H100 SXM5 or B200 SXM6?
Both are dedicated, both are monthly, both open without an identity check. What separates them is memory, bandwidth and $5,729 a month — and which of those matters depends entirely on what you intend to run.
- $6,061/moH100 SXM5, one card
- $11,790/moB200 SXM6, one card
- 80 vs 180 GBmemory per card
- 2.4×bandwidth gap
The short answer
Which one, and when
- Pick the H100 SXM5
- when 80 GB holds your model. It saves $5,729/month, and memory you do not use is memory you paid for.
- Pick the B200 SXM6
- when your model needs more than 80 GB on a single card. That is a threshold, not a preference — below it, the extra $5,729/month buys you nothing.
- Neither, if
- your model does not fit on a single card of either. Sharding costs synchronisation at every layer — see the interconnect guide before buying a multi-GPU node.
Side by side
| H100 SXM5 | B200 SXM6 | Difference | |
|---|---|---|---|
| Architecture | Hopper | Blackwell | Hopper is older |
| Memory per card | 80 GB HBM3 | 180 GB HBM3e | 100 GB in favour of the B200 SXM6 |
| Memory bandwidth | 3,350 GB/s | 8,000 GB/s | 2.39× in favour of the B200 SXM6 |
| Form factor | SXM on an HGX board | SXM on an HGX board | Same |
| Node sizes | 4×, 8× | 4×, 8× | Up to 1440 GB in one node |
| From, per month | $6,061 | $11,790 | $5,729/month apart |
Price at every node size
| Node | H100 SXM5 | B200 SXM6 | Total VRAM | |
|---|---|---|---|---|
| 4 × GPU | $6,061/mo | $11,790/mo | 320 GB / 720 GB | H100 SXM5 B200 SXM6 |
| 8 × GPU | $11,509/mo | $22,351/mo | 640 GB / 1440 GB | H100 SXM5 B200 SXM6 |
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 H100 SXM5
Nothing. Every model the H100 SXM5 holds on one card, the B200 SXM6 holds too — the difference between them is speed, not capability.
Only the B200 SXM6
- Qwen 3 235B-A22B (MoE) — 4-bit (AWQ, GPTQ), ~59 tok/s
- Mixtral 8×22B (MoE) — FP8, ~17 tok/s
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.
| Card | $/GB of VRAM, per month | $/TB per second, per month | Better at |
|---|---|---|---|
| H100 SXM5 | $75.76 | $1,809 | Neither |
| B200 SXM6 | $65.50 | $1,474 | Both ratios |
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.
H100 SXM5 against B200 SXM6, in short
Is the H100 SXM5 worth the extra money over the B200 SXM6?
Which one is faster for LLM inference?
Do both come as multi-GPU nodes?
Can I rent either without an identity check?
Other comparisons
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.