Hackernews posts about FP64

  1. Bolt Graphics Targets FP64 HPC Workloads with Zeus GPU (www.hpcwire.com)
    1 points by rbanffy 5 days ago | discuss
  2. 4-bit floating point FP4 (www.johndcook.com)
    98 points by chmaynard 18 days ago | 79 comments
  3. Show HN: realistic_blas: Exact infinite-precision LA, useful errs, f64 fast path (github.com)
    2 points by timschmidt 10 days ago | discuss
  4. Sol-RL: FP4 Explore, BF16 Train (nvlabs.github.io)
    2 points by gmays 24 days ago | discuss
  5. Nvidia's moat is not what it used to be
    6 points by florianleibert 24 days ago | 11 comments
  6. 15 years of FP64 segmentation, and why the Blackwell Ultra breaks the pattern (nicolasdickenmann.com)
    202 points by fp64enjoyer 3 months ago | 80 comments
  7. Nvidia Blackwell Ultra (GB300): -97% INT8/FP64,+50% FP4 Dense,+55% VRAM,+114% At (resources.nvidia.com)
    8 points by Sweepi 8 months ago | 3 comments
  8. AMD Hints at Big FP64 Increases in MI430X GPU as Ozaki Underwhelms (www.hpcwire.com)
    5 points by rbanffy about 2 months ago | discuss
  9. FP64 Floating-Point Emulation in INT8 (arxiv.org)
    3 points by matt_d 7 months ago | discuss
  10. SC25: Estimating AMD's Upcoming MI430X's FP64 and the Discovery Supercomputer (chipsandcheese.com)
    2 points by rbanffy 5 months ago | discuss
  11. SC25: Estimating AMD's Upcoming MI430X's FP64 and the Discovery Supercomputer (chipsandcheese.com)
    2 points by rbanffy 5 months ago | discuss
  12. SC25: Estimating AMD's Upcoming MI430X's FP64 and the Discovery Supercomputer (chipsandcheese.com)
    2 points by rbanffy 5 months ago | discuss
  13. Show HN: 1B Embeddings
    4 points by INVARIAN about 1 month ago | 2 comments
  14. Show HN: HyperAI GPU Leaderboard – A benchmark comparison site for AI workloads (hyper.ai)
    2 points by Ada_trying 3 months ago | 2 comments
  15. Research Acceleration Engine: The Arrival of Excellent GPU Rental Service
    1 points by chlor over 1 year ago | discuss
  16. Flux runs 2x faster on 10GB VRAM with Nvidia FP4 (blackforestlabs.ai)
    3 points by anjneymidha over 1 year ago | discuss
  17. Custom FP4 CUDA Kernel – 129 Tflops on DGX Spark with Pre-Quantized Weight Cache (forums.developer.nvidia.com)
    2 points by vkaufmann 2 months ago | 1 comments
  18. Testing Nvidia's FP4: Running 70B LLMs on a Single RTX 5090 with Real Benchmarks (ai.gopubby.com)
    2 points by Aedelon about 2 months ago | discuss
  19. SageAttention3: Microscaling FP4 Attention. 5x Speed up (arxiv.org)
    2 points by 0xjunhao 11 months ago | discuss
  20. Large Language Model Training Using FP4 Quantization (arxiv.org)
    2 points by t55 over 1 year ago | discuss
  21. Tflops Gap: Why FP4 Moe Kernel Engineering Matters on Blackwell (huggingface.co)
    1 points by apsys 4 months ago | discuss
  22. Huawei unveils Atlas 950 SuperCluster – promises 1 ZettaFLOPS FP4 performance (www.tomshardware.com)
    1 points by buyucu 8 months ago | discuss
  23. Ask HN: If types are good why shouldn't every var be a custom type?
    2 points by jjallen over 1 year ago | 3 comments
  24. C++ ranges/views vs. Rust iterator
    2 points by bijan7 8 months ago | 1 comments
  25. Show HN: NVFP4 on Desktop Blackwell – 122B MoE on a Single RTX PRO 6000 31 tok/s
    2 points by jcartu about 2 months ago | discuss
  26. Show HN: Fin-primitives Zero-panic, decimal-precise trading types for Rust (crates.io)
    1 points by Shmungus about 2 months ago | discuss
  27. Student loans in default to be referred to debt collection, Dept of Ed says (apnews.com)
    48 points by mystraline about 1 year ago | 91 comments
  28. Student loans in default to be referred to debt collection (apnews.com)
    14 points by everybodyknows about 1 year ago | 2 comments
  29. One Nuclear War Can Ruin the Whole Climate (www.wsj.com)
    10 points by Bostonian 12 months ago | 3 comments
  30. Making Bubbles (justinsecurity.medium.com)
    9 points by mooreds almost 2 years ago | 2 comments
  31. Maximizing Density and Efficiency with Fermyon and GKE (www.fermyon.com)
    9 points by bacongobbler about 1 year ago | discuss
  32. Okta to Cut 3% of Workforce to Reallocate Resources Toward Growth Priorities (www.wsj.com)
    6 points by durron over 1 year ago | 1 comments
  33. Inside the AI labs training China's humanoid robots (www.ft.com)
    4 points by hhs about 2 months ago | discuss
  34. Anthropic and IBM Partner in Bid for AI Business Customers (www.wsj.com)
    3 points by doener 7 months ago | 1 comments
  35. Global Preview Event Presentation – Microsoft Flight Simulator 2024 [video] (www.youtube.com)
    3 points by amichail over 1 year ago | discuss