Dissecting the NVIDIA Blackwell Architecture with Microbenchmarks

2025-07-14

AI Summary

NVIDIA’s Blackwell architecture lacks systematic microarchitectural analysis, hindering informed software optimization and hardware co-design. Method: We design and deploy a comprehensive microbenchmark suite targeting memory hierarchy behavior, SM execution pipeline characteristics, and fifth-generation Tensor Cores—supporting FP4/FP6 precision—to conduct the first empirical, low-level characterization of Blackwell. We perform comparative latency, throughput, cache access pattern, and power measurements across RTX 5080 and H100 PCIe platforms. Contribution/Results: Our analysis reveals critical trade-offs in scheduler enhancements, L2 cache policy adjustments, and energy-efficiency improvements—and identifies performance regressions in specific workloads. We quantitatively bound Blackwell’s generational improvements over Hopper and derive portable, application- and compiler-aware optimization guidelines. These findings provide an empirical foundation and methodology for next-generation GPU software-hardware co-design.

Technology Category

• Machine Learning: Hardware-aware ML
• Search and Optimization: Evaluation and Analysis
• Humans and AI: Other Foundations of Human Computation & AI

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Web performance, measurement, and characterization
• Security and Privacy: Large-scale security measurements
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

The rapid development in scientific research provides a need for more compute power, which is partly being solved by GPUs. This paper presents a microarchitectural analysis of the modern NVIDIA Blackwell architecture by studying GPU performance features with thought through microbenchmarks. We unveil key subsystems, including the memory hierarchy, SM execution pipelines, and the SM sub-core units, including the 5th generation tensor cores supporting FP4 and FP6 precisions. To understand the different key features of the NVIDIA GPU, we study latency, throughput, cache behavior, and scheduling details, revealing subtle tuning metrics in the design of Blackwell. To develop a comprehensive analysis, we compare the Blackwell architecture with the previous Hopper architecture by using the GeForce RTX 5080 and H100 PCIe, respectively. We evaluate and compare results, presenting both generational improvements and performance regressions. Additionally, we investigate the role of power efficiency and energy consumption under varied workloads. Our findings provide actionable insights for application developers, compiler writers, and performance engineers to optimize workloads on Blackwell-based platforms, and contribute new data to the growing research on GPU architectures.

Problem

• Analyze NVIDIA Blackwell architecture's microarchitectural performance features
• Compare Blackwell with Hopper architecture for generational improvements
• Investigate power efficiency and energy consumption under varied workloads

Innovation

• Microbenchmark analysis of NVIDIA Blackwell architecture
• Comparison with Hopper using RTX 5080 and H100
• Study of FP4 and FP6 tensor core precisions