RTX 5090 AI workstation, 32GB of VRAM on a standard circuit.
The JNCS AI 5090 is a single-GPU AI workstation: one GeForce RTX 5090 on a Ryzen 9000 processor and an ASUS ProArt X870E board. It is sized for local LLM inference, image and video generation, and fine-tuning of smaller models, and it draws under 850W at full load.
- 32GBVRAM, GeForce RTX 5090
- 8–16cores, Ryzen 9000
- 64–128GBDDR5-6000
- 15A, 120Vstandard circuit
RTX 5090 workstation specifications
Fixed platform parts are listed as-is; ranges are the options offered in the configurator.
| GPU | MSI GeForce RTX 5090 32G: 32GB GDDR7, 575W. One card. |
|---|---|
| Processor | AMD Ryzen 9000 on AM5: Ryzen 7 9800X3D or 9850X3D (8 cores), Ryzen 9 9900X or 9900X3D (12 cores), or Ryzen 9 9950X3D (16 cores, 32 threads, 5.7 GHz boost) |
| Motherboard | ASUS ProArt X870E-CREATOR WIFI (ATX, dual-channel DDR5) |
| Memory | 64GB (2×32GB), 96GB (2×48GB), or 128GB (2×64GB) DDR5-6000, matched dual-channel kit |
| Storage | Samsung 9100 PRO PCIe 5.0 NVMe, 1TB to 8TB per drive, up to two drives. Optional: up to two Seagate Exos 7200 RPM enterprise hard drives, 12TB to 28TB each, as independent volumes or a RAID 1 or RAID 0 pair |
| Cooling | Noctua NH-D15 G2 chromax.black, dual-tower air cooler |
| Power supply | Lian Li 1200W 80 Plus Platinum, ATX 3.1, native 12V-2x6 |
| Case | Fractal Design Define 7, solid steel or dark tempered glass side panel |
| Operating system | Windows 11 Pro, installed and updated |
What 32GB of VRAM runs
Total VRAM sets the largest model that can be loaded on the GPU.
Model weights occupy roughly parameters × bits per weight ÷ 8. A 70-billion-parameter model at 4-bit quantization is about 35GB of weights; with context and runtime overhead it requires approximately 48GB of VRAM. Context length and batch size add to every figure below.
Local LLM inference
A 30-billion-parameter model at 4-bit quantization fits in 32GB of VRAM. Smaller models fit at higher precision or with longer context.
Image and video generation
Diffusion models for image and video generation run on a single high-VRAM GPU; 32GB leaves room for larger resolutions and batch sizes than 16GB or 24GB cards.
Fine-tuning smaller models
Parameter-efficient fine-tuning of smaller models fits on one card. Full training of large models is a multi-GPU workload.
See the multi-GPU build →When to step up
A 70B-class model needs approximately 48GB at 4-bit. That calls for an RTX PRO 5000 or 6000 Blackwell card, offered in the JNCS AI PRO.
See the RTX PRO build →Power and circuit requirements
Estimated draw at full load is approximately 795W to 845W depending on the processor, on the 1200W Platinum unit. That is within the 1,440W continuous limit of a standard 15A, 120V circuit, so no electrical work is required.
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.
RTX 5090 AI workstation questions
Is one RTX 5090 enough for local LLMs?
Can a second RTX 5090 be added to this build?
How does the RTX 5090 compare with the RTX PRO 4500 Blackwell?
How much system memory should an RTX 5090 workstation have?
What circuit does the RTX 5090 workstation need?
How is an AI workstation ordered and paid for?
Related AI workstation builds
RTX PRO 6000 Blackwell workstation
One or two RTX PRO Blackwell GPUs, up to 192GB VRAM, Threadripper 9000.
View →Multi-GPU AI workstation
Up to four RTX PRO Blackwell GPUs, up to 384GB VRAM, Threadripper PRO 9000.
View →Local LLM workstation guide
How much VRAM a model needs, and which build provides it.
View →Configure an RTX 5090 AI workstation.
Select the processor, memory, and storage in the configurator. Fit and power are validated on every selection, and the quote request takes about a minute.
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.