proomt

Search

Search posts, papers, and topics

All posts

Hacker News front pageFaizan A Khattak, Mantas Mikaitis, Carlo J. Graziani2 min readpaperadvanced

Accurate Models of AMD Matrix Cores

Read on arxiv.orgHN 79·11 comments

Summary

The authors reverse‑engineer AMD’s CDNA matrix cores, build per‑architecture MATLAB models that match hardware bit‑for‑bit on 10 M random tests, and use them to compare AMD vs NVIDIA tensor‑core accuracy.

  • AMD matrix units differ in accumulator width, rounding, subnormal handling, etc., and are undocumented.
  • Custom test vectors expose each numerical feature; iterative random testing refines a software model until it reproduces hardware results exactly.
  • Validated models achieve bit‑level reproducibility across MI100, MI210/250, and MI300A/300X GPUs.
  • Demo applications show measurable accuracy gaps between AMD matrix cores and NVIDIA tensor cores.

Accurate software models let developers predict numerical error, write portable kernels, and benchmark hardware without needing proprietary documentation—critical for high‑performance ML and scientific code.

7/10
architectureperformancegpuhardware architecturefloating pointmatrix multiplication

Related reading

  1. Benchmarking Wild vs Mold

    Reproduces Mold’s linker benchmarks on a 16‑core Ryzen, shows that configuration (filesystem, delete‑output, fork) explains most of the Wild vs Mold speed gap, and notes recent Mold releases and upcoming Wild tweaks that close the gap.

    Lobstersgithub.io4 minHN461lobste.rs52

  2. NVIDIA Vera Rubin NVL72 Delivers Leading Performance in MLPerf Inference v6.1 Debut

    NVIDIA’s Vera Rubin NVL72 AI inference system shows up to 3.7× higher throughput than the prior GB300 NVL72 on MLPerf v6.1 benchmarks (Qwen3‑VL, DeepSeek‑R1), achieves 99% scaling efficiency across 288 GPUs, and benefits from software optimizations (NVFP4 precision, kernel fusion, disaggregated serving). The post is a product announcement with concrete benchmark numbers but limited technical dept…

    Nvidianvidia.com4 min

  3. SiliconBench: Speed, Memory, and Fidelity for LLM Serving on Unified-Memory Desktops

    SiliconBench benchmarks nine Apple‑Silicon LLM serving engines on unified‑memory desktops, measuring speed, memory usage, and output fidelity across Qwen3, Qwen3.5, and Gemma‑4 models. It finds vllm‑metal doubles throughput at modest concurrency, memory budgets often fail to preserve headroom, and only three stacks satisfy all fidelity and model‑coverage requirements, with tensor‑parallel scaling…

    Hugging Face Daily Papersarxiv.org1 minpaper

  5. Sample Count Is Not Enough: Candidate-Generation Strategy Shapes the Energy and Performance of LLM Test-Time Scaling

    Increasing the number of LLM candidates (N) improves reasoning accuracy, but the way those candidates are generated (batch size vs sequential calls) dramatically affects latency, GPU‑hours, and energy. On A100 GPUs, eight serial 1‑candidate calls consume ~5× more energy and take ~6× longer than a single 8‑candidate batched call, even though total candidate count is identical. The authors recommen…

    Hugging Face Daily Papersarxiv.org2 minpaper

  6. Article: Architecting Secure and Scalable Facial Verification Systems

    A real‑world post‑mortem of a high‑volume face verification service that moved from a naïve synchronous API to an async, layered pipeline (edge validation, preprocessing, decoupled detection/verification, decision engine) to achieve 8.5k rpm, p99 < 1.8 s, 30 % cost savings, and strict privacy controls.

    InfoQinfoq.com15 min