HopperLive Marketplace Asks Available

NVIDIA H100

H100 SXM5 is the Hopper datacenter GPU: 80 GB HBM3, 3.35 TB/s, fourth-generation Tensor Cores, and FP8 through the Transformer Engine. A typical SXM node quote is 26 vCPUs, 116 GB of host RAM, and 80 GB of VRAM. It is still the default card for 30B–70B training and production FP8 serving.

Deploy NVIDIA H100 on Marketplace →All Hardware Fleet

Flexible Compute Modes

Deploy NVIDIA H100 Your Way: Docker Pods or Root KVM VPS

NovaPod lets you decide how you want to harness GPU power. Whether you need an instant pre-cached container stack, a full root virtual machine with dedicated public IP, or a multi-node NVLink cluster, pay only for what you run with second-by-second crypto billing.

Docker GPU PodCold Boot < 15s

Fast, Lightweight AI Containers

Pure GPU compute without hypervisor overhead. Boot straight into PyTorch, vLLM, ComfyUI, or JupyterLab.

1-second precision metering
37+ Pre-cached ML framework templates
Zero idle fee when instance is stopped

Deploy as Docker Pod

Root KVM VPSPersistent NVMe

Full Root Virtual Machine

Dedicated operating system (Ubuntu 24.04 LTS / Windows GPU Server) with complete sudo root SSH access from minute one.

Dedicated public IPv4 (all ports open)
Pause anytime; pay only $0.00014/GB/hr for disk
Run Docker-in-Docker, k3s, custom drivers

Deploy as Root VPS

Multi-GPU Cluster900 GB/s NVLink

Supercomputing & Distributed Training

Pool up to 8× NVIDIA H100 on a unified node fabric with Quantum-2 InfiniBand RDMA for multi-node scale.

Zero-loss all-reduce gradient exchange
PyTorch FSDP, DeepSpeed Zero-3 ready
Native GPUDirect Storage & RDMA fabric

Configure Cluster

Hardware Specifications

Architecture: NVIDIA Hopper
VRAM: 80 GB HBM3
Bandwidth: 3.35 TB/s
Tensor Cores: 528, 4th generation
CUDA cores: 16,896
FP64: 34 TFLOPS
FP32: 67 TFLOPS
TF32: 989 TFLOPS with sparsity
FP16 Tensor: 1,979 TFLOPS with sparsity
FP8 Tensor: 3,958 TFLOPS with sparsity
INT8: 3,958 TOPS with sparsity
NVLink: 900 GB/s
MIG: Up to 7 instances
Reference host: 26 vCPUs, 116 GB RAM
Form factor: SXM5 or PCIe
TDP: 700 W SXM5, about 350 W PCIe
Multi-node: 400 Gb/s class fabric on serious jobs

Still the default datacenter GPU

H100 SXM5 is 80 GB of HBM3 at 3.35 TB/s, 16,896 CUDA cores, fourth-generation Tensor Cores, and FP8 through the Transformer Engine. FP32 is 67 TFLOPS. FP64 is 34 TFLOPS, about three times an A100, which is why some simulation codes still ask for Hopper specifically. Board power on SXM5 is 700 W. The PCIe card is about half that and has no NVLink.

The catalog host next to this GPU is 26 vCPUs and 116 GB of RAM. That is one training process and its dataloader, not an eight-GPU chassis. Multi-GPU training wants SXM boards on NVLink inside the node, and a 400 Gb/s class fabric with GPUDirect RDMA once you leave the node.

What fits in 80 GB

A 70B model in FP8 fits on one H100 with on the order of 10 GB left for KV cache. That is enough for moderate batch and moderate context. A 30B model fits in BF16. FP16 70B does not fit. Long context or a large batch pushes you to H200 at 141 GB.

Decode at batch size one is bandwidth-bound. H100's 3.35 TB/s is about 1.6× an A100 80GB, which is most of the inference gap people feel before they look at FLOPS. FP8 on the Transformer Engine is roughly a 1.6× step over FP16 on attention and MLP layers when the software path is actually on.

How a job is launched here

Open the marketplace, take an H100 offer, and start from a vLLM, PyTorch, or CUDA template. Billing is the hourly ask divided into seconds. There is no 20-minute minimum and no separate spot product unless the offer itself says the host can reclaim the card.

A QLoRA fine-tune of a 70B model on a modest dataset is often a working day on one H100. A full-parameter fine-tune wants several GPUs and checkpoints. Frameworks you should expect to work are PyTorch, TensorFlow, JAX, vLLM, TensorRT-LLM, and SGLang, on a CUDA 12 image.

SXM, PCIe, and clusters

Choose SXM when GPUs in the node must talk over NVLink. Choose PCIe when you are serving one model and the host is a normal server. Mixing them in one training job is how you get a slow rank.

Past one node, all-reduce time dominates unless the network is built for it. Plan InfiniBand or a comparable RDMA fabric, NCCL that matches the topology, and a checkpoint interval you can afford to lose. Eight GPUs on NVLink is a different design from eighty GPUs on ethernet.

What the Numbers Mean in Production

Bandwidth vs A100 80GB: ~1.6×
Training vs A100: About 3–4×
70B FP8: One GPU
FP64: 34 TFLOPS

Architectural Generation Comparison

Pick on the constraint. H100 is the compute default. H200 is the memory upgrade. A100 is the cheaper Ampere card.

	H100	H200	A100 80GB
Architecture	Hopper	Hopper	Ampere
VRAM	80 GB HBM3	141 GB HBM3e	80 GB HBM2e
Bandwidth	3.35 TB/s	4.8 TB/s	2.0 TB/s
FP8	Yes	Yes	No
Menu host	26 vCPU · 116 GB	44 vCPU · 182 GB	14 vCPU · 100 GB

Storage & Tooling Infrastructure

High-Speed Storage, Native Ports & Crypto Rails

Every NVIDIA H100 instance on NovaPod connects directly to high-throughput infrastructure designed to eliminate I/O bottlenecks and onboarding delays.

25 Gb/s Shared Network Storage

Mount large model checkpoints and training sets across one or multiple instances simultaneously. Data outlives single pods and stays accessible when instances stop.

JupyterLab, VS Code & SSH

Connect from your local terminal with standard SSH keys or open web-based IDEs directly in browser tabs without managing complex reverse proxy configs.

Non-Custodial Multi-Chain Rails

Deposit USDT, USDC, SOL, ETH, or BNB to your dedicated personal deposit address. Confirmed on-chain in seconds with zero mandatory identity verification.

Target Workloads for NVIDIA H100

Training and fine-tuning

FP8 mixed precision is why H100 still beats A100 on 30B and larger runs. A modest 70B QLoRA on one card is often a working day.

7B–70B training and full fine-tunesLoRA at scaleMultimodal models that fit 80 GBFSDP and DeepSpeed on SXM

Production inference

A 70B model in FP8 fits on one SXM with some KV-cache room. FP16 70B does not fit.

vLLM, TensorRT-LLM, SGLang70B FP8 servingSpeculative decodingModerate batch and context

Diffusion and HPC

Batch image and video jobs, plus FP64 work that is about three times an A100.

Flux and video diffusionMolecular dynamicsCFD and other FP64 codesBatch, not a single preview

Multi-GPU training

Stay inside one SXM node on NVLink until the batch no longer fits. Past that, all-reduce wants a real fabric.

NVLink inside the nodeGPUDirect RDMA between nodesNCCL that matches the topologyCheckpoints you can afford to lose

When to Choose NVIDIA H100

Pick H100 for FP8 and 30B and up

The Transformer Engine is the reason to pay more than an A100.

Pick A100 if you do not need FP8

Sub-30B training, BERT-scale jobs, and many inference paths are cheaper on Ampere.

Pick H200 if you run out of memory

Same compute family, 141 GB of HBM3e when the KV cache or a colocated model does not fit.

Pick PCIe only for one GPU

SXM5 is 700 W with NVLink. PCIe is lower power and has no NVLink. Do not mix them in one training job.

SXM5 versus PCIe

SXM5 is 700 W with NVLink inside the node. PCIe is about 350 W and easier to mix into an existing server. Train on SXM. A single inference worker can be PCIe.

Multi-node

Serious all-reduce wants a 400 Gb/s class fabric and GPUDirect RDMA, not the datacenter ethernet you use for SSH. Stay on one node until the batch no longer fits.

Frequently Asked Questions

How much does an H100 cost here?

The marketplace card shows the hourly ask. We charge that ask per second. Hyperscaler on-demand rates for the same silicon are often several dollars per GPU-hour. Compare the live ask, not a screenshot from another site.

What is the cheapest path?

The lowest ask on a card you can actually launch, for a job that fits in 80 GB. A reclaimable offer saves money only if a preemption is acceptable. Production serving wants a card that stays.

Can I rent it by the hour?

Yes. The meter is finer than that: seconds, with no minimum and no reserved-instance contract on a normal deploy.

How fast does it start?

A ready image on a listed offer is a container start, not a hardware order. If the image is huge and not cached, the pull dominates. The GPU itself is already in the host.

SXM or PCIe?

SXM5 is 700 W with NVLink, for multi-GPU training. PCIe is about half the power and fine for a single inference worker.

Does it do FP8?

Yes. Hopper's Transformer Engine runs FP8 mixed precision. That is the main gap versus A100, and it is why a transformer training step can be several times faster when the software path is actually on.

Will 70B fit?

In FP8, yes, with on the order of 10 GB left for cache. That is moderate batch and moderate context. FP16 70B does not fit. Long context belongs on H200.

Do you support clusters?

Multi-GPU starts as more than one H100 on NVLink. Multi-node needs a 400 Gb/s class fabric and GPUDirect RDMA. Ask for that when one node is full. Ethernet used for SSH is not that fabric.

Which regions?

Whatever the offer lists. Capacity moves. The marketplace is the inventory, not a city written on this page.

Which frameworks?

PyTorch, TensorFlow, JAX, vLLM, TensorRT-LLM, Triton, and SGLang on a CUDA 12 image, with cuDNN and NCCL. Bring your own image if you already have one.

Is it worth it over A100?

When the job is FP8 or bandwidth-bound, yes. H100 is about 1.6× the memory bandwidth and adds FP8. For a 13B LoRA, A100 is usually the better dollar.

Rent or buy?

Rent unless the card would sit near full use for a year and you already pay for power and racks. Newer parts arrive faster than a purchased H100 depreciates.

Alternative Hardware Options

Ready to Deploy NVIDIA H100?

Launch an instance on the live marketplace in under 15 seconds. Fund your balance with cryptocurrency and pay per second with zero commitments.

Launch on Marketplace →