Previously, we had a technical analysis on the APU in the first evaluation of the Ruilong 2200G/2400G. The entire CPU architecture is basically the same as that of the ZEN. The part of the moving tool is only the surrounding cache. For example, to insert the Vega GPU in , Slow down the three into 4MB, but in order to make up for the slow losses, AMD smartly increased the speed of the two slow cycles.
The Ruilong APU also uses updated power management and acceleration frequency algorithms to make the CPU stronger and more intelligent. In addition, the biggest difference between the new product and the previous generation of the same specification product is the use of a native quad-core die. Quadruple-size sharp dragons are obtained by shielding two cores from two CCX quad-cores respectively.
Ruilong AMD Ryzen architecture diagram
One generation Ruilong structure is two CCX designs, the middle is connected by the IF bus, the advantage is to EPYC/Ryzen ThreadRippe 1950x, down to R3 1200, only need to produce this one Die (core) components, greatly reducing the development of the film Cost, and each generation of Intel has to develop several die Die, this is why Ruilong can present to our consumers in such a cost-effective attitude.
However, there are problems with it. The CPU access between two CCXs of Ruilong can only pass through the IF bus. The delay is much higher than the delay between the four CPUs in CCX. This results in some software not adapting to the Ruilong CPU. The performance also caused the Ruilong CPU to depend a bit on memory performance.
Different from a generation of Ruilong, the Ruilong APU is a native quad-core Die. It is a different tape. Therefore, as the flagship of the APU, the R5 2400G can increase the frequency accordingly. To make Ryzen 7.
Then we use the 2400G as an example to see how the different architectures of the Ruilong APU and the generation of Ruilong will cause performance differences. What are the differences between them?
In addition, due to the high price of graphics cards today, many consumers have the idea of starting the APU to transition to the end of the tidal wave, so we will also add a game performance test after adding the graphics card alone.
Test product
Test product description:
The product comparison selected R5 1500X/1400/2400G. All three are quad-core and eight-thread design. Only R5 2400G is a native quad-core design with a nuclear display. The R5 1500X has the largest three-times of 16MB between the three, and the R5 1400 also has the 8MB of the third-level cache, but the R5 2400G is only 4MB..
The L3 cache is designed to read missed data after the L2 cache. In the CPU with L3 cache, only about 5% of the data needs to be called from memory, which further improves the CPU efficiency. The three-level cache size theoretically has a certain influence on CPU game performance.
In terms of selling price, R5 1500X and R5 2400G are priced the same, but R5 1400 is cheaper by 300 yuan.
Test Methods
In order to verify the difference between the performance of different CPU architectures, we will test all three CPUs locked to 3.8 GHz, so that we can see the performance difference between products at the same frequency.
In addition, taking into account that many consumers do not like overclocking or overclocking, we will use the CPU's default frequency to test the comparison once more, and see how the new power consumption and acceleration frequency algorithm introduced by the Ruilong APU will give the product much improvement. Also give a reference to consumers who prefer to use the CPU by default.
testing platform
Colorful Tomahawk C.AB350M-HD Magic Version V14
Test instruction:
The test is mainly divided into two parts, which are theoretical performance test and game test. The test basically covers all the CPU running sub-software, popular games. The test platform uses the mainstream of the Rainbow Tomahawk C.AB350M-HD magic sound version V14A motherboard with Zhiqi Flare X flame gun series DDR4 memory, stable platform performance.
In order to avoid bottlenecks in the game test, we use the NVIDIA GTX1080 public version of the graphics card. The games are tested at 1080p, full HD quality.
CPU theoretical performance test
CPU-Z is a well-known CPU detection software, is a software that detects the highest degree of CPU usage, comes with Benchmark can accurately determine the CPU's multi-core performance and single-core performance.
Test summary:
The CPU-Z performance test compares the number of CPU cores and the main frequency. The score increases linearly with the increase of the frequency. The R5 2400G eats the dividend with the default higher frequency under the default, the highest single and double core scores, and the same frequency. The difference between the three scores is not significant.
The CineBench R15 is a persuasive CPU and graphics test system that uses the Cinema 4D special effects software engine developed for the film and television industry to test purely CPU-based rendering of a high-precision 3D scene, supporting up to 256 logic cores. In addition, the new version also enhances the inspection of shaders, anti-aliasing, shadows, lights, and reflection blurring, and makes more accurate detection of CPU performance.
Test summary: Single-core performance has little difference under the same frequency. In the multi-core performance testing, the R5 2400G smart setting makes it really catch up to the default performance of the R5 1500X, but the R5 1500X with 16MB L3 cache has obvious advantages at the same frequency. .
X264 FHD Benchmark can effectively test CPU's floating point performance and the actual transcoding ability of multimedia audio and video.
Test summary: The default R5 2400G eats the frequency of the dividend, the same frequency under the three differences are not.
Fritz Chess Benchmark is chess software that can be used to test CPU performance.
Test summary: The difference in R4 2400G's native 4-core architecture has a large impact on chess scores. Even if the level 3 cache is half as low as R5 1400, it is also about 4% ahead of the same frequency. The default is even more of a race.
Winrar is commonly used compression / decompression software, comes with a benchmark performance test can test the CPU's compression capabilities.
Test summary: Both the native quad-core design and the junior grade ease have a relatively large impact on the compression performance of the file. At the same frequency, the R5 1500X and R5 2400G lead the R5 1400 by about 18% and 12%, respectively.
3Dmark is a futuremark company's software designed to measure graphics performance. Now it has released 3Dmark99, 3Dmark2001, 3Dmark2003, 3Dmark2005, 3Dmark2006, 3Dmark vantage, 3Dmark 11 and The new 3DMark. Now 3Dmark is more than a measurement card. Performance software, which has gradually transformed into a software to measure the performance of the whole machine.
3DMark's test physics score can be basically calculated CPU actual game performance.
Test summary: In the 3DMark test, the role of the third-level cache is pulled up, and the R5 2400G is at the bottom in both the default state and the intra-frequency comparison.
CPU benchmark performance test summary:
By default, the R5 2400G compensates well for the three slow losses by increasing the frequency and architecture differences. The overall performance is similar to that of the R5 1500X, which is significantly ahead of the R5 1400. But if everyone is at the same frequency, the R5 1500X is still the best performer, but the lead is not large.