Skip to content
Bitpute
ToolsModelsGPUsCloudLearn
Learn · Concepts

Why VRAM matters more than anything else for local AI

Learn

By Bitpute · Published 12 July 2026 · Updated 24 July 2026 · How we estimate · Sources · Editorial policy · Version history · Report an error

On this page

For running LLMs on your own machine, VRAM is the hard wall — it decides what you can run at all, and its bandwidth decides how fast. Raw compute barely enters into it. Here's why.

Concept guideLast reviewed: July 2026

People shopping for an AI GPU instinctively look at teraflops. For inference — actually running a model — that's the wrong number. Two things about your VRAM decide almost everything: its capacity and its bandwidth.

1Capacity decides what you can run

A model has to fit — weights + overhead + KV cache — inside VRAM. If it doesn't, you have three options, all bad: spill part of it to system RAM (a large slowdown), run on CPU (much slower still), or simply can't load it. So capacity is a yes/no gate: a 24 GB card runs models a faster 12 GB card physically cannot, no matter how quick the 12 GB card's cores are.

The uncomfortable truth
For local LLMs, more VRAM usually beats a faster GPU. A slower card with more memory runs bigger, smarter models; a quicker card with less memory just runs the small ones faster.

2Bandwidth decides how fast

Token generation is memory-bandwidth bound: to produce each token, the GPU reads the active weights from VRAM once. So the speed ceiling is simply how fast it can read memory:

Decode ceiling (tok/s)ceiling = memory_bandwidth (GB/s) ÷ active_weight_bytes (GB)

That's why memory bandwidth (GB/s), not core count, is the spec that predicts tokens/second. It also explains why lower quantization is faster — fewer bytes to read per token. (Full derivation in the methodology.)

3What happens when you run out

None are a substitute for fitting in VRAM. That's why sizing first — before you buy or download — saves the most pain. Start with the VRAM calculator.

4The practical VRAM tiers

VRAMWhat it comfortably runs (Q4_K_M)
12 GBUp to ~14B models — great for chat & coding
16 GB14B comfortably, 32B tight
24 GB32B models, or 70B with offload
48 GB70B-class models on a single card

Pick your card by the model size you actually want to run, then check bandwidth for speed — in that order.

Check what fits → Best GPUs for local AI Buy vs rent

Related: what VRAM is.

5Common questions

Why does VRAM matter more than any other GPU spec for LLMs?

Because it is a hard limit. If a model does not fit, it either fails to load or falls back to system RAM at a large speed penalty. Every other specification only matters once the model fits.

What happens when you run out of VRAM?

Either an out-of-memory error, or the runtime offloads layers to system RAM. The second keeps the model running but collapses decode speed, because offloaded layers cross PCIe on every token.

How much VRAM do I actually need?

It depends on parameter count and quantization. Usable capacity is about 95% of nameplate, so a 24 GB card gives roughly 22.8 GB — and KV cache for your context comes out of the same budget.