{"slug": "nvidia-s-opt-in-mxfp8-kernel-hits-2-5x-in-a-kimi-k3-microbenchmark-on-one-vr200", "title": "NVIDIA's opt-in MXFP8 kernel hits 2.5x in a Kimi K3 microbenchmark on one VR200", "summary": "NVIDIA merged an opt-in TensorRT-LLM path on November 16 that runs eligible Kimi K3 shared experts with packed MXFP8 weights on SM107 hardware, with contributor-reported gains reaching 2.5x in an eager prefill microbenchmark on one VR200 GPU. The merged pull request benchmarks a single shared-expert layer, and NVIDIA says end-to-end latency and accuracy were not measured; the 2.5x figure is the best eager prefill result, at 8,192 tokens.", "body_md": "# NVIDIA's opt-in MXFP8 kernel hits 2.5x in a Kimi K3 microbenchmark on one VR200\n\n**The TensorRT-LLM change is off by default and targets replicated shared experts on SM107.**\n\n        By [RuntimeWire Staff](https://runtimewire.com/author/runtimewire-staff)\n        · Published \n\nNVIDIA merged an opt-in TensorRT-LLM path on November 16th that runs eligible Kimi K3 shared experts with packed MXFP8 weights on SM107 hardware, with contributor-reported gains reaching 2.5x in an eager prefill microbenchmark on one VR200 GPU.\n\nThe merged pull request benchmarks a single shared-expert layer. NVIDIA says end-to-end latency and accuracy were not measured. The 2.5x figure is the best eager prefill result, at 8,192 tokens.\n\nWhat the kernel changes…", "url": "https://wpnews.pro/news/nvidia-s-opt-in-mxfp8-kernel-hits-2-5x-in-a-kimi-k3-microbenchmark-on-one-vr200", "canonical_source": "https://runtimewire.com/article/nvidia-s-opt-in-mxfp8-kernel-hits-2-5x-in-a-kimi-k3-microbenchmark-on-one-vr200", "published_at": "2026-11-17 18:59:34+00:00", "updated_at": "2026-11-17 20:08:34.264377+00:00", "lang": "en", "topics": ["artificial-intelligence", "large-language-models", "ai-infrastructure", "ai-tools"], "entities": ["NVIDIA", "TensorRT-LLM", "Kimi K3", "VR200", "SM107", "MXFP8"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/nvidia-s-opt-in-mxfp8-kernel-hits-2-5x-in-a-kimi-k3-microbenchmark-on-one-vr200", "markdown": "https://wpnews.pro/news/nvidia-s-opt-in-mxfp8-kernel-hits-2-5x-in-a-kimi-k3-microbenchmark-on-one-vr200.md", "text": "https://wpnews.pro/news/nvidia-s-opt-in-mxfp8-kernel-hits-2-5x-in-a-kimi-k3-microbenchmark-on-one-vr200.txt", "jsonld": "https://wpnews.pro/news/nvidia-s-opt-in-mxfp8-kernel-hits-2-5x-in-a-kimi-k3-microbenchmark-on-one-vr200.jsonld"}}