Page 120 – Electronics-Lab.com

ams OSRAM TOPLED SFH 2240 A01 Photodiode

Posted on 18 November, 202218 November, 2022 by mixos

ams OSRAM TOPLED® SFH 2240 A01 Photodiode provides a spectral sensitivity adapted to human eye sensitivity (Vλ). The SFH 2240 A01 features a fast switching time, high linearity, and a low-temperature coefficient of spectral sensitivity. The device is AEC-Q102 qualified in a clear silicone package designed for in-cabin sensing applications.

Features

Clear silicone package
ESD: 1.5kV acc. to ANSI/ESDA/JEDEC JS-001 (HBM)
Spectral sensitivity adapted to human eye sensitivity (Vλ)
AEC-Q102 Qualified
Low temperature coefficient of spectral sensitivity
High linearity
Fast switching time

Relative Spectral Sensitivity

more information: https://www.osram.com/ecat/com/en/class_pim_web_catalog_103489/prd_pim_device_18386448/

Weller Tools introduces new smart soldering platform

Posted on 18 November, 202222 November, 2022 by mixos

Weller Tools has introduced its new smart soldering platform WXsmart. The platform consists of a 2-channel soldering station and a hot air module with an air and vacuum requirements, from pico to micro and ultra soldering are part of this platform. A calibration unit as well as a soldering tip holder make it a complete platform.

Weller Tools states that with the Total Process Control (TPC), the WXsmart offers maximum connectivity and traceability at the highest quality. The platform fully supports all IoT standards. The hand soldering station also has cyber security to prevent costly downtimes.

Main benefits:

Connect – Weller WXsmart offers maximum connectivity, supporting IoT standards for full traceability with mobile-enabled, remote-control access.
Smart – The all-in-one station acts as the brain of the workbench by consolidating all soldering applications into one station. The platform fully supports Smart Soldering 4.0 with intelligent tips and tools, saving cost, time and space.
Control – Total Process Control (TPC) to manage the complete soldering process from tip to station.
Safe – Built-in cybersecurity to prevent costly downtimes.
Auto Calibration – Captures accurate data and documentation by automatically re-adjusting its soldering calibration for strong auditing.
Modularity – Backwards compatibility of tips and tools for soldering, desoldering and hot-air applications, ensures the security of the all-in-one station investment.

The platform is comprised of the following units:

WXsmart Soldering Station WXsmart: Powerful two-channel station with 300 W, ESD-safe touch screen and a functional benchtop controller (preheating plated, solder baths, filtration and calibraiton unit)

Hot Air Module WXair: WXair is a self-contained rework module that can operate a standalone unit with two channels: one for vacuum and one for hot air. In combination with the WXsmart soldering station, it is an all-in-one rework station with four channels. This benchtop control operates as the brain of a workbench by consolidating all soldering applications into one station. The platform fully supports Smart Soldering 4.0 with intelligent tips and tools saving cost, time and space.

WXMPS MS Smart Soldering Iron: With an ergonomic pencil design, this iron fits comfortably in the operator’s hands, providing precision for small soldering applications under the microscope. Additionally, it is ideal for micro and standard components, covering a wide range of applications for electronics, medical devices, sensor technology, aerospace, automotive and military. The WXMPS MS is a unique two-in-one smart soldering iron that allows the operator to use two different tip series with one soldering iron. The Pico Tip Series RTPS MS for component sizes from 0402 to 01005 (XS-S) as well as Micro Tip Series RTMS MS for component sizes from 0504 to 1608 (M-L) are compatible with the same soldering iron WXMPS MS. The WXUPS handles component sizes up to 1612 and bigger (XL-XXL).

These soldering pencils strategically use Weller´s active (cartridge) tip technology for optimised heat transfer and maximum power with 40W performance. The rapid heat-up time (within 3 seconds for Pico tips) delivers excellent heat transfer and recovery time during the soldering process. Short tip-to-grip distance enables precise work.

Smart Tips: With a new generation of RT MS Smart Tips, Weller meets these increasing demands and more. These new Smart Tips include an integrated chip in every soldering tip, which offers many new features to the user. The tips provide full traceability from a unique, individual serial number on every tip as well as automatic tip identification because the WXsmart display shows each tip description. Additionally, Weller’s Tip Lock functionality provides full process control and maximum precision via the calibrated smart tips.

WCU Measurement and Calibration Device: Thid device captures accurate data and documentation by automatically readjusting the soldering calibration for strong auditing.

Soldering Tip Holder WCTH: The WCTH is a clever storage unit for Weller cartridges that helps organising the workbench. Sixteen tips can be stored when not in use.

OKW’s Smart New MINI-DATA-BOX For Miniaturised Go-Anywhere Electronic Devices

Posted on 18 November, 2022 by mixos

OKW’s smart new MINI-DATA-BOX is available with or without flanges for mounting the unit.

OKW has extended its wide range of sensor enclosures with the launch of tough, elegant new MINI-DATA-BOX. Its different versions, sizes, colours and optional flanges offer 32 possible permutations – all available as standard.

Innovative and compact MINI-DATA-BOX is ideal for applications including IoT/IIoT, automation, security and surveillance technology, environmental monitoring, measurement and control, smart logistics, peripherals, interfaces and ICT. Its small size also makes it suitable for wearable electronics.

MINI-DATA-BOX can be specified in two shapes: Square and Edge (rectangular), either with or without flanges for screws and cable ties, enabling fast mounting to walls, ceilings, rails and masts.

Each enclosure offers plenty of flat surfaces for interfaces. The lids have distinctive bevelled corners – a ‘diamond cut’ design which reduces weight and further enhancesthe housings’ stylish, modern aesthetics.

The lid and base feature an all-round tongue–and–groove joint for added strength. They are assembled with tamperproof Torx screws, a key requirement for medical electronics. IP 65 ingress protection is optional. Inside, there are fixing supports for PCBs and components in both the top and bottom sections. Alternatively, potting of electronic assemblies is possible.

MINI-DATA-BOX can be specified in plan sizes 40 x 40 mm, 50 x 50 mm, 40 x 60 mm and 50 x 70 mm – all of which are available in two heights (15 and 20 mm). The enclosures are moulded from UV-stable and flame-resistant ASA+PC-FR (UL 94 V-0)in a choice of two standard colours, traffic white (RAL 9016) and anthracite grey (RAL7016). Custom colours are available on request.

Accessories include seals for IP 65 ingress protection and non-slip feet for desktop use.

OKW can supply MINI-DATA-BOX fully customised. Services include CNC machining, EMC shielding, lacquering, laser marking, printing of legends and logos, and installation/assembly of accessories.

For more information, view the OKW website: https://www.okw.co.uk/en/Plastic-enclosures/Mini-Data-Box.htm

NXP’s MCX N Advanced Microcontrollers Delivers best-in-class performance and Power Efficiency

Posted on 18 November, 202218 November, 2022 by Emmanuel Odunlade

Many people seem to be talking a lot about edge computing these days but the fact remains that the word still means different things to different people having different needs within the edge computing space. It happens most times that some applications even require additional optimizations out of their hardware, maybe a bit of extra focus on its connectivity, its power, or its performance, which makes edge computing platforms look architecturally different from one another.

To address some challenges that this diversity often creates for designers, NXP Semiconductors has announced its intentions to launch a family of 4 new MCUs that target different needs within the edge computing space — something for everyone designing for Edge AI. One of these is the MCX N Advanced microcontrollers designed for high-performance energy-efficient edge AI workloads.

“Developers are increasingly looking to push the boundaries of what’s possible at the edge as they create new devices that can better anticipate and automate in smart homes, smart factories, and smart cities. This requires advanced MCUs [Microcontroller Units] that are more efficient, simplify edge intelligence, and do all of that securely,” NXP’s Rafael Sotomayor explains. “As we look to the future of MCUs, the MCX N series delivers the balance between power and performance for tomorrow’s IoT and industrial applications.”

The MCX N series designed for secure and intelligent applications promises low power consumption and a significant performance boost to edge AI performance. The series is offered in two models namely the MCX N94x and MCX N54x, both of which use a heterogeneous multi-core design with an Arm Cortex-M33 core and a streamlined core designed for control functions. The company also introduced an in-house neural processing unit architecture into the mix, which they claim to offer a third-fold performance gain for machine learning workloads.

“The multicore design delivers improved system performance and reduced power consumption by enabling smart, efficient distribution of workloads to the analog and digital peripherals,” says NXP’s CK Phua. “As a result, the MCUs consume less than 45μA/MHz active current, less than 2.5μA in power-down mode with the real-time clock (RTC) enabled and 8kB SRAM retention, and less than 1μA in deep power-down mode with the RTC active and 8kB SRAM.”

The two models offer equivalent performance but while the N94x chips offer peripherals designed for analog and motor control workloads, the N54x chips focus on high-speed USB with PHY, SD card, and smart card interfaces for IoT. Both however feature 2MB of flash and 512 kB of SRAM as well as a broad range of peripherals such as an integrated Digital Signal Processor, an EdgeLock secure subsystem, hardware root-of-trust, secure boot, and cryptographic acceleration.

The new MCUs should support the MCUXpresso software suite as well as the company’s development environment for machine learning projects. There would also be a number of RTOS (Real Time Operating System) to choose from.

The new MCUs are due to sample early next year and no particular date has been picked for general availability yet. Other useful details can be found on their website.

Meet the Raspberry Pi CM4-Powered “EdgeBox RPI-200 Rugged Industrial Edge Controller” Designed for Industrial IoT

Posted on 18 November, 2022 by Emmanuel Odunlade

We recently found out that Seeed Studio is offering a set of rugged industrial edge computing controllers designed for a variety of industrial applications: the EdgeBox RPI-200 series.

EdgeBox RPI-200 is a complete industrial controller equipped with a flexible and powerful processor capable of performing real-time operation and multi-process. It is powered by AI-enabled Raspberry Pi Compute Module 4 running at up to 1.5GHz and equipped with up to 4GB of RAM and 16GB eMMC.

The system also features rich IO resources and multiple industrial interfaces that achieve data transmission in a variety of situations. It comes with extended SSD storage, an enhanced HDMI 2.0 connection for better visual presentation as well as several communication capabilities such as an RJ45 Gigabit Ethernet and full-featured wireless interconnectivity (built-in 2.4GHz and 5GHz WiFi, Bluetooth 5.0 and Bluetooth Low Energy support). There is also a mini PCIe slot to support the installation of optional 4G cellular LoRaWan Gateway or ZigBee modules. Other features include 2x USB 2.0 ports, one RS485 port, and another RS232 port.

The industrial edge computing controller is small, compact, and lightweight. It is fitted in a rugged and maintenance-reducing high-grade structure with a full aluminum heatsink enclosure for maximum thermal management. It also supports 35mm DIN and wall mount which make it convenient to use where space is limited or in extreme conditions.

The EdgeBox also has great support for a variety of industrial software and platforms, including Codesys, Node-Red, MQTT, OPC UA, etc. It is perfect for high-end applications such as smart automation and control systems, IoT Gateways, 4G LTE Route, software-defined Programmable Logic Controller (PLC), Industrial controllers, edge computing controllers, etc.

Specifications Include:

Quad-core Raspberry Pi CM4 running at up to 1.5GHz
Up to 4GB RAM
Up to 32GB eMMC
Built-in 2.4GHz/5GHz IEEE 802.11 b/g/n/ac WiFi
Bluetooth 5.0, Bluetooth Low Energy (BLE)
1x GbE RJ45 port
Mini PCIe slot support 4G LTE/LoRaWAN Gateway or ZigBee module
1x HDMI 2.0 ports with 4k@60fps
1x M.2 socket support 2242 NVMe SSD
2x USB 2.0 Type-A
2x isolated DI and 2x isolated DO
1x RS-485 and 1x RS-232
1x Real Time Clock (RTC)
1x Watchdog Timer
1x optional UPS unit (Uninterrupted Power Supply)
Support for AWS and Azure Cloud Services
Power Supply: DC 12V to 36V
Dimension: 124 mm x 76 mm x 35 mm
Operating Temperature: -20°C to +60°C
OS: Debian Linux
Certification: RoHS, CE, FCC, TELEC, UKCA

The EdgeBox RPI-200 series is a family of 4 modules. The lowest version comes with 1GB RAM and 8GB eMMC storage and it is the only one without WiFi or Bluetooth. Next to that is the variant with 2GB RAM and 8GB eMMC, while the one ahead of it comes with 4GB RAM and 16GB eMMC. The top variant then comes with 4GB RAM and up to 32GB eMMC memory. All four models cost $229, $239, $269, and $279 respectively.

You can visit the Seeed Studio’s store for more details including purchase links and the corresponding shipping costs.

MeLe’s new PCG02 “Pro” Model accommodates a 2280 NVMe SSD and has Impressive Capabilities

Posted on 18 November, 202222 November, 2022 by Emmanuel Odunlade

The past few years have seen MeLe expand its PCG02 line of PC Stick, going from the MeLe PCG02 which first launched with an Ethernet port, to the orange stick preloaded with Ubuntu and to several other models which were based on new Intel processors with Apollo Lake and Gemini Lake variants. The company has now introduced a new “Pro” model that is the first in the series to come with support for an optional M.2 2280 PCIe NVMe SSD and the absence of an HDMI male port.

MeLe PCG02 Pro is one desktop PC that you can take with you anywhere you go. It weighs just 182 grams and measures 146 by 61 by 20 mm; as big as an iPhone 14 Pro but small enough to easily slide into your pocket without any struggles.

The fanless computer is offered in two variants which are powered differently by the Celeron J4125 Gemini Lake processor or the Celeron N5105 Jasper Lake Intel processor. It comes with 8GB of LPDDR4 memory and a storage interface that can be configured with 128GB/256GB/512GB of eMMC. The MeLe PCG02 Pro also packs flexible peripherals such as 2x USB 3.2 ports, 1x GbE RJ45 port, and 2x 4K@60Hz HDMI ports which are capable of driving two independent displays. It is further equipped with WiFi 5 or Wifi 6 connectivity as well as Bluetooth 4.2 or Bluetooth 5.2 depending on the particular variant and comes with an M.2 NVMe/SSD Storage in the mini PC stick that is only available in the Gemini Lake model.

The two variants of the computer ship pre-installed with Windows 11 Pro but they should also support most Linux distributions such as Ubuntu and Debian.

Specifications Include:

CPU: Intel Celeron J4125 quad-core Gemini Lake Refresh processor at up to 2.7 GHz (J4125) or Intel Celeron N5105 quad-core Jasper Lake processor at up to 2.9 GHz (N5105)
GPU: Intel UHD 600 graphics at 250/750 MHz (J4125) or Intel UHD 605 graphics at 450/800 MHz (N5105)
8GB LPDDR4 2133MHz / 2933MHz (J4125 / N5105)
128GB to 512GB eMMC flash
1x microSD card reader of up to 2TB
M.2 NVMe or SATA SSD support up to 4TB (J4125 only)
1x Gigabit Ethernet LAN port
WiFi 5 (J4125) and WiFi 6 (N5105)
Bluetooth 4.2 (J4125) and Bluetooth 5.2 (N5105)
2x 4K@60Hz HDMI 2.0 ports
1x 3.5mm earphone jack
Digital audio output via HDMI
1x USB Type-C (5Gbps)
2x USB 3.2 (5Gbps)
CMOS Reset
1x Kensington Lock slot
1x Power button
2x WiFi antenna
Power: 12V/2A via USB Type-C (supports USB PD3.0)
Dimension: 146 mm x 61 mm x 200 mm
Weight: 182 grams
Operating System: Windows 11 Pro

Fanless Tech is selling the Celeron J4125 variant of the Pro PC for around $229 and up while the Celeron N5105 variant, which delivers a higher CPU and graphics performance as well as better wireless capabilities, starts at $249. Please note that the varying prices depending on the size of the eMMC flash storage that they are delivered with.

Amazon also sells all variants of the mini PC but they come with a $30 off coupon for customers in the United States.

Ukama Provides Private Cellular Data Network for Anyone Anywhere

Posted on 18 November, 202222 November, 2022 by Emmanuel Odunlade

A great team of hardware and software engineers passionate about making life easier for people has provided a way for individuals and enterprises to create a custom 4G LTE-based cellular network with a modern stack. The team designed a decentralized cellular-data network that can be deployed and controlled by anyone anywhere.

“Ukama makes cellular internet accessible to everyone — whether you are looking for free, faster streaming on your smartphone at home, or bringing cellular data connectivity to a rural area for the first time, or rolling out a secure industrial IoT network,” says Kashif C. Ali, CEO and founder of Ukama. “It is a complete solution that includes SIMs (and support for eSIMs), a cellular base station, distributed and embedded cellular Core, a management console, and cloud services.”

Ukama is very easy to use and set up. All you need is electrical power and the internet and you can have free cellular data for your smartphone or IoT devices in no time. Ukama networks can enable applications across various sectors: warehouses, hospitals, manufacturing lines, ports etc.

“With Ukama, you can use all of your SIM devices (and even equip other devices with cellular adapters that take SIMs) without being tied to a big telco — you are the telco, albeit one that relies on your ISP for the backend connection to the rest of the internet,” the company further explained.

Some of the Key Benefits that come with using Ukama include:

It is a reliable and secure network that you can control. You don’t have to deal with poor network coverage or lack of service that are usually associated with other cellular data networks.
It gives you access to run a cellular data network anywhere for yourself and for others in your apartment, your neighborhood, or your locality.
You can decide to run your network for others either as a commercial service or for free.
It allows you to develop and deploy faster and more efficient IoT devices based on private cellular networks.
You can access global roaming if you are outside the coverage area of your own network.

Ukama also boasts of some unique features which stand it out amongst existing SoC-based and SDR-based networks. See comparison below

Features	Ukama	Others
Global eSIM	Yes	No
Distributed Core	Yes	No
Evolved Packet Core (EPP)	Embedded	Runs on GPP or requires additional hardware
Ease of Use	High	Moderate or Low
Power Consumption	Low	From low to medium to high
Managed Service available	Yes	No
Open hardware	Yes	Not all

Ukama could also be a good alternative for WiFi for many devices. The only difference there is between the two is that, unlike WiFi, Ukama provides an official way to share and bill anyone who connects to your Ukama network especially if you don’t want them hogging all the bandwidth.

Ukama is offered in four variants namely the Ukama Home Node, Ukama Tower Node, Ukama Amplifier Unit, and the UkamaTRX module. There’s also a table showing the differences between some of their hardware features.

Features	Home Node	Amplifier Node	Amplifier Unit	TRX Module
Highest Active Devices	8	128	None	8
RF connectors	2x SMA	7x SMA	3x SMA	5x U.FL
GPS module	None	Yes	None	None
Main Processor	1.6 GHz MIPS quad-core	1.6 GHz MIPS quad-core	500 MHz ARM single-core	1,6 GHz MIPS quad-core
Secondary GPP	None	Intel quad-core 1.9 GHz	None	None
Storage	16GB	32GB	2GB	16GB
RAM	2GB	2GB	128GB	2GB
Ethernet	1 Gbps	1 Gbps	100 Gbps	1 Gbps
Debug port	None	None	None	Serial
Cooling	Passive heat sinks	Passive heat sinks	Passive heat sinks	None available
Power Consumption	Less than 30W	Less than 45W	Less than 20W	Less than 25W
Power Input	12V DC barrel jack	PoE+	PoE	12V DC two pins
Temperature	0 to 50°C	-40 to +55°C	-40 to +55°C	-40 to +55°C
Dimension	8.2 x 4.8 x 1.3 inches	6.1 x 8.4 x 3.0 inches	4.6 x 9.4 x 4.0 inches	4.5 x 1.2 x 4.6 inches
Weight	1.36 kg	2.04	1.45 kg	0.68
Ruggedness	Indoor only	Outdoor	Outdoor	Workbench only

Some accessories and tools also available for sale include an Indoor 4G/5G monopole whip antenna, a 1-meter Ethernet cable, a data eSIM, and an outdoor high-gain antenna.

Prices for all are available on the crowdsupply product page where the product was launched. An $8 shipping fee applies to those in the United States while others elsewhere are to pay $18 for shipping.

Other valuable details on Ukama including support and documentation as well as global roaming rates can also be found on the Crowd supply page.

Solder Party’s FlexyPin Adapter offers a swift way to bring the Uno form-factor to the Raspberry Pi Pico

Posted on 18 November, 202218 November, 2022 by Abhishek Jadhav

A Sweden-based consumer electronics and computers retail specialist, Solder Party has released its FlexyPin Adapter that gives the Raspberry Pi Pico an Uno form factor. With the help of FlexyPins, the board allows the Raspberry Pi Pico to be used with Uno shields without having to solder the Pico in place.

For castellated modules that have a pitch as low as 1.27mm, a connector pin called the FlexyPin has been designed. Since the footprint drills would be so close together, one might be able to use it with a 1.00mm pitch but it would require a more expensive PCB. The well-formed and gold-plated FlexyPins enable easy removal and insertion of modules without the hassle of soldering. They work great in many use cases such as testing, flashing, and others where the castellated modules are involved.

The Raspberry Pi Pico boards make use of 3.3V logic, whereas the Arduino Uno is 5V. The Raspberry Pi Pico is not 5V safe, hence the compatibility between an adapted Raspberry Pi Pico and an original Arduino is not one-to-one. In addition, it exposes fewer analog pins, which could provide a challenge for some shields, however, the pin-mapping was chosen specifically to support an 8-bit 3.5-inch 480×320 parallel LCD shield without any issues. The LCD shield is driven by making use of the RPi’s PIO along with the bb_spi_lcd library.

To push against the castellated pin headers of a Raspberry Pi Pico or wireless-capable Raspberry Pi Pico W, the FlexyPins which are available as an extra are soldered into the board. Furthermore, along with an extra Serial Wire Debug (SWD) connector which is added to the bottom of the board, the pins are brought out to the Arduino Uno-format headers.

From Solder Party’s Lectronz store, the Raspberry Pi Pico to Uno FlexyPin adapter can be ordered for $1.50. Additionally, the soldering kit includes a reset button, a set of socket headers, and an optional pack of 100 FlexyPins. Under the CERN Open Hardware License v1.2, the design files are available on GitHub.

IoT startup Everactive explores the potential of batteryless IoT development kit for all

Posted on 18 November, 202218 November, 2022 by Abhishek Jadhav

California-based IoT startup Everactive has announced the release of an ultra-low-power semiconductor and wireless IoT development kit with the ability to operate without a battery. This enables developers to leverage their onboard power to operate in distributed edge locations and also easily scale their Internet of Things applications without the constraints of any battery or power supply.

Everactive has included a development kit with two of the company’s patented batteryless IoT devices, each with a sensor suite that can simultaneously measure temperature, pressure, and a triaxial accelerator. The IoT development kit uses a low-power indoor photovoltaic harvester as the only power source which makes it a batteryless IoT device. The hardware platform is capable of measuring and wirelessly transmitting data down to every 15 seconds.

“Our core breakthroughs in ultra-low-power wireless systems have overcome both the battery problems and networking issues that have plagued the IoT,” said Dr. Benton Calhoun, co-founder, and co-CTO at Everactive. “With that as the foundation, we then built an entire data delivery system specifically designed to manage the massive influx of new real-world data sets only made possible by batteryless technology. Ultimately, it is that data— and the derived insights— that matters to our developers and their end users.”

Everactive is achieving batteryless IoT hardware design through its patented energy harvesting solution that generates all the required power from multiple low-level energy harvesting sources, including indoor solar, thermal gradients, RF, vibration, and more. For seamless wireless communication, the company developed a mix of standard-compliant and proprietary integrated radios for broad data rates of up to several Mbps and communication ranges of kilometers.

Everactive Evernet is a wireless networking solution designed for high-density batteryless devices that takes advantage of the company’s always-on ultra-low-power receive that reduces power consumption while supporting thousands of devices with millisecond latency and 250+ meter non-line-of-sight range.

For more information on the product, you can consider reserving a spot for the evaluation kit. Everactive has started accepting orders for the batteryless IoT development kit and will begin shipping in November. According to public information, the hardware kit is sold at $599 which includes 2x ENV+ Eversensor, 1x USB Evergateway, and unlimited user accounts.

IBM has launched Artificial Intelligence Unit– Ready to replace CPU for deep learning training

Posted on 18 November, 202218 November, 2022 by Abhishek Jadhav

IBM believes that computing power and AI models have grown exponentially, but the supporting hardware to train them and run them on servers in the cloud or edge devices hasn’t advanced. IBM Research AI Hardware Center has decided to launch an application-specific integrated circuit for artificial intelligence that can be programmed to run any type of deep learning task– Artificial Intelligence Unit.

According to IBM, several manufacturers are looking for flexibility and high-precision CPUs that are suited for general-purpose software applications. But for deep learning models, high precision is something that puts the system at a disadvantage when it comes to training and running them, which requires massively parallel AI operations.

IBM took two paths to develop an application-specific integrated circuit that would be able to replace the central processing unit for deep learning tasks. One is to embrace lower precision as the company believes AI chips do not require ultra-precise because they are not calculating trajectories for landing a spacecraft on the moon. The second is that an AI chip should streamline AI workflows.

IBM Artificial Intelligence Unit is an upgraded and scaled version of the company’s proven AI accelerator built into its Telum chip. The hardware features 32 processing cores that resemble the AI core embedded in the Telum chip that currently powers the company’s z16 system. One of the other differences is that the new IBM AIU is faster with smaller 5nm transistors compared to 7nm in the Telum.

IBM led its efforts in the space of artificial intelligence with a mission to improve AI hardware efficiency by 2.5 times every year. By 2029, IBM aims to train and run AI models one thousand times faster than they could three years ago. IBM expects the Artificial Intelligence Unit to be one of the most significant landmarks in the journey of AI acceleration.

IBM is yet to provide internal details of the Artificial Intelligence Unit, but the enterprise-grade, industry-scale AI hardware is optimized for many types of matrix and vector multiplication operations. The company plans to reveal more details as time goes by.