ai-acceleratorvendor
FuriosaAI RNGD
FuriosaAI's Renegade inference accelerator, with 48 GB of HBM3 on the vendor product page.
Good forData-center LLM and multimodal inference
- Power consumption
- 180 W
Specifications
- Maker
- FuriosaAI
- Region
- KOREA
- Class
- ai-accelerator
- Status
- shipping
- Claim level
- vendor
- Product
- Data-center inference NPU
- SKU
- RNGD
- Memory
- 48 GB
FuriosaAI’s product page lists RNGD at 48 GB of HBM3 and a 180 W power profile, and it describes the card as in production for LLM and multimodal inference. The 512 TFLOPS figure on that page is FP8, eight processing elements at 64 TFLOPS each, so it is not recorded as peak FP16. Customers are pointed to a Furiosa Access program to evaluate and deploy the card. Throughput charts on the page are marketing comparisons and are not entered as specifications.
Name meaning:Furiosa pronounces RNGD as Renegade.