Multi-Chip GPU QoS and Chiplet Isolation for Cloud VMs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing graphics processing units (GPUs) face challenges in efficiently partitioning hardware resources among virtual machines while ensuring fair resource allocation and data isolation for security purposes, particularly in cloud computing environments where customers rent partial hardware resources.
Innovation Solution
Implementing a multi-chip GPU architecture with chiplet isolation and quality of service (QoS) mechanisms to compartmentalize compute partitions, ensuring each partition receives a fair share of resources and maintains data isolation through encryption and dedicated memory paths.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a single GPU is virtually partitioned among multiple virtual machines, then hardware utilization efficiency is improved, but resource allocation fairness and data isolation are compromised
Solution Approach 1:
The patent divides a single GPU into multiple independent chiplets (compute partitions), each assigned to a specific virtual machine. This physical segmentation enables true hardware-level isolation while maintaining high utilization through virtualization management, resolving the contradiction between efficiency and isolation.
Solution Approach 2:
The patent introduces a virtualization layer with QoS mechanisms and isolation hardware as an intermediary between the physical GPU and virtual machines. This mediator ensures fair resource allocation and data isolation while maintaining efficient hardware utilization through coordinated management.
2Adaptability or versatility
If hardware resources are shared among virtual machines, then cloud service provider flexibility is improved, but security isolation between customers is worsened
Solution Approach 1:
By physically segmenting the GPU into isolated chiplets, each customer's virtual machine receives dedicated hardware resources with guaranteed isolation. This enables flexible resource allocation across multiple customers while maintaining strong security boundaries, as each chiplet operates independently with controlled access.
Solution Approach 2:
The patent implements different isolation quality levels for different resource access paths. Critical data paths have strong isolation guarantees through dedicated hardware, while less sensitive operations can share resources, providing both security and flexibility simultaneously.
3Reliability
If compute partitions are isolated within multi-chip GPU, then security and fault protection are improved, but device complexity is worsened
Solution Approach 1:
The patent uses segmentation to create isolated chiplets, which actually simplifies the overall system architecture by distributing complexity across independent modules. Each chiplet can be designed and verified separately, reducing the complexity burden compared to a monolithic design while achieving superior isolation.
Solution Approach 2:
The isolated chiplet architecture provides multiple functions: security isolation, fault containment, thermal management, and scalability. By designing for isolation as a fundamental principle, the patent achieves these multiple benefits simultaneously without proportionally increasing complexity.
Data Source
Figure 1
Figure 2A
Figure 2B~2C
AI summary
One embodiment provides a graphics processor comprising a system interface and a plurality of chiplets coupled with the system interface. Each of the plurality of chiplets is configurable to be assigned to a partition of a plurality of partitions. A chiplet of the plurality of chiplets includes an interface to a memory device, a cache memory coupled with the interface to the memory device, and a graphics core cluster coupled with the interface to the memory device and the cache memory, the graphics core cluster including a plurality of graphics cores configured to execute graphics and compute workloads.