Skip to content

DEV Community

Ruitong

Measuring cross-accelerator LLM equivalence. Comparing NVIDIA, AMD, and Ascend hardware.

Beijing, China Joined on Aug 15, 2026 https://ruitong.io

Ruitong

Aug 16

Ascend vs CUDA: Same Qwen3-8B Model, Same Seed, 15% Token Agreement

#ai #deeplearning #llm #machinelearning

1 min read

Ruitong

Aug 15

The Same LLM, the Same Weights, the Same Prompt — Different Hardware Gives Different Answers

#llm #inference #hardware #gpu

3 min read

loading...