Inferbase
Beta
NVIDIA A100 SXM 40GB
Datacenter
The original A100 introduced the Ampere architecture with 3rd-gen Tensor Cores and 40GB HBM2e memory. Widely deployed across major cloud providers for AI and scientific computing.
VRAM
40 GB
Memory
HBM2e
Bandwidth
1555 GB/s
TDP
400W
Suitable Workloads
Medium Language Models
Inference for models up to 70B parameters
Distributed Training
Multi-node training with fast interconnects
Enterprise Deployment
Designed for 24/7 datacenter operations
Key Highlights
- Built on Ampere architecture
- 432 AI accelerator cores for tensor operations
- 6,912 compute cores for parallel processing
- SXM4 form factor
Memory
VRAM40 GB
Memory TypeHBM2e
Memory Bandwidth1555 GB/s
Compute Performance
FP32 (Single Precision)156 TFLOPS
FP16 (Half Precision)312 TFLOPS
BF16312 TFLOPS
INT8624 TOPS
Architecture
ArchitectureAmpere
Compute Cores6,912
AI Accelerators432
Release Year2020
Power & Physical
TDP400 W
Max Power400 W
Form FactorSXM4
PCIe GenerationGen4
Interconnect
Multi-GPU SupportYes
Interconnect Bandwidth600 GB/s
Pricing
MSRP$10,000
Notes
Original A100
Models That May Fit on NVIDIA A100 SXM 40GB
Estimates based on INT8 quantization. Actual fit depends on framework and batch size.
Related Tools
Added Jan 25, 2026
Last updated: Jan 25, 2026
Build Your AI Stack with Confidence
Explore models, compare pricing and benchmarks, and right-size your infrastructure — all in one place.