GPU//discrete

A standalone GPU on its own PCB with dedicated GDDR/HBM memory.

A standalone GPU on its own PCB with dedicated GDDR/HBM memory.

Connected to the host CPU via PCIe or NVLink.

Delivers maximum compute throughput for ML training, rendering, and HPC.

Power draw ranges from 75W (entry) to 700W+ (datacenter)

Data must be explicitly transferred between host RAM and device memory; adds latency but offers massive bandwidth on-card.

Hand size; 450 W; for car cameras, heavier data streams, and datacenter workloads.