Win10 system DX12 API performance new and old graphics card comparison measured efficiency amazing

Microsoft's Windows 10 new system has been officially released. For an operating system, the degree of hardware performance is a common concern for users. What is the experience of Windows 10 for ordinary notebook users? Can Windows 10 let the hardware in the hands of enthusiasts take it to the next level? Can the legendary DX12 really improve graphics performance?

I believe that many of my friends have upgraded their computers to Win10. The interactive interface and functions are not discussed. Win10 integrated DX12 is quite attractive for gamers. Because according to Microsoft's demonstration, even if you don't talk about new special effects, the efficiency of execution alone is several times higher than that of DX11. However, Microsoft's official demo is just a word, what is the actual situation? Not long ago, FutureMark added a Driver Overhead test to the new version of 3DMark, which is specifically designed to test the API performance of DX12, which also allows us to have an intuitive concept of DX12 performance improvement.

Windows10 has been released for a long time, but the DX12 test program has been so far, except for FutureMark's new "API Overhead Function" test and UE4 engine-based DX12 demo DEMO. We may not be able to find other DX12 programs available. In addition, "API overhead function" and UE4 DX12 demo DEMO is mainly aimed at the improvement of DX12 efficiency, but not the increase of screen effects, so just want to meet the eyeball friends, and then wait. Of course, before that, let's first experience the super high execution efficiency brought by DX12.

To fully support DX12, these conditions must be available

▲Not all versions of Win10 are integrated with DX12, it is best to run Dxdiag to confirm

▲Driver model WDDM2.0 can support DX12, here you can also see that 9 series N card can support DX12.1

compatible and perfect support DX12 is different two concepts, simply put, now mainstream graphics can be Compatible with DX12, but to fully support all DX12 effects, it must be GCN 1.1 and above or the GeForce 900 series N card. Of course, even if it is only compatible with DX12, you can enjoy the efficiency improvement brought by DX12.

Does that as long as the video card is OK? of course not. In addition to the graphics card support, the driver, operating system and the program itself are required. Drivers, NVIDIA, AMD and Intel have updated the graphics driver, you just need to use the latest version of OK, I really do not want to update, check if the driver is WDDM 2.0 (run Dxdiag can be seen under the display page) Needless to say, the system can only use Win10 (as far as whether the DX12 is stripped out and used on Win7 is completely ok, it is still unknown); in terms of program, 3DMark should be updated to 1.5.915 to be normal in the official version of Win10. Test the project using the API overhead feature.

“API overhead function” What is the measurement?

In fact, the official description of FutureMark is very clear. The API overhead function test item is not suitable for the comparison performance between different graphics cards. It mainly tests the efficiency and efficiency of the system in the GPU drawing operation under different APIs. The higher the API, the more drawing operations can be invoked while consuming the same system resources.

The whole test principle is also very simple, that is, the drawing operation is always increased until the frame update frame rate drops to 30fps, and how many drawing operations are called at this time, the more the API is executed, the more efficient the execution is. high. This test includes DX11 single-threaded, DX11 multi-threaded, Mantle (only supports A card) and DX12 a total of 4 sub-tests, and the difference in execution efficiency of various APIs is clear at a glance.

Measured! DX12 is highly intriguing

▼Test Platform

Processor: Intel Core i7 4770K

Board: ASUS B85M-G PLUS

Memory: Kingston DDR3 1600 4GB× 2

Graphics: GeForce GTX970; GeForce GTX760; Core Graphics

Hard Disk: Kingston HyperX SAVAGE SSD 240GB

Monitor: DELL U2410

Power: Huntkey MVP600

1, high-end discrete graphics card measured

For high-end graphics cards such as GTX970, the efficiency improvement brought by DX12 is amazing

but multi-threaded DX11, the score has doubled compared to single-threaded DX11, but after using DX12, this gap is a joke — — DX12 draws more than 1.6 million per second drawing operation, than multi-threaded DX11 mode It’s 6 times faster! Of course, the N card does not support Mantle. In fact, you don't have to care. After all, DX12 is a favorite of all mainstream graphics cards. With DX12, who still cares about AMD, AMD is not ready to continue playing.

In addition, from the curve on the right side of the figure, you can see an interesting phenomenon. When using DX12, the CPU temperature is the lowest in the middle, and then slowly rises. This can actually prove that the DX12's pressure on the CPU is lower when the load has reached the limit of the DX11 environment. Later, because the frame rate of the DX12 frame is reduced to 30fps, the CPU temperature becomes higher. —— This also proves that DX12 takes up less system resources under the same load.

2, the mainstream of the previous generation alone measured

GTX760 as the previous generation of graphics cards, even if you can not fully support DX12 effects, you can enjoy the high efficiency of DX12

We know that the 9 series N card can perfectly support DX12.1, and the 7 series old graphics card can only be compatible with DX12. However, this does not matter, from the test results, the GTX760 can also enjoy the high efficiency of DX12, and the increase is even higher than the 9-series graphics card, reaching 11 times! Of course, the GTX760's multi-threaded DX11 mode is not much faster than the single-threaded DX11 mode itself.

From the curve, the performance of GTX760 is somewhat different from that of GTX970. It is that the GPU temperature is lower in the middle of the test, and the CPU temperature is relatively stable. It seems that the GPU architecture has obvious influence on the efficiency of DX12.

3. Entry-level core graphics card

The core graphics card also has more than twice the efficiency improvement

Although compared with the mainstream, the core graphics card The performance is much weaker, but it also benefits from DX12 —— even it does not perfectly support all DX12 effects. In DX12 mode, the core graphics card has also achieved more than double the efficiency improvement, and the score is higher than the GTX970's single-threaded DX11 mode, which is indeed very impressive.

The curve is also well interpreted. After all, the core graphics card is in the CPU, so the temperature curves of the two are almost coincident, and the temperature is also showing a steady upward trend, and the overall performance is relatively stable.

Summary:

From the test point of view, DX12's improvement in execution efficiency is indeed very obvious, which may be attributed to its further full utilization of multi-core CPUs and GPUs. However, it should be noted that the API overhead function test cannot be used to simply measure the performance of the graphics card. After all, this is only a call to check the drawing operation, and does not involve various image effects, materials, textures, and parts occupied by the memory. The factors have a greater impact on the speed of DX12 games.

So, in future DX12 games, you may find that their hardware requirements are still very high. It is still not easy for mainstream graphics cards to run in high-quality mode, but you will also notice that In the case of DX11, the DX12 game will have a significant increase in the complexity of the scene details, which is also the value of upgrading DX12.

Special experience: DX12 DEMO using UE4 engine

The DX12DEMO made by CryZEN based on UE4 engine is officially launched. We can also experience the effect of DX12 game in advance.

With GTX970, 1920× 1080P resolution, DEMO will run at 60fps, which shows that it has locked the maximum frame rate, but when running in DX11 mode, the frame rate will drop. Up to 40fps. When using the GTX760, you can also run DEMO (it is not mandatory to support DX12 full-effect support), the speed is in the range of 50-60fps, it is also smooth, and occasionally there are tiny tensons. Of course, here we also tried the core graphics card, but it can also run, but in addition to the screen effects are almost nothing, the running speed is also calculated by 0. Xfps, this speed has no meaning.