
Rewriting My Triton FlashAttention Kernel in CuTe DSL, One Optimisation at a Time
Table of Contents What Triton did under the hood The setup CuTe in four ideas Round 0: a...
Tag archive

Table of Contents What Triton did under the hood The setup CuTe in four ideas Round 0: a...
Jensen Huang’s rumored return to Nvidia’s CEO chair isn’t just a leadership story—it’s a catalyst for re‑examining the GPU compute stack that powers modern AI. This editorial dissects Nvidia’s hardware‑software contract, surfaces common engineering missteps, and offers a concrete CUDA‑Python blueprint for future‑proofing your workloads.
My laptop's RTX 3050 now generates 235 tokens per second with TinyLlama-1.1B, using an inference...
Read what the OOM error says, then cut the biggest consumer: activations, optimizer state or fragmentation. An ordered checklist with GB math and code.
Your CUDA Code Didn't Change. Your Risk Register Just Did: What Nvidia's CEO Exit Means for...
The 8GB VRAM Breakthrough: How PagedAttention and Dynamic KV-Cache Quantization Unlock 64k...
A local model feels quick when one person is chatting. What happens when four requests arrive...
🎯 The Setup I've got a single RTX 4080 (16GB VRAM), 64GB system RAM, and a stubborn desire...
Over the past couple of weeks, upstream researchers showed that consumer GeForce RTX 40, 30, and 20...
A longer prompt arrives. The model still fits, but the wait grows—and every new token has more...

Yesterday, Jensen Huang resigned as Nvidia's CEO, effective immediately. The stock fell as much as...
Introduction While setting up CUDA for an RTX 5060 GPU on Windows 11 + WSL2, I ran into a...