HPE Apollo System Platforms
Two purpose-built platforms optimised for different HPC and AI workload profiles.
Why HPE Apollo for HPC & AI?
Purpose-Built Density
Apollo chassis are purpose-engineered for maximum server density — the Apollo 2000 fits up to 80 servers in a 42U rack, delivering 4 times the compute density of traditional rack servers in the same physical footprint.
AI Training at Scale
The Apollo 6500 Gen10 Plus with NVLink 2.0 enables GPU-to-GPU data transfer at 300 GB/s, dramatically reducing AI training time for large neural network models that require fast inter-GPU communication.
Direct Liquid Cooling
The Apollo 2000 supports optional Direct Liquid Cooling (DLC) for CPU and memory components — enabling higher processor TDPs in space-constrained environments and significantly reducing data centre cooling costs at scale.
Flexible CPU/GPU Architecture
Apollo systems support Intel Xeon Scalable and AMD EPYC processors. The Apollo 6500 supports both SXM2 (NVLink-connected) and PCIe GPU module configurations, providing flexibility across different NVIDIA GPU generations and workloads.
HPC Cluster Ready
Apollo systems integrate with HPE Performance Cluster Manager, Lustre and GPFS parallel file systems, InfiniBand and high-speed Ethernet fabric options — the complete building blocks for large-scale HPC cluster deployments.
Optimised for Thermal Efficiency
Dense GPU configurations generate substantial heat. Apollo chassis are purpose-designed with enhanced airflow and liquid cooling options to maintain performance at sustained GPU workloads without thermal throttling.
HPE Apollo Use Cases
Artificial Intelligence & Deep Learning
Train large language models, convolutional neural networks, and transformer architectures at scale. The Apollo 6500 with NVLink 2.0 enables GPU cluster training without PCIe bandwidth bottlenecks.
Genomics & Life Sciences
Genome sequencing, protein folding, and drug discovery workloads require massive parallelism. Apollo 2000 density maximises cluster nodes within HPC data centre footprints subject to space constraints.
Financial Risk & Quantitative Analysis
Monte Carlo simulations, risk analytics, and real-time derivatives pricing benefit from Apollo 2000 scale-out density and low-latency interconnect between nodes.
CFD & Scientific Simulation
Computational fluid dynamics, seismic processing, and climate modelling consume thousands of cores. Apollo 2000 delivers the node count needed for large-scale simulation jobs.
Frequently Asked Questions — HPE Apollo Systems
What is the difference between HPE Apollo 2000 and Apollo 6500?
The HPE Apollo 2000 Gen10 Plus is designed for high-density CPU-based scale-out computing — fitting up to 80 independent servers in a 42U rack for HPC clusters, genomics, and financial modelling workloads. The HPE Apollo 6500 Gen10 Plus is designed for GPU-intensive AI and HPC workloads — it accommodates up to 8 high-power GPUs per server node with NVLink 2.0 GPU-to-GPU interconnect at 300 GB/s, enabling fast AI training and deep learning inference.
What GPUs does HPE Apollo 6500 Gen10 Plus support?
The HPE Apollo 6500 Gen10 Plus supports NVIDIA Tesla T4 (inference-optimised), NVIDIA Quadro RTX 6000 (professional visualisation and AI), and NVIDIA Quadro RTX 8000 GPUs. Both SXM2 (NVLink-connected, for maximum GPU-to-GPU bandwidth) and PCIe Gen3 module configurations are available. The SXM2 configuration with NVLink 2.0 delivers up to 300 GB/s GPU-to-GPU bandwidth — essential for large AI training workloads.
What is Direct Liquid Cooling on HPE Apollo systems?
HPE Apollo Direct Liquid Cooling (DLC) is an option for the Apollo 2000 Gen10 Plus that routes liquid coolant directly to the processor and optionally the memory subsystem — removing heat at the source rather than relying on air circulation. This enables higher processor TDPs in environments where air cooling is insufficient, reduces data centre cooling energy consumption, and allows higher rack density configurations in thermally constrained data centres.
Is HPE Apollo suitable for AI inference as well as training?
Yes — both workloads are supported. AI training workloads typically use the Apollo 6500 with SXM2 NVIDIA GPUs connected via NVLink for maximum training throughput. AI inference workloads can use the Apollo 6500 with NVIDIA Tesla T4 PCIe GPUs — T4 is optimised for inference, supporting INT8 and FP16 precision with excellent performance per watt. The Apollo 2000 can also run inference workloads at scale in a CPU-based configuration.
Can Servnet supply HPE Apollo systems in the UK?
Yes — Servnet is an authorised HPE reseller and can supply HPE Apollo 2000 Gen10 Plus and Apollo 6500 Gen10 Plus systems for HPC and AI deployments in the UK. We can help design cluster configurations, advise on GPU options and interconnect fabric, and provide competitive pricing. Contact sales@servnetuk.com or call 0800 987 4111 for an Apollo consultation.
Build Your HPC or AI Cluster
Servnet is an authorised HPE reseller. Our team can design and quote Apollo HPC and AI cluster configurations for UK research, financial, and enterprise AI deployments.
Compare HPE Apollo with other vendors
Servnet is vendor-neutral. Explore alternative and complementary platforms in the same category.