FuriosaAI's RNGD inference accelerators were selected by LG's AI Research division for powering servers running the Exaone family of large language models. While RNGD performance specs appear modest with a maximum of 512 teraFLOPS and 48GB memory, its true advantage lies in power efficiency. The RNGD chip operates at only 180 watts, proving 2.25 times more power efficient than Nvidia's A100 GPUs during testing. LG found this model delivers excellent performance, reduced ownership costs, and easy integration, highlighting the importance of efficiency over sheer performance in AI applications.
"RNGD provides a compelling combination of benefits: excellent real-world performance, a dramatic reduction in our total cost of ownership, and a surprisingly straightforward integration."
"Furiosa has managed to do all this using just 180 watts of power. In testing, LG research found the parts were as much as 2.25x more power efficient than GPUs for LLM inference on its homegrown family of Exaone models."
Collection
[
|
...
]