Local AI Hardware / GPUs for local AI
NVIDIA Tesla V100 32GB (PCIe) for local AI
Sold prices for the NVIDIA Tesla V100 32GB (PCIe) are tracked every day on GPU Sold (https://gpusold.com/gpu/tesla-v100-32).
32 GB of HBM2 with high bandwidth, passively cooled.
- Memory
- 32 GB
- Memory bandwidth
- 900 GB/s
- Board power
- 250 W
- Architecture
- Volta
Before you buy a used GPU for AI
- Memory first. The model and its context have to fit in VRAM (or it spills to system RAM and slows to a crawl). Use the VRAM calculator.
- Bandwidth second. Token generation speed is mostly limited by memory bandwidth, not compute.
- Check memory temperatures under load on GDDR6X cards (3090, 3090 Ti). Worn thermal pads are common on ex-mining cards.
- Plan power and slot spacing before buying two or more cards. Transient spikes on high-end cards trip undersized power supplies.