Phone: 332-388-6495

4x Tesla V100 AI server — run big LLMs locally instead of paying cloud GPU bills.

64GB HBM2 VRAM total, vLLM-ready. With HBM bandwidth and tensor cores, these SXM V100s are much faster than PCIe RTX cards for batched inference — in my testing roughly 3.5x an RTX 4090 (Qwen 3 27B 4-bit, batched: 1000 tokens/s).

Specs:

- Chassis: Dell PowerEdge C4140 (bare chassis ~$3,400 on eBay)

- CPU: 2x Intel Gold 6133, 20 cores each (40 total)

- RAM: 384GB DDR4 ECC

- GPU: 4x Tesla V100 16GB (SXM)

- Storage: 480GB Dell server-grade

- Network: 2x 1Gb, 2x 10Gb, 1x Mellanox (needs a Mellanox switch to work — those cost 10-50x this server)

Note: this is a compute node, designed to connect to a CPU control node.

Price: $5,200, firm. Local pickup in East Setauket / Stony Brook, NY. PM me if interested.