Event ·Multi-source

HyperAccel Bertha 4nm AI芯片量产

韩国AI芯片初创企业HyperAccel的Bertha 500已在三星4nm制程节点实现量产。该芯片拥有768 TFLOPS的FP8算力，支持多种数据格式，集成256MB片上SRAM缓存。Bertha 500的吞吐量、成本效益和能效分别宣称达到NVIDIA H100的2倍、19倍和12倍。

Current read

HyperAccel的Bertha 500芯片量产，性能对标H100，数据中心推理场景的新选择。

9 sources73° AI heatLast updated 2026/9/9 08:47:06

Evidence

9 sources

Primary sourceIT之家

2026/9/9 08:47:06View source

Related source 1Tech in Asia

2026/9/10 05:10:34View source

Related source 2The Business Times: Tech

2026/9/8 01:59:16View source

Related source 3NVIDIA AI

2026/9/8 11:00:07View source

Related source 4AWS Machine Learning Blog

2026/9/8 16:21:57View source

Related source 5Aravind Srinivas

2026/9/8 17:10:38View source

Related source 6shao__meng

2026/9/8 23:28:19View source

Related source 7Decoder

2026/9/10 12:23:16View source

Related source 8VTV Công nghệ

2026/9/10 20:52:00View source

Conflict check

Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.

Open the story