Tiamat: A 98-to-134ms/Step Transformer-Based Diffusion Model Processor Supporting Classifier-Free Guidance for Image Generation
Abstract
A$16 n m$FinFET transformer-based diffusion model processor chip is fabricated for supporting class-conditional DiT-XL/2 and text-to-image PixArt-a with 98 ms and 134 ms generation time per step with 7.37TOPS and 1477 mW at 400 MHz. Classifier-free guidance batching reduces external memory access (EMA) for weights by 50%, and weight-reordered quantization brings another 37% reduction. Hybrid microscaling blocking reduces featuremap buffer size by 37%, while approaching floating-point quality.
Fields of Study & Subjects
Advanced Neural Network Applications
Low-power high-performance VLSI design
Generative Adversarial Networks and Image Synthesis
Publication details
- Journal / source
- Digest of technical papers/Digest of technical papers - IEEE International Solid-State Circuits Conference
- Pages
- 54–56
- Publication date
- 15 February 2026
- Language
- EN