Rent NVIDIA A40 GPU by the hour
The NVIDIA A40 is a datacenter GPU built on the Ampere architecture with compute capability 8.6 and 45 GB of VRAM. On Cloudj, it delivers about 29.9 TFLOPS of single-precision performance and about 536 GB/s of memory bandwidth within a 300 W power envelope. You can deploy it with per-hour billing, full SSH access, and ready templates including PyTorch/CUDA, vLLM, Ollama, and ComfyUI.
Specifications
VRAM per GPU
45 GB
Architecture
NVIDIA Ampere
Compute capability
8.6
FP32 compute
≈ 29.9 TFLOPS
Memory bandwidth
≈ 536 GB/s
Power limit
300 W
Values reported by the infrastructure provider for machines currently listed; real performance varies by machine.
Available configurations
Live stock and prices, updated continuously. The final price is confirmed before you order.
Rent nowWhat the A40 is good for
Large Language Model Inference
Run models that require up to 45 GB of memory using pre-configured vLLM or Ollama templates.
Deep Learning Fine-Tuning
Fine-tune neural networks in PyTorch/CUDA with ample VRAM for larger batch sizes and weights.
Generative Image Workflows
Execute complex diffusion pipelines and hold multiple checkpoints in memory simultaneously using ComfyUI.
Datacenter Compute Tasks
Utilize about 29.9 TFLOPS of single-precision compute and about 536 GB/s of memory bandwidth for simulation and rendering workloads.
Is it the right GPU for you?
Choose the A40 when your workload requires 45 GB of VRAM on a datacenter Ampere architecture, which exceeds standard consumer memory capacities. Opt for a smaller consumer card if your models fit into less memory, or move to a higher-bandwidth datacenter GPU if your training jobs are heavily bottlenecked by memory throughput.