https://gowda.ai/tags/cuda/Recent content in CUDA on Thamme GowdaHugoen-usThu, 13 Aug 2026 10:00:00 +0000https://gowda.ai/posts/2026/08/llm-fast-inference/Thu, 13 Aug 2026 10:00:00 +0000https://gowda.ai/posts/2026/08/llm-fast-inference/A practical look at high-quality W4A8 inference: why naive symmetric int4 wastes a code, how Tahoma combines asymmetric weights and MSE clipping with dynamic FP8 activations, and how the resulting runtime compares with vLLM on one H100.