Friday 6 January 2017

Geography of a GPU.


The GPU is not an essential part to a computer. It can be added to improve visual performance and to enhance rendering speeds. The GPU is essential a different intended purpose CPU. GPU's are expensive, for a good one at least. If you were to build your own PC, a rule of thumb in to spend one third of your budget on the GPU. GPU's can be used for improving visual performance by achieving higher FPS in games, or for quick rendering of videos for example.

Some main producers of GPU's are EVGA, who make the Nvidia GPU's and Radeon who make AMD GPU's. AMD tends to be cheaper in general. For example, their new flagship RX 480 costs around the 250 euro range while the new flagship EVGA 1080 cost an impressive 600 and up to buy. EVGA does have a good selling point, most a of their new cards come VR ready, and even some f their old ones do too.

The GPU starts off with a new design being thought of. This is called the PRS. This is a very costly process and will involve a lot of people. These would be engineers, designers and coders. They then begin planning where the "building blocks" will go on the chip of the GPU. This is done with code. The aim of this is to reduce delay time between signals. They test the code and it goes through a process called synthesis, which will compile the code. One it's done. it will give instructions to each transistor on the chip. This is called a netlist. The chip is then tested is then tested. If the computer finds anything wrong with the chip, the team backtracks to find the problem and then fix it. The design of the physical board and chip-set is now considered. This is almost all automatic. The way the chip is made is by taking a picture of each layer of the transistors, then making the chip using these layers. The next step is to make the chip, and this uses a thing called photo-lithography. This is when all the connections on the chip are made. First, special material called photo resist is applied to the wafer. Then UV light is directed at it with a template covering it. The material that the template covered allows the material to stay, Then the wafer is chemically altered, while the material that stays does not allow the chemicals to touch  the wafer. This done is along with an oxide layer. The wafer is then hit with phosphorus ions mixed with silicone to allow energy transfer. Around this phosphorus, silicon dioxide is used to insulate it. Then, a thin layer of silicone dioxide is placed on top of the phosphorus. This will help it not to short out. Trenches are then drilled and copper is placed inside of them. Excess copper is then taken away. Many layers of copper is usually necessary, for example, one to provide power, another to connect one component to another, etc. The chips are then placed onto a PCB, and a cooler is placed onto it. 

http://icdn5.digitaltrends.com/image/evga107001-640x427-c.jpg
  

No comments:

Post a Comment

Note: only a member of this blog may post a comment.