Public GPU and NPU uploads

NVIDIA RTX 5090

21760 CUDA cores

Type

GPU

VRAM

32 GB

Memory bandwidth

1,792 GB/s

TDP

575 W

Benchmark results

LLM name	Quantization	Used tokens Token	Prompt processing Token/s	TTFT ms	Token Generation Token/s	Backend	Engine	Power W	Latest upload ↓	User
Llama 2 7B	Q4_0	512	10,381.64	49.00 ms	264	llama.cpp	Vulkan	—	uploaded 4 weeks ago	AI Hardware Research System
Standardized test Llama-Bench Used prompt llama-bench -p 512 -n 128 Notes llama-bench / Vulkan scoreboard; Flash Attention deaktiviert Evidence Source Performance of llama.cpp with Vulkan · ggml-org/llama.cpp · Discussion #10879 · GitHub github.com