Specialty market
20-01-2021 |
19-01-2021 |
18-01-2021 |
15-01-2021 |
14-01-2021 |
20-01-2021 |
January 13, 2020|
08-01-2021 |
December 29, 2020|
07-01-2021 |
December 17, 2020|
March 12, 2020|
23-11-2020 |
June 8, 2020|
May 8, 2020|
April 8, 2020|
20-11-2020 |
September 12, 2020|
30-11-2020 |
September 10, 2020|
July 10, 2020|
15-01-2021 | | Passed
recent,
It will use small chips to replace large-scale monolithic semiconductors. According to the patent, AMD will take a thorough approach that has never used chiplets before. Share on-chip resources.
Most chip designs break down the system into core components such as CPU and GPU. Nevertheless, as the technology becomes more and more complex, AMD will break the GPU itself into small chips and then connect them.
GPU is a large parallel system
The fast access speed is the fact that the GPU is always integrated on a piece of silicon. However, silicon technology has reached its limit. For various reasons, the company cannot simply expand the scale of the chip. In addition, when the circuits are divided, the access time increases, and the size of the bus generally decreases.
According to the patent, AMD will solve this problem by using GPU chiplets that use its own L1 and L2 caches, but all chiplets in the design can access the L3 cache. The chiplets will be connected using a dedicated bus called HBX Passive Cross Link, allowing the chiplets to access low-level caches on other chiplets as fast as they can access their own local caches.
Orient vertically instead of horizontally. Considering that silicon chips can only be so large, such a design decision makes sense. Even chips designed with small chips still face similar size challenges (ie packaging and placement).
Recently, there has been a lot of discussion around small chips and how companies such as AMD can use them to produce next-generation devices. In short, a small chip is a small piece of semiconductor that contains functional circuits for making larger circuits. Traditional semiconductors integrate all circuit functions on a single chip, but small chips spread the functions across multiple smaller chips.
Use wires or traces. Designs using small chips can also integrate SMD components, such as capacitors and resistors, for filtering and current limiting purposes. Chiplets are still a new technology and only a few companies use them, but their advantages will soon make them the dominant technology in the semiconductor industry.
The advantage of chiplets is only
technology. When manufacturing semiconductors, the possibility of a single transistor failure is very small, and a device containing one million transistors has a relatively high job opportunity.
However, modern devices with extremely small feature sizes may have billions of transistors, which greatly increases the failure rate of the die. Therefore, chip designers have two options to improve wafer yield. Either reduce the total number of transistors (degrade the performance of the device), or increase the price of each device.
The price can be increased, but many modern devices are aimed at the consumer market, so excessive prices will make this technology unusable for commercialization. Instead, designers are looking for small chips to solve their problems.
Chiplets essentially spread the high number of transistors across multiple devices. If one of the billion transistors fails, then 99 of the 100 chiplets are available, and a single device containing one billion transistors will fail. Since many defects in silicon wafers are point defects (meaning that a specific area of the wafer is unavailable), using smaller dies can produce more usable devices, thereby increasing yield.
Chiplets also allow customers to use chiplets as building blocks to create customized devices. In a situation similar to Apple M1 processor development, customers can choose to combine the CPU with various functional blocks for their applications. The use of building blocks also reduces the price of such designs, because semiconductor designers can focus on developing standard glue logic components similar to the 4000 and 7400 series.
As foundries face increasing difficulties in reducing the size of transistors, small chips will become the key to future semiconductor technology.
? Are they easy to customize? The answers to these questions are not yet clear, but it is clear that small chips will continue to exist.
Robin Mitchell is an electronic engineer who has been in electronics since he was 13 years old. After obtaining a Bachelor of Arts degree from the University of Warwick, Robin entered the field of online content creation, developing articles for professionals and manufacturers, news and projects the same. Currently, Robin runs MitchElectronics, a small electronics company that produces educational kits and resources.
by
| November 7, 2017
| April 22, 2020
| September 28, 2020
information
Manage newsletters
Our application
Social profile
All news and general inquiries:
Tel
+86 400-003-5559
