Integrated CDC-PHY Compression for Faster GPU Links

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current graphics processing units (GPUs) face challenges in efficiently processing graphics data due to limitations in parallel processing capabilities and synchronization of threads, particularly in SIMT architectures, which can lead to inefficiencies in graphics data operations.

Innovation Solution

Implementing lossless and lossy automatic hardware compression techniques in graphics-to-graphics network links using GPUs, which optimize data processing by enhancing parallel processing and thread synchronization through dedicated circuitry and compression algorithms.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Loss of information

If lossless compression is applied to graphics data, then data accuracy is maintained, but processing time and computational overhead increase

Engineering Contradiction:
Improvedata accuracyVSAvoidprocessing time
Core Design Contradiction:
Loss of informationVSLoss of time

Solution Approach 1:

The compression system is divided into separate modules: an entropy coder module for lossless compression, a quantization module for lossy compression, and a selector that chooses between them. This segmentation allows the system to apply different compression strategies to different data types or importance levels, optimizing both accuracy and processing time by not applying lossless compression uniformly to all data.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system dynamically selects between lossless and lossy compression modes based on data characteristics and performance requirements. The selector module can adaptively choose the appropriate compression type for different graphics data streams, allowing the system to maintain high accuracy when needed while reducing processing time when acceptable quality can be achieved with lossy compression.

Inventive Principle:
Principle #15Dynamics

2Productivity

If parallel processing is increased in SIMT architecture, then processing throughput improves, but thread synchronization complexity and overhead increase

Engineering Contradiction:
Improveprocessing throughputVSAvoidthread synchronization complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The parallel processing system divides graphics data into independent processing streams that can be handled by separate thread groups. By segmenting the work into smaller, more manageable units with clearer synchronization points, the system achieves high throughput while reducing the complexity of coordinating thousands of threads, as each segment can be processed with simpler synchronization protocols.

Inventive Principle:
Principle #1Segmentation

3Productivity

If hardware compression circuitry is added to GPUs, then data processing efficiency improves, but device complexity and manufacturing cost increase

Engineering Contradiction:
Improvedata processing efficiencyVSAvoiddevice complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The compression circuitry is designed as a multi-functional module that can perform both lossless and lossy compression operations, as well as handle different data formats and compression standards. This universal design allows a single hardware addition to provide multiple compression capabilities, reducing the overall complexity compared to implementing separate dedicated circuits for each compression type while still achieving significant data processing efficiency improvements.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Data Source

PatentEP4618545A1Lossless and lossy automatic hardware compression in graphics-to-graphics network links
Publication Date: 2025.09.17 INTEL CORP
  • EP4618545A1 patent drawingFigure 1
  • EP4618545A1 patent drawingFigure 2A
  • EP4618545A1 patent drawingFigure 2B~2C

AI summary

An apparatus to facilitate lossless and lossy automatic hardware compression in graphics-to-graphics network links is disclosed. The apparatus includes compressor/decompressor circuitry (CDC) integrated with physical layer (PHY) intellectual property (IP) hardware circuitry for a graphics processor unit (GPU)-to-GPU communication link communicably coupling a first GPU to one or more other GPUs, the CDC to: receive a data message from the first GPU, wherein the data message is in an uncompressed format; determine that a compression process is to be applied to the data message; apply the compression process to the data message to generate a compressed data message; and cause a GPU link IP hardware circuitry that comprises the PHY IP hardware circuitry to transmit the compressed data message over the GPU-to-GPU communication link.