精选理由
NVIDIA 的 Blackwell Ultra 训练 DeepSeek-V3 671B 比上一代快 3 倍，每 GPU 跑到 1648 TFLOPS，优化太狠了。
NVIDIA Blackwell Ultra 在预训练 671B 参数的 DeepSeek-V3 模型上达到每 GPU 1648 TFLOPS 的性能。这一数字约为上一代 GPU 性能的 3 倍。该成果通过 Megatron-Core、TorchTitan 和 JAX 等框架的协同设计与持续软件优化实现。Blackwell Ultra 在该基准测试中刷新了世界纪录。
原文 · NVIDIA AI
NVIDIA Blackwell Ultra just set a world record for pre-training DeepSeek-V3 671B, achieving 1,648 TF...
NVIDIA Blackwell Ultra just set a world record for pre-training DeepSeek-V3 671B, achieving 1,648 TFLOPs per GPU. This is about 3x the earlier delivered performance of the previous generation, and is a result of our extreme co-design and continuous software optimization across popular frameworks including Megatron-Core, TorchTitan, and JAX. 💬 16 🔄 22 ❤️ 165 👀 7653 📊 36 ⚡