A100 SXM4 or A100 PCIE: which should I pick?

The listed memory, FP32 and bandwidth figures are the same: 40 GB, 15.6 TFLOPS and 1313 GB/s. The SXM4 runs at 400 W instead of 250 W, so pick it for sustained heavy jobs and the PCIE card when either is fine and you want the cheaper live offer.

Is 40 GB enough to fine-tune a 70B model?

Not comfortably. 40 GB handles LoRA on 7B–13B models and QLoRA on mid-size models. For 70B-class fine-tuning look at 80 GB and larger cards such as the H100 SXM or H200 NVL.

Does the A100 SXM4 support BF16?

Yes. Ampere with compute capability 8.0 supports BF16, which is the default precision for most current LLM training and fine-tuning recipes.

How is an A100 SXM4 server billed on Cloudj?

Per minute from your account balance while it runs, at the price locked when you deploy, plus bandwidth per GB. A stopped instance is charged for storage only. There are no monthly GPU plans.

GPU Cloud · NVIDIA · Ampere

Rent an NVIDIA A100 SXM4 GPU with 40 GB VRAM

Name: NVIDIA A100 SXM4 cloud GPU
Brand: Cloudj
Availability: InStock

Rent A100 SXM4 GPUs by the hour on Cloudj: 40 GB VRAM on NVIDIA Ampere with about 1313 GB/s memory bandwidth and a 400 W power limit for sustained AI work. Built for teams that run long training or serving jobs and want the SXM module's thermal headroom, it offers 15.6 TFLOPS FP32 and compute capability 8.0. Pick a live configuration, start from the Ubuntu 22.04 + CUDA or PyTorch + JupyterLab template and connect over SSH with your key.

Rent A100 SXM4 All GPUs

A100 SXM4 specifications

VRAM per GPU

40 GB

Architecture

NVIDIA Ampere

Compute capability

FP32 compute

≈ 15.6 TFLOPS

Memory bandwidth

≈ 1313 GB/s

Power limit

400 W

Values reported by the infrastructure provider for machines currently listed; real performance varies by machine.

Live A100 SXM4 configurations and prices

From $0.46/h

incl. 32 GB disk · Live price, updated every 15 minutes

Deploy

What you can run on the A100 SXM4 cloud GPU

Long-running training jobs

The 400 W power limit suits multi-hour BF16 training where clocks have to hold steady from the first epoch to the last.

Production-style LLM serving

Host a 13B model in 16-bit or a quantized 30B-class model behind vLLM or SGLang and test throughput before you commit to infrastructure.

Instruction tuning with Unsloth

Run LoRA or QLoRA on 7B–13B models with longer sequences than a 24 GB card allows.

Research reproduction

Reproduce papers that were benchmarked on A100-class hardware on the same Ampere architecture and memory size.

Is the A100 SXM4 the right GPU for your workload?

Choose the 40 GB A100 SXM4 when your training or inference workloads exceed consumer GPU memory limits and require datacenter memory bandwidth. Opt for a higher-VRAM datacenter GPU if your unquantized models exceed 40 GB, or select a smaller workstation or consumer card for lightweight inference and development.

A100 SXM4 vs similar cloud GPUs

GPU	VRAM per GPU	FP32 compute	Power limit
NVIDIA A100 SXM4	40 GB	≈ 15.6 TFLOPS	400 W
NVIDIA A100 PCIE	40 GB	≈ 15.6 TFLOPS	250 W	Details →
NVIDIA A40	48 GB	≈ 29.9 TFLOPS	300 W	Details →
NVIDIA L40S	48 GB	≈ 73.3 TFLOPS	350 W	Details →
NVIDIA Tesla P40	24 GB	≈ 9.4 TFLOPS	250 W	Details →

Guides for A100 SXM4

A100 cloud Europe guide: renting A100 GPUs in the EUA100 cloud Europe guide: PCIe vs SXM4 on the GPU list, what the 40 GB versions handle, searching EU codes and when an A100 beats an H100 on value.Rent A100 USA: PCIe and SXM4 options billed by the minuteRent A100 USA hosts for training and batch work: PCIe vs SXM4 in practice, multi-GPU listings, and a cost estimate that includes storage while stopped.A100 cloud in Asia: Singapore, Japan and Korea searchesPlan an A100 cloud Singapore Japan search: the SG, JP, KR and HK codes, the A100 variants listed, 48 GB alternatives in the region and moving data from Vietnam.AI training GPU USA checklist: planning multi-day rented runsAn AI training GPU USA checklist for multi-day runs: choose 1x to 8x machines, check the availability date, set a checkpoint cadence and track per-minute cost.

All articles →

A100 SXM4 cloud GPU: frequently asked questions

Rent the A100 SXM4 cloud GPU by the hour

Pick a live configuration, choose a template and connect over SSH with your key. Billing stops when you destroy the instance.

Rent A100 SXM4 All GPUs