A100 PCIe 40 GB
Local inference speed
llama.cpp decode speed (tok/s) - estimated
| Workload | tok/s |
|---|---|
| LLaMA 3 8B Q4 | 182.6 |
| LLaMA 3 8B F16 | 82.9 |
Estimated - Estimated from memory bandwidth (llama.cpp decode) · updated 2026-11-22
Stacks for the A100 PCIe 40 GB
Complete local setups that run on 40 GB
Run Qwen3.6-27B with vision, 256K context, and hybrid thinking locally
AI pair programming in your terminal with local models, fully private
A private AI assistant on Windows with document RAG - no CLI, no Docker, all desktop GUI
A private AI coding assistant in your editor
Prototype and deploy multi-agent RAG applications visually with local models
A100 PCIe 40 GBLink to this section
The A100 PCIe 40 GB is an NVIDIA GPU built on the Ampere architecture, released 2020-06-22. For running AI locally, the numbers that matter are its 40 GB of HBM2e and 1560 GB/s of memory bandwidth. VRAM decides which models fit at all; bandwidth sets how fast they generate text.
What you can run on 40 GBLink to this section
At Q4_K_M quantization (the usual local default), 40 GB holds models up to roughly 63B parameters, leaving headroom for context. On this card you can run, among others:
- deepseek-v4-flash-0731-spark - 60.3B parameters
- HyperNova 60B 2605 - 58.7B parameters
- Llama-3_3-Nemotron-Super-49B-v1_5 - 49.9B parameters
- G9v3-39A5B - 39B parameters
- K2 Horizon MoVA 36B A4B - 37.4B parameters
Larger models need a higher-VRAM card, a second GPU, or CPU offload (which is much slower). Check any specific model with the VRAM calculator, or see the full picture on what can I run.
Local LLM speed (LLaMA 3, llama.cpp)Link to this section
Single-stream token-generation throughput - estimated from memory bandwidth:
| Model (quant) | Speed on A100 PCIe 40 GB |
|---|---|
| Llama 3 8B (Q4_K_M) | 182.6 tok/s |
| Llama 3 8B (F16) | 82.9 tok/s |
| Llama 3 70B (Q4_K_M) | ✗ won't fit |
Because decode is memory-bandwidth bound, the 1560 GB/s figure is the best single predictor of chat speed on this card. Estimates are calibrated against measured RTX-40-series cards and are typically within ~15%.
Memory and powerLink to this section
- VRAM: 40 GB HBM2e (5120-bit bus)
- Bandwidth: 1560 GB/s
- TDP: 250 W - a 600 W+ power supply is recommended
- Process: 7 nm
- Interface: PCIe 4.0 x16
Quantization and contextLink to this section
Quantization trades a little quality for a lot of VRAM. On 40 GB you can fit roughly a 63B model at Q4_K_M, about a 34B model at the higher-quality Q8, or a smaller model at full FP16. Longer context windows also consume VRAM (the KV cache grows with context length), so leave a few GB of headroom if you plan to use large prompts or many concurrent requests. For most chat and coding use, Q4_K_M on this card is the sweet spot between speed, quality, and the 40 GB budget.
How it comparesLink to this section
Similar cards for local AI, by VRAM and 8B-Q4 speed:
| GPU | VRAM | Bandwidth | Llama 3 8B Q4 |
|---|---|---|---|
| A100 PCIe 40 GB | 40 GB | 1560 GB/s | 182.6 tok/s |
| Jetson AGX Orin 32 GB | 32 GB | 204.8 GB/s | 24 tok/s |
| NVIDIA RTX A6000 | 48 GB | 768 GB/s | 89.9 tok/s |
| NVIDIA L40S | 48 GB | 864 GB/s | 101.1 tok/s |
Bottom lineLink to this section
The A100 PCIe 40 GB is best for llm-inference, image-gen, comfyui. With 24 GB+ it comfortably handles most open models, including 30B-class at Q4. If you need more, compare with Jetson AGX Orin 32 GB and NVIDIA RTX A6000.
SourcesLink to this section
- Specifications: RightNow GPU Database (TechPowerUp data)
- Benchmarks: GPU-Benchmarks-on-LLM-Inference (basis for the bandwidth estimate)
Specs and benchmarks last checked 2026-11-22. Verify current pricing before buying.
Frequently asked
Quick answers to common questions
How much VRAM does the A100 PCIe 40 GB have?
The A100 PCIe 40 GB has 40 GB of VRAM with 1560 GB/s memory bandwidth.
What local AI models can run on the A100 PCIe 40 GB?
The A100 PCIe 40 GB with 40 GB VRAM can run many models depending on quantization. Models up to ~61B params may fit at Q4_K_M. Use our VRAM calculator to check specific models.
Is the A100 PCIe 40 GB good for local AI inference?
A100 PCIe 40 GB is best for llm-inference, image-gen, comfyui. With ample VRAM it handles most open models well.
Where can I buy the A100 PCIe 40 GB?
Check our buy links above for the best current prices on Amazon, Newegg, and B&H. Prices vary by retailer and availability.
How does the A100 PCIe 40 GB compare to other GPUs?
A100 PCIe 40 GB has 40 GB VRAM and 1560 GB/s bandwidth. This puts it in the high-end category, suitable for most open models. Browse our hardware directory for side-by-side comparisons.
What power supply do I need for the A100 PCIe 40 GB?
The A100 PCIe 40 GB has a TDP of 250W. A standard quality PSU of 650W+ should suffice. Always check the manufacturer's recommendations for your specific build.
Nearby options
Similar hardware and models that fit