Raspberry Pi 5 with 16GB RAM Announced at $350
The Raspberry Pi Foundation has released the Raspberry Pi 5, a new single-board computer available in a 16GB RAM variant priced at $350, with additional 2GB, 4GB, and 8GB options. The device features a 2.4GHz quad-core Arm Cortex-A76 processor, USB 3.0 ports, improved Gigabit Ethernet with PoE capability, and custom in-house silicon (RP1 southbridge) that significantly enhances I/O performance. The upgrade delivers 2-3× CPU performance improvement and substantially faster graphics, camera, and peripheral support compared to the Raspberry Pi 4, though it requires new cases due to its redesigned form factor.
The Raspberry Pi 5 represents a significant redesign of the popular single-board computer platform, introducing the first full-size Raspberry Pi to use custom silicon designed in-house. The 16GB variant is priced at $350, with smaller RAM configurations also available. Key improvements include a 2.4GHz 64-bit quad-core Arm Cortex-A76 processor delivering 2-3× performance gains over the Pi 4, a VideoCore VII GPU with OpenGL ES 3.1 and Vulkan 1.2 support, dual 4Kp60 HDMI output with HDR, and dual-band Wi-Fi. The custom RP1 southbridge chip doubles aggregate USB bandwidth, triples MIPI camera/display interface bandwidth to 1.5Gbps, doubles SD card performance via SDR104 support, and introduces a PCIe 2.0 x1 interface for high-bandwidth peripherals. The device maintains the standard 40-pin GPIO header and adds a real-time clock with battery backup and a power button, though the significant redesign means existing Raspberry Pi 4 cases are incompatible.
What different sources said
- Hacker NewsCenter
Raspberry Pi 5 – 16 GB, $350
Related
AI Self-Correction Capabilities Remain Key Challenge, Says Raindrop Founder
Ben Hylak, founder and CTO of AI monitoring company Raindrop, identified AI's ability to fix its own mistakes as a critical remaining problem for the technology sector. Hylak is a former engineer at SpaceX and Apple who now leads efforts to help companies improve baseline AI model performance. The advancement of AI self-correction capabilities is considered essential for broader AI deployment and reliability.
AI Companies Pursue State-Level Regulations as Federal Action Stalls
Major AI firms are actively shaping state-level regulations rather than waiting for federal legislation, as Congress and the White House have not yet established a national AI safety framework. While these companies publicly support a unified national standard, they are simultaneously engaging with individual states to influence policy. This shift reflects growing frustration with the pace of federal action and raises questions about whether fragmented state rules will ultimately undermine the industry's stated preference for uniform standards.
Researchers Develop Low-Cost Foundation Model Training Method at $1,500
Researchers at Sapient Intelligence developed HRM-Text, a foundation language model that can be trained from scratch for approximately $1,500, compared to the millions typically required. The approach uses a Hierarchical Recurrent Model architecture that trains on instruction-response pairs rather than raw text, achieving performance competitive with much larger models on industry benchmarks. This development could democratize AI model training for enterprises that currently cannot afford the computational costs of building foundation models.