GPU × use-case guide · Embeddings

Is the L40 a good GPU for embeddings?

The L40 is a ada NVIDIA GPU with 48 GB GDDR6 (864 GB/s), 362 TFLOPS BF16 / 733 TFLOPS FP8, and a 300 W TDP. Embeddings workloads care most about maximum batch throughput on small encoder models — pure tokens-per-dollar. Here's how the L40 measures up.

What models fit on a single L40?

Weights only, reserving ~25% of the 48 GB for KV cache, activations and fragmentation. ✓ = fits on one card.

Largest single-card fit: Qwen 2.5 14B at BF16, Gemma 2 27B at FP8, Qwen 2.5 72B at INT4. Bigger models need tensor-parallel across 8 cards.

L40 for embeddings, specifically

Embeddings is throughput-bound rather than VRAM-bound, so the L40's 864 GB/s of bandwidth and 362 TFLOPS matter more than raw capacity. Because it runs offline, batch aggressively to push tokens-per-dollar down. Size it precisely on the calculator.

L40 pricing across providers

Verdict

At 48 GB, the L40 is a solid mid-to-high-tier choice for embeddings: single-card up to Qwen 2.5 14B (BF16) or Gemma 2 27B (FP8), and cost-effective at ~$0.99/hr.