six times to achieve correct reflection and scattering effects.

Because the light source is sunny, only one sun light source is needed. This scenario poses a huge challenge to the dynamic shadow implementation capability of the video card: Because the scene span is too large, according to Futuremark, even two 2048x2048 texture operations are still not enough.

In the third scenario, only one light source is used, so the speed is also the fastest in the three scenarios. The third scenario is the only winner of the competition between ATi and NVIDIA's flagship graphics card. X800Pro is still far higher than 6800 Standard Edition.


Test Results

CPU Test

CPU Test has been introduced since 3D Mark2000 and is a "traditional Test project ". The CPU test still follows the game in the previous scenarios. It only reduces the resolution and submits all Vertex Shader to the CPU. In this way, we can minimize the workload of graphics cards and increase the CPU load to the maximum.

CPU score = (CPU Test 1 × CPU Test 2) ^ 0.5 × 1500
In 3D Mark 05, CPU Test uses Game3 and Game1 (in sequence ). In the Game3 scenario test, the CPU not only undertakes Vertex Shader's work, but also computes the track of the spacecraft flying in the canyon. In order to ensure that the computing workload is the same in different systems, the flight routes of the spacecraft are the same.

CPU Test1 uses the D Lite algorithm, which is a more advanced algorithm, mainly considering possible virtual or physical multi-core processors in the future, the CPU can be processed by a maximum of two threads at the same time.

CPU Test2 uses a 2-frame/second Fixed rendering mode, but does not include AI processing.

● 3D Features Test

Although these scores are not included in the final scores of 3D Mark, they can reflect the processing speed of the video card itself, because they are all purely theoretical speed tests, it is rare in other testing software. We will focus on Pixel Shader and Vertex Shader tests.

Pixel Shader

The most complex material Test is the Game3 Test, so it is listed separately for Pixel Shader testing. The test reflects the sunlight in the rock. The filling speed of such a scenario is very fast, and it is also very dependent on the memory bandwidth.

Vertex Shader

The Vertex Shader test of 3D Mark05 is divided into two modes: simple and complex. In simple mode, only one light source shines on six sea monster models, and each sea monster model has about 1 million vertices, therefore, the workload is quite large. 3D Mark05 has fully switched to Shader Model2, but even Vertex Shader1 can complete this test well.


Material Test

In the complex mode, Vertex Shader tests a large area of grass. The surface material and bending degree of each grass are independent, which is very similar to the actual grass. The swing of the grass is calculated by the CPU, but the software optimizes it to minimize the impact caused by CPU differences, more reflects the Vertex Shader computing power of the video card. In the picture, the grass leavesLensFar away, which makes it uncomfortable to enjoy grass at close distance, but this can avoid the influence of filling rate on performance.


Material Test

● Block test
Batch Test)
Block testing is a newly added test item in 3D Mark 05. It is used to perform rendering tests on a very simple scenario that has not been optimized. It is designed to address a weakness in the current drive. Over the past few years, game developers have always been rendering "blocks (Batch)" as much as possible. However, in practical applications, small block rendering not only achieves the same effect, it can also achieve fast speed. This test was proposed as early as 3D Mark2001, and by 2005, more developers proposed to join this test. There are 6 models in this test, each of which consists of a 128-layer 128x182 square, including 8, 32, 512, 2048, and 32768 triangles.


Block Test

Today, graphics card drivers generally optimize the tests of the last two "blocks". The smaller the block size (the smaller the number of triangles contained), the slower the rendering speed. Therefore, this test is a test of the drive rendering capability to some extent.


Test Results

Online Result Browser
The Online Result Browser (ORB) service is the most distinctive service provided by FutureMark. After you upload your test results to the FutureMark server, the system automatically compares the scores uploaded by others. However, to log on to the ORB system, you must have a 3D Mark score. ORB rankings are currently the most competitive overclocking rankings, because the ORB service is open from 3D Mark2001 to 3D Mark 05.


Overall score

Related Article

Contact Us

The content source of this page is from Internet, which doesn't represent Alibaba Cloud's opinion; products and services mentioned on that page don't have any relationship with Alibaba Cloud. If the content of the page makes you feel confusing, please write us an email, we will handle the problem within 5 days after receiving your email.

If you find any instances of plagiarism from the community, please send an email to: info-contact@alibabacloud.com and provide relevant evidence. A staff member will contact you within 5 working days.

A Free Trial That Lets You Build Big!

Start building with 50+ products and up to 12 months usage for Elastic Compute Service

  • Sales Support

    1 on 1 presale consultation

  • After-Sales Support

    24/7 Technical Support 6 Free Tickets per Quarter Faster Response

  • Alibaba Cloud offers highly flexible support services tailored to meet your exact needs.