NVIDIA H200 SXM 141GB vs NVIDIA H100 SXM5 80GB

The H200 and the H100 deliver near-identical FP8 throughput (1,979 vs 1,979 TFLOPS dense). The H200 also carries 61 GB more memory (141 GB vs 80 GB).

H200: Hopper, 2024 H100: Hopper, 2022

Memory

141 vs 80 GB

61 GB more for the H200

Bandwidth

4.8 TB/s vs 3.4 TB/s

H200 moves data faster

TDP

700 W vs 700 W

Same power budget

See rental pricing from $1.73/hr Download comparison sheet Printable version →

NVIDIA H200 SXM 141GB

Full specs →

What is the NVIDIA H200?

The H200 is a GPU from NVIDIA on the Hopper architecture, an accelerator for AI training, inference and high-performance computing. It is an SXM module, the mezzanine format used in dense multi-GPU servers. H200 full specifications →

Can I rent the H200 in the cloud?

Yes. Cloud providers tracked by Flopper rent the H200 by the hour, and the pricing page compares their current rates and the price history side by side. H200 rental prices →

How does the H200 compare with other accelerators?

This page sets the H200 against the H100. The H200 is the newer generation of the two. Both are NVIDIA parts, and both ship as an SXM module, so they fit the same class of system. Flopper also compares it head to head with the B300 SXM 262GB and A100 SXM4 80GB, across every published precision, memory and power. H200 vs B300 SXM 262GB →

Where do the H200 figures come from?

Every H200 figure on this page traces to NVIDIA's own datasheet or specification table. How Flopper verifies specifications →

NVIDIA H100 SXM5 80GB

Full specs →

What is the NVIDIA H100?

The H100 is a GPU from NVIDIA on the Hopper architecture, an accelerator for AI training, inference and high-performance computing. It is an SXM module, the mezzanine format used in dense multi-GPU servers. H100 full specifications →

Can I rent the H100 in the cloud?

Yes. Cloud providers tracked by Flopper rent the H100 by the hour, and the pricing page compares their current rates and the price history side by side. H100 rental prices →

How does the H100 compare with other accelerators?

This page sets the H100 against the H200. The H100 is the older generation of the two. Both are NVIDIA parts, and both ship as an SXM module, so they fit the same class of system. Flopper also compares it head to head with the A100 SXM4 80GB, Instinct MI300X 192GB and B300 SXM 262GB, across every published precision, memory and power. H100 vs A100 SXM4 80GB →

Where do the H100 figures come from?

Every H100 figure on this page traces to NVIDIA's own datasheet or specification table. How Flopper verifies specifications →

Specifications

H200

H100

Architecture

HopperHopper

Launch Year

20242022

Form Factor

SXMSXM

Memory

141 GB80 GB

Memory Bandwidth

4.8 TB/s3.4 TB/s

TDP

700 W700 W

Process Node

4nm4nm

Spec

H200

H100

Architecture HopperHopper

Launch Year 20242022

Form Factor SXMSXM

Memory 141 GB80 GB

Memory Bandwidth 4.8 TB/s3.4 TB/s

TDP 700 W700 W

Process Node 4nm4nm

Performance (TFLOPS)

H200

H100

34 TFLOPS 34 TFLOPS

67 TFLOPS 67 TFLOPS

494 TFLOPS

989 TFLOPS sparse

494 TFLOPS

989 TFLOPS sparse

990 TFLOPS

1,979 TFLOPS sparse

990 TFLOPS

1,979 TFLOPS sparse

990 TFLOPS

1,979 TFLOPS sparse

990 TFLOPS

1,979 TFLOPS sparse

1,979 TFLOPS

3,958 TFLOPS sparse

1,979 TFLOPS

3,958 TFLOPS sparse

FP6

1,979 TOPS

3,958 TOPS sparse

1,979 TOPS

3,958 TOPS sparse

Tensor Core

67 TFLOPS

Precision

H200

H100

34 TFLOPS 34 TFLOPS

67 TFLOPS 67 TFLOPS

494 TFLOPS

989 TFLOPS sparse

494 TFLOPS

989 TFLOPS sparse

990 TFLOPS

1,979 TFLOPS sparse

990 TFLOPS

1,979 TFLOPS sparse

990 TFLOPS

1,979 TFLOPS sparse

990 TFLOPS

1,979 TFLOPS sparse

1,979 TFLOPS

3,958 TFLOPS sparse

1,979 TFLOPS

3,958 TFLOPS sparse

FP6

1,979 TOPS

3,958 TOPS sparse

1,979 TOPS

3,958 TOPS sparse

Tensor Core

67 TFLOPS

Figures are dense. Where a vendor publishes a figure it sits under the dense one and never decides a winner. Hover any precision for its definition, or read the GPU glossary.

FLOPS by Precision

What actually differs

The H200 is an H100 with a larger, faster memory and nothing else: every FLOPS row above matches, so it pays only where memory is the limit. Training and prefill-heavy jobs that fit in 80 GB run at the same speed on either part and belong on the cheaper H100 hour.

Serving is where the H200 earns its rate. Decode is bound by bandwidth, and 4.8 TB/s against 3.35 TB/s generates tokens about 1.4 times faster on the same model. Capacity matters more: 141 GB holds a 70B model in FP8 with room for long-context KV cache, a replica that otherwise takes two H100s and a tensor-parallel split.

Price the gap on demand, since both floors are spot or community capacity: the H200 starts at $2.09 per GPU-hour against $1.79, a premium the bandwidth alone repays. Compare live rates on the H200 pricing page or step up to B200 vs H200.

New to the terms? The glossary explains FP64, FP32, TF32, BF16, FP16, FP8, INT8, Sparsity and TFLOPS. For how the generations relate, see NVIDIA architectures explained.

Frequently asked questions

Is the H200 faster than the H100?

Only where memory is the limit. Both run 990 TFLOPS BF16 and 1,979 TFLOPS FP8 dense, so training and prefill finish at the same speed. Token generation is limited by memory bandwidth, where the H200 moves 4,800 GB/s against 3,350 GB/s, about 1.4 times faster on a model both can hold.

How much more memory does the H200 have than the H100?

The H200 carries 141 GB of HBM3e against 80 GB of HBM3 on the H100 SXM, about 1.8 times as much. A serving replica that needs two H100s for weights and KV cache often fits on one H200. MIG slices grow too: up to seven at 18 GB each, against 10 GB.

Is the H200 worth the extra cost over the H100?

For memory-bound serving, usually. On-demand rates start at $2.09 per GPU-hour for the H200 and $1.79 for the H100, about 17 percent more for about 43 percent more bandwidth. For training, or for models that fit comfortably in 80 GB, the H100 delivers the same compute for less.

Where can I rent an H200 or an H100?

Flopper lists 57 H200 listings across 31 providers, from $2.00 to $10.60 per GPU-hour, and 67 H100 SXM5 80GB listings across 34 providers, from $1.41 to $12.29. Both floors are spot or community capacity; on-demand starts at $2.09 for the H200 and $1.79 for the H100.

Where to Rent

NVIDIA H200 SXM 141GB

Rent from

$2.09 / GPU-hr

Lowest of 54 live listings across 31 providers · median $4.47

See all H200 rental prices →

Price history and every provider on the pricing page. Some links are affiliate links.

NVIDIA H100 SXM5 80GB

Rent from

$1.73 / GPU-hr

Lowest of 66 live listings across 34 providers · median $3.85

See all H100 rental prices →

Price history and every provider on the pricing page. Some links are affiliate links.

Get Comparison Updates

New GPUs added weekly. Be the first to see how they compare.

Open in Full Compare Tool → H200 full specs H100 full specs Estimate cluster cost in the calculator →

NVIDIA H200 SXM 141GB vs NVIDIA H100 SXM5 80GB

NVIDIA H200 SXM 141GB

What is the NVIDIA H200?

Can I rent the H200 in the cloud?

How does the H200 compare with other accelerators?

Where do the H200 figures come from?

NVIDIA H100 SXM5 80GB

What is the NVIDIA H100?

Can I rent the H100 in the cloud?

How does the H100 compare with other accelerators?

Where do the H100 figures come from?

Specifications

Performance (TFLOPS)

FLOPS by Precision

What actually differs

Frequently asked questions

Is the H200 faster than the H100?

How much more memory does the H200 have than the H100?

Is the H200 worth the extra cost over the H100?

Where can I rent an H200 or an H100?

Where to Rent

NVIDIA H200 SXM 141GB

NVIDIA H100 SXM5 80GB

Related comparisons

Get Comparison Updates

Stay Updated on GPU Specs