Comparison · More memory, more bandwidth, same architecture

# H200 SXM5 or H100 SXM5?

Both are dedicated, both are monthly, both open without an identity check. What separates them is memory, bandwidth and $2,344 a month — and which of those matters depends entirely on what you intend to run.

- $8,405/mo H200 SXM5, one card
- $6,061/mo H100 SXM5, one card
- 141 vs 80 GB memory per card
- 1.4× bandwidth gap

## The short answer

Which one, and when

- **Pick the H200 SXM5:** when your model needs more than 80 GB on a single card. That is a threshold, not a preference — below it, the extra $2,344/month buys you nothing.
- **Pick the H100 SXM5:** when 80 GB holds your model. It saves $2,344/month, and memory you do not use is memory you paid for.
- **Neither, if:** your model does not fit on a single card of either. Sharding costs synchronisation at every layer — see the [interconnect guide](https://gpuserver.io/guides/nvlink-vs-pcie) before buying a multi-GPU node.

## Side by side

**H200 SXM5 compared with H100 SXM5**

|  | [H200 SXM5](https://gpuserver.io/gpu/h200) | [H100 SXM5](https://gpuserver.io/gpu/h100-sxm5) | Difference |
|---|---|---|---|
| Architecture | Hopper | Hopper | Same generation |
| Memory per card | 141 GB HBM3e | 80 GB HBM3 | 61 GB in favour of the H200 SXM5 |
| Memory bandwidth Predicts generation speed | 4,800 GB/s | 3,350 GB/s | 1.43× in favour of the H200 SXM5 |
| Form factor | SXM on an HGX board | SXM on an HGX board | Same |
| Node sizes | 4×, 8× | 4×, 8× | Up to 1128 GB in one node |
| From, per month | $8,405 | $6,061 | $2,344/month apart |

## Price at every node size

**Monthly price of both cards at each available node size**

| Node | H200 SXM5 | H100 SXM5 | Total VRAM |  |
|---|---|---|---|---|
| 4 × GPU | $8,405/mo | $6,061/mo | 564 GB / 320 GB | [H200 SXM5](https://gpuserver.io/configure?c=h200-x4) [H100 SXM5](https://gpuserver.io/configure?c=h100-sxm-x4) |
| 8 × GPU | $15,945/mo | $11,509/mo | 1128 GB / 640 GB | [H200 SXM5](https://gpuserver.io/configure?c=h200-x8) [H100 SXM5](https://gpuserver.io/configure?c=h100-sxm-x8) |

## What each one holds on a single card

At 8k context, best precision that fits. This is the difference that decides a purchase — not the specification sheet.

### Only the H200 SXM5

- **Qwen 3 235B-A22B (MoE)** — 4-bit (AWQ, GPTQ), ~35 tok/s
- **Mixtral 8×22B (MoE)** — 4-bit (AWQ, GPTQ), ~20 tok/s

### Only the H100 SXM5

Nothing. The H200 SXM5 holds everything the H100 SXM5 does.

**The same model on both: Llama 3.3 70B.** On the H200 SXM5, roughly 31 tokens/second at FP8; on the H100 SXM5, roughly 43. Single-stream and deliberately conservative — with continuous batching the aggregate is several times higher on both. Treat the *ratio* as the useful number, not the absolute.

## Cost per GB and per TB/s

Two ratios that cut through the specification sheet. The first tells you what memory costs; the second what speed costs.

**Cost per gigabyte of VRAM and per terabyte per second of bandwidth**

| Card | $/GB of VRAM, per month | $/TB per second, per month | Better at |
|---|---|---|---|
| [H200 SXM5](https://gpuserver.io/gpu/h200) | $59.61 | $1,751 | Both ratios |
| [H100 SXM5](https://gpuserver.io/gpu/h100-sxm5) | $75.76 | $1,809 | Neither |

These ratios rank cards; they do not choose one. A card that is cheaper per gigabyte is worthless if it has too few gigabytes to hold your model at all — capacity is a threshold, not a slope. Use them to break a tie between two cards that both fit.

## H200 SXM5 against H100 SXM5, in short

Is the H200 SXM5 worth the extra money over the H100 SXM5?

It costs $2,344 more a month — 39% — for 61 GB more memory and 43% more memory bandwidth. It also holds 2 model(s) on a single card that the H100 SXM5 cannot.

Which one is faster for LLM inference?

Generation speed is bounded by memory bandwidth, because each token requires re-reading the active weights. The H200 SXM5 has 4,800 GB/s against 3,350 GB/s, so the H200 SXM5 leads by roughly 43% on a model both of them hold.

Do both come as multi-GPU nodes?

H200 SXM5: 4×, 8×. H100 SXM5: 4×, 8×. SXM cards live on an HGX board of four or eight — never two, never ten — and are joined by NVLink rather than PCIe.

Can I rent either without an identity check?

Yes, both, on the same terms. No document, no selfie, no phone number and no company registration, at any level of spend. Payment is in cryptocurrency, each invoice gets its own address, and root access arrives under 5 minutes after the first confirmation.

## Other comparisons

- [H100 SXM5 vs A100 SXM4  The generational jump, on the same board Compare](https://gpuserver.io/compare/h100-sxm5-vs-a100-sxm4)
- [H100 SXM5 vs H100 PCIe  Same card, different fabric Compare](https://gpuserver.io/compare/h100-sxm5-vs-h100-pcie)
- [B200 SXM6 vs H200 SXM5  Blackwell against Hopper, at the top Compare](https://gpuserver.io/compare/b200-vs-h200)

## Both are racked and waiting, in all 6 data centres.

Whichever you pick, it is monthly, paid in crypto, opened without an identity check and delivered in under 5 minutes.

[Open the configurator](https://gpuserver.io/configure) [Read the guides](https://gpuserver.io/guides)

---

Source: https://gpuserver.io/compare/h200-vs-h100-sxm5/. This file is generated from the same data as the website; if a figure here differs from a page, the page is authoritative and this file is stale — the canonical source is https://gpuserver.io/.