The Dual RTX 5090 Threadripper Pro Workstation is purpose‑built for professionals who require flagship GPU performance, high PCIe bandwidth and long‑term expandability. By using the proven WRX80 platform with DDR4 memory rather than newer DDR5 alternatives, this system delivers comparable real‑world throughput for GPU‑accelerated workloads at a significantly lower cost — making it the smart choice for AI, GPU compute, simulation and rendering workloads where sustained performance and value matter most.
Built and tested in the UK by Punch Technology, every system comes with a 3-year parts and labour warranty and next‑day delivery to UK mainland addresses, backed by over 30 years of workstation building experience and our status as an AMD Elite System Builder.
The system uses fast NVMe storage for rapid application loading, dataset access and project scratch space. The WD Black SN7100 provides high sustained read and write performance, making it well suited to AI pipelines, rendering caches and large project files, while leaving further M.2 and SATA expansion available for growth.
At the heart of this system is a dual NVIDIA RTX 5090 configuration. Each GPU delivers 32GB of GDDR7 memory and exceptional CUDA, AI and rendering performance, enabling accelerated workflows in machine learning, 3D rendering, scientific computation and simulation. A hybrid air and liquid cooling configuration is used intentionally to prevent airflow restriction between physically large GPUs, ensuring stable clocks, reduced thermal throttling and quieter operation during sustained workloads.
The Phanteks ENTHOO PRO 2 chassis is designed for extreme expansion and airflow. With support for multiple full‑length GPUs, large radiators and extensive internal airflow, it provides the thermal headroom and space required for high‑power accelerator configurations, while remaining fully serviceable and upgrade‑friendly.
Connectivity is tailored for professional and infrastructure environments. Dual 10GbE wired networking (Intel X710) delivers fast, reliable access to shared storage, render farms and data servers — a deliberate choice over wireless that ensures maximum stability and bandwidth for sustained data‑intensive workloads. The WRX80 platform further provides an abundance of PCIe lanes for additional GPUs, storage controllers or high‑speed networking cards.
Ideal for:
- AI researchers and developers
- Machine learning and LLM inference
- 3D rendering and ray tracing
- Visual effects and content creation
- Scientific simulation
- Transcoding and digital audio
- Running multiple virtual machines
This system is ideally suited for AI researchers, developers, studios and professionals who need a reliable, cost‑efficient dual RTX 5090 Threadripper Pro workstation with the ability to expand well beyond the limitations of standard desktop systems.
Full Specification
| Processor | |
| CPU | AMD Ryzen Threadripper Pro 5975WX |
| Cores / Threads | 32 Cores / 64 Threads |
| Frequency | Base 3.6GHz / Max Boost 4.5GHz |
| CPU Cooler | Arctic Freezer 4U‑M |
| Memory & Storage | |
| RAM | 64GB DDR4 3200MHz DIMM |
| Configuration | Quad Channel — 4 × 16GB |
| Maximum Supported | Up to 1TB (8 DIMM slots) |
| Primary SSD | 1TB WD Black SN7100 NVMe PCIe 4.0 |
| Secondary Drive | None (expandable — additional M.2 and SATA bays available) |
| Graphics & Display | |
| GPU | 2 × NVIDIA RTX 5090 32GB GDDR7 |
| Display Outputs | Via GPU — multiple DisplayPort & HDMI (up to 8 displays) |
| Motherboard & Connectivity | |
| Motherboard | ASRock WRX80D8‑2T |
| Chipset | AMD WRX80 |
| Wired Networking | Dual 10GbE (Intel X710) |
| Wireless | None (wired-only for maximum stability) |
| Audio | Not included |
| Expansion & Power | |
| PCIe Slots | 7 × PCIe Gen 4 x16 (configuration dependent) |
| Power Supply | 2000W FSP Cannon Pro, ATX 3.0 / PCIe Gen 5 |
| Chassis | Phanteks ENTHOO PRO 2 — Black |
| Software & Support | |
| Operating System | Windows 11 Professional (clean install) |
| Input Devices | No keyboard or mouse included |
| Warranty | 3 Years Parts & Labour — Return to Base (RTB) |




FAQ
Is this system suitable for CPU‑only workloads?
This system is optimised for GPU‑accelerated workloads. Customers primarily focused on CPU‑bound applications may wish to consider newer TRX50 or WRX90 platforms.
Is this system suitable for AI and LLM workloads?
Yes. The dual RTX 5090 configuration provides 64GB of total GDDR7 GPU memory across two cards, making this system well suited to LLM inference, AI model training, machine learning pipelines and GPU-accelerated data processing. The WRX80 platform’s high PCIe lane count also allows future GPU or NVMe expansion without bottlenecking.
Why use DDR4 rather than DDR5?
For GPU-accelerated workloads, the primary bottleneck is GPU memory bandwidth rather than system RAM speed. DDR4 at this capacity and channel configuration delivers comparable real‑world performance to DDR5 for the vast majority of AI, rendering and simulation tasks, while keeping the overall system cost meaningfully lower. The platform supports up to 1TB of RAM for memory-intensive workloads.
Why use mixed GPU cooling?
Hybrid air and liquid cooling prevents airflow restriction between physically large GPUs, improving sustained performance and leaving room for future expansion.
What does the warranty cover?
This system is covered by a 3-year parts and labour Return to Base (RTB) warranty. This means that in the event of a component failure, you return the system to Punch Technology and we repair or replace the faulty parts and return it to you at no additional cost. Our team is also available for ongoing technical support throughout the warranty period.
Can the system be expanded?
Yes. Additional GPUs, storage and networking cards can be added. Higher‑power PSU options are available on request. Contact our team to discuss a custom configuration.







What others are saying
There are no contributions yet.