What is a stream processing unit?

The Stream processor (Sp,stream Processor) unit is the name of a common scalar shader within the unified architecture GPU.

The Flow processing unit is a new universal rendering unit, which combines the previous vertex shader with the pixel shader, and is the next generation of graphics rendering specification after pixel pipelines (pixel pipeline) and vertex pipelines (vertex pipeline). The flow processing unit can not only complete vs (Vertex Shader, vertex Shader) operation, also can complete PS (Pixel Shader, pixel shader) operation, and can be composed of arbitrary vs/ps proportional to the needs, so as to give developers a broader space to play. In short, the rendering pattern of the rendered pipeline/vertex unit, which was previously made up of a fixed proportion, is now replaced by any proportional rendering of the pipeline/vertex unit, which consists of a stream processing unit.

The Flow-processing unit is part of the rendering pipeline, and a complete rendering pipeline consists of a stream processor and a texture-map processor. This is the new concept unified rendering architecture introduced by DirectX 10 (DirectX 10 is not yet installed on XP, but some software has this feature). Using the unified rendering architecture can effectively and fully utilize the video card idle resources, no longer need the previous generation of graphics cards to rely on the rendering pipeline to execute commands queued and executed one by one. Nvidia GeForce 8800GTX is the world's first PC desktop graphics support DirectX 10, the core code G80, but also the end of the graphics pipeline concept of the first graphics card, the introduction of a new flow processing unit concept, is the graphics history of a major innovation.

The flow processing unit directly affects the processing capacity, because the flow-processing unit is the core of the video card, also known as the brain. The more the number of flow processing units, the more processing power, the general proportional relationship, but this is limited to Nvidia's own core or AMD's own core comparison category. The comparison between Nvidia and AMD flow-processing units is comparable, that is, Nvidia's 1 flow-processing units are equivalent to AMD's 5 flow-processing units (as development may vary).

The streaming processor directly maps the multimedia graphics data stream to the stream processor for processing, which is programmable and not programmable. The bandwidth between the stream memory and the SRF is the bandwidth between the 2GB/S,SRF and the processor registers is 32gb/s,alu Cluster (ALU cluster) The bandwidth between the registers and the ALU is 544gb/s, the proportional relationship between the three kinds of bandwidth is 1:16:272.

Effect

Each stream processor has a dedicated high-speed unit for decoding and executing streaming data. The onboard cache is a typical unit with a stream processor that can quickly enter and read data to complete the next rendering.

Flow processor How much of the graphics performance is decisive, can be said that the low end of high school graphics card in addition to the core of the main difference is the number of flow processors, but one thing to note is that the number of NV and AMD graphics flow processors are not comparable, their two graphics core architecture is different, Can not compare the flow processor to see how much performance, under normal circumstances, NV's graphics flow processor number will be significantly less than AMD, to from the flow of the processor to see how much performance, can only own compared with their own, such as 3850 compared with 3450, 8600 compared with 8800.

Of course, just like your CPU frequency is the same as the truth. The general graphics flow processing will affect the video and high-definition video decoding function, but the most important is your video card core. Now the best core should be G92. But that's not enough, Your graphics architecture also determines performance. Like the difference between a professional video card and a game card, even though everything is exactly the same, it's not a concept. 8800GTS is not as good as a common G92 core professional graphics card performance.

This is a parameter of the graphics card, 2006 by NVIDIA company first proposed a concept, namely formerly used to say two graphics card parameter pixel pipelines (pixel rendering pipeline) and vertex pipelines (vertex shading unit), referred to as SP, Its role is to deal with the signal transmitted by the CPU, directly into the display can be recognized digital signals.

In general, the more flow processors, the stronger the performance of graphics, such as a 640-stream processor graphics card, such as the 80-stream processor graphics card higher than a few grades.