随着人工智能技术的飞速发展，AI作图已经成为了一个备受关注的领域。近日，Stability AI发布了最新的研究报告，该报告详细对比了英伟达与英特尔在AI作图训练方面的性能差异。报告指出，Intel Gaudi 2芯片在AI作图训练中的表现令人瞩目，其性能超越了英伟达的H100芯片。
以Stability AI的2B参数MMDiT模型为例，在24深度的混合精度BFloat16设置下，Gaudi 2芯片可以达到每秒1254张图片的训练速度。相比之下，英伟达的H100-80GB计算卡只能达到每秒595张图片的训练速度，而A100-80GB计算卡则为每秒381张图片。这意味着Gaudi 2芯片的性能比H100计算卡提高了55%，是A100计算卡的2.43倍。
这一消息对于人工智能和芯片行业来说都具有重要意义，它不仅展示了英特尔在芯片技术上的进步，也为AI作图领域的发展提供了新的可能。随着技术的不断进步，我们可以期待在未来看到更多高效能的芯片被应用于人工智能领域，推动行业的进一步发展。
英文标题：Stability AI Reports: Intel Gaudi 2 Chip Outperforms Nvidia H100 in AI Image Generation Training
英文关键词：AI Image Generation, Chip Performance, Technology Comparison
英文新闻内容：
In a recent report, Stability AI, the developer of the popular Stable Diffusion AI image generation model, has revealed that Intel’s Gaudi 2 chip outperforms Nvidia’s H100 in AI image generation training. The report focused on comparing the efficiency of Nvidia and Intel computing cards in AI image generation tasks.
Using Stability AI’s 2 billion parameter MMDet model as an example, the Gaudi 2 chip achieved a training speed of 1,254 images per second at a depth of 24 with mixed-precision BFloat16. In contrast, Nvidia’s H100-80GB card managed only 595 images per second, and the A100-80GB card achieved 381 images per second. This indicates that the Gaudi 2 chip is 55% faster than the H100 card and nearly 2.43 times faster than the A100 card.
This news is significant for both the AI and semiconductor industries, showcasing Intel’s advancements in chip technology and opening up new possibilities for the development of AI image generation. As technology continues to advance, we can expect to see more high-performance chips applied to the AI field, driving further industry progress.
【来源】https://www.itheat.com/view/45945.html
Views: 1