Designing Next Generation 5G Wireless Infrastructure Using RISC-V
summary
LeapFrog presents a tile/chiplet architecture of clustered LPUs — a sea of RISC-V cores with a custom ISA and vector/matrix units that implement L1 plus optional L2/L3 on-chip. Claim shows 10x performance/W/cost versus legacy RAN and Intel FlexRAN, and power targets of about a third of typical embedded base stations (and under 0.1x versus FlexRAN). Design scales from small cells to mMIMO, supports disaggregated or inline L1, and is demoed with Ventana compute chiplets and a Mamer productization path.
AI-written summary · about
chapters (14)
- 0:00Introduction to LeapFrog
- 1:06Team and background
- 2:10Market drivers for 5G
- 3:54Challenges in 5G infrastructure
- 4:57RISC-V-based chiplet architecture
- 6:20AI-enhanced wireless processing
- 7:24LeapFrog Processing Unit design
- 8:45Disaggregating layer one
- 9:47Power efficiency and DU/CU integration
- 10:50Open RAN and integration models
- 11:55Software stack overview
- 12:56Go-to-market strategies
- 13:59Applications beyond wireless
- 14:22Summary and Q&A
related talks
15mRISC-V CPUs for the New Era of Heterogeneous Computing
11mA RISC V ISA Extension for Ultra Low Power IoT Wireless Signal Processing
13mWhy RISC-V is Inevitable
29mAccelerating the RISC-V Revolution: Unleashing Custom Silicon with Revolutionary Design Platforms
18mRISC-V Performance Delivered
10mRevolutionizing RISC V Based Application Design Possibilities With Global Foundries
12mGRVI Phalanx Update: Plowing The Cloud With Thousands Of RISC V Chickens
02h 18mRISC-V 101
