GPU × use-case guide · Translation
Is the B200 NVL (pair) a good GPU for translation?
The B200 NVL (pair) is a blackwell NVIDIA GPU with 360 GB HBM3e (16000 GB/s), 4500 TFLOPS BF16 / 9000 TFLOPS FP8, and a 1800 W TDP. Translation workloads care most about high batch throughput and cost efficiency across many short-to-medium requests. Here's how the B200 NVL (pair) measures up.
What models fit on a single B200 NVL (pair)?
Weights only, reserving ~25% of the 360 GB for KV cache, activations and fragmentation. ✓ = fits on one card.
| Model | BF16 | FP8 | INT4 |
|---|---|---|---|
| Llama 3.1 8B | ✓ | ✓ | ✓ |
| Qwen 2.5 14B | ✓ | ✓ | ✓ |
| Gemma 2 27B | ✓ | ✓ | ✓ |
| Mixtral 8x7B (MoE) | ✓ | ✓ | ✓ |
| Llama 3.3 70B | ✓ | ✓ | ✓ |
| Qwen 2.5 72B | ✓ | ✓ | ✓ |
| Llama 3.1 405B | — | — | ✓ |
Largest single-card fit: Qwen 2.5 72B at BF16, Qwen 2.5 72B at FP8, Llama 3.1 405B at INT4. Bigger models need tensor-parallel across 4 cards.
B200 NVL (pair) for translation, specifically
Translation is throughput-bound rather than VRAM-bound, so the B200 NVL (pair)'s 16000 GB/s of bandwidth and 4500 TFLOPS matter more than raw capacity. Because it runs offline, batch aggressively to push tokens-per-dollar down. Size it precisely on the calculator.
B200 NVL (pair) pricing across providers
| Provider | On-demand $/hr | Reserved $/hr |
|---|---|---|
| lambda | $11.49 | — |
| coreweave | $14 | $10.5 |
Verdict
With 360 GB, the B200 NVL (pair) is a data-center-class card that comfortably handles translation for models up to Qwen 2.5 72B at full precision on a single card — a strong pick if your budget supports ~$11.49/hr.
See full B200 NVL (pair)specs & pricing, size your model on the calculator, or compare every GPU on the GPU list.