Configurable GPU Chiplets for Scalable Processing Without Large Dies
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The increasing demands on graphics processing units (GPUs) lead to larger and more expensive dies due to the need for additional physical resources, making manufacturing and scalability challenging.
Innovation Solution
The GPU is partitioned into multiple dies, or GPU chiplets, which can be configured to function as a single GPU or multiple GPUs, allowing flexible and cost-effective resource allocation based on operating modes, with a configurable number of chiplets assembled to implement varying generations of technology.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the physical resources of GPU are increased to meet growing demands, then processing capability is improved, but die size and manufacturing cost increase
Solution Approach 1:
The GPU is divided into multiple independent shader engine dies, each capable of performing graphics processing functions. These separate dies can be manufactured independently and then combined through interconnect technology to form a complete GPU system, thereby achieving high processing capability without requiring a single large die
2Productivity
If the physical resources of GPU are increased to meet growing demands, then processing capability is improved, but manufacturing cost increases
Solution Approach 1:
By segmenting the GPU into multiple shader engine dies, each die can be manufactured using standard fabrication processes for smaller, more cost-effective dies. The modular architecture allows for economies of scale in manufacturing individual dies while reducing the complexity and cost associated with manufacturing a single large, complex die with all resources integrated
3Productivity
If the physical resources of GPU are increased, then processing capability is improved, but device complexity increases
Solution Approach 1:
The GPU architecture is segmented into multiple independent shader engine dies with standardized interfaces and communication protocols. This modular design reduces overall system complexity by allowing each die to be designed, tested, and optimized independently, while the interconnect technology provides a systematic framework for integrating these modules without creating unmanageable complexity
Data Source
AI summary
A graphics processing unit (GPU) of a processing system is partitioned into multiple dies (referred to as GPU chiplets) that are configurable to collectively function and interface with an application as a single GPU in a first mode and as multiple GPUs in a second mode. By dividing the GPU into multiple GPU chiplets, the processing system flexibly and cost-effectively configures an amount of active GPU physical resources based on an operating mode. In addition, a configurable number of GPU chiplets are assembled into a single GPU, such that multiple different GPUs having different numbers of GPU chiplets can be assembled using a small number of tape-outs and a multiple-die GPU can be constructed out of GPU chiplets that implement varying generations of technology.