B200 SXM vs MI300X
Both run on the same nodes, the same templates and the same tiers here — so the choice comes down to memory, bandwidth and price. Every figure below is from the catalogue or the vendor datasheet.
- B200
- $1.69/h180 GB · Blackwell · 16 available
- MI300X
- $0.99/h192 GB · CDNA 3 · 16 available
- Price gap
- 41%On spot, per GPU-hour
- Over 100 h
- $70.00Difference for the same GPU-hours
What actually separates them.
Four differences that change a decision, computed from the numbers rather than asserted.
Bigger models fit on the MI300X
192 GB against 180 GB. At FP16 that is about 76B parameters on one GPU, against 72B.
The MI300X costs 41% less per hour
$0.99 against $1.69 on spot — $70.00 of difference over 100 GPU-hours.
Token generation is faster on the B200
7,700 GB/s against 5,300 GB/s. Inference reads the whole model from memory for every token, so bandwidth, not FLOPS, usually sets the ceiling.
More memory per dollar on the MI300X
194 GB per dollar per hour against 107 GB. The usual tie-breaker when both cards fit the job.
Every figure we hold on both cards.
A row is marked only when both values exist and one is genuinely better. B200 leads on 3 rows, MI300X on 6.
| B200 SXM | MI300X | |
|---|---|---|
| GPU memoryDecides which models fit at all | 180 GB HBM3e | 192 GB HBM3 |
| Memory bandwidthSets token throughput on inference | 7,700 GB/s | 5,300 GB/s |
| FP32 | 75 TFLOPS | 163 TFLOPS |
| 8-bit float (FP8)Halves the memory a model needs, where the stack supports it | Supported | Supported |
| Architecture | Blackwell (2025) | CDNA 3 (2023) |
| Board powerLower is cheaper to run at scale — for us, not for your bill | 1000 W | 750 W |
| GPU-to-GPU linkOnly matters for multi-GPU training | NVLink | PCIe only |
| Node sizes | 1×, 2×, 4×, 8× | 1×, 2×, 4×, 8× |
| Spot price | $1.69/h | $0.99/h |
| On-demand price | $2.49/h | $1.45/h |
| Memory per dollarGPU memory per $1 of spot time, per hour | 107 GB | 194 GB |
| 30-day reclaim rateShare of spot instances reclaimed | 5–10% | <5% |
| Available now | 16 GPUs | 16 GPUs |
Specifications come from each vendor's datasheet — linked on the B200 page and the MI300X page. Prices are ours, per GPU-hour, billed per minute.
The same models, on one GPU of each.
Weights only, with 20% of the memory kept free for activations and the KV cache. A cross means one GPU is not enough — the model page has the multi-GPU arithmetic.
| Model | Precision | B200 180 GB | MI300X 192 GB |
|---|---|---|---|
| Llama 3.1 8B8B parameters | FP16 · 16 GB | Fits | Fits |
| FP8 · 8 GB | Fits | Fits | |
| INT4 · 4 GB | Fits | Fits | |
| Qwen2.5 32B32B parameters | FP16 · 64 GB | Fits | Fits |
| FP8 · 32 GB | Fits | Fits | |
| INT4 · 16 GB | Fits | Fits | |
| Llama 3.3 70B70B parameters | FP16 · 140 GB | Fits | Fits |
| FP8 · 70 GB | Fits | Fits | |
| INT4 · 35 GB | Fits | Fits | |
| Mixtral 8x22B141B parameters | FP16 · 282 GB | Needs 2× node | Needs 2× node |
| FP8 · 141 GB | Fits | Fits | |
| INT4 · 71 GB | Fits | Fits |
The same 100 GPU-hours on each.
Compute only, billed per minute. Storage is $0.08 per GB-month on either card.
| B200 | MI300X | |
|---|---|---|
| One hour, spot | $1.69 | $0.99 |
| One day, spot | $40.56 | $23.76 |
| 100 GPU-hours, spot | $169.00 | $99.00 |
| 100 GPU-hours, on-demand | $249.00 | $145.00 |
| A month, reserved | $1,350.50 | $766.50 |
Take the B200 when…
- Training and full fine-tuning — Enough memory for optimiser states and activations on models a smaller card can only run in inference.
- Serving large models — A 70B model at FP8 on a single GPU, with room for a long context window.
- Multi-GPU jobs over NVLink — GPU-to-GPU traffic stays off the PCIe bus, which is what makes tensor and pipeline parallelism worth it.
Take the MI300X when…
- Training and full fine-tuning — Enough memory for optimiser states and activations on models a smaller card can only run in inference.
- Serving large models — A 70B model at FP8 on a single GPU, with room for a long context window.
- Anything that can checkpoint — On the spot tier this model is 32% below its own on-demand price, with a 2-minute notice before a reclaim.
B200 or MI300X?
Which is cheaper, the B200 or the MI300X?
The MI300X, at $0.99 per GPU-hour on spot against $1.69 — 41% less. On-demand: $1.45 against $2.49.
Which one should I take for inference?
Take the one that holds the model with room for the KV cache, then the one with more memory bandwidth. Here that is the MI300X on capacity (192 GB) and the B200 on bandwidth.
Can I run both on the same job?
Not inside one instance — a node holds GPUs of a single model. You can run two instances in parallel, on the same persistent volume in turn, or split the work: a big card for the model that needs capacity, a cheap one for the parts that do not.
Other comparisons: the full list · all 47 models
Try both for less than the price of deciding.
An hour on each costs $2.68 on spot. Nothing renews, and your disk moves between them.
Billed per minute from the moment the instance is reachable. Minimum credit $40, no subscription.