Tag

#amd

48 posts2 followers

Explore Hashnode

Alternatives

Shraddha bhatOct 17 · 9 min read

The $8.2 Billion Bet on the Third Dimension: How AMD Just Changed the AI Race Forever

Imagine walking into a room you’ve never seen before and instantly knowing where the exit is, how far the table is from the wall, and what it would feel like to pick up the coffee mug on the counter.

aiinsights1909.hashnode.dev

Servers99Oct 16 · 8 min read

KNOD Explained: Offloading Linux Network Traffic to AMD GPUs

Modern data centers are routinely pushing link speeds of 100Gbps to 400Gbps, creating a massive bottleneck for server infrastructure. While network interfaces can ingest data at blistering speeds, pro

servers99.hashnode.dev

Gayatri PanseOct 13 · 23 min read

Epoch 1: Try It, Time It, Then Buy It

TL;DR vLLM vs llama.cpp: on a 20–24 GB card, vLLM fit 1.83 users at 32K context. llama.cpp fit ~13, because it needs 4× less KV-cache memory for Gemma 4. On AMD, four concurrent requests is the swee

gradientdecent.hashnode.dev

MIG serversOct 8 · 5 min read

AMD EPYC 9006 "Venice" CPUs: Architecting the Agentic AI Stack

Modern AI infrastructure has rapidly transitioned from static, monolithic training and inference clusters to highly dynamic Agentic AI pipelines. Unlike traditional models that rely on predictable com

migservers.hashnode.dev

ThrottleOct 6 · 4 min read

The FP8 trap: my GPU bill dropped 47% because the model was printing "!!!!!!"

I had one hour on an AMD MI300X and one question: what does a token actually cost on it? One GPU, vLLM's ROCm build, Qwen2.5 at 7B, 32B and 72B. Thirty-two requests in flight, 256 output tokens max, p

throttle.hashnode.dev

Vedant JadhavOct 3 · 7 min read

What 57 LLM Benchmark Runs on an AMD Instinct MI300X Taught Me About VRAM, Throughput, and Quantization

How much VRAM does a model actually need? How does model size affect generation speed? Does quantization always reduce memory? Does lower precision always make inference faster? I wanted to answer the

3 vedantjadhav.hashnode.dev

Josh GreenAug 1 · 5 min read

A 30B Model Crawled on My Mini PC, and the Weights Were Innocent

There is a comfortable assumption in the local LLM world: if a model only activates a few billion parameters per token, it will run fast on modest hardware. Sparse mixture-of-experts models are sold o

1 joshgreen.hashnode.dev

Manu ShuklaJul 28 · 14 min read

ROCm vs CUDA in 2026: can enterprises actually escape CUDA lock-in?

ROCm vs CUDA in 2026: can enterprises actually escape CUDA lock-in? Summary. CUDA has been NVIDIA's real moat for 18 years, and 2026 is the first year that moat has visible cracks. AMD's ROCm 7, relea

ecorpit.hashnode.dev

DustinVKJul 12 · 19 min read

I Gave My Coding Agent a Free tier

There is a TL;DR section near the bottom if you just want the summary of setting it all up minus the story. Why host a model? The world of AI coding is moving at lightspeed in 2026. Every week I learn

undefinedbehavior.hashnode.dev

Brian KingJul 9 · 9 min read

Installing ROCm & Llama.cpp for CPUs & APUs.

Abstract. I document my complete workflow for installing AMD ROCm drivers on Ubuntu Desktop 24.04 LTS and building the llama.cpp library with HIP (GPU) backend support for my AMD Strix Halo APU with i

solodev.app

Search Hashnode