NVIDIA

A100 80GB

Practical deployment guidance for this machine, using conservative fit estimates instead of marketing-style claims.

Memory
80 GB VRAM
System RAM
256 GB
Bandwidth
1935 GB/s
Preferred backend
vLLM

Hardware notes

  • Server-class accelerator with enough headroom for serious 70B deployments.
  • Designed for higher concurrency and more stable long-context serving.

Comfortable fits

Llama 3.1 8B Instruct

comfortable

Q4 at 8k context is estimated at 6.44 GB total and 90-150 tok/s.

Phi-4

comfortable

Q4 at 8k context is estimated at 10.86 GB total and 90-150 tok/s.

Qwen2.5 Coder 14B Instruct

comfortable

Q4 at 8k context is estimated at 11.3 GB total and 90-150 tok/s.

Gemma 3 27B IT

comfortable

Q4 at 8k context is estimated at 20.6 GB total and 73-123 tok/s.

Borderline fits

This machine either fits the small set comfortably or falls straight into no-fit territory.

Backend support

These are the runtimes currently associated with this hardware profile.

vLLMllama.cpp

Use the live calculator

The calculator lets you change context length, runtime, quantization, and concurrency instead of relying on a fixed profile.

Open calculator