{"author":"Sweepi","children":[{"author":"Sweepi","children":[],"created_at":"2025-08-28T08:41:30.000Z","created_at_i":1756370490,"id":45049880,"options":[],"parent_id":45049879,"points":null,"story_id":45049879,"text":"[+114% Attention acceleration]\nAny idea how they got +50% FP4 from the same silicon? "Firmware" improvements? \nOr did they found a way to disable the INT8 and FP64 units and re-use them e.g. as overspill registers?\nAny other ideas why INT8/FP64 is down -97% on the same chip? QA/certification issues?

In case you you want to compare the complete specs, I would post them here, but since hn supports less formatting than early 2000s bb-forums, check it here: https://www.forum-3dcenter.org/vbulletin/showpost.php?p=1380...","title":null,"type":"comment","url":null},{"author":"nabla9","children":[{"author":"aurareturn","children":[],"created_at":"2025-08-28T09:25:07.000Z","created_at_i":1756373107,"id":45050149,"options":[],"parent_id":45049993,"points":null,"story_id":45049879,"text":"Certainly comparing Blackwell's FP4 performance to H100 FP16, no?","title":null,"type":"comment","url":null}],"created_at":"2025-08-28T09:01:21.000Z","created_at_i":1756371681,"id":45049993,"options":[],"parent_id":45049879,"points":null,"story_id":45049879,"text":"96% TCO and energy savings for 65 racks eight-way HGX H100 air-cooled versus 1 rack GB200 NLV72 liquid-cooled with equivalent performance on GPT-MoE-1.8T real-time inference throughput.

Big if true. Energy and cooling costs can represent up to 30-40% of the total cost of setting up and running an AI data center.","title":null,"type":"comment","url":null}],"created_at":"2025-08-28T08:41:30.000Z","created_at_i":1756370490,"id":45049879,"options":[],"parent_id":null,"points":8,"story_id":45049879,"text":null,"title":"Nvidia Blackwell Ultra (GB300): -97% INT8/FP64,+50% FP4 Dense,+55% VRAM,+114% At","type":"story","url":"https://resources.nvidia.com/en-us-blackwell-architecture/blackwell-architecture-technical-brief"}