Local LLM Hardware

Local LLM on NVIDIA A100 80GB

80GB VRAM · HBM2e · 2039 GB/s memory bandwidth · 400W TDP · released 2020

At a glance

VRAM80 GB
MemoryHBM2e
Memory bandwidth2039 GB/s
TDP400 W
Released2020
2026 price (used / new)~$15000
Tierdata-center
Best forproduction serving, 80GB fits most models comfortably

What models fit on NVIDIA A100 80GB?

Memory bandwidth is the bottleneck for LLM token generation. The NVIDIA A100 80GB's 2039 GB/s bandwidth gives roughly the throughput below at Q4_K_M quantization. Numbers are conservative and assume a single request (batch=1), 2048-token context.

ModelQuantizationFits?Approx tokens/sec
Llama 3.2 1BQ4_K_MYes (0.8 GB)~367 tok/s
Llama 3.1 8BQ4_K_MYes (~5 GB)~122
Llama 3.1 8BQ8_0Yes (~8 GB)~92
Qwen 2.5 14BQ4_K_MYes (~9 GB)~71
Mistral Nemo 12BQ4_K_MYes (~8 GB)~82
Qwen 3 32BQ4_K_MYes (~20 GB)~37
DeepSeek R1 distilled 32BQ4_K_MYes (~20 GB)~37
Llama 3.1 70BQ4_K_MYes (~42 GB)~20
Llama 4 Scout 109B (MoE)Q4_K_MYes (~64 GB)~16

Token estimates use a simple bandwidth heuristic (~tokens/sec ≈ bandwidth_GB/s × model_efficiency). Real numbers vary with model architecture, batch size, and KV-cache size. Treat as ballpark.

Build recommendations for NVIDIA A100 80GB

Best model to download first

Llama 3.1 8B for everyday chat and coding; Qwen 3 32B for top quality on 24GB. Start with ollama pull llama3.1:8b.

Recommended inference backend

Use Ollama for general use or Mullama if you need a drop-in Ollama alternative with native bindings for 6 languages. For production serving on multi-GPU setups, see vLLM.

Other GPUs to consider

Sources

[1] nvidia.com/a100