Deep learning workstation, from one GPU to four.
A deep learning workstation is a balance: enough GPU memory for the model, enough GPUs for the throughput you need, and a processor, memory, and storage that keep the GPUs supplied. Our three builds run from a single RTX 5090 to four RTX PRO Blackwell GPUs, with every GPU on a full PCIe 5.0 x16 link.
- 1–4GPUs per system
- 384GBVRAM, maximum
- 96CPU cores, maximum
- 1TBECC DDR5, maximum
Choosing a configuration
The three builds, and the work each suits.
| Build | GPUs and VRAM | Platform | Suited to |
|---|---|---|---|
| JNCS AI 5090 | 1× RTX 5090, 32GB | Ryzen 9000, 8–16 cores, 64GB to 128GB DDR5 | Prototyping, computer vision, inference, models that fit in 32GB |
| JNCS AI PRO | 1–2 RTX PRO Blackwell, 32GB to 192GB | Threadripper 9000, 24–64 cores, 64GB to 512GB ECC | Larger models, longer runs, two experiments at once |
| JNCS AI MAX | 1, 2, or 4 RTX PRO Blackwell, 32GB to 384GB | Threadripper PRO 9000, 16–96 cores, 128GB to 1TB ECC | Multi-GPU training, shared group machines, the largest models in a tower |
Keeping the GPUs supplied
CPU cores for data loading
Decoding images, augmenting batches, and tokenizing text run on the processor. Threadripper and Threadripper PRO provide 16 to 96 cores, so data loading keeps pace with several GPUs.
PCIe 5.0 x16 per GPU
Multi-GPU training moves gradients between cards every step. Threadripper 9000 provides 80 PCIe 5.0 lanes and Threadripper PRO 9000 provides 128, so every GPU in the AI PRO and AI MAX runs at a full x16 link.
Registered ECC memory
ECC corrects single-bit memory errors, which matters for datasets cached in RAM and for runs that last days. The AI PRO and AI MAX use registered ECC DDR5 on four and eight channels.
Fast local datasets
Samsung 9100 PRO PCIe 5.0 NVMe drives, 1TB to 8TB each, in several M.2 slots, with optional 12TB to 28TB Seagate Exos hard drives for raw data and archives.
GPU capacity per system
One, two, or four of the same card in the AI PRO and AI MAX. The AI 5090 runs one GeForce RTX 5090 (32GB).
| GPU | 1 GPU | 2 GPUs | 4 GPUs |
|---|---|---|---|
| RTX PRO 4500 Blackwell, 32GB, 200W | 32GB | 64GB | 128GB |
| RTX PRO 5000 Blackwell, 48GB, 300W | 48GB | 96GB | 192GB |
| RTX PRO 5000 Blackwell, 72GB, 300W | 72GB | 144GB | 288GB |
| RTX PRO 6000 Blackwell Workstation Edition, 96GB, 600W | 96GB | 192GB | 384GB |
Operating system and software
Every system ships with Windows 11 Pro installed and updated, with the GPU drivers in place. PyTorch runs on Windows directly, and PyTorch, TensorFlow, and JAX all run with GPU support under WSL 2, which provides a Linux environment on the same machine. If your group installs its own operating system, tell us in the notes on the quote request.
Power and circuit requirements
A single-GPU system, or two RTX PRO 4500 or 5000 cards, runs on a standard 15A, 120V circuit. Two RTX PRO 6000 cards, or any four-GPU configuration, require a dedicated 20A, 110–120V circuit, with a 30A, 220–240V circuit recommended. Four RTX PRO 6000 cards draw approximately 2,850W and require a dedicated 30A, 220–240V circuit. The configurator shows the requirement for every configuration as you build it.
Assembly, burn-in, and delivery
Each system is assembled by hand at our shop in Fairport, NY, by a shop that has built computers for more than thirty years.
Fit and power validated
The configurator checks physical fit, memory, PCIe slots, and power on every selection. The power supply is sized automatically to the GPU count and estimated draw, with a 12V-2x6 cable for every GPU.
Burn-in under load
Every system is stress-tested under sustained full load and firmware-updated before it leaves the bench, so the first long job is not the first test.
Windows 11 Pro installed
Windows 11 Pro is installed and updated before shipment.
Warranty and support
Every build carries a 3-year warranty and lifetime technical support from the shop that built it.
Quoting and payment
The configurator shows a price for every part and an estimated total, including assembly, burn-in, and testing, as the build is configured. GPU and memory pricing changes frequently, so a technician confirms GPU availability and final pricing by email, with lead time, typically within 1-2 business days. There is no cost and no obligation to request a quote.
Payment for AI workstations is by ACH bank transfer; credit cards are not accepted for these systems. An invoice with bank details is issued once the quote is approved. Systems ship across the United States from Fairport, NY, with local pickup available.
Deep learning workstation questions
How many GPUs does a deep learning workstation need?
Is an RTX 5090 good for deep learning?
Why Threadripper instead of a desktop processor?
Can it run Linux?
How much system RAM does a deep learning workstation need?
How is an AI workstation ordered and paid for?
Related AI workstation pages
RTX 5090 AI workstation
One GeForce RTX 5090, 32GB VRAM, Ryzen 9000, standard 15A circuit.
Read →RTX PRO 6000 Blackwell workstation
One or two RTX PRO Blackwell GPUs, up to 192GB VRAM, Threadripper 9000.
Read →Multi-GPU AI workstation
Up to four RTX PRO Blackwell GPUs, up to 384GB VRAM, Threadripper PRO 9000.
Read →LLM fine-tuning workstation
VRAM for full fine-tuning, LoRA, and QLoRA, by model size.
Read →For universities and labs
Spec sheets and RFQs quoted as written, NDAs, matched sets.
Read →Image and video generation
VRAM for diffusion models, and one generation queue per GPU.
Read →Configure a deep learning workstation.
Select a build, the GPU and count, then the processor, memory, and storage. Fit, PCIe slots, and power are validated on every selection.
Have an exact parts list or a department spec sheet? Send the whole list. For more than four GPUs or rackmount systems, see AI servers and larger systems.