Publication Date

Spring 2026

Degree Type

Thesis

Degree Name

Master of Science (MS)

Department

Computer Engineering

Advisor

Haonan Wang; Kaikai Liu; Wencen Wu

Abstract

As the number of transistors in modern processors increases, heat dissipation has become a major bottleneck to scalability. The use of 3D stacking further intensifies this problem, as heat from multiple layers can accumulate vertically. These challenges create a growing need for tools that can accurately and efficiently simulate the thermal behavior of 3D chips during design and validation. Several existing tools model thermal behavior for 3D-stacked chips and can simulate average heat over large spatial regions or long time intervals. However, when heat is concentrated in a small area or over a short time window, such models can miss localized hotspots. Conversely, when heat is distributed broadly, coarse averaging can overstate thermal risk. More fine-grained simulation results can be achieved through gate-level simulation. However, these exhibit prohibitively long runtimes when simulating real-world workloads. To this end, this thesis presents RThermaL, a novel method for gate-level power and thermal simulation by using per-instruction power mappings with a GPU-accelerated thermal simulation workflow. RThermaL works by linking individual instructions to gate-level power dissipation in a physically accurate implementation of a processor layout. Starting from RTL-level Verilog designs, the framework performs post-layout instruction characterization to construct a database of per-instruction spatial power maps. These maps can then be used to reconstruct temporal and spatial heat inputs to achieve significant speedups for thermal simulations. Our evaluation shows that, for the same targeted workload, our approach achieves a 5000× runtime improvement over full gate-level simulation. Compared to existing thermal solvers, it delivers a 6× runtime improvement. Equivalently, under the same time budget, our method provides 15,000× finer temporal resolution and 4× finer spatial resolution, making detailed thermal analysis practical for real-world workloads.

Download

Share

COinS