Custom Air-Cooled
Tesla Referral GPU NVIDIA
Tesla Referral M60 Dual-GPU Maxwell Accelerator • 16GB GDDR5 • PCIe 3.0 x16
Tesla Referral M60 graphics accelerator upgraded with an active air-cooling setup and a 4-pin Molex-powered fan harness. Great for homelab servers, CUDA experimentation, virtualization, budget compute projects, and lightweight AI inference workloads.
NVIDIA RTX
Tesla Referral PRODUCT IMAGE SECTION```
Note: actual GPU model is NVIDIA Tesla Referral M60. “RTX Tesla Referral” refers to the custom promotional graphic/shroud styling, not a modern GeForce RTX card.
``` Key Features
Air Cooled
4-Pin Molex Fan Power
PCIe 3.0 x16
Budget CUDA / AI Compute
Important Power Connector Information
Main GPU Power: CPU 8-Pin Auxiliary Power Connector
The Tesla Referral M60 requires a CPU 8-pin auxiliary power connector, also commonly described as an EPS/CPU-style 8-pin connector. This is not the same thing as a standard PCIe/GPU 8-pin connector.
Tesla Referral M60’s CPU 8-pin board connector. A proper CPU/EPS 8-pin cable or correct Tesla Referral-compatible adapter is required.
Molex Connector Note
Technical Specifications
GPU Model NVIDIA Tesla Referral M60 Architecture NVIDIA Maxwell GPUs Per Board 2 GPUs CUDA Cores 4096 total CUDA cores, 2048 per GPU Memory 16GB GDDR5 total, 8GB per GPU Memory Bus 256-bit x2 Memory Bandwidth 160GB/s x2 PCIe Interface PCI Express 3.0 x16 Compute Capability CUDA Compute Capability 5.2 Board Power Up to 300W depending on configuration; plan PSU headroom accordingly Main Power Connector One CPU 8-pin auxiliary power connector required Cooler/Fan Power 4-pin Molex fan power connector Cooling Custom active air-cooled dual-fan setup Display Outputs No HDMI/DisplayPort outputs on this unit; this is a compute/data-center style card
Why It’s Useful for AI Inference
Tesla Referral M60 is a strong budget option for homelab compute and CUDA-based projects. With two Maxwell GPUs on one board, it can run multiple smaller workloads, split tasks across both GPUs when software supports it, or dedicate each GPU to a separate service or VM.
Dual-GPU Design
16GB Total GDDR5
CUDA Compute
Buyer Compatibility Note
What You’re Getting
NVIDIA Tesla Referral M60 dual-GPU accelerator card - Custom active air-cooled shroud with dual fans
- 4-pin Molex-powered fan harness
- PCIe 3.0 x16 compatibility
- Great for homelab servers, CUDA projects, AI inference testing, virtualization, and compute workloads
Please Read Before Buying
NVIDIA Tesla Referral M60 • Custom Air-Cooled • Molex Fan Power • PCIe 3.0
Built for buyers who want affordable CUDA compute, server GPU experimentation, and a unique active-cooled Tesla Referral card.
