ProjectsJuly 29, 2026 · updated August 29, 2026

Hawk Point NPU LLM

Evidence & reproducibility

Integrity 83% · 5/6 evidence signals

GOODVERIFIED

The repository documents CPU/NPU first-token correctness checks, hardware validation tooling, and measured decode experiments without copying mutable benchmark numbers into the portfolio.

Machine-readable checks

✓ Pinned repository commit exists
✓ CPU/NPU correctness path documented
✓ Benchmark conditions retained with implementation

Evidence

CPU/NPU first-token argmax validation
Explicit hardware-validation tooling
Benchmark conditions and measurements retained with the implementation

Environment

accelerator: AMD XDNA1 / AIE2
runtime: XRT + MLIR-AIE / IRON
execution: CPU orchestration with native NPU kernels

Last verified: 2026-08-29Verified commit 12a84482 ↗

ReproductionFollow the repository hardware-validation workflow at the pinned commit.

Experiment ledger

1 recorded experiment

EXP-2026-08-02-NPU-VALIDATION · 2026-08-02
Pinned native NPU validation baseline
PASS
Hypothesis: The documented native path can be evaluated against a CPU reference without hiding host orchestration.Pinned the repository state that documents first-token CPU/NPU correctness validation and hardware-aware decode measurements.
- CPU/NPU first-token argmax validation is documented in the repository.
- Host orchestration remains explicit in the architecture boundary.
Experiment commit 12a84482 ↗

Problem

The project asks how much useful transformer work can actually be moved onto first-generation AMD XDNA NPUs instead of treating the accelerator as a black box.

Target environment

• AMD XDNA1 / Ryzen AI Phoenix and Hawk Point• AIE2 array• MLIR-AIE / IRON• XRT runtime

Architecture

Tokenization and orchestration stay on the CPU. The native path targets the AIE2 array directly, with decoder work such as projections, RMSNorm, RoPE, attention, KV caches, residuals, and SwiGLU implemented as AIE2 kernels.

Inference boundary

Tokenizer + API
CPU orchestration
XRT / IRON
AIE2 decoder kernels
Sampling + stream

Current result

The repository includes terminal chat, selectable checkpoints, hybrid NPU/CPU layer offload, hardware validation tools, and an OpenAI-compatible API with streaming responses.

Validation

On the documented Hawk Point test system, the native path records matching CPU/NPU first-token argmax results and measured decode benchmarks. Exact numbers and test conditions remain in the repository so they can evolve with the implementation without turning this page into a stale benchmark claim.

Limitations

The project deliberately keeps host orchestration visible. A result is only described as NPU-accelerated when the execution path and correctness evidence support that statement.

Stack

Python
C++ AIE2 kernels
MLIR-AIE / IRON
XRT
AMD XDNA1 / Ryzen AI
OpenAI-compatible HTTP API

View the source, measurements, and hardware notes on GitHub.

Related projects

XDNA NPU Miner

Experimental

PythonAIE2MLIR-AIEXRTRyzen AI

★ 0 stars⑂ 0 forks◌ 45 commits◉ 0 issues

C++Updated Aug 12, 2026

An early-stage research project exploring whether suitable cryptocurrency mining kernels can be mapped honestly and measurably onto AMD XDNA1 NPUs.

Read case study View project

Samsung Odyssey G5 Firmware Research

Research

PythonSPARC V88051FirmwareReverse Engineering

★ 17 stars⑂ 0 forks◌ 20 commits◉ 0 issues

PythonUpdated Aug 27, 2026Release v1.0.0

Reverse engineering and reproducible tooling for the Samsung Odyssey G5 G55C firmware, including a minimal patch that exposes an existing hidden factory page.