GPU × use-case guide · Text-to-speech

Is the L40S a good GPU for text-to-speech?

The L40S is a ada NVIDIA GPU with 48 GB GDDR6 (864 GB/s), 362 TFLOPS BF16 / 733 TFLOPS FP8, and a 350 W TDP. Text-to-speech workloads care most about low-latency streaming synthesis; small models, so value cards win. Here's how the L40S measures up.

What models fit on a single L40S?

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.

L40S for text-to-speech, specifically

Text-to-speech is throughput-bound rather than VRAM-bound, so the L40S's 864 GB/s of bandwidth and 362 TFLOPS matter more than raw capacity. Because it's latency-sensitive, run it at low batch sizes on a fast framework (vLLM ≈ 85% MFU) rather than maximising batch. Size it precisely on the calculator.

L40S pricing across providers

Verdict

At 48 GB, the L40S is a solid mid-to-high-tier choice for text-to-speech: single-card up to Qwen 2.5 14B (BF16) or Gemma 2 27B (FP8), and cost-effective at ~$1.09/hr.