Rabbit friends, the new Antutu review PC version has been officially released, and it has been launched on the official website of Antutu and major channels, welcome everyone to download and experience.
Similar to the Android version test process, the Antutu PC version test items also include CPU, GPU, Memroy (memory and storage) and UX (user experience), and the computer performance score is finally obtained by simulating a variety of complex scenarios. In addition, in addition to the traditional x86-64 architecture, the Antutu evaluation PC version also supports the new AI PC of the Arm64 architecture, and the evaluation system of the two versions is completely consistent, and the scores can be compared horizontally. When downloading and installing, please pay attention to choose the version that matches your computer's CPU architecture.
Users can download the software from Antutu's official website or other cooperative channels. (The size of the X86-64 version installation package is 832.8MB, and the size of the ARM64 version installation package is 780.2MB), run the installer, and follow the prompts.
Performance testing
1. After installation, click Test Now, and the software will perform a detailed performance test on your computer's CPU, GPU, MEM, and UX.
2. During the test, the page will display the corresponding test progress, and after each test project, there will be a green checkmark in the upper right corner, and the test progress will be displayed in real time in the center area of the page.
The GPU test part is based on UE4 Unreal Engine and Direct 11, and two 3D test scenarios were created: "Four Seasons" and "Coastline 2.0", the former is a high-voltage test scenario for high-end GPUs, and the latter is a stress test scenario for low-end GPUs.
In the MEM part, the results can show bandwidth and latency. As we all know, games need to read and write memory frequently, and if the memory latency is too high, the time to load the game will increase, resulting in the game screen lagging, so the bandwidth and latency displayed by the results are more conducive to the user's understanding of the specific situation of their device ROM.
The final UX part, unlike mobile, Secure, xml and other tests are refactored based on Windows APIs.
Once the test is complete, the overall performance score of the PC will be displayed at the top of the results page, and the corresponding scores for different hardware will be displayed at the bottom, so that users can compare different hardware models.
Stress test
In addition to hardware testing, Antutu also provides a stress test to test the stability, heat dissipation performance, and power consumption of the computer through long-term high-load calculations.
Other testing tools
The toolbox is currently open for screen dead pixel test, color bar test, and grayscale test. The GPU performance test, GPU ray tracing test, and GPUDLSS test are currently in internal testing and will be released for everyone to experience as soon as possible.
Antutu AI test
The software can test the AI processing energy efficiency of your computer's CPU and GPU, and will also add support for NPU in the future to accurately evaluate the AI capabilities of your computer, so that you can better plan your computer's AI use scenarios.
In the CPU test part, the computing power, data processing speed, and instruction execution efficiency of the CPU are evaluated by running specific AI tasks and algorithms, ranging from simple mathematical operations to complex logical reasoning and large language model Q&A, so as to more comprehensively show the performance of the CPU in different AI applications. Since CPUs are the core components of computers, they play a vital role in AI processing.
The GPU test is basically the same as the CPU test, which evaluates the GPU's computing power, processing speed, and instruction execution efficiency through specific AI tasks and algorithms. In some specific scenarios, GPUs can accelerate AI computing, which has great advantages for processing large-scale data and load models.
The specific test thresholds are as follows:
1. Must be on WIndows10 or above.
2. Reserve at least 10GB of free hard disk space.
3. The CPU supports at least AVX2 instruction set.
4. Sufficient memory space is required, for example, when using the Orion Star Model 14B, at least 16GB of RAM is required.
5. The correct GPU driver needs to be installed
6. You need to install a software environment that supports OpenCL or CUDA.
Here's how to use it:
1. Users can download the installer of Antutu AI large model evaluation software from Antutu's official website or other reliable channels.
2. Run the installer and follow the prompted steps.
3. After opening the software, the software will automatically match the appropriate large model resources, click Download and install.
4. Wait for the installation to complete.
5. After the large model is installed, users can choose to perform CPU test or GPU test according to their own needs.
6. Click the start test button, the software will automatically detect whether the device meets the test requirements, and if so, the test will be carried out.
If there is an anomaly during the test, it may be related to the available memory and available memory limitations. Before starting the test, we recommend that you close unnecessary background apps to free up more memory, or restart the device to clear up the temporary memory footprint.
In addition, the software checks the resource integrity of the large model before testing, and if the file is damaged or saved in the wrong location, it will also cause the test to be impossible. The solution is to exit and restart the program to download the large model resource package again.
When you complete the test with the Antutu AI large model evaluation software, the results page will display detailed information, such as the highest score in the tested item at the top of the page, which reflects the best performance of the computer in the CPU and GPU and in the test.
A score below 100,000 indicates that the device may struggle to efficiently handle complex AI tasks, while a score between 100,000 and 500,000 indicates that the device performs well and can handle most common AI tasks. A score of more than 500,000 indicates that your device has excellent performance in AI processing and is suitable for AI applications with high processing requirements.
At the same time, the results page will clearly show whether the large model used in this test is the Orion Star Large Model 14B or the Tongyi Qianwen Large Model 7B.
Of course, the results page also shows your computer's CPU and GPU models, so you can have a more specific reference when comparing the performance of different devices.
If an anomaly occurs during testing, the results page provides instructions to help you understand the problem and take action to resolve it. For example, the video memory is insufficient, the graphics card is not installed, and the drivers are incompatible.
It should be noted that the "speed: xx tokens/s" displayed on the result page is the average of Primpt inference and generation speed, not the inference speed or generation speed of a single large model. The higher this speed value, the more efficient your device is at handling AI tasks.
Since the PC version of Antutu AI large model evaluation is still in the testing stage, the algorithm will continue to be optimized in the future, so there may be certain problems in the process of using the experience, if there is a bug, you are welcome to comment and browse feedback, we will solve it as soon as possible.
(9041225)