MemryX MX3 edge AI accelerator delivers up to 5 TOPS, is offered in die, package, and M.2 and mPCIe modules

MemryX MX3 EVB

Jean-Luc noted the MemryX MX3 edge AI accelerator module while covering the DeGirum ORCA M.2 and USB Edge AI accelerators last month, so today, we’ll have a look at this AI chip and corresponding modules that run computer vision neural networks using common frameworks such as TensorFlow, TensorFlow Lite, ONNX, PyTorch, and Keras. MemryX MX3 Specifications MemryX hasn’t disclosed much performance stats about this chip. All we know is it offers more than 5 TFLOPs. The listed specifications include: Bfloat16 activations Batch = 1 Weights: 4, 8, and 16-bit ~10M parameters stored on-die Host interfaces – PCIe Gen 3 I/O and/or USB 2.0/3.x Power consumption – ~1.0W 1-click compilation for the MX-SDK when mapping neural networks that have multiple layers Under the hood, the MX3 features MemryX Compute Engines (MCE) which are tightly coupled with at-memory computing. This design creates a native, proprietary dataflow architecture that utilizes up to 70% […]

Official Raspberry Pi M.2 HAT+ launched for $12

Raspberry Pi M2 HAT+ on Raspberry Pi 5

The official Raspberry Pi M.2 HAT+ is finally out for $12. The add-on board allows users to connect M.2 M-key peripherals, mainly NVMe SSDs, but also AI accelerators, to their Raspberry Pi 5 leveraging the PCIe connector on the(relatively) new SBC. We have to stress “official” because it’s been possible to do the exact same thing with third-party boards from PineBerry (now PineBoards), Waveshare, Pimoroni, and Geekworm for about half a year. I also had the opportunity to review the GEEKWORM X1001 and Waveshare M.2 PCIe HAT+ with Cytron MAKERDISK SSDs last month. But let’s have a look at what the official Raspberry Pi M.2 HAT+ has to offer. Raspberry Pi M.2 HAT+ M Key specifications: M.2 M-key socket for 2230 or 2242 modules Single-lane PCIe 2.0 interface (500 MB/s peak transfer rate) routed via Raspberry Pi PCIe FFC connector. (Note: PCIe 3.0 should also work fine on most Raspberry […]

Pineboards adds four more Raspberry Pi 5 PCIe HAT+ boards with PCIe x4 slot, mPCIe socket, Coral Dual Edge TPU support, and audio ports

Pineboards Raspberry Pi 5 HAT

Pineboards, previously known as PineBerry, has launched four new Raspberry Pi HAT+ with a PCIe interface: the Hat uPCIty Lite, HatDrive! Piano, Hat mPCIe, and Hat Ai! Dual. The Polish company decided to change the name from Pineberry to Pineboards since the “berry” name implied they were manufacturing single-board computers, while Pineboards, not to be confused with Pine64 boards, apparently does not :). Nevertheless, let’s have a look at the four new HAT+ boards. Hat uPCIty Lite – PCIe x4 slot and ATX power supply Specifications: Supported SBCs – Raspberry Pi 5 and other SBCs with a compatible 16-pin PCIe FPC connector and form factor PCIe 16-pin PCIe FPC connector (input) PCIe X4 slot (output) to connect PCIe cards such as NVIDIA or AMD graphics cards; Note: only PCIe x1 Gen2 and Gen3 are supported Misc – 12V and PWR LEDs Power Supply 12V/8A via ATX power connector 12V/8A via […]

Arduino announces the Pro 4G Module and Portenta Mid Carrier board at Embedded World 2024

Arduino Pro Portenta Mid carrier and 4G Module

Arduino has announced two new products that are part of the Arduino Pro family at Embedded World 2024: the Arduino Pro 4G Module and Arduino Portenta Mid Carrier board. Arduino Pro 4G module Two variants of the mini PCIe module are available: one global and one for the EMEA and Southeast Asia markets with similar interfaces, but based on different Quectel modules, and the global model also supports GNSS. Specifications: Both also support 3G/2G fallback if 4G LTE network connectivity is out of range. The 4G LTE module works with Arduino Pro carrier boards featuring a mini PCIe connector such as the Max Carrier and the new Mid Carrier which we will be covering below. More information can be found on the documentation website and product page. The Arduino Pro 4G module global goes for $65 while the EMEA version sells for 29 Euros. Those prices are without antennas which […]

Axelera Metis PCIe Arm AI evaluation kit combines Firefly ITX-3588J mini-ITX motherboard with 214 TOPS Metis AIPU PCIe card

Metis PCIe Arm Evaluation Kit

Axelera has announced the general availability of several Metis PCIe AI Evaluation Kits that combine the company’s 214 TOPS Metis AIPU PCIe card with x86 platforms such as Dell 3460XE workstation and Lenovo ThinkStation P360 Ultra computers, Advantech MIC-770v3 or ARC-3534 industrial PCs, or the Firefly ITX-3588J mini-ITX motherboard powered by a Rockchip RK3588 octa-core Cortex-A76/A55 SoC. We’ll look into detail about the latter in this post. When Axelera introduced the Metis Axelera M.2 AI accelerator module in January 2023 I was both impressed and doubtful of the performance claims of the company since packing a 214 TOPS Metis AIPU in a power-limited M.2 module seemed like a challenge. But it was hard to check independently since the devkits were not available yet although the company only started their early-access program in August last year. Now, anybody with an 899 Euros and up budget can try out their larger Metis […]

PCIe 7.0 to support up to 512GB/s bidirectional transfer rates

PCIe 7.0 Speed 512GB/s

The PCI-SIG first unveiled the PCIe Express (PCIe) 7.0 specification at US DevCon in June 2022 with claims of bidirectional data rates of up to 512GB/s in x16 configuration, and the standard is now getting closer to the full release in 2025 with the release of the specification version 0.5. PCIe 7.0 increases data transfer speeds to 128 GT/s per pin doubling the 64 GT/s of PCIe 6.0 and quadrupling the 32 GT/s of PCIe 5.0, delivering up to 256 GB/s in each direction in x16 configuration, excluding encoding overhead. In other words, the total maximum bandwidth of a PCIe 7.0 x1 interface (32GB/s) would be equivalent to PCIe Gen3 x16 or PCIe Gen4 x8 as shown in the table below. PCIe 7.0 highlights: 128 GT/s raw bit rate and up to 512 GB/s bidirectionally via x16 configuration PAM4 (Pulse Amplitude Modulation with 4 levels) signaling Doubles the bus frequency […]

BeagleY-AI SBC features TI AM67A vision processor with 4 TOPS AI accelerators

Texas Instruments AM67A single board computer

The BeagleBoard.org Foundation’s BeagleY-AI is an open-source hardware, credit card-sized SBC powered by a Texas Instruments AM67A quad-core Cortex-A53 vision processor with various programmable blocks capable of delivering up to 4 TOPS for AI algorithms. The board ships with 4GB RAM, relies on a microSD card slot for storage and OS booting, implements gigabit Ethernet, WiFi 6, and Bluetooth 5.4 connectivity, and can drive up to three displays via micro HDMI, OLDI (LVDS), and MIPI DSI interfaces. The BeagleY-AI also comes with two MIPI CSI camera interfaces, four USB 3.0 ports, a USB Type-C port, and a 40-pin GPIO header for expansion. We can also see a 16-pin PCIe FPC connector that looks somewhat similar to the 20-pin PCIe connector on the Raspberry Pi 5 but officially supports PCIe Gen3 x1. BeagleY-AI specifications: SoC – Texas Instruments AM67A (J722S) “vision processor” CPU Quad-core 64-bit Arm Cortex-A53 processor at 1.4GHz Arm […]

Efinix Titanium Ti375 FPGA offers quad-core hardened RISC-V block, PCIe Gen 4, 10GbE

Efinix Titanium Ti375 RISC-V FPGA SoC

Efinix Titanium Ti375 SoC combines high-density, low-power Quantum compute fabric with a quad-core hardened 32-bit RISC-V block and features a LPDDR4 DRAM controller, a MIPI D-PHY for displays or cameras, and 16 Gbps transceivers enabling PCIe Gen 4 and 10GbE interfaces. The Titanium Ti375 also comes with 370K logic elements, 1.344 DSP blocks, 2,688 10-Kbit SRAM blocks, and 27,53 Mbits embedded memory, as well as DSP blocks optimized for computing and AI workloads, and XLR (eXchangeable Logic and Routing) cells for logic and routing. Efinix Titanium Ti375 specifications: FPGA compute fabric 370,137 logic elements (LEs) 362,880 eXchangeable Logic and Routing (XLR) cells 27,53 Mbits embedded memory 2,688 10-Kbit SRAM blocks 1,344 embedded DSP blocks for multiplication, addition, subtraction, accumulation, and up to 15-bit variable-right-shifting Memory – 10-kbit high-speed, embedded SRAM, configurable as single-port RAM, simple dual-port RAM, true dual-port RAM, or ROM FPGA interface blocks 32-bit quad-core hardened RISC-V block […]

Exit mobile version
UP 7000 x86 SBC