OwnAI Compute
Book a call Sign in
GPU configuration

GPU infrastructure, compared honestly

Every option below is supported in our server platforms. Select a model and workload to see what each GPU delivers for your organisation, then compare several side by side.

Prices in US at the enterprise multiplier (2×)
GPUClassVRAMBandwidthECCPowerFits Llama 3.1 8B?Recommended countExpected tokens/sHardware price
GeForce RTX 5090
Blackwell · PCIe, 2–3.5 slot (board-partner dependent)
Professional / high-end consumer GPU 32 GB1,792 GB/sNo575 W Yes (2 GPUs) 2 11,524 $9,600
$4,800 each
GeForce RTX 5080
Blackwell · PCIe, 2–3 slot
Professional / high-end consumer GPU 16 GB960 GB/sNo360 W Yes (4 GPUs) 4 11,661 $9,200
$2,300 each
RTX PRO 6000 Blackwell Server Edition
Blackwell · PCIe, dual-slot passive
Enterprise workstation & server GPU 96 GB1,792 GB/sYes600 W Yes (1 GPU) 1 6,402 $17,000
$17,000 each
NVIDIA L40S
Ada Lovelace · PCIe, dual-slot passive
Enterprise workstation & server GPU 48 GB864 GB/sYes350 W Yes (2 GPUs) 2 5,556 $30,000
$15,000 each
NVIDIA H100 NVL
Hopper · PCIe, dual-slot passive, NVLink bridge
Data-centre accelerator 94 GB3,938 GB/sYes400 W Yes (1 GPU) 1 14,069 $56,000
$56,000 each
NVIDIA H200 NVL
Hopper · PCIe, dual-slot passive, NVLink bridge
Data-centre accelerator 141 GB4,800 GB/sYes600 W Yes (1 GPU) 1 17,149 $64,000
$64,000 each

Recommended count is the smallest number of GPUs meeting memory, 20% peak concurrency at 10 tokens/s per user and daily demand of 20 prompts per employee. Illustrative until benchmarked.

Side-by-side comparison

GeForce RTX 5090
Professional / high-end consumer GPU
RTX PRO 6000 Blackwell Server Edition
Enterprise workstation & server GPU
NVIDIA H200 NVL
Data-centre accelerator
ArchitectureBlackwellBlackwellHopper
VRAM32 GB GDDR796 GB GDDR7 with ECC141 GB HBM3e with ECC
Memory bandwidth1,792 GB/s1,792 GB/s4,800 GB/s
FP16 tensor (dense)209 TFLOPS251 TFLOPS835 TFLOPS
ECC memoryNoYesYes
NVLinkNoNoYes
Power575 W600 W600 W
Form factorPCIe, 2–3.5 slot (board-partner dependent)PCIe, dual-slot passivePCIe, dual-slot passive, NVLink bridge
CoolingActive (axial fans); needs 3-slot spacing and strong chassis airflowPassive; server chassis airflowPassive; data-centre airflow
Support termsConsumer warranty; not covered by enterprise driver and support programmesEnterprise drivers, 3-year warranty, vendor supportEnterprise support and 5-year NVIDIA AI Enterprise option
Trade-offsBest inference per dollar; 32 GB without ECC limits model size and reliability for continuous service.Large ECC memory in a standard PCIe server; the usual choice for 70B-class models without HBM pricing.Largest memory per PCIe card: fewer GPUs for 70B+ models and long contexts.
Supported models (INT4, per GPU)Llama 3.2 3B, Llama 3.1 8B, Gemma 3 12B, Qwen2.5 14BLlama 3.2 3B, Llama 3.1 8B, Gemma 3 12B, Qwen2.5 14B, Qwen2.5 32B, Llama 3.3 70B, gpt-oss 120B (MoE, 5B active)Llama 3.2 3B, Llama 3.1 8B, Gemma 3 12B, Qwen2.5 14B, Qwen2.5 32B, Llama 3.3 70B, gpt-oss 120B (MoE, 5B active)
Expected concurrent sessions (8B, 8k ctx, 1 GPU)2481122
Hardware price (each)$4,800$17,000$64,000
SourceManufacturer pageManufacturer pageManufacturer page

Professional / high-end consumer

GeForce RTX 5090 and 5080. Excellent inference per dollar for 8B–14B models and small teams. Trade-offs: no ECC memory, consumer warranty, active coolers that need 3-slot spacing and strong airflow, and 16–32 GB that caps model size and concurrency. We cap deployments at four cards.

Enterprise workstation and server GPU

RTX PRO 6000 Blackwell and L40S. ECC memory, passive cooling for dense servers, enterprise drivers and multi-year support. 48–96 GB per card runs 32B–70B-class models in standard PCIe servers at a fraction of HBM pricing.

Data-centre accelerators

H100 NVL and H200 NVL. HBM3/HBM3e bandwidth of 3.9–4.8 TB/s and up to 141 GB per card for the highest throughput on 70B+ models and long contexts, with NVLink pairing. Premium cost, power delivery and facilities requirements.

Rejoining the server...

Rejoin failed... trying again in seconds.

Failed to rejoin.
Please retry or reload the page.

The session has been paused by the server.

Failed to resume the session.
Please retry or reload the page.