Godson3A6000 processor evaluation catches up with Intel's 10th generation 4-core processor

Because of its strategic importance, the chip industry has been given priority attention. However, constrained by a series of measures taken by western countries in recent years, China's chip industry has been suppressed in various aspects, and authorization bans, supply interruptions and other situations are common, forming the commonly known "choke" phenomenon. Faced with increasingly severe environmental challenges, domestic chip enterprises have always been unable to break through the blockade of key core technologies and have to seek survival under the existing rule system. However, this strategy is only a stopgap measure.

As a key component of electronic hardware, chip production and design involve extremely high technical difficulties. In addition to lithography machine manufacturing, chip design is the key to the development of the industry. To truly break through existing constraints, independent chip architecture is the top priority. At present, desktop computers are dominated by X86 instruction set architecture, and mobile terminals are dominated by ARM. At present, the development of chip architecture of domestic enterprises mostly depends on foreign authorization, which makes it almost impossible for China's chip industry to respond effectively when encountering supply interruption and other problems. Therefore, to truly realize domestic independent chips, it is necessary to break through the limitations of the underlying instruction system.

Good news At present, domestic Godson Zhongke has made a zero breakthrough in chip architecture, and successfully launched the LoongArch architecture in 2020, marking that domestic chips officially have their own instruction set architecture system.

one Introduction to Godson 3A6000

Released in 2020 After LoongArch autonomous instruction set architecture, Godson Zhongke also bases on LoongArch architecture processor 3A 5000 series The processor was successfully taped, and after about three years of development, Godson3A6000 was also officially released this year.

The Godson 3A6000 processor still uses the LoongArch autonomous instruction set architecture, which uses the same 12nm process as the 3A5000 processor. However, through architecture optimization design, the fixed-point and floating-point scores have been significantly improved. The Godson3A6000 processor core has been upgraded from the LA464 of the Godson3A5000 to the LA664. The name 664 represents 6 transmit 64 bits. Each Godson3A6000 processor has four LA664 cores, which support 128 bit vector processing extension instructions (LSX) and 256 bit advanced vector processing extension instructions (LASX).

The previous generation Godson3A5000 processor has four cores, and the Godson3A6000 processor supports synchronous multithreading technology (SMT2) for the first time. The principle is known as hyper threading, which makes the core threads of Godson3A6000 processor reach four cores and eight threads. Like the previous generation 3A5000, the Godson3A6000 processor is still the main frequency of 2.5Ghz. The three-level cache is 16MB, and a dual channel DDR4-3200 memory controller is integrated internally.

This time, we also got a desktop computer equipped with Godson 3A6000. This time, the Godson 3A6000 is still soldered and packaged on the motherboard. The host provides dual 8GB 3200MHz memory, 256GB solid state disk and RX 580 graphics card, and the system is Unison UOS. Most of the current UniTrust UOS and LoongArch architectures have been adapted Office software Compared with that of Godson3A5000 Software The ecology has really grown a lot. As a daily office host, the software support is enough to meet the needs of most users.

Previously, according to the test conducted by Saixi Laboratory of China Electronics Technology Standardization Research Institute during the production of Godson3A6000, the test results show that the overall performance of Godson3A6000 processor is equivalent to the 10th generation Core quad core processor launched by Intel in 2020. So in this test, we will use Intel Core i3-10105 to make a comparison. At the same time, we will add the performance of the third generation of the AMD Ruilong 3 3100 processor. The basic parameters of the three processors are as follows.

In terms of parameters, In addition to the disadvantage in frequency, the other parameters of the Godson 3A6000 processor have indeed caught up with those of Intel and AMD entry-level processors three years ago. As for the performance difference, whether it can really match the performance of Intel Core i3-10105, a 10 generation 4-core 8-thread processor, let's take a look at the measured results, This time, we mainly compare the benchmark test and actual experience. The test platform uses the same version of UniTrust UOS system, and the hardware configuration is identical except for the motherboard and CPU.

two Benchmarking

Unixbench test:

First, let's take a look at the benchmark test. We will use three test software: SPEC CPU 2006, Stream, and Unixbench. First, use UnixBench to test. This software is a performance testing tool for Unix like (Unix, BSD, Linux) systems, and is widely used to test the performance of Linux system hosts. It can test the performance of system call, reading and writing, process, graphical test, etc. It is also a software to test the whole machine in all aspects.

UnixBench's test results show that Godson3A6000 processor has a multithreading score of 8032.1 and a single thread score of 2021.3. Intel Core i3-10105 has a single core score of 2423.1 and a multi-core score of 2423.1. AMD Reelong 3 3100 single core score 2938.2, multi-core score 10169. As you can see from the test results, Godson 3A6000 processor is better than Intel Core i3-10105 in the multi-core performance of Unixbench The performance is better, but there is still a certain gap for single core.

SPEC CPU 2006：

Next, we will conduct a comparative test of SPEC 2006. SPEC 2006 is a large CPU performance test project, focusing on testing the system's processor, memory subsystem and compiler. Be able to test the most basic fixed-point performance int and floating point performance fp of CPU.

According to the test results, the Godson3A6000 processor scored 43 in single threaded fixed-point performance int, 54.7 in single threaded floating-point performance fp, 39 in single threaded fixed-point performance int and 46.7 in single threaded floating-point performance fp in SPEC2006. AMD Reelong 3 3100 single core scored 41.1, and single thread floating point performance fp scored 57.3. In this test, the Godson 3A6000 processor outperformed the Intel Core i3-10105 in both fixed-point performance int and floating point performance fp, and the single threaded fixed-point performance int even surpassed the AMD Reelong 3 3100 by a little.

In the multithreading test, the Godson3A6000 processor scored 139 in the multithreaded fixed-point performance int, 153 in the multithreaded floating-point performance fp, 95.8 in the multithreaded fixed-point performance int, and 101 in the multithreaded floating-point performance fp of the Intel Core i3-10105. The Ruilong 3-3100 multithreading score was 133, and the multithreading floating-point performance fp score was 154. The fixed-point performance int of Godson3A6000 exceeded that of Intel Core i3-10105 and AMD Reelong 3 3100 in the multithreading test, and the floating point performance fp was only 1 less than that of AMD Reelong 3 3100.

Stream：

Stream is the mainstream memory bandwidth test program in the industry, and the test behavior is relatively simple and controllable. This program requires very little CPU computing power and has a great pressure on CPU memory bandwidth. As the number of processor cores increases, the memory bandwidth does not increase linearly. Therefore, memory bandwidth is more and more important for improving the processing power of multi-core processors.

We tested the performance of single thread, 4-thread and 8-thread in Stream respectively. Among them, Godson 3A6000 outperformed two Intel and AMD processors in this software. In particular, the single thread performance is the best except for the Scale mode, which shows that this generation of Godson 3A6000 is better at single thread optimization. However, the AMD Ruilong 3 3100, which had outstanding performance in the first two tests, encountered a bottleneck in the Stream test, which was more consistent with the performance of the Ruilong processor at that time, and was always poor in memory.

In terms of benchmark test performance, Godson 3A6000 is indeed more advantageous than Intel Core i3-10105 on the whole, and even SPEC 2006 and Stream are more advantageous than AMD Reelong 3 3100. These software can make domestic processors perform better in Tongxin UOS system, which is also the importance of ecosystem, Only when the ecology is in place can the processor play a stronger role.

three Application test

In addition to the above three benchmark software, we also measured the performance of Godson3A6000 on common office software.

WPS file:

Then we first use WPS to open 50MB and 200MB documents respectively, and record the time when the document was opened. To ensure more accurate test data, we will open each document three times on average.

The actual test shows that, The Godson 3A6000 takes the shortest time to open a 50MB document, only 0.8s, while a 200MB document takes the longest, 2.12s, but it is also very close to the Intel Core i3-10105. On the whole, the Godson 3A6000 has greater advantages in processing small capacity documents.

WPS PPT：

Next, we will use PPT to test. Open 50MB PPT respectively. One PPT contains pictures+videos, and the other contains text+pictures. Similarly, each PPT will open the average value three times.

It takes 1.68s for Godson 3A6000 to open the 50MB text+picture PPT and 1.8s for the 50MB picture+video PPT. AMD Ruilong 3 3100 takes 1.54s to open 50MB text+picture PPT, and 1.56s to open 50MB picture+video PPT. Intel Core i3-10105 takes 1.14s to open the 50MB text+picture PPT and 1.32s to open the 50MB picture+video PPT. In the PPT open test, Godson 3A6000 is the most time-consuming of the three processors.

WPS Excel ：

Let's test it Excel Table The average value of each table will be opened 3 times.

In the WPS Excel opening test, it took 2.61s for Godson 3A6000 to open a 50MB plain text table, 3.15s for AMD Ruilong 3 3100 to open a 50MB plain text table, and 2.35s for AMD Ruilong 3 3100 to open a 50MB plain text table. Intel Core i3-10105 takes 2.74s to open a 50MB plain text table, and 3.21s to open a 50MB text+picture table. In this test, although the Godson 3A6000 is not as fast as the AMD Reelong 3 3100, it takes less time than the Intel Core i3-10105.

Video on:

Finally, let's test the opening speed of video playback. We choose the default mode under UniTrust's UOS operating system Video player And open 4K and 1080P mp4 videos of the same size, record the opening time of the three platforms respectively, and The average value will be opened 3 times 。

According to the test results, it takes 1.23s for Godson 3A6000 processor to open 4K video and 1.19s for 1080 video. It takes 1.42s to open 4K video and 1.27s to open 1080 video for AMD Reelong 3 3100. Intel Core i3-10105 takes 1.67s to open 4K video and 1.37s to open 1080 video. Among the three processors, the Godson 3A6000 processor takes the shortest time.

four Game experience

Surprisingly, the Dragon Core 3A6000 can play games, such as《 Dead cell 》、《 Torchlight 》As well as the old version of Tomb Raider, and even Loongson has adapted World of Warcraft nostalgic clothes, but unfortunately Blizzard has launched the mainland market, so it is impossible to experience.

In terms of games, among the adapted games, the Dragon Core 3A6000 has been able to play relatively smoothly. For example, the whole experience of Dead Cell is smooth, and there are no adverse reactions such as flash back and crash.

The frame number of Torch Light is expected to be about 40 frames, which is lower than that of Dead Cell, but it can also run smoothly. However, during the test, it was also found that the eight cores of Torch Light are not full, and the utilization rate is not high. In terms of multithreading, further optimization is needed, but it has been played, and I believe that this day will not be too far away.

five New breakthroughs, new challenges

The release of Godson 3A6000 is undoubtedly a milestone. Through the LoongArch autonomous instruction set architecture and SMT synchronous multithreading and multi-faceted optimization, this processor has made a series of breakthroughs. The measured results can also prove that, whether in benchmark testing or office experience, Godson 3A6000 Tongxin's performance on the UOS system has been comparable to, or even partially surpassed, Intel Core i3-10105, which has raised the performance of domestic self-developed chips to a higher level. What's important is that this is an achievement achieved on the completely domestic instruction set architecture that is autonomous and controllable. Of course, there are still some problems with Godson 3A6000. For example, it is found during testing that multithreading still needs to be optimized, core utilization is not high, and there is still much room for progress in software compatibility experience. These are the directions that need to be worked on in the future.

at present LoongArch And domestic system Market acceptance also needs time and resources to promote. In today's increasingly complex global industry chain, a complete semiconductor ecosystem can not be established by a single enterprise or single product. It requires the improvement of supply chain, the enrichment of software ecology, and the expansion of application scenarios. This must be a long and arduous process. It cannot be achieved overnight. It needs stable capital investment, continuous talent reserve and advanced research and development.

Moreover, in the context of rapid technological development, competition is also accelerating. The global chip giants are constantly innovating, adopting more advanced production processes and design methods, such as 7nm, 5nm and even more advanced. In order not to be left behind quickly, domestic manufacturers must accelerate the pace of research and development, invest more resources, and expand technical cooperation and talent introduction worldwide.

The semiconductor industry is a highly integrated and complex field, and every small progress may bring a huge breakthrough. We have reason to believe that, with the joint efforts of all parties at home and abroad, domestic chips will not only be a "substitute", but also a "leading role" with competitiveness and influence in the world. But it all takes time, money and patience. LoongArch's autonomous instruction set architecture has built a bridge for fully controllable and autonomous domestic chips. Godson 3A6000 has taken a new step in the commercialization of domestic chips, and waiting for the full rise of the domestic chip industry will be a long and expectable process.

This article is an original article. If it is reproduced, please indicate the source: Godson3A6000 processor evaluation catches up with Intel's 10th generation 4-core processor https://diy.zol.com.cn/838/8382163.html