Skip to content
Updated minutes ago
nvidia

H200 SXM

nvidia · hopper · 141 GB HBM3e · 700W TDP

VRAM

141 GB

BF16 TFLOPS

990

Bandwidth

4800 GB/s

From

$2.69/hr

Calculate ROI with this GPU →

Spec Sheet

VRAM141 GB HBM3e
Memory Bandwidth4800 GB/s
BF16 TFLOPS990
FP16 TFLOPS990
FP8 TFLOPS1979
INT8 TOPS1979
TDP700W
InterconnectNVLINK
NVLink Bandwidth900 GB/s
Max per Node8
PCIe Gen5
CUDA Compute Capability9
Tensor CoresYes

Pricing by Provider

ProviderOn-DemandReservedSpotBadge
lambda$3.49/hr$2.69/hr-Cheapest
tensordock$3.80/hr-$2.90/hr
coreweave$4.25/hr$3.19/hr-
runpod$4.69/hr--

Pricing History

coreweave
$5.49/hr 0.0%
2024-09-012025-03-01
Low: $5.49High: $6.99
lambda
$3.99/hr 0.0%
2024-09-012025-03-01
Low: $3.99High: $5.49

Compatible Models (278)

Training Capabilities

Estimated GPU count for full fine-tuning (AdamW, BF16) and QLoRA

Model SizeFull Fine-TuneQLoRA
7B model1 GPU1 GPU
13B model2 GPUs1 GPU
70B model10 GPUs1 GPU

Energy Efficiency

Estimated tokens/second per Watt for popular models

Mistral 7B
0.94 t/s/WFP8
Qwen 2.5 7B
0.90 t/s/WFP8
Llama 3.1 8B
0.85 t/s/WFP8
DeepSeek V3
0.19 t/s/WFP8
Llama 3.1 70B
0.10 t/s/WFP8
Qwen 2.5 72B
0.09 t/s/WFP8

Similar GPUs

GPUVRAMBF16 TFLOPSBW (GB/s)From
H2096 GB1484000$0.99/hr
GH20096 GB9904000$2.99/hr
H100 NVL94 GB8353938$3.09/hr
H100 NVL 94GB (per GPU pair)188 GB16707876$5.49/hr
H100 SXM80 GB9903350$1.89/hr

Methodology Note

Performance estimates for the H200 SXMare based on InferenceBench's roofline performance model with CUDA kernel-level optimization including FlashAttention v2 and PagedAttention. Memory calculations account for model weights (141 GB HBM3e available), KV-cache allocation, and activation memory. Throughput predictions use the H200 SXM's rated 4800 GB/s memory bandwidth and 990 BF16 TFLOPS compute capacity as roofline ceilings, with empirical correction factors per GPU architecture (hopper). See our full methodology.

Frequently Asked Questions

How many AI models can run on H200 SXM?

The H200 SXM can run 278 AI models from our database within a single node. Compatible models range across various parameter sizes depending on the quantization precision (BF16, FP8, INT4). Smaller models fit on a single GPU while larger models may require multi-GPU setups up to 8x H200 SXM.

What is the H200 SXM inference throughput?

The H200 SXM delivers 990 BF16 TFLOPS and 1979 FP8 TFLOPS with 4800 GB/s memory bandwidth. Actual inference throughput (tokens/sec) depends on the model size, precision, and batch size. Use our calculator for model-specific throughput estimates.

How much does H200 SXM cost per hour?

The H200 SXM is available starting from $2.69/hour via lambda. Prices vary by provider and pricing tier (on-demand, reserved, spot). Compare pricing across all providers in the table above.