✅Runs easilyfp16 · 4K ctx · batch 1
Run Qwen3-8B on L40S
At fp16 precision, Qwen3-8B (8.0 B parameters) needs roughly 20.2 GB of VRAM. L40S has 48 GB, leaving 27.8 GB of headroom — plenty of room for longer contexts and larger batch sizes. Expected throughput ≈ 112 tokens/sec on a single card.
42%
VRAM
Runs easily
Estimated VRAM (fp16)
20.2 GB
of 48 GB on L40S· 27.8 GB free
Tok/sec
112
TFTT
597ms
Power
134W
Rent on GPUniq marketplace
L40S
Live prices from verified providers · billed hourly, no commitment.
Memory breakdown (fp16)
Model Weights
16.00 GBKV Cache
0.63 GBActivations
2.50 GBFramework Overhead
1.04 GBQuantization comparison
Lower precision = less VRAM with a small quality trade-off. Quality order: FP16 > INT8 > INT4.
| Precision | VRAM | Utilisation | Fits on L40S? |
|---|---|---|---|
| FP16 (full precision) | 20.2 GB | 42% | ✅Runs easily |
| INT8 (8-bit) | 10.9 GB | 23% | ✅Runs easily |
| INT4 / Q4 (4-bit) | 6.3 GB | 13% | ✅Runs easily |
Frequently asked
- Can L40S run Qwen3-8B?
- Yes. Qwen3-8B needs ≈ 20.2 GB VRAM at fp16 and L40S provides 48 GB. Expected throughput is 112 tokens/sec per GPU.
- How much VRAM does Qwen3-8B use?
- About 20.2 GB at fp16, 6.3 GB at INT4 (for a 4K context, batch size 1). Longer contexts add to KV cache size; larger batches increase activations.
- What's the fastest way to run Qwen3-8B on L40S?
- Use a production inference engine like vLLM, SGLang, or TensorRT-LLM with paged attention — they cut KV cache 2–3× vs the naive estimate and batch multiple requests efficiently. For L40S, enable Flash Attention 2.
- Where can I rent a L40S?
- GPUniq aggregates live L40S offers from verified providers. You can deploy an instance in about a minute and pay hourly with no commitment.
Other GPUs for Qwen3-8B
Try it live
Chat with Qwen3-8B — no setup
Send a prompt and see how Qwen3-8B responds directly in our chat. No installation, no GPU required to test.
Want to tweak sequence length, batch size, or fine-tuning? Open the full calculator →