GPU Upgrade Inquiry: Dell Precision T5810 (685W PSU) for AI Model Running, Training & Finetuning

Hello Dell Community,

I am looking to upgrade the graphics card in my Dell Precision Tower 5810 to support local AI workloads, specifically training and finetuning Large Language Models (LLMs).

Current System Specifications:

• Workstation: Precision Tower 5810

• CPU: Intel(R) Xeon(R) CPU E5-1680 v4 @ 3.40GHz

• RAM: 32GB (4x8GB) DDR4-2400MHz (Quad-Channel)

• Current GPU: AMD FirePro W7100

• Power Supply (PSU): 685W

• OS: Windows 10

My Goal: I need to install a modern NVIDIA GPU (to utilize CUDA) with a high VRAM capacity (ideally 12GB or 16GB or 24GB) for AI training tasks. I am currently considering the NVIDIA RTX 4060 Ti (16GB) or the RTX 3060 (12GB).

Questions for the experts:

1. Power Constraints: With the 685W PSU, what is the maximum TDP (wattage) recommended for a single GPU in Slot 2 (PCIe 3.0 x16)?

2. Cabling: I currently see 6-pin power leads. Does Dell offer an official dual 6-pin to 8-pin adapter (Part #D92C9 or similar) that is safe for the high-sustained power draw of AI training?

3. Physical Clearance: Are there specific height or length restrictions I should be aware of regarding the front fan shroud or the side panel crossbar when choosing a modern "triple-fan" card?

4. Compatibility: Are there any known BIOS/UEFI limitations on the T5810 that would prevent it from recognizing NVIDIA 40-series cards?

Thank you for your guidance!