Multi-GPU BMC Monitoring via USB-NIC and I2C Path Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional Baseboard Management Controllers (BMCs) in information handling systems are limited by the use of a single management path (I2C) for GPU monitoring and management, reducing the functionality available for GPU system operations.
Innovation Solution
A BMC engine is configured to utilize both a USB-NIC path for first log collection and inventory operations during initialization, and an I2C path for sensor monitoring during runtime, with the USB-NIC path also used for firmware updates when requested, enhancing the monitoring and management capabilities of multiple GPUs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional BMC devices use only the legacy I2C path for GPU monitoring and management operations, then the system maintains simplicity in path selection, but the functionality and speed of GPU system operations are significantly reduced
Solution Approach 1:
The patent makes the USB-NIC path multi-functional by using it for both firmware update operations and other GPU monitoring/management operations. This allows the fast path to serve multiple purposes, significantly enhancing BMC device functionality without adding complex path management logic, as the path selection remains simple while the capabilities expand.
Solution Approach 2:
The system dynamically selects between I2C and USB-NIC paths based on the operation type and availability. The USB-NIC path is used when available and needed for fast operations, while falling back to I2C when USB-NIC is unavailable or not needed, providing adaptive functionality without permanent complexity.
2Speed
If conventional BMC devices use the fast USB-NIC path for all GPU operations, then the speed of operations is improved, but the path becomes unavailable for other components and consistency is lost
Solution Approach 1:
The patent segments the operation types and assigns them to appropriate paths: USB-NIC path is used for firmware updates and log collection operations, while I2C path handles sensor monitoring. This segmentation ensures that the fast USB-NIC path is used selectively for operations that benefit from its speed, while maintaining reliable and consistent path availability for all GPU operations.
3Adaptability or versatility
If conventional BMC devices limit USB-NIC path usage to firmware updates only, then path availability for other components is maintained, but the functionality and speed of other GPU operations are reduced
Solution Approach 1:
The patent extends the USB-NIC path functionality beyond firmware updates to include log collection and other GPU monitoring operations. This multi-functional usage significantly improves productivity and efficiency of GPU operations by leveraging the fast USB-NIC path for multiple critical tasks without compromising its availability for firmware updates.
Solution Approach 2:
The system dynamically determines when to use the USB-NIC path based on operation type and path availability, allowing flexible and efficient GPU operations while maintaining reliability for other components.
Data Source
AI summary
A Baseboard Management Controller (BMC)/multi-Graphics Processing Unit (GPU) monitoring/management system includes a computing device having a GPU system with a plurality of GPUs, and a BMC device coupled to the GPU system via a Universal Serial Bus-Network Interface Controller (USB-NIC) path and an Inter-Integrated Circuit (I2C) path. The BMC device performs first log collection operations and inventory operations via the USB-NIC path with the GPU system during initialization of the computing device. During runtime of the computing device that follows the initialization of the computing device, the BMC device performs sensor monitoring operations with the GPU system via the I2C path. During runtime of the computing device, the BMC device performs second log collection operations with the GPU system via the USB-NIC path, and will perform firmware update operations with the GPU system via the USB-NIC path in response to receiving a firmware update request.