Neural Inference | AI & Hardware

The rig that started it all. Built in 2017, "The Veteran" represents the Pascal era of GPU compute. Despite its age, it serves as a crucial baseline for testing high-efficiency quantized models (GGUF) that can run on older consumer hardware.

GPUNVIDIA GTX 1080

VRAM8GB GDDR5X

CPUIntel Core i7-8700K

RAM64GB DDR4

Storage1TB NVMe Gen3

OSUbuntu 22.04 LTS

Shop This Rig ↗

Performance Log

Date	Model	Hardware	Params	Context	Prompt Eval	Token Gen
2026-02-01	Gemma-3-4.3B	NVIDIA GTX 1080	Q4_K_M	4096 tk	1259.1 t/s	46.3 t/s
2026-02-01	Gemma-3-4.3B (32k Context)	NVIDIA GTX 1080	Q4_K_M	32768 tk	984.0 t/s	41.1 t/s
2026-02-01	Gemma-3-4.3B (64k Context)	NVIDIA GTX 1080	Q4_K_M	65536 tk	791.8 t/s	36.8 t/s

SPONSORED TESTS// AD_SLOT: 1234567890 // FORMAT: AUTO

Performance Analysis

Trend Visualization

Historical Archive

Filter:

Showing 3 of 3 runs

Date	Model	Quant	Context	VRAM	Prompt (t/s)	Gen (t/s)
2026-02-01	Gemma-3-4.3B	Q4_K_M	4,096	5.5GB	1259.1	46.3
2026-02-01	Gemma-3-4.3B (32k Context)	Q4_K_M	32,768	5.9GB	984.0	41.1
2026-02-01	Gemma-3-4.3B (64k Context)	Q4_K_M	65,536	6.5GB	791.8	36.8