Rent the L40S if…
you serve models up to about 30B parameters at FP8, run image or video generation, rendering or transcoding, and want the lower monthly price.
Deploy an L40S · $429/mo
NVIDIA L40S48 GB GDDR6 · Ada Lovelace
vs
NVIDIA A10080 GB HBM2e · Ampere
NVIDIA L40S vs A100: which GPU should you rent?
Two answers to the same budget: 48 GB of GDDR6 with FP8, RT cores and video engines on the L40S, against 80 GB of HBM2e, NVLink and MIG on the A100 80 GB.
- L40S, dedicated$429per month · $0.59/h effective
- A100, dedicated$649per month · $0.89/h effective
- Monthly saving−34%the L40S costs $220 less a month
The short answer
Which one should you rent?
For inference and graphics, the L40S is usually the better buy: it has FP8, which the A100 lacks, slightly more dense FP16 throughput, ray-tracing cores and video encoders, for $429 a month. For training and fine-tuning the A100 wins: 80 GB instead of 48, 2.4× the memory bandwidth, NVLink between GPUs and MIG partitioning, for $649. The deciding question is usually memory: if your model and its optimizer state or KV cache need more than 48 GB, rent the A100.
- Memory. 80 GB of HBM2e at 2 TB/s on the A100 against 48 GB of GDDR6 at 864 GB/s on the L40S.
- Precision. FP8 on the L40S's fourth-generation tensor cores, the format modern inference stacks target; BF16, FP16 and TF32 only on the A100.
- Multi-GPU. NVLink and MIG on the A100; PCIe only and no MIG on the L40S, which is meant to work alone or in loosely coupled groups.
- Graphics and video. RT cores and three NVENC encoders with AV1 on the L40S for rendering and transcoding; the A100 has neither.
Rent the A100 if…
you fine-tune or train and need 80 GB, high memory bandwidth, NVLink between GPUs or MIG partitions.
Deploy an A100 · $649/moSpecifications
L40S vs A100: the numbers side by side
NVIDIA's published figures for each part, tensor throughput without sparsity. The last column is how the L40S compares, as a multiple of the A100.
| Specification | NVIDIA L40S | NVIDIA A100 | L40S vs A100 |
|---|---|---|---|
| Architecture | Ada Lovelace · 2023 | Ampere · 2020 | — |
| GPU memory | 48 GB GDDR6 | 80 GB HBM2e | 0.6× |
| Memory bandwidth | 864 GB/s | 2 TB/s | 0.42× |
| FP16 / BF16 tensor (dense) | 362 TFLOPS | 312 TFLOPS | 1.2× |
| FP8 tensor (dense) | 733 TFLOPS | Not supported | Only the L40S |
| FP4 tensor (dense) | Not supported | Not supported | — |
| Maximum power | 350 W | 400 W | 0.88× |
| Form factor | PCIe Gen4 x16 | SXM4 | — |
| GPU-to-GPU link | PCIe only | NVLink 3 · 600 GB/s | — |
| Multi-Instance GPU (MIG) | No | Up to 7 instances | — |
| ECC memory | Yes | Yes | — |
| Video engines | 3 NVENC · 3 NVDEC (AV1) | 5 NVDEC · no NVENC | — |
| CUDA compute capability | 8.9 | 8.0 | — |
| Dedicated host here | 12 vCPU · 96 GB RAM · 1 TB NVMe | 16 vCPU · 128 GB RAM · 2 TB NVMe | — |
| Price per month here | $429 | $649 | 0.66× |
Sources: NVIDIA datasheets and product pages for each GPU. Tensor figures are dense (NVIDIA often prints them doubled “with sparsity”). Prices are this site's flat monthly rates for a dedicated server with the host listed, in USD excluding VAT.
Value
What each dollar buys
Price per month divided into the two resources most jobs run out of first — memory and memory bandwidth — and what the same GPU costs kept on for a month at the market-median hourly rate.
Per GB of GPU memoryL40S $8.94 · A100 $8.11per month, dedicated host included
Per TB/s of bandwidthL40S $497 · A100 $318per month — what decode-heavy inference pays for
Effective hourly rateL40S $0.59/h · A100 $0.89/hflat monthly price over 730 hours
Same GPU on demand, 24/7L40S $1,124 · A100 $1,285market-median hourly rate × 730 h
Model sizes
What fits on one L40S and one A100
The largest model whose weights fit on a single GPU, keeping 10% of the memory for the runtime, activations and a small KV cache. Long contexts and large batches need more — size them with the GPU memory calculator.
| Weights precision | NVIDIA L40S | NVIDIA A100 | Typical use |
|---|---|---|---|
| 16-bit (BF16 / FP16) | up to ~22B parameters | up to ~36B parameters | Full-precision inference; the base weights of a LoRA fine-tune |
| 8-bit (FP8 / INT8) | up to ~43B parameters | up to ~72B parameters | FP8 serving with vLLM or TensorRT-LLM on GPUs that support it; INT8 weights elsewhere |
| 4-bit (GPTQ, AWQ, Q4) | up to ~79B parameters | up to ~130B parameters | Quantised inference on a budget; QLoRA fine-tuning |
Parameters in billions, weights only: 2 bytes per parameter at 16-bit, 1 at FP8 or INT8, about 0.55 at 4-bit (GPTQ, AWQ, GGUF Q4). Mixture-of-experts models need memory for all their experts, not just the active ones.
Prices
L40S and A100 price per month, by term
Per dedicated GPU server, in USD excluding VAT — CPU cores, RAM, NVMe, 20 TB of transfer and 24/7 support included.
| Term | NVIDIA L40S | NVIDIA A100 | Difference per month |
|---|---|---|---|
| Monthly | $429 | $649 | $220 |
| 3 months · −5% | $408 | $617 | $209 |
| 6 months · −10% | $386 | $584 | $198 |
| 12 months · −15% | $365 | $552 | $187 |
| Dedicated host | 12 vCPU · 96 GB RAM · 1 TB NVMe · 10 Gbps | 16 vCPU · 128 GB RAM · 2 TB NVMe · 25 Gbps | included |
L40S vs A100: questions answered
Is the L40S faster than the A100?
For FP8 inference and graphics, often: it supports FP8, has slightly more dense FP16 throughput and adds RT cores. For memory-bound work — training, large batches, long contexts — the A100's 2 TB/s of HBM2e and 80 GB usually make it faster.
Can the L40S fine-tune LLMs?
Yes, within 48 GB: LoRA on 13B-class models at 16-bit, QLoRA on 70B models with small batches. Full fine-tunes and large batches are A100 or H100 territory.
L40S or A100 for Stable Diffusion and FLUX?
The L40S: 48 GB is plenty for SDXL and FLUX pipelines, FP8 speeds them up, and its encoders handle the video output. Most image pipelines leave the A100's extra memory unused.
Are the L40S and the A100 in stock?
17 of 40 L40S units were free at the last stock update; 21 of 48 A100 units were free at the last stock update (18 September 2026). Both are provisioned automatically, in under 10 minutes after payment.
Can I rent the L40S or the A100 without KYC?
Yes. Both are rented with an email address and paid from a prepaid balance in Bitcoin, Ether or USDT — $429 and $649 a month, dedicated, with no identity check and no card.
Other GPU comparisons
Both are in stock, by the month
Dedicated, single-tenant and live in under 10 minutes: the L40S at $429 or the A100 at $649 a month, paid in BTC, ETH or USDT with no KYC.