Compare GPUs

Compare 2-4 public GPU options using the same database-backed price, inventory and specification records.

Select up to four GPUs. Comparison URLs are noindex to avoid duplicate search pages.

GPU NVIDIA H200 SXM 141GB GPU accelerator

NVIDIA H200 SXM 141GB

 NVIDIA H100 SXM 80GB GPU accelerator

NVIDIA H100 SXM 80GB
Monthly rentalUSD 2,345/monthUSD 1,860/month
Hourly equivalent~ USD 3.21/hour~ USD 2.55/hour
VRAM141 GB HBM3e80 GB HBM3
ArchitectureHopperHopper
Bandwidth4,800 GB/s3,350 GB/s
Form factorSXM moduleSXM module
InterfacePCIe Gen5 x16 host interfacePCIe Gen5 x16 host interface
TDP700 W700 W
FP3267 TFLOPS dense67 TFLOPS dense
TF32494.5 TFLOPS dense / 989 TFLOPS sparse494.5 TFLOPS dense / 989 TFLOPS sparse
BF16989.5 TFLOPS dense / 1979 TFLOPS sparse989.5 TFLOPS dense / 1979 TFLOPS sparse
FP16989.5 TFLOPS dense / 1979 TFLOPS sparse989.5 TFLOPS dense / 1979 TFLOPS sparse
FP81979 TFLOPS dense / 3958 TFLOPS sparse1979 TFLOPS dense / 3958 TFLOPS sparse
INT81979 TOPS dense / 3958 TOPS sparse1979 TOPS dense / 3958 TOPS sparse
MIGUp to 7 MIGs @ 18 GBUp to 7 MIGs @ 10 GB
NVLink900 GB/s NVLink900 GB/s NVLink
AvailabilityLimited availabilityAvailable
FrameworksPyTorch, TensorFlow, JAX, Hugging Face, vLLM, TensorRT, ONNX Runtime, CUDAPyTorch, TensorFlow, JAX, Hugging Face, vLLM, TensorRT, ONNX Runtime, CUDA
Model compatibility guidanceRecommended for quantised inference; review context and concurrencyRecommended for quantised inference; review context and concurrency
ActionConfigure & rentConfigure & rent