AI Work 100

• You're running 7B–70B class open-weight models locally for a small team.
• Single or dual RTX 5090 (32 GB GDDR7 each) gives you enough headroom and you don't need NVLink coherence.
• Office colocation — a quiet tower workstation on a standard wall outlet.

• You need to serve more than ~10 concurrent users in production — step up to AI PRO 500.
• Your fine-tunes need 4+ GPUs or you want NVLink coherent memory — that's AI PRO 500 or AI Server territory.
• The workload is heavily I/O-bound on bulk storage rather than GPU memory.

• ·Individual developers and researchers running 7B–70B class models locally
• ·Small-team RAG prototypes — up to 10 concurrent users on a 13B model
• ·LoRA / QLoRA fine-tuning on consumer-class GPUs
• ·Cost-sensitive entry into on-prem AI — quiet office deployment

Two RTX 5090s give 64 GB of GPU memory across both cards. Comfortably runs 7B–13B at FP16, and 30B–70B with INT4 / INT8 quantization on a single card. Tensor parallelism across the two cards is over PCIe (no NVLink).

Tower workstation on a standard 240 V wall outlet. Air-cooled, sub-55 dBA under load. 2.5 GbE on-board with optional 10 GbE add-in. No special HVAC or rack needed.

Step up to AI PRO 500 when you need 4 GPUs, ECC memory, or quieter office operation under sustained departmental load. Step to the EMARQUE AI Server (RTX PRO 6000 SE config) for production rackmount deployments.