Here is the Graphic Card Benchmark Performance Rating 2019
1, TDP difference
The two laptop graphics cards are the same model. If the TDP is different, the performance will be different. For example, the RTX2060 mobile version of the TDP is between 80W and 90W, and the performance gap is about 5%. Another example is the MX150 graphics card, which is also a low-power version. The frequency difference is also large. It is impossible to subdivide to sort all the frequencies and then sort them. This will cause the graph to be too redundant and complicated, and on the other hand, it has no meaning. .
2, version difference
The same GPU, the public and non-public version of the performance is different, there are differences between the non-public version. Separate labels with large differences. For example, GT1030, D5 and D4 versions, D5 memory frequency is usually about 6000MHz, D4 memory frequency is only 2000MHz, and the number of texture units in the GPU has also been reduced, from 32 to 24, which results in its performance only GT1030 D5 About half of the original GT1030 D5 performance has been very bad, this D4 version is even worse.
3, the performance value of the nuclear display has been converted, but not enough to reflect the actual game situation.
The speed-speed space graphics ladder diagram is more based on running statistics, but in the actual game, the nuclear display with a small gap between the running points and NVidia alone will be significantly faster.
For example: MX230 alone, in theory, only 15% higher than Iris plus G7, Vega10, but in some games, 30 to 100% faster. DOTA2, MX230 can run 60 frames in 1080P+ high quality, and only 30 frames in nuclear display. There are three main reasons:
(1), alone does not account for the memory.
(2), the nuclear display will also be limited by the CPU itself TDP, especially the mobile version.
(3), the actual game is better for N card optimization.
4, comprehensive trade-offs
The sorting of the speed space graphics card ladder diagram is a comprehensive trade-off, considering the running points, the actual game, and even considering the actual situation of the market. Such as GTX1660, the opponent is RX590, many projects have won each other. The RX560XT parameter is the same as the RX470D, but it will be slightly higher. The reason is not how much optimization the kernel has made, but only one that can be bought on the market. Its core frequency and memory frequency are higher than the average RX470D. .
Similarly, the RX580 (2048SP) and the RX570 are essentially the same card, but the former is generally more frequent, so the position in the figure is also slightly higher, which is also considered the actual situation of the market.
5, on the mobile version of the naming
The digital model on the previous notebook graphics card has an “M” behind it, which means a mobile version, such as NV’s GT830M, GTX950M, etc., AMD’s R9 M290X, HD8750M, etc., but NV starts with the 10th generation N graphics card. On the mobile version of the model, this M was canceled, which led to “chaos”. For example, the GTX1060 6G, the desktop graphics card and the notebook graphics card are exactly the same, but the parameters are different, why is this happening? ?
The reason is that the previous graphics cards have higher energy consumption due to the limitations of the process and micro-architecture design, and cannot meet the cooling standards of notebooks.
For example: GTX960, stream processor is 1024, and GTX960M stream processor is only 640, the frequency is also lower, because the notebook’s heat dissipation is far less than the desktop, this is also helpless, only shrinking can reduce power consumption, control temperature. Because the specification shrinks too much, many people know the phrase “multiple M performance cuts half”.
By the tenth generation of N-cards, the energy consumption ratio has been greatly improved. The specifications of notebook graphics cards are basically the same as those of desktop computers. The main difference is that some frequencies are reduced to control power consumption, and performance is not shrinking. Therefore, NV cancels this. M, so as not to misunderstand as “multiple M performance cut half.” However, the performance of the mobile version and the desktop version are different after all, and some differences are still very large. Therefore, in order to avoid misleading the readers of this article, I have noted the word “mobile” in the figure to distinguish it from the desktop graphics card.
6, on the N card algebra division
Strictly speaking, the nine generations and ten generations in the picture are not accurate. Since the release of the first graphics card in 1995, Nvidia has more than 10 generations of products. The better argument is: This is the first generation of N-cards since the adoption of the “unified architecture”. For example, GTX1080 can be regarded as the tenth generation N-card with a unified architecture. However, the picture cannot be written too detailed due to the limited area. Too much text will look cumbersome.
The 11th generation N card (starting with RTX20) uses the Turing architecture, the 10th generation (GTX10) uses the Pascal architecture, and Turing has more than two techniques: ray tracing (RT) and deep learning supersampling (DLSS). ). There are currently very few games that support RT, but there are already some support for DLSS, such as “Jesus Survival”, “Ark: Survival Evolution”, “Final Fantasy 15”, “Killer 2”, Tomb Raider: Shadow. At present, many games run smoothly at 1920X1080 resolution, and the performance of one to 2K resolution drops by a factor of one. When it is 4K, it becomes a cartoon, and DLSS can make the game run smoothly and quickly at high resolution.