Saved in:
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2501.10187 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866913811602079744 |
|---|---|
| author | Canakci, Burcu Liu, Junyi Wu, Xingbo Cheriere, Nathanaël Costa, Paolo Legtchenko, Sergey Narayanan, Dushyanth Rowstron, Ant |
| author_facet | Canakci, Burcu Liu, Junyi Wu, Xingbo Cheriere, Nathanaël Costa, Paolo Legtchenko, Sergey Narayanan, Dushyanth Rowstron, Ant |
| contents | To match the blooming demand of generative AI workloads, GPU designers have so far been trying to pack more and more compute and memory into single complex and expensive packages. However, there is growing uncertainty about the scalability of individual GPUs and thus AI clusters, as state-of-the-art GPUs are already displaying packaging, yield, and cooling limitations. We propose to rethink the design and scaling of AI clusters through efficiently-connected large clusters of Lite-GPUs, GPUs with single, small dies and a fraction of the capabilities of larger GPUs. We think recent advances in co-packaged optics can enable distributing AI workloads onto many Lite-GPUs through high bandwidth and efficient communication. In this paper, we present the key benefits of Lite-GPUs on manufacturing cost, blast radius, yield, and power efficiency; and discuss systems opportunities and challenges around resource, workload, memory, and network management. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2501_10187 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Good things come in small packages: Should we build AI clusters with Lite-GPUs? Canakci, Burcu Liu, Junyi Wu, Xingbo Cheriere, Nathanaël Costa, Paolo Legtchenko, Sergey Narayanan, Dushyanth Rowstron, Ant Hardware Architecture Artificial Intelligence Distributed, Parallel, and Cluster Computing To match the blooming demand of generative AI workloads, GPU designers have so far been trying to pack more and more compute and memory into single complex and expensive packages. However, there is growing uncertainty about the scalability of individual GPUs and thus AI clusters, as state-of-the-art GPUs are already displaying packaging, yield, and cooling limitations. We propose to rethink the design and scaling of AI clusters through efficiently-connected large clusters of Lite-GPUs, GPUs with single, small dies and a fraction of the capabilities of larger GPUs. We think recent advances in co-packaged optics can enable distributing AI workloads onto many Lite-GPUs through high bandwidth and efficient communication. In this paper, we present the key benefits of Lite-GPUs on manufacturing cost, blast radius, yield, and power efficiency; and discuss systems opportunities and challenges around resource, workload, memory, and network management. |
| title | Good things come in small packages: Should we build AI clusters with Lite-GPUs? |
| topic | Hardware Architecture Artificial Intelligence Distributed, Parallel, and Cluster Computing |
| url | https://arxiv.org/abs/2501.10187 |