The Memory Wall: The Invisible Cap on Every LLM
Same prompt, same model, same GPU. One returns in half a second. The other takes twelve. The reason isn't more compute. The model doesn't "think harder" — its weights are frozen at inference time. The reason is bandwidth. Every token in your prompt has a Key and a Value vector cached in GPU memory. About 16 KB per token, per layer. Multiply by ~96 layers and you're at 1.5 MB per token. A 100K-token prompt: 150 GB sitting in memory. An H100 has 80 GB. Your cache is twice the slot size. But even when it fits, attention has to read every prior K/V from memory on every new token. That's bandwidth-bound, not compute-bound. The H100 does 989 TFLOPS but only 3.35 TB/s of memory bandwidth. Compute keeps doubling. Bandwidth doesn't. Wulf and McKee called this "the memory wall" in 1995 — three decades before transformers ran straight into it. The field has been fighting back: Flash Attention keeps K/V in SRAM. Sliding window goes O(N) instead of O(N²). KV quantization stores at 4 bits instead of 16. Multi-query attention shares K/V across heads — ~96× less cache. Every long-context model uses all four. Simultaneously. The wall is bandwidth, not compute.

The Strange Math That Predicts (Almost) Anything

I Tested the Cheapest Path to 96GB of VRAM

the true reason C++ always wins

Why DeepSeek V4 Has Everyone Freaking Out

KV Cache: The Invisible Trick Behind Every LLM

99% Of Your Kernel Modules Never Load. Your Distro Packs Them Anyway

Yann LeCun's $1B Bet Against LLMs

Is RAG Still Needed? Choosing the Best Approach for LLMs

What I Learned From Implementing LLM Architectures From Scratch (And How to Get Started)

An Insanely Elegant LLM Architecture Breakthrough Just Dropped

DeepSeek Gave LLMs a Real Memory (It's Not RAG)

Expensive RTX 5090 for LLMs? NO. Use This Instead. (SXM2 + Z8 G4, #RACERRRZ)

Mastering LLM Inference Optimization From Theory to Cost Effective Deployment: Mark Moyou

Llama.cpp Just Merged MTP And You Should Be Using It.

Why Inference is hard..

Train Your Brain to Never Forget (5 Feynman Habits)

This Local LLM Looked Smart Until I Saw What It Made Up

They solved AI’s memory problem!

Yann LeCun Says LLMs Have 2 Years Left…

