GPU instances catalogue
Accelerated compute for AI, graphics and HPC
Eleven NVIDIA accelerator models — from H200 to RTX Pro 6000 — catalogued with NVIDIA-published specifications and delivered as partner-sourced capacity with written scope.
- Models catalogued
- 11 NVIDIA GPUs
- Workloads
- AI · graphics · HPC
- Delivery model
- Partner-sourced
Workload guide
Choose by workload, not by badge
Most selection decisions come down to memory, scale-up interconnect and whether the job is AI, graphics or both. These criteria narrow the eleven models — a scoping conversation settles the final choice.
Train and fine-tune
Memory capacity and interconnect decide how a model has to be split.
- How much GPU memory the largest model or batch needs
- Whether training must scale across multiple GPUs
- Precision and framework requirements for the training stack
Serve and infer
Throughput per watt and per card decides how a fleet is shaped.
- Concurrent requests, latency targets and traffic peaks
- How many models or versions must run side by side
- Rack density, power and hosting constraints
Design, render and visualise
Frame-buffer size and RT cores decide what a desktop or render node can hold.
- Scene, frame-buffer and working-set size
- Interactive desktops versus batch render nodes
- Remoting and workstation software support
Starting points, not performance claims — scoping confirms the right fit for your workload.
The catalogue
All eleven GPU models
Every card links to its model page — NVIDIA-published specifications, workload guidance and the shared partner-sourced delivery model.
NVIDIA H200
Hopper-class acceleration with the largest HBM3e buffer in the catalogue, for AI workloads that are constrained by memory.
- GPU memory
- 141 GB HBM3e
- Memory bandwidth
- 4.8 TB/s
- FP8 tensor
- 3,958 TFLOPS
NVIDIA H100
The proven Hopper accelerator for LLM training and inference, with FP8 Transformer Engine support and 900 GB/s NVLink scale-up.
- GPU memory
- 80 GB HBM3
- Memory bandwidth
- 3.35 TB/s
- FP8 tensor
- 3,958 TFLOPS
NVIDIA A100
A proven training-and-HPC accelerator with an 80 GB memory pool per GPU and mature tooling across the AI ecosystem.
- GPU memory
- 80 GB HBM2e
- Memory bandwidth
- 2,039 GB/s
- FP16 tensor
- 624 TFLOPS
NVIDIA L40S
One card for generative AI, 3D graphics, rendering and video — the pragmatic choice where workloads refuse to stay in one lane.
- GPU memory
- 48 GB GDDR6 ECC
- Memory bandwidth
- 864 GB/s
- FP8 tensor
- 1,466 TFLOPS
NVIDIA A30
A right-sized Ampere accelerator for teams stepping up from CPU inference or low-end GPUs into real AI capacity.
- GPU memory
- 24 GB HBM2
- Memory bandwidth
- 933 GB/s
- Max power
- 165 W
NVIDIA A2
Entry-level, single-slot Ampere acceleration for inference and edge workloads that do not justify flagship hardware.
- GPU memory
- 16 GB GDDR6
- Memory bandwidth
- 200 GB/s
- Max TDP
- 40–60 W
NVIDIA L4
The efficiency play: dense video processing and AI inference in a low-profile card built for many-servers, many-streams deployments.
- GPU memory
- 24 GB GDDR6
- Memory bandwidth
- 300 GB/s
- Max power
- 72 W
NVIDIA RTX 8000
A 48 GB professional graphics platform for visualisation, design and graphics rendering on proven Turing-era tooling.
- GPU memory
- 48 GB GDDR6 ECC
- Memory bandwidth
- 672 GB/s
- CUDA cores
- 4,608
NVIDIA RTX A6000
48 GB of ECC professional graphics memory on Ampere — a workhorse for AI-assisted design, rendering and visualisation.
- GPU memory
- 48 GB GDDR6 ECC
- Memory bandwidth
- 768 GB/s
- CUDA cores
- 10,752
NVIDIA RTX 6000 Ada
The current-generation professional graphics flagship: Ada throughput, 48 GB ECC memory, and strong AI capability in one card.
- GPU memory
- 48 GB GDDR6 ECC
- Memory bandwidth
- 960 GB/s
- CUDA cores
- 18,176
NVIDIA RTX Pro 6000
The Blackwell-generation professional flagship — 96 GB of GDDR7 and current-generation AI/graphics capability in workstation or server editions.
- GPU memory
- 96 GB GDDR7 ECC
- Memory bandwidth
- 1,792 GB/s
- CUDA cores
- 24,064
How to read this catalogue
Verified specs, honest commercial notes
- Specifications are taken from NVIDIA's published product pages and datasheets — every model page states its variant caveats.
- Tensor Core throughput figures are NVIDIA's sparsity-accelerated ratings; memory, bandwidth and power figures are as published.
- Availability, configuration and commercial terms are confirmed in writing per engagement. This catalogue describes what we can scope — not stock we hold.
Delivery model
How Scalenta delivers GPU capacity
Partner-sourced, scoped and operated with the same discipline as the rest of the infrastructure business.
Partner-sourced capacity
Accelerator capacity is sourced through verified data center and hardware partners — no facility-ownership claims, no stock promises.
Scoped per engagement
Model, variant, GPU count, interconnect and host configuration are confirmed in writing before anything is committed.
Operated with 24/7 support
Deployments are covered by 24/7 support and SLA-backed managed services scoped to your environment.
FAQ
Catalogue questions
Does Scalenta own the GPUs in this catalogue?
How do I choose between the eleven models?
Can I combine GPUs from different generations?
What does pricing look like?
Request a GPU consultation
Tell us the workload — model, memory and throughput expectations — and we will map it to the right accelerator with a written scope. No obligation.