Untitled attachment
https://files.mastodon.social/media_attachments/files/114/117/641/758/742/096/original/0d46c5a38e1d72c1.png
Bob
@Bob88488488
This is the craziest part: NVIDIA uses the B200 (2.5× compute and 2.4× bandwidth) with FP4, yet its throughput is only barely better than DeepSeek using the H800 with FP8 (21k vs. 14.8k). That’s insane.
Rafael Dominguez
@somoscode
·
2 mar.
Imagine what these guys could do with the latest chips! 🤦🏻♂️
Bobinas P4G is a social network. It runs on GNU social, version 2.0.1-beta0, available under the GNU Affero General Public License.
All Bobinas P4G content and data are available under the Creative Commons Attribution 3.0 license.