Designed to accelerate training and inference of large AI models with its expanded memory capacity
Delivers exceptional performance for complex scientific simulations and research applications
Powers NVIDIA DGX A100 systems delivering outstanding AI performance for enterprise applications
Enhanced memory bandwidth and capacity make it ideal for large-scale data analytics applications
Model | Memory | Memory Type | Memory Bandwidth | FP16 TFLOPS | INT8 TOPS | TDP | Price/Hour |
---|---|---|---|---|---|---|---|
NVIDIA H20
H20
|
96GB | HBM3 | 4.0 TB/s | 148 TFLOPS | 296 TOPS | 400W | $0.75/hr |
NVIDIA H100
|
141 GB | HBM3 | 3.35 TB/s | 1979 | 3958 | 700W | $1.80/hr |
NVIDIA L20
L20
|
48GB GDDR6 | HBM3 | 864 GB/s | 59.35 TFLOPS | 239T | 275W | $0.65/hr |
NVIDIA H200
Latest
|
141 GB | HBM3E | 4.8 TB/s | 2000 | 4000 | 700W | $2.50/hr |
NVIDIA A6000
|
48GB | GDDR6 | 768 GB/s | 75.6 TFLOPS | 38.7TFLOPS | 300W | $1.19/hr |
NVIDIA A100
|
80 GB | HBM2E | 1555GB/s | 624 | 1248 | 400W | $0.9/hr |
Model | Memory | Memory Type | Memory Bandwidth | CUDA Cores | Boost Clock | TDP | Price/Hour |
---|---|---|---|---|---|---|---|
RTX 5090
RTX 5090
|
32GB | Other | 1792 GB/s | 18874 | 2.7 GHz | 575W | $1/hr |
RTX 4090
RTX 4090
|
24GB | GDDR6X | 1008 GB/s | 16384 | 2.52 GHz | 450W | $0.35/hr |
RTX 3090
RTX 3090
|
24 GB | GDDR6X | 1TB/s | 10752 | 2 GHz | 350W | $0.25/hr |