What is Hardware Acceleration? What does it mean to enable hardware acceleration?

DXVA is a general term for hardware acceleration, not only data decoding (including H261, MPEG1, MPEG2, H263, MPEG4), but also Alpha Blending and Picture Resampling. However, the biggest impact on CPU utilization is decoding, especially HDTV decoding. For low-end systems, letting the GPU take care of the hardware decoding is the way to go.

However, the use of hardware decoding is also limited by many limitations in the whole function, especially the post-processing of ** data after decompression (such as deinterlace, sharpening, demosaicing).Therefore, for those who pursue image quality, they would rather give up hardware decoding, of course, at the cost of upgrading the entire system. More on that later.

How does the system work when hardware decoding takes effect? Now we have two processors, CPU and GPU. They exchange data via the PCI AGP PCIe bus.

Raw data is read out from the file system (directshow's source filter) and compressed data is separated (separator). Put it in system memory. GPU in a daze.

3.The CPU asks the GPU to start hardware decoding, and now the CPU enters a daze and the GPU starts to get busy. Of course, the CPU will regularly check how busy the GPU is.

5.**The data cannot be displayed immediately after it has been decoded, because post-processing is required, such as deinterlace, 3:2pulldown, etc.

The GPU then uses its own post-processing circuitry to process it. The CPU is still in a daze.

As you can see above, when the hardware decoding takes effect, the CPU can really be idle, but in order for it to take effect, some conditions must be met. The most critical reason is that compared with the CPU > system memory and GPU > graphics card video memory, the PCI, AGP, PCIE bus is a turtle speed. PCI is 133MB s, AGP8X is, PCIe16X is 4GB S, and CPU-system memory, GPU memory is several times faster.

If you are interested, you can compare the memory bus speed of different eras with the AGP PCIe bus. Because the bus on which the CPU and GPU interconnect is slow, once the data (whether before or after decompression) is handed over to the GPU, the GPU is responsible for processing it until it is finally displayed.

Of course, you may ask, maybe the CPU is slow, maybe the GPU does some of the work on the turtle bus, or is it faster than the whole CPU? Perhaps, but in the DirectShow architecture, the filters (splitters, decoders, post-processing, renderers) that represent all aspects of data processing need to negotiate an acceptable data storage method, if a filter says I can only play in the system memory, everyone has to play in the memory, and the GPU and video memory have to be abandoned.

It's hardware optimization, you don't need to know it in detail! For example, the graphics operation of the CPU is carried out on the GPU of the graphics card to realize parallel computing and improve the speed!

Hardware acceleration is the use of hardware modules to replace software algorithms to make full use of the inherent fast characteristics of hardware grinding and land search. The computer displays that it will be faster to use hardware acceleration, but when the projector or screenshot is found that the display area is black, then you need to turn off hardware acceleration (hardware acceleration is operated in the opposite direction).

1. Turn off graphics hardware acceleration:

1) Right-click on the desktop, screen resolution.

2) Click Advanced settings.

3) Click Troubleshooter, change the settings, and drag the pointer from full to none.

4), click OK.

2. Turn off the hardware acceleration of the sound card

3) Turn off all advanced sound effects under the Advanced tab.