ABI: A tightly integrated, unified, sparsity-aware, reconfigurable, compute near-register file/cache GPU architecture with light-weight softmax for deep learning, linear algebra, and Ising compute
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , |
|---|---|
| Format: | Preprint |
| Publié: |
2026
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866914598610796544 |
|---|---|
| author | Raman, Siddhartha Raman Sundara Kulkarni, Jaydeep P. |
| author_facet | Raman, Siddhartha Raman Sundara Kulkarni, Jaydeep P. |
| contents | We present a tightly integrated and unified near-memory GPU architecture that delivers 6 to 16 times speedup and 6 to 13 times energy savings across Convolutional Neural Networks, Graph Convolutional Networks, Linear Programming, Large Language Models, and Ising workloads compared to MIAOW GPU. The design includes a custom sparsity-aware near-memory circuit providing about 1.5 times energy savings, and a lightweight softmax circuit providing about 1.6 times energy savings. The architecture supports reconfigurable compute up to INT16 with dynamic resolution updates and scales efficiently across problem sizes. ABI-enabled MI300 and Blackwell systems achieve about 4.5 times speedup over baseline MI300 and Blackwell. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_14262 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | ABI: A tightly integrated, unified, sparsity-aware, reconfigurable, compute near-register file/cache GPU architecture with light-weight softmax for deep learning, linear algebra, and Ising compute Raman, Siddhartha Raman Sundara Kulkarni, Jaydeep P. Hardware Architecture We present a tightly integrated and unified near-memory GPU architecture that delivers 6 to 16 times speedup and 6 to 13 times energy savings across Convolutional Neural Networks, Graph Convolutional Networks, Linear Programming, Large Language Models, and Ising workloads compared to MIAOW GPU. The design includes a custom sparsity-aware near-memory circuit providing about 1.5 times energy savings, and a lightweight softmax circuit providing about 1.6 times energy savings. The architecture supports reconfigurable compute up to INT16 with dynamic resolution updates and scales efficiently across problem sizes. ABI-enabled MI300 and Blackwell systems achieve about 4.5 times speedup over baseline MI300 and Blackwell. |
| title | ABI: A tightly integrated, unified, sparsity-aware, reconfigurable, compute near-register file/cache GPU architecture with light-weight softmax for deep learning, linear algebra, and Ising compute |
| topic | Hardware Architecture |
| url | https://arxiv.org/abs/2602.14262 |