HPC & AI foundation · Foundation to advanced

Designing a Hyperion Platform: CPU, Memory, GPU, Storage and Fabric

A workload-first method for selecting cores, memory capacity and bandwidth, accelerator topology, NVMe, shared storage and RDMA across the Hyperion X-Series.

A useful HPC or AI specification begins with the working set and execution path, not a favourite processor or an assumed GPU count. Hyperion systems deliberately span CPU-led, single-GPU, multi-GPU and multi-node designs, so the same family can be sized against measured workload constraints.

This guide shows how to turn a representative run into an architecture decision and how to avoid the common error of buying one fast component while starving it elsewhere.

Start with a workload envelope

Record the largest realistic input, peak resident memory, read/write volume, desired concurrency, wall-clock target and acceptable queue time. Separate latency-sensitive online work from throughput-oriented batch work. For AI, record model parameter count, numerical precision, context length, batch size, activation memory and whether training, fine-tuning or inference is required. For simulation, capture mesh or matrix size, solver type, MPI ranks, OpenMP threads and checkpoint pattern.

Run a baseline if any suitable machine exists. CPU utilisation, memory high-water mark, storage throughput, GPU utilisation and transfer time reveal the real bottleneck. Treat vendor peak figures as component capabilities, not application performance guarantees.

CPU and memory are a coupled decision

Core count is valuable only while memory bandwidth, cache locality and software licensing allow scaling. AxiKeystone and AxiVector favour large memory and CPU density for sparse solvers, compilation, genomics, Monte Carlo and data preparation. NUMA-aware placement matters: bind ranks and threads to the memory local to their CPU domain, and test one socket or NUMA region before increasing parallelism.

Capacity headroom should include the operating system, page cache, workflow engine, temporary decompression, multiple concurrent users and failure recovery. When datasets must remain in memory, 1.15TB per node can remove repeated storage passes; when the algorithm streams predictably, bandwidth and layout can matter more than raw capacity.

GPU fit, topology and utilisation

VRAM is a hard fit constraint; aggregate VRAM is not automatically one address space. A single 96GB GPU in AxiAnvil simplifies large-model serving. Four 24GB GPUs in AxiRelay favour replicas and independent services. AxiForge, AxiForge Max and the cluster platforms support distributed training, but the framework must partition data or model state and use NCCL correctly.

Profile host-to-device copies, kernel occupancy, launch overhead and input-pipeline stalls. NVIDIA recommends representative profiling, coalesced memory access and minimising unnecessary transfers. A faster accelerator will remain idle if CPU decoding, data augmentation or storage cannot feed it.

Storage tiers and data lifecycle

Use local NVMe for the active working set, shuffle, checkpoints and temporary data; protected NAS or object storage for source datasets, shared artefacts and retention; and a separate backup target for recovery. Estimate sustained rather than headline throughput, including small-file metadata behaviour. Genomics and computer vision frequently benefit from staging many small files into larger archives or indexed formats.

Capacity planning must include replicas, snapshots, model versions, experiment artefacts and growth. Define which data is authoritative and which is reproducible scratch so backup cost follows business value.

When a fabric becomes necessary

Multi-node applications need communication that justifies the additional operational layer. AxiVector uses direct 100GbE RoCEv2 for two-node MPI; AxiLattice, AxiBastion and AxiTitan use switched fabrics for more flexible scheduling. GPUDirect RDMA can allow compatible devices to exchange data without staging through ordinary host copies, but drivers, topology, RDMA configuration and application libraries all need validation.

Measure strong scaling for a fixed problem and weak scaling as the problem grows. If communication consumes the saved compute time, a larger single node is the better engineering choice.

A repeatable selection gate

  1. Benchmark one representative workload and capture resource traces.
  2. Identify hard fit constraints: RAM, VRAM, local scratch, network or software licences.
  3. Select the smallest baseline that clears those constraints with operational headroom.
  4. Test the target software stack in containers or controlled environments.
  5. Validate performance, failure recovery, power, cooling and export-compliance assumptions before formal quotation.

Hyperion configurations are baselines, not substitutes for workload validation. Axiotech can use benchmark evidence to adjust CPU, memory, accelerator, storage and fabric together.

Primary technical references

References are provided for software architecture and implementation planning. Validate the versions, licences, support matrix and regulated-use requirements applicable to the final deployment.

Relevant Hyperion platforms

Hyperion X1140

AxiFoundry

A professional development and proof-of-concept node for CUDA, AI, data science and software engineering.

One 4U node; 96 AMD EPYC cores; 384GB ECC DDR5; one NVIDIA RTX PRO 4000 24GB; enterprise NVMe and remote management.

Hyperion X1440

AxiRelay

A dense inference, RAG, CI and multi-service node where several independent GPU workloads must run concurrently.

One 4U node; 96 AMD EPYC cores; 768GB ECC DDR5; four NVIDIA RTX PRO 4000 24GB GPUs providing 96GB aggregate VRAM.

Hyperion X1160

AxiAnvil

A large-memory single-GPU platform for private LLM inference, retrieval-augmented generation, medical imaging and large scientific models.

One 4U node; 96 AMD EPYC cores; 768GB ECC DDR5; one NVIDIA RTX PRO 6000 96GB GPU; enterprise NVMe.

Hyperion X1260

AxiForge

A professional two-GPU training, visualisation and applied-research server for larger datasets and models.

One 4U node; 96 AMD EPYC cores; 768GB ECC DDR5; two NVIDIA RTX PRO 6000 96GB GPUs; 192GB aggregate VRAM and high-capacity NVMe scratch.

Hyperion X1460

AxiForge Max

The maximum single-node Hyperion platform for AI training, molecular modelling, generative engineering and large accelerated workflows.

One 4U node; 128 AMD EPYC cores; 1.15TB ECC DDR5; four NVIDIA RTX PRO 6000 96GB GPUs providing 384GB aggregate VRAM.

Hyperion X1140-C

AxiKeystone

A CPU- and memory-led HPC platform for simulation, genomics, analytics, compilation and workloads with large in-memory working sets.

One 4U node; 128 AMD EPYC cores; 1.15TB ECC DDR5; one NVIDIA RTX PRO 4000 24GB; enterprise U.2 NVMe.

Hyperion X1150-C

AxiGovernor

A controls-engineering and virtual-commissioning platform for TwinCAT, PLC, motion, simulation and private engineering AI.

One 4U node; 96 AMD EPYC cores; 768GB ECC DDR5; one NVIDIA RTX PRO 5000 48GB; engineering storage and remote management.

Hyperion X1250

AxiOverseer

A production-intelligence node for machine vision, telemetry, predictive maintenance and low-latency factory AI.

One 4U node; 96 AMD EPYC cores; 768GB ECC DDR5; two NVIDIA RTX PRO 5000 48GB GPUs; protected local/NAS data services.

Hyperion X2160

AxiVector

A tightly coupled two-node simulation and mathematical-computing platform for MPI, CFD, FEA and optimisation.

Two 4U nodes; 256 AMD EPYC cores; 2.3TB aggregate ECC DDR5; two 96GB GPUs; direct 100GbE RoCEv2 RDMA between nodes.

Hyperion X3140

AxiLattice

A shared departmental CPU-compute facility for research pipelines, simulation, genomics and queued multi-user work.

Three 4U nodes; 384 AMD EPYC cores; 3.4TB aggregate ECC DDR5; three professional GPUs; switched 100GbE RDMA, 25GbE storage and Slurm.

Hyperion X2260

AxiCrucible

A two-node multi-user AI workgroup for clinical research, production intelligence, training and resilient service placement.

Two 4U nodes; four NVIDIA RTX PRO 6000 96GB GPUs; 100GbE RDMA; shared protected storage; KVM and Kubernetes-ready infrastructure.

Hyperion X3460

AxiBastion

A sustained departmental AI/HPC platform for pharmaceutical modelling, large training campaigns and shared accelerated research.

Three 4U nodes; 384 AMD EPYC cores; 3.4TB ECC DDR5; twelve NVIDIA RTX PRO 6000 96GB GPUs; 100GbE compute and 25GbE storage fabrics.

Hyperion X4460

AxiTitan

The flagship sovereign AI and research-computing platform for organisation-scale shared services and the largest Hyperion workloads.

Four 4U nodes; 512 AMD EPYC cores; 4.6TB ECC DDR5; sixteen NVIDIA RTX PRO 6000 96GB GPUs; more than 1.5TB aggregate VRAM; 100GbE compute and dedicated storage fabrics.

Configuration and quotation

Validate this workload on Hyperion

Final architecture and price depend on representative code and data, concurrency, storage, networking, site infrastructure, component availability and export compliance.

Request Formal Quotation