AMD MI300X
PulseAugur coverage of AMD MI300X — every cluster mentioning AMD MI300X across labs, papers, and developer communities, ranked by signal.
1 day(s) with sentiment data
-
Indian languages cost 4.4x more tokens than English for LLMs
A recent analysis reveals that tokenizing Indian languages for large language models can be significantly more expensive than English, with some languages costing over four times as many tokens per sentence. This increa…
-
AMD's AITER kernels slow Gemma 4 performance in vLLM
A technical guide details an experiment testing AMD's AITER kernels within the vLLM framework for serving the Gemma 4 12B model on an AMD MI300X GPU. The results indicated that enabling AITER, which provides tuned GPU k…
-
Gemma 4 models: fp8 format overtakes bf16 for larger sizes on AMD MI300X
A technical guide demonstrates how to serve Google's Gemma 4 models of various sizes, from 2B to 31B parameters, on an AMD MI300X GPU using vLLM. The analysis reveals that for smaller models (2B parameters), bfloat16 (b…
-
FP8 Quantization Cost Savings Misleading Due to Model Output Issues
A developer discovered that using FP8 quantization with vLLM on AMD MI300X GPUs, specifically for the Qwen2.5 model, led to a significant 47% reduction in GPU costs. However, this cost saving was misleading, as the mode…
-
New tools and research tackle GPU optimization for AI workloads
Several research papers and a new open-source tool address challenges in optimizing AI workloads on GPUs. COMPASS-ABS aims to reduce fragmentation in shared GPU clusters for deep learning training, improving resource ut…
-
Top GPU Neoclouds 2026: CoreWeave, Nebius, Lambda, Crusoe, Groq Ranked
A 2026 ranking of GPU neocloud providers highlights CoreWeave as the sole Platinum-rated option, commanding premium pricing for high-end NVIDIA GPUs. Nebius leads in publishing on-demand pricing for the latest B300 chip…
-
Kog's new engine boosts AI inference speed on existing GPUs · 4 sources tracked
French startup Kog has developed a new inference engine designed to significantly accelerate AI model performance on existing datacenter GPUs, such as the AMD MI300X and NVIDIA H200. The company's software-driven approa…
-
Developer runs 122B Qwen LLM locally on AMD MI300X GPU
A developer has successfully deployed the Qwen3.5-122B large language model locally on AMD MI300X GPU infrastructure. This setup utilizes vLLM and SGLang for inference, enabling an autonomous agent built with OpenClaw. …
-
MoonMath AI open-sources HIP attention kernel for AMD MI300X, beating AITER v3
MoonMath AI has open-sourced a new bf16 forward attention kernel for AMD's MI300X GPU, written in HIP. This kernel reportedly outperforms AMD's own AITER v3 across various configurations, achieving up to a 1.26x speedup…
-
MoonMath AI open-sources AMD MI300X attention kernel outperforming AITER v3 · 3 sources tracked
MoonMath AI has released an open-source HIP attention kernel for AMD's MI300X GPU, which reportedly outperforms AMD's own AITER v3. The kernel achieves speedups of up to 1.26x by optimizing memory placement and using on…
-
Qwen3 32B fine-tuning fails on AMD MI300X
A fine-tuning attempt of the Qwen3 32B model on AMD MI300X hardware encountered significant issues, leading to wasted resources and a lack of learning. The process reportedly consumed $10 in GPU credits before it was re…
-
Kog AI achieves 3,000 tokens/s LLM inference on standard GPUs
Kog AI has launched a tech preview of its Kog Inference Engine (KIE), demonstrating significantly faster real-time LLM inference speeds on standard datacenter GPUs. The engine achieves 3,000 output tokens per second on …
-
Triton MoE kernel achieves high performance on AMD, NVIDIA
A new fused Mixture-of-Experts (MoE) dispatch kernel, written entirely in Triton, achieves 89-131% of the performance of Stanford's Megablocks library. This kernel notably runs on AMD MI300X hardware without any code mo…
-
MachinaCheck uses specialized agents to ensure manufactured parts are machinable
MachinaCheck is a novel multi-agent system designed to bridge the gap between CAD design and CNC manufacturing, ensuring parts are machinable before production. This system utilizes specialized agents that parse geometr…
-
Zyphra's TSP strategy boosts LLM training throughput by 2.6x
Zyphra has developed a new technique called Tensor and Sequence Parallelism (TSP) designed to optimize the training and inference of large transformer models. This hardware-aware strategy combines aspects of Tensor Para…
-
MachinaCheck automates CNC manufacturability analysis using on-premise AI
A new system called MachinaCheck has been developed to automate the manufacturability assessment of CNC parts, reducing the process from an hour to 30 seconds. This multi-agent AI system leverages the Qwen 2.5 7B Instru…