Hackernews posts about FP8 GEMM

1. 01
   

   newShow HN: InstinctFlash – High-Performance Serving Runtime for Robotics Modelsgithub.com

   ▲ 27guanming071712d💬 4savesaved
2. 02
   

   newShow HN: Halo 3's Guardian, playable in the browser, as my personal websiterunboli.com

   ▲ 29runshouse18d💬 9savesaved
3. 03
   

   newDeepGEMM: clean and efficient FP8 GEMM kernels with fine-grained scalinggithub.com

   ▲ 391mfiguiere586d💬 67savesaved
4. 04
   

   newDeepSeek Open Source DeepGEMM – FP8 GEMM Library(300 lines for 1350+ FP8 TFLOPS)twitter.com

   ▲ 4helloericsf586d💬 1savesaved
5. 05
   

   newFP8 GEMM Optimization on AMD CDNA4 Architecturerocm.blogs.amd.com

   ▲ 4skidrow109ddiscusssavesaved
6. 06
   

   newFP8 GEMM Optimization on AMD CDNA4 Architecturerocm.blogs.amd.com

   ▲ 3skidrow110ddiscusssavesaved
7. 07
   

   newDeep Dive into 4-Wave Interleave FP8 GEMMrocm.blogs.amd.com

   ▲ 3skidrow110ddiscusssavesaved
8. 08
   

   newFP8 GEMM Optimization on AMD CDNA4 Architecturerocm.blogs.amd.com

   ▲ 2skidrow107ddiscusssavesaved
9. 09
   

   newFP8 GEMM Optimization on AMD CDNA4 Architecturerocm.blogs.amd.com

   ▲ 1skidrow108ddiscusssavesaved
10. 10
    

    newFP8 GEMM Optimization on AMD CDNA4 Architecturerocm.blogs.amd.com

    ▲ 1skidrow109ddiscusssavesaved
11. 11
    

    newShow HN: Fine-tuned Llama 3.2 3B to match 70B models for local transcriptsbilawal.net

    ▲ 31phantompeace398d💬 8savesaved
12. 12
    

    newTransformers 2.0: What Ilya and Sam Might Have Missed

    ▲ 5metehan777729d💬 1savesaved
13. 13
    

    newShow HN: NVFP4 on Desktop Blackwell – 122B MoE on a Single RTX PRO 6000 31 tok/s

    ▲ 2jcartu209ddiscusssavesaved
14. 14
    

    newExplosive GEMM: arbitrarily large FP error can be incurred in the GEMM operationgithub.com

    ▲ 1statusfailed287ddiscusssavesaved
15. 15
    

    newOzaki Scheme II: A GEMM-oriented emulation of FP matrix multiplication using INTarxiv.org

    ▲ 2matt_d519ddiscusssavesaved
16. 16
    

    newShow HN: Run open-weight OCR, VLM and vision models behind one APIwww.vlmrun.com

    ▲ 5dr_blueberry30ddiscusssavesaved
17. 17
    

    newShow HN: CLI for image/video to ASCII artgithub.com

    ▲ 1kevinwong141ddiscusssavesaved
18. 18
    

    newShow HN: Doom running in OpenSCAD at 10-20 FPSwww.mikeayles.com

    ▲ 2mikeayles253d💬 1savesaved
19. 19
    

    newShow HN: Lean64 – doom64 style FPS on Lean 4github.com

    ▲ 2threepts82ddiscusssavesaved
20. 20
    

    newShow HN: KiDoom – Running DOOM on PCB Traceswww.mikeayles.com

    ▲ 362mikeayles313d💬 49savesaved
21. 21
    

    newTell HN: OpenAI's Codex CLI is currently free to use

    ▲ 6davidpolberger244d💬 1savesaved
22. 22
    

    newShow HN: Realtime, expressive AI personas that you can video callplayground.keyframelabs.com

    ▲ 4kraddypatties314d💬 2savesaved
23. 23
    

    newShow HN: I spent 2.5 years developing my 3D portfolio website, all under 5 MBshees.dev

    ▲ 3sheesdev63d💬 4savesaved
24. 24
    

    newShow HN: I built a PHP-to-native compiler (written in Rust); now it runs DOOMgithub.com

    ▲ 3nahime184d💬 3savesaved
25. 25
    

    newStera: Open-Source Infra That Turns iPhones into Spatial Data for World Models

    ▲ 2satpal142d💬 1savesaved
26. 26
    

    newShow HN: I built an MCP to help build Cursor for Video Editingsmithery.ai

    ▲ 2misbahsy496ddiscusssavesaved
27. 27
    

    newShow HN: VerityNgn–Open-source AI that fact-checks YouTube videoshotchilianalyticsllc.mintlify.app

    ▲ 1ajjcoppola262ddiscusssavesaved

· edit