A Comparison of the Cerebras Wafer-Scale Integration Technology with Nvidia GPU-based Systems for Artificial Intelligence
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866913737804349440 |
|---|---|
| author | Kundu, Yudhishthira Kaur, Manroop Wig, Tripty Kumar, Kriti Kumari, Pushpanjali Puri, Vivek Arora, Manish |
| author_facet | Kundu, Yudhishthira Kaur, Manroop Wig, Tripty Kumar, Kriti Kumari, Pushpanjali Puri, Vivek Arora, Manish |
| contents | Cerebras' wafer-scale engine (WSE) technology merges multiple dies on a single wafer. It addresses the challenges of memory bandwidth, latency, and scalability, making it suitable for artificial intelligence. This work evaluates the WSE-3 architecture and compares it with leading GPU-based AI accelerators, notably Nvidia's H100 and B200. The work highlights the advantages of WSE-3 in performance per watt and memory scalability and provides insights into the challenges in manufacturing, thermal management, and reliability. The results suggest that wafer-scale integration can surpass conventional architectures in several metrics, though work is required to address cost-effectiveness and long-term viability. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2503_11698 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | A Comparison of the Cerebras Wafer-Scale Integration Technology with Nvidia GPU-based Systems for Artificial Intelligence Kundu, Yudhishthira Kaur, Manroop Wig, Tripty Kumar, Kriti Kumari, Pushpanjali Puri, Vivek Arora, Manish Hardware Architecture Cerebras' wafer-scale engine (WSE) technology merges multiple dies on a single wafer. It addresses the challenges of memory bandwidth, latency, and scalability, making it suitable for artificial intelligence. This work evaluates the WSE-3 architecture and compares it with leading GPU-based AI accelerators, notably Nvidia's H100 and B200. The work highlights the advantages of WSE-3 in performance per watt and memory scalability and provides insights into the challenges in manufacturing, thermal management, and reliability. The results suggest that wafer-scale integration can surpass conventional architectures in several metrics, though work is required to address cost-effectiveness and long-term viability. |
| title | A Comparison of the Cerebras Wafer-Scale Integration Technology with Nvidia GPU-based Systems for Artificial Intelligence |
| topic | Hardware Architecture |
| url | https://arxiv.org/abs/2503.11698 |