Technical Analysis of Nvidia GB10's Memory Subsystem Architecture from CPU Perspective
By
ingve
An everything bagel for the brain. Substantive, layered, well-seasoned.
Summary
This article provides a technical analysis of Nvidia GB10's memory subsystem from the CPU perspective. GB10 is a collaboration between Nvidia and Mediatek that integrates Nvidia's Blackwell architecture into an integrated GPU. The article focuses on the memory subsystem design decisions and tradeoffs required to feed the powerful compute resources, including 48 Blackwell SMs (matching RTX 5070 core count) and 20 CPU cores (10 Cortex X925 + 10 Cortex A725). The analysis examines how Nvidia and Mediatek approached memory subsystem design for this high-performance integrated solution.
Key quotes
· 5 pulledGB10 is a collaboration between Nvidia and Mediatek that brings Nvidia's Blackwell architecture into an integrated GPU.
GB10's GPU has 48 Blackwell SMs, matching the RTX 5070 in core count.
The CPU side has 10 Cortex X925 and 10 Cortex A725 cores and is therefore quite powerful.
Feeding all of that compute power requires a beefy memory subsystem, and can lead to difficult tradeoffs.
Analyzing GB10's memory subsystem from the CPU side will be the focus of this article.
You might also wanna read
How high-bandwidth memory became a critical bottleneck for AI chip performance
The article examines how high-bandwidth memory (HBM) chips, particularly those made by Micron Technology, have become a critical bottleneck

Intel's Crescent Island GPU targets AI inference with up to 480GB LPDDR5X memory
Intel has revealed new details about its Crescent Island GPU at Computex 2026. The GPU is based on the Xe3P architecture and targets AI infe
