gpu on Hacker News

157 titles from 125 authors in the last 30 days, with 149 distinct links and 31 Show HN.

Counts titles, it does not confirm facts. Open the source before concluding anything.

Projects on this term

The score comes from our model.

• Show HN: UEFI pre-driver Nvidia GPU unlocker for Windows · score 8.5

  nv-uefi-unlocker is a Rust-based UEFI DXE pre-driver tool that performs a full firmware transaction to unlock NVIDIA GPU performance (e.g., FP32 GEMM +67-75% on RTX 20/30/40 series) before Windows loads. It chains SEC2 scrubbing, signed GSP-RM boot, a V67 four-write payload, readback verification, and FLR, then hands control back to the Windows Boot Manager. The project is remarkable for its deep reverse-engineering of proprietary NVIDIA firmware internals and its reproducible asset pipeline from pinned public inputs.

• Show HN: Algeon – Compose GPU graph, vector, and ML algorithms in SQL · score 8.0

  Algeon is a GPU-accelerated query engine built on Apache DataFusion that exposes NVIDIA cuGraph, cuVS, and cuML algorithms as SQL table functions. It enables composing graph, vector, and ML pipelines directly within SQL statements over Parquet data, with results feeding joins and aggregations in the same query. The project is technically ambitious, integrating GPU execution into a relational engine, but remains experimental with limitations around single-GPU memory and Parquet-only sources.

• Show HN: Capy-flash – a 125B MoE model on two 16 GB consumer GPUs · score 8.0

  Capy-flash is a C++ inference engine that serves a 125B MoE model (Qwen3.8-Flash-Next) in NVFP4 on two 16 GB consumer GPUs by offloading routed experts to CPU RAM and using tensor parallelism over PCIe without peer-to-peer. It includes agent-oriented features like context caching, expert residency tracking, and MTP speculative decoding, with measured performance competitive with larger setups.

• Show HN: Zero-shot classifier on scale-to-zero GPUs, one forward pass per call · score 7.8

  Reflex is a GGUF-native Rust and CUDA inference engine optimized for cold starts, compiling all CUDA kernels ahead of time to achieve ~448ms first result on scale-to-zero GPUs. It targets single-tenant, one-shot inference (classification, extraction) where startup latency dominates cost, and includes a serverless sidecar for OpenAI-compatible HTTP.

Recent titles

1. GPUlayer: Train PyTorch on every GPU you can reach-faster than PyTorch Lightning
   2026-11-01 · gpulayer.dev · HN discussion
2. WiCi: Connecting GPUs over WiFi
   2026-11-01 · wici.ai · HN discussion
3. Speeding up GPU copies by permuting iteration spaces
   2026-10-31 · futhark-lang.org · HN discussion
4. GPUCost – Cloud GPU Prices Normalized by VRAM and LLM VRAM Sizer
   2026-10-31 · gpu-cost-r34p.vercel.app · HN discussion
5. Sysinfo: Fixing performance to retrieve process GPU usage
   2026-10-31 · blog.guillaume-gomez.fr · HN discussion
6. I helped fix sleep-wake hangs on Linux with AMD GPUs (2025)
   2026-10-31 · nyanpasu64.gitlab.io · HN discussion
7. I took my Rust GUI from 135 MB to 30 MB by ditching the GPU
   2026-10-31 · trystan-sarrade.com · HN discussion
8. The N64's GPU Is a Technical Marvel [video]
   2026-10-30 · youtube.com · HN discussion
9. Chinese AI chipmaker Iluvatar CoreX says new GPU matches Nvidia's H200
   2026-10-30 · scmp.com · HN discussion
10. GPUs aren't infrastructure; they're fuel
    2026-10-30 · minhanhday.substack.com · HN discussion
11. Fiber lasts decades, GPUs last three years: the AI boom isn't the dot-com boom [video]
    2026-10-30 · youtube.com · HN discussion
12. Global GPU prices rise 35% in just three months
    2026-10-30 · videocardz.com · HN discussion
13. Dell workstation with 96GB Nvidia GPU may cost less than graphics card alone
    2026-10-30 · videocardz.com · HN discussion
14. Advanced Matrix Multiplication Optimization on Nvidia GPUs
    2026-10-29 · salykova.github.io · HN discussion
15. Show HN: Kinoloom – A local AI film studio for an 8 GB GPU
    2026-10-29 · github.com · HN discussion
16. Grace Blackwell needs a new scheduler: 27% more usable GPUs from the same racks
    2026-10-29 · skypilot.ai · HN discussion
17. GeForce vs. Workstation GPUs: Nvidia's Half Rate, Priced
    2026-10-29 · airealist.ai · HN discussion
18. GPU or PSU coil whine: a test log and 47 owner reports across 44 models
    2026-10-29 · techfuelhq.com · HN discussion
19. GPU database HeavyAI (formerly MapD) is now open source
    2026-10-28 · twitter.com · HN discussion
20. Writing GPU kernels in Rust: how I made speakrs 60× faster than pyannote
    2026-10-28 · avencera.com · HN discussion
21. Supercomputer for sale with 9,600 A100 GPUs and 4,200 EPYC CPUs, free shipping
    2026-10-28 · tomshardware.com · HN discussion
22. Show HN: Pandora M1, a distributed GPU monitor prototype
    2026-10-28 · blog.christianperone.com · HN discussion
23. Show HN: Algeon – Compose GPU graph, vector, and ML algorithms in SQL
    2026-10-28 · github.com · HN discussion
24. GPU vs. CPU. How to Cut Live Streaming and AI Processing Costs?
    2026-10-28 · red5.net · HN discussion
25. Hyperbolic – Open-Access GPU and AI Cloud
    2026-10-28 · hyperbolic.ai · HN discussion
26. Show HN: UEFI pre-driver Nvidia GPU unlocker for Windows
    2026-10-28 · github.com · HN discussion
27. .aamt – one GPU file format for model weights, agent memory and worlds
    2026-10-27 · asamanthinks.com · HN discussion
28. Slughorn – GPU Vector Graphics
    2026-10-27 · slughorn.io · HN discussion
29. Show HN: SSH zydra.sh, a GPU Linux desktop in your terminal, with Doom
    2026-10-27 · zydra.sh · HN discussion
30. Show HN: Llmrun – which open LLMs run on your GPU, and how good they are
    2026-10-27 · llmrun.dev · HN discussion

All trends