Analysis of AMD's NUMA Modes and Uniform Memory Access in Modern Server Architectures
By
zdw
Hand-rolled, kettle-boiled, baked to perfection. Worth every minute at the bakery.
Summary
This article examines AMD's NUMA (Non-Uniform Memory Access) modes, specifically focusing on NPS0 (Nodes Per Socket 0) mode which presents a dual-socket system as a single monolithic entity rather than subdividing it. The analysis explores how uniform memory access functions as interconnects become increasingly non-uniform in modern server architectures, with AMD designating NUMA modes using the NPS prefix. The content discusses the technical implications of memory access patterns in multi-socket server systems as core and memory controller counts increase.
Key quotes
· 4 pulledNUMA, or Non-Uniform Memory Access, lets hardware expose affinity between cores and memory controllers to software.
AMD designates their NUMA modes with the NPS (Nodes Per Socket) prefix.
NPS0 is a special NUMA mode that goes in the other direction. Rather than subdivide the system, NPS0 exposes a dual socket system as a single monolithic entity.
It's a reflection of how non-uniform interconnects get as core and memory controller counts keep going up.
You might also wanna read
Understanding the Super Nintendo's Hardware Architecture: Clock Signals and Component Synchronization
A deep-dive technical analysis of the Super Nintendo Entertainment System (SNES) hardware architecture, focusing on how its components work
Research Directions for Overcoming Memory and Interconnect Challenges in Large Language Model Inference Hardware
This article discusses the technical challenges of Large Language Model (LLM) inference, highlighting how the autoregressive Decode phase ma
arxiv.org·4mo ago
Technical Analysis of Nvidia GB10's Memory Subsystem Architecture from CPU Perspective
This article provides a technical analysis of Nvidia GB10's memory subsystem from the CPU perspective. GB10 is a collaboration between Nvidi
chipsandcheese.com·5mo ago
VectorWare Launches as First GPU-Native Software Company
VectorWare announces itself as the first GPU-native software company, arguing that the fundamental technology shift is from CPU to GPU archi
vectorware.com·7mo ago
Understanding Modern GPU Architecture for Machine Learning: H100 and B200 Technical Analysis
This article provides a technical deep dive into modern GPU architecture, specifically focusing on NVIDIA GPUs like H100 and B200 used for m
jax-ml.github.io·9mo ago
AMD Re-Releases Ryzen 7 5800X3D at Lower $350 Price Point Amid High RAM Costs
AMD is re-releasing its former flagship Ryzen 7 5800X3D CPU at a reduced price of $350 (down $100 from its 2022 launch price) starting June
gizmodo.com·3h ago