H100 PCIe vs SXM vs NVL: Which H100 GPU Is Fastest and Most Cost-Effective for Fine-Tuning LLMs?

XiongjieDai/GPU-Benchmarks-on-LLM-Inference

Multiple NVIDIA GPUs or Apple Silicon for Large Language Model Inference?

**An Edge-First Generalized LLM LoRA Fine-Tuning Framework for Heterogeneous GPUs**

A Blog post by QVAC on Hugging Face

Benchmarking Tesla GPUs - esologic

Decommissioned enterprise tesla GPUs are widely available and extremely cheap. In this post, I benchmark tesla GPUs to see if they are useful

Asus ROG Flow X13 GV302X 13.3" Ryzen 9 7940hs 4.0ghz 16GB RAM 512gb SSD

Discover the blend of performance, portability, and versatility with the ASUS ROG Flow X13. This notebook is designed to meet the demands of gaming enthusiasts and creative professionals. At its core is the powerful AMD Ryzen 9 7940HS processor, complemented by the NVIDIA GeForce RTX 4070 graphics card, ensuring smooth gameplay and efficient multitasking.

InferenceMAX™: Open Source Inference Benchmarking

NVIDIA GB200 NVL72, AMD MI355X, Throughput Token per GPU, Latency Tok/s/user, Perf per Dollar, Tokens per Provisioned Megawatt, DeepSeek R1 670B, GPTOSS 120B, Llama3 70B

leobeeson/llm_benchmarks

A collection of benchmarks and datasets for evaluating LLM.

qualcomm/nexa-sdk

Run frontier LLMs and VLMs with day-0 model support across GPU, NPU, and CPU, with comprehensive runtime coverage for PC (Python/C++), mobile (Android & iOS), and Linux/IoT (Arm64 & x86 Docker). Supporting OpenAI GPT-OSS, IBM Granite-4, Qwen-3-VL, Gemma-3n, Ministral-3, and more.

NVIDIA DGX Station for Windows Puts a Trillion-Parameter AI Supercomputer on Every Enterprise Desk

News Summary: NVIDIA announces DGX Station for Windows — the world’s most powerful deskside AI supercomputer for developing and running agents on Windows — built on the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip, coming in Q4 this year. DGX Station brings frontier AI agents to Windows — enabling enterprise developers, researchers, engineers, designers and data scientists to build and deploy AI across the workflows and applications their business runs on. DGX Station will support NVIDIA OpenShell on Windows, built on new Windows security and containment primitives. TAIPEI, Taiwan, June 01, 2026 (GLOBE NEWSWIRE) - NVIDIA GTC Taipei - NVIDIA today announced NVIDIA DGX Station™ for Windows , the world’s most powerful deskside AI supercomputer designed to build, run and connect always-on AI agents to Windows applications and workflows, capable of running frontier AI models of up to 1 trillion parameters locally. Historically, heavy-duty enterprise AI workloads —

Why compute might get 10x more expensive in coming years

If a human-level software engineer that could run on an H100 equivalent, at current market rates for software engineers, that H100 should rent for over $250k a year. That’s 15x today’s spot price.

GPU and TPU Comparative Analysis Report

Introduction

Modern GPU Programming For MLSys — Modern GPU Programming For MLSys

Machine learning systems sit at the heart of modern AI workloads. In these systems, performance often comes down to the quality of a small number of GPU kernels. Attention kernels, LLM prefill and decode kernels, low-precision block-scaled GEMMs, fused MoE layers, and other large fused kernels all directly shape end-to-end speed in both training and serving.

Modern GPU Programming For MLSys — Modern GPU Programming For MLSys

Machine learning systems sit at the heart of modern AI workloads. In these systems, performance often comes down to the quality of a small number of GPU kernels. Attention kernels, LLM prefill and decode kernels, low-precision block-scaled GEMMs, fused MoE layers, and other large fused kernels all directly shape end-to-end speed in both training and serving.

vik on Twitter / X

Photon, our inference engine, isn't fast just because of GPU kernels. A lot of the speedup comes from engine-level work: request scheduling, prefix caching, image processing, all tuned to keep the GPU saturated. https://t.co/3M7eFcFKo5— vik (@vikhyatk) May 2, 2026

LLMs Can Now Write GPU Kernels That Beat torch.compile - Break AI Scaling Limits in 7 Days

We're now seeing multi-agent systems that take your PyTorch code and produce CUDA or Triton kernels with 2x to 14x speedups over torch.compile(mode='max-autotune-no-cudagraphs'). Not on toy benchmarks. On real models like Llama-3.1-8B, Whisper, and Stable Diffusion. Learn proven techniques to shift the scaling law intercept and achieve 10-50% performance gains.

Benchmarks | EXO

Transparent benchmarks for LLMs tested on real hardware. Coming soon.