# Hugging Face — Faster Training and Inference: Habana Gaudi®2 vs Nvidia A100 80GB

- Company: Hugging Face (huggingface.co)
- Announced: 2022-12-14
- Category: research-paper
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://huggingface.co/blog/habana-gaudi-2-benchmark
- Record: https://forck.live/items/2121-faster-training-and-inference-habana-gaudi-2-vs-nvidia-a100-80gb
- Subject: Platform
- Models affected: BERT, Stable Diffusion, T5-3B

Habana Gaudi2 accelerator is about twice as fast as Nvidia A100 80GB for both training and inference, based on benchmarks including BERT pre-training, Stable Diffusion inference, and T5-3B fine-tuning. The source describes how to access Gaudi2 via Intel Developer Cloud and presents throughput comparisons.

## Evidence

Verbatim from https://huggingface.co/blog/habana-gaudi-2-benchmark:

> Spoiler alert - Gaudi2 is about twice faster than Nvidia A100 80GB for both training and inference!

---

Record: https://forck.live/items/2121-faster-training-and-inference-habana-gaudi-2-vs-nvidia-a100-80gb
Catalogue: https://forck.live/llms.txt
Current issue: https://forck.live/feed.md