Qualcomm had two main announcements at Embedded World 2024: the ultra-low-power Qualcomm QCC730 WiFi microcontroller for battery-powered IoT devices and the Qualcomm RB3 Gen 2 Platform hardware and software solution designed for IoT and embedded applications based on the Qualcomm QCS6490 processor that we’re going to cover today.
The kit is comprised of a QCS6490 octa-core Cortex-A78/A55 system-on-module with 12 TOPS of AI performance, 6GB RAM, and 128GB UFS flash connected to the 96Boards-compliant Qualcomm RBx development mainboard through interposer, as well as optional cameras, microphone array, and sensors.
Qualcomm QCS6490/QCM6490 IoT processor
Specifications:
- CPU – Octa-core Kryo 670 with 1x Gold Plus core (Cortex-A78) @ 2.7 GHz, 3x Gold cores (Cortex-A78) @ 2.4 GHz, 4x Silver cores (Cortex-A55) @ up to 1.9 GHz
- GPU – Adreno 643L GPU @ 812 MHz with support for Open GL ES 3.2, Open CL 2.0, Vulkan 1.x, DX FL 12
- DSP – Hexagon DSP with dual HVX and 4K HMX
- VPU – Adreno 633 VPU up to 4K60 decode for H.264/H.265/VP9, Up to 4K30 encode for H.264/H.265; Support for HDR10 and HDR10+ playback
- AI – 6th gen Qualcomm AI Engine that combines Compute Hexagon DSP with dual Hexagon Vector, eXtensions (HVX), Hexagon Co-processor (Hexagon CP) 2.0 and Hexagon Tensor accelerator for up to 12 TOPS of AI performance
- Memory
- Dual channel non-PoP LPDDR5 up to 3200 MHz
- Dual channel non-PoP LPDDR4X up to 2133 MHz
- Storage – UFS 2.x/3.1, two-lane HS gear 4, SD v3.0, eMMC 5.1, PCIe two-lane NVMe
- Display
- DPU – Aderno 1075 DPU
- DisplayPort
- 1x 4-lane DSI DSC1.2, D-PHY 1.2, or C-PHY 1.0; VESA DSC 1.2
- Camera
- Spectra ISP 570L – Triple 14-bit image signal processing (ISP) + two lite ISP 22 + 22 + 22 MP, 64 MP/30 fps
- 5x 4-lane CSIs (4/4/4/4/4) D-PHY 1.2 or C-PHY 1.2
- 64 MP / 36 + 22 MP / 3×22 MP at 30fps ZSL 192 MP non-ZSL
- Connectivity
- Modem (QCM6490 only)
- 2G/3G/4G/5G – mmWave and sub-6 GHz bands (Rel. 15)
- 3.7 Gbps DL, 2.5 Gbps UL, 400 MHz mmW, 100 MHz sub-6
- Wi-Fi 6 (802.11ax) and Wi-Fi 6E (6 GHz) with Uplink/Downlink MU-MIMO, 4K QAM, 160MHz channels (5 & 6 GHz)
- Bluetooth 5.2 and FM supported
- GNSS – GPS, GLONASS, NavIC, BeiDou, Galileo, QZSS, and SBAS
- Modem (QCM6490 only)
- USB – USB 3.1 Type-C with DisplayPort, USB 2.0
- PCIe – 2x PCIe interfaces
- Process – 6nm
While it’s the first time I’ve heard about the QCS6490, we already mentioned the 5G modem-equipped QCM6490 in our article about the Fairphone 5 smartphone. While the QCM6490 supports Android “with long-term support for OS upgrades, security updates, and enterprise”, the QCS6490 found in the RN3 Gen 2 platform runs “Qualcomm Linux” with an LTS kernel and an IoT software stack. As an IoT processor, the QCS6490 gets a 15-year longevity period.
Qualcomm RB3 Gen 2 Platform
There are two versions of the Qualcomm RB3 Gen 2 Platform with the Vision Kit including cameras and the Core Kit with the main board and a heatspreader.
Both share the following specifications:
- SoC – Qualcomm QCS6490 octa-core AI SoC as described above
- System Memory – 6 GB LPDDR4x (uMCP package)
- Storage
- 128 GB UFS Flash (uMCP package)
- MicroSD card slot
- PCIe expansion socket for NVMe SSD
- Display
- Full-size HDMI connector
- USB Type-C supporting DP alt mode
- mini-DP connector
- DSI expansion
- Up to 2x simultaneous displays
- Camera
- Core Kit – 2x C-PHY/D-PHY 30-pin expansion ports on interposer board
- Vision Kit – 1x IMX577 D-PHY 12 MP, 1x OV9282 D-PHY 1 MP with bracket, plus additional D-PHY and GMSL-capable expansion ports
- Audio
- Core Kit – 1x DMIC, 2x digital audio amplifiers, I2S/Soundwire/DMIC expansion on low-speed connectors
- Vision Kit – 4x DMIC, 2x digital audio amplifiers, I2S/Soundwire/DMIC expansion on low-speed connectors
- Networking
- USB – 1x USB 3.0 Type-C port, 1x USB 2.0 w/OTG port, 2x USB 3.0 Type-A ports, 1x USB 3.0 on high-speed expansion
- PCIe
- 1x PCIe Gen 3 2-lane to expansion connector
- Optional 1x PCIe Gen 3 1-lane on expansion connector
- Sensors
- Core Kit – IMU onboard (ICM-42688), additional expansion
- Vision Kit – IMU (ICM-42688), Pressure sensor (ICP-10111), Mag sensor/compass(AK09915), additional expansion
- Expansion – Low-speed and High-speed connectors for 96boards Mezzanines
While the press release mentions support for Linux only, the product brief lists both Android and Linux and several SDKs: the Qualcomm Intelligent Multimedia Product SDK (for Linux), Qualcomm Intelligent Robotics Product SDK, Qualcomm Neural Processing SDK, and Hexagon SDK. Qualcomm Linux is currently available for private preview, is planned for wider availability to developers in the coming months, and is maintained by Foundries.io which Qualcomm just acquired. The recently announced Qualcomm AI Hub with a library of pre-optimized AI models is also compatible with the new platform
Compared to the Qualcomm RB3 robotics platform introduced 5 years ago with a Snapdragon 845 SoC, the RB3 Gen 2 delivers a 10x increase in on-device AI processing, supports quadruple 8MP+ camera sensors, computer vision, and integrates Wi-Fi 6E connectivity. Qualcomm expects the RB3 Gen 2 to be integrated into robots, drones, industrial handheld devices, industrial and connected cameras, AI edge boxes, intelligent displays, and more.
Qualcomm RB3 Gen 2 platform is available for pre-order now on Thundercomm for $399 (Core Kit) and $599 (Vision Kit) with a 12V wall power supply, a USB Type-C cable, mini speakers, a setup guide, and a pick tool for setting switches, and the Vision Kit also adds a mounting bracket for the high-resolution and low-resolution CSI cameras part of the kit. More details may be found on the product page.
Jean-Luc started CNX Software in 2010 as a part-time endeavor, before quitting his job as a software engineering manager, and starting to write daily news, and reviews full time later in 2011.
Support CNX Software! Donate via cryptocurrencies, become a Patron on Patreon, or purchase goods on Amazon or Aliexpress
It would be cool, if they’d supply SoC for sbc makers…imagine yet another Raspberry Pi killer. That would shake Rockchip and co 😁
They did release a few 96boards SBCs before such as the Dragonboard 410c: https://www.cnx-software.com/2015/03/12/qualcomm-dragonboard-410c-96boards-development-board/
However the price/performance ratio was not ideal, and Qualcomm does not like to publicly release documentation either…
too bad for us
how does this compare to the $70 Beagle board just announced which also has 4TOPS for AI tasks? This one seems much more expensive.
The Qualcomm solution is much more powerful with Cortex-A78 class cores, 12 TOPS of AI performance, more memory, faster UFS storage, and more interfaces including some with higher speed (e.g. PCIe Gen3 x2), etc… but it’s also pricier.
There’s now a Qualcomm RB3 Gen 2 *Lite* platform based on the Qualcomm QCS5430 processor with lower frequencies and an Adreno 642L GPU.
Not quite sure why it’s needed as the kits are only $10 cheaper.
https://www.thundercomm.com/product/qualcomm-rb3-gen-2-lite/#versions