New memory integrates clock driver directly into the memory modules
Offers superior performance for heavy computational loads
Micron Technology has announced a new lineup of clock driver DDR5 memory under the Crucial brand, promising improved speed, stability, and performance.
This new generation of memory includes two form factors: CUDIMM (clocked unbuffered dual inline memory modules) for desktops and CSODIMM (clocked small outline dual inline memory modules) for laptops.
The clock driver is integrated directly into the memory modules, setting a new benchmark for high-performance computing.
Speed and stability with DDR5 clock driver integration
With this launch, Micron becomes the first company to release JEDEC-standard DDR5 CUDIMM and CSODIMM modules.
Regular DDR5 memory relies on the CPU's clock driver to maintain performance. However, Micron’s new design places the clock driver on the memory module itself, with the company claiming that the result is a more reliable memory operation and smoother performance, especially under heavy computational loads.
Both modules offer speeds of up to 6,400 MT/s (megatransfers per second), putting them 15% faster than non-clock-driver-based DDR5 and more than twice the speed of DDR4.
The new DDR5 memory modules have been validated for use with Intel Core Ultra processors (Series 2), with Intel certifying these modules for capacities up to 64GB, with plans to support even higher configurations - up to 256GB- on future AI-powered PCs and workstations.
The Crucial DDR5 CUDIMM and CSODIMM modules are available for immediate purchase in 16GB capacities through Crucial’s website, backed by a limited lifetime warranty.
Higher-capacity options, including 64GB modules, are expected to hit the market through distribution channels during the first half of 2025. These larger-capacity modules will allow users to build systems with total memory capacities of up to 256GB, providing more flexibility for intensive tasks.
“As AI takes flight, a memory paradigm shift is needed to keep pace with unprecedented system performance requirements,” said Dinesh Bahal, corporate vice president and general manager of Micron’s Commercial Products Group.
“Micron is shipping the industry’s first JEDEC-standard, commercially available DDR5 CUDIMM and CSODIMM solutions to power fast, out-of-the-box speeds for AI PCs and high-end workstations. With this new category, we are arming the ecosystem with next-generation memory solutions to future-proof today’s devices for tomorrow’s AI workloads."
Western Digital unveils new HDD storage options for businesses
32TB UltraSMR and 26TB ePMR CMR are the world's highest capacity ePMR HDDs
Built for scalability, efficiency, and long lasting with 5 year warranty
Western Digital has unveiled new advancements in hard drive technology which it says align with the burgeoning needs of AI and data-intensive applications.
The company has recently launched the new 32TB UltraSMR HDD and 26TB ePMR CMR HDD.
These new hard drives, equipped with innovative features like OptiNAND and ArmorCache offer scalable, efficient, and cost-effective solutions for hyperscalers, cloud service providers (CSPs), and enterprises.
Cutting-edge 11-disk design for maximum efficiency
The 32TB Ultrastar DC HC690 SMR HDD is the industry’s highest-capacity HDD to date, using cutting-edge UltraSMR technology to support deep content storage at scale. On the other hand, the 26TB Ultrastar DC HC590 CMR HDD offers a drop-in replacement for enterprises looking to grow their data capacity without compromising on performance or reliability.
The 32TB UltraSMR HDD offers up to 257 MiB/s sequential performance and using just 5.5W of power while idle making it ideal for "warm" and "cold" data storage tiers, helping businesses optimize their energy consumption while scaling their storage needs. Meanwhile, the 26TB CMR HDD features a sustained transfer rate of up to 288 MiB/s and consumes just 5.6W in idle mode, making it a reliable option for mission-critical environments.
Their 11-disk design is the first of its kind to be commercially available and increases the storage capacity without increasing the physical size of the drives, making them highly efficient for modern data centers.
Western Digital integrates the new HDDs into the Ultrastar Data60 and Data102 hybrid storage platforms which can accommodate up to 102 drives, offering up to 3.26 petabytes of raw capacity. The company also uses IsoVibe and ArcticFlow to reduce vibration and optimize cooling.
The 26TB WD Gold SATA HDDs offer a decent balance for system integrators and resellers. Drawing from the Ultrastar HDD platform, these drives are tailored for small and medium businesses who require reliable, high-performance storage, and feature up to 2.5 million hours mean time between failure (MTBF) and a five-year warranty.
“Our CMR and UltraSMR technology isn't just breaking records—it’s giving customers the efficiency and TCO benefits they’ve been asking for, and we’re not done yet,” said Ravi Pendekanti, SVP of Product Management, HDD Business Unit, Western Digital.
“By delivering the industry’s highest ePMR capacities available today, we are ensuring that our customers have the storage efficiency, scalability, reliability and unmatched value they need to stay ahead."
VPN usage in Pakistan has never been so high, the government is set to introduce a new policy that could potentially ban the use of "unregistered" services starting today, November 30, 2024.
People in Pakistan have been turning en masse to VPN services throughout 2024 as they look for a way to bypass growing censorship online. WhatsApp is just the latest social media platform currently blocked in the country. The block came two days after authorities restricted Bluesky amid a surge in popularity worldwide. X, Facebook, and Instagram also can't be accessed without a VPN.
I've spoken with some experts to understand what's at stake and discover everything you might need to know to retain internet freedoms in the country.
How to avoid disruption in Pakistan
Set up a manual VPN
Contact VPN support teams
Rotate VPN providers
If you experience some issues downloading or using your VPN app in the next days or weeks, Gytis Malinauskas, Head of Legal at Surfshark suggests setting up a manual VPN connection. You'll likely find the details on how to do this on specific devices on your provider's website such as this step-by-step guide from Surfshark. Remember, you'll still need a VPN account to do this.
If this doesn't work, contact your provider's support team for assistance. "Our team is continuously working to provide our users with privacy and security solutions even in regions with challenging internet landscapes," Malinauskas explains.
Both Malinauskas and Simon Migliano, Head of Research at Top10VPN, also recommend rotating through multiple VPN providers.
Are free VPNs a solution?
Potentially. I suggest checking out our best free VPN page to get all the most secure freebies – PrivadoVPNand Proton VPN are today's top picks.
"It's likely to be a bit of a lottery, at least at first," Migliano told TechRadar. "Based on the recent test runs of Pakistan's new firewall, I'm expecting VPN blocks to be much more widespread from November 30. But as we've seen in countries with a much more mature internet censorship regime, such as China and Iran, it's near-impossible to block all the VPNs, all the time."
It's worth keeping in mind that the legal grounds for using a VPN to bypass restrictions in Pakistan after November 30 aren't yet clear. Commenting on this point, Migliano said: "Using an unregistered VPN in Pakistan after the deadline will be a risky proposition, depending on how strictly the authorities enforce the new rules."
Alternatives to using a VPN in Pakistan
Tor Browser
Proxy Services
SmartDNS service
VPNs may be the most widespread and easy-to-use circumventing tools out there, but certainly aren't the only way to bypass internet blocks, bans, and other geo-restrictions.
Tor Browser is probably the best VPN alternative you could use. Completely free to use, Tor is even considered a more secure option as it reroutes your internet connections (and data in transit) through at least three encrypted layers. This boosted encryption, however, can considerably slow down your connection speed.
It's near-impossible to block all the VPNs, all the time
Simon Migliano, Top10VPN
Proxy services are another effective method of spoofing your IP address to trick your ISP (internet service provider) into thinking you're browsing from a completely different country within a couple of clicks.
The downside here is that, unlike VPNs, proxies don't encrypt your internet connections, making them less secure overall. As Pakistan's firewall is expected to boost online surveillance, too, I recommend keeping this method as a last resort – especially now as it's not clear the risk of bypassing the government's imposed online restrictions.
You can also use a SmartDNS service to bypass geo-restriction online with ease. Again, these services do not offer the same level of protection as VPNs, meaning they don't encrypt your data. Worse still, a SmartDNS doesn't even change your IP but simply reroutes information about your location instead.
Why is Pakistan restricting VPN usage?
The government first shared plans to regulate the use of VPNs back in August to curb VPN misuse. Authorities later deemed unregistered VPNs a "security risk" for Pakistan as they can be used to access "sensitive data."
Under the new policy, using a VPN is still permitted for legitimate purposes such as banking, foreign missions, corporate enterprises, universities, IT companies, call centers, and freelance professionals.
This is why the Pakistan Telecommunication Authority (PTA) has been urging businesses and freelance workers to register their VPN services by the November 30 deadline to avoid disruption.
The debate is still ongoing, however, on whether or not commercial VPN usage should also be blocked. That said, VPNs have often been a target during the year, with residents facing temporary connectivity issues starting in February (when the ex-Twitter was first restricted).
Do you know?
(Image credit: Shutterstock)
The best VPN apps are security software that encrypts your internet connections to prevent snooping. They also spoof your real IP address location to boost anonymity and grant access to otherwise geo-restricted content.
Not only a new stringent policy, VPN censorship is also expected to worsen as the government beefs up its filter technology.
As Al Jazeera reported, Pakistan's government has implemented a China-like firewall, which was first tested in mid-July. This system not only can allow authorities to better monitor online traffic but also control the use of popular apps more efficiently.
Most notably, the new Pakistan firewall "has the ability to block VPNs" – a Ministry of Defence official, familiar with the new deployments, told Al Jazeera.
At the time of writing, it's still early to determine the full impact all this has on people using VPNs to keep accessing the free web. Experts, however, are worried by this surge in internet censorship.
"Attempts to tighten online surveillance always raise concerns about digital privacy and freedom of expression," Gytis Malinauskas, Head of Legal at Surfshark, told TechRadar. "This could also worsen the digital divide and inequalities between those who can access unfiltered information and those who cannot."
from TechRadar - All the latest technology news https://ift.tt/LcnZg1y
Raspberry Pi has launched its 5-based Compute Module 5 (CM5)
Users can get an upgraded I/O board and other accessories for it
Available to buy from just $45, with multiple configurations
Raspberry Pi has unveiled its latest modular variant of its single-board computer.
Essentially a stripped-back version of the Raspberry Pi 5, the Compute Module 5 (CM5) comes a decade after the first Compute Module, launched in 2014.
CEO Eben Upton said he recognized not all applications are suited to the Pi 5’s design and peripherals, plus users want more control over how and where they set up their SBCs, hence the stripped-down Compute Module.
Raspberry Pi CM5
The simplified design and abundance of configuration options, both in terms of components and hardware accessories, are designed to cater to the needs of Raspberry Pi commercial customers.
The CM5's compact form factor and lack of traditional ports make it ideal for the industrial and embedded sectors, which accounted for around three-quarters (70-80%) of Raspberry Pi sales in 2023.
It uses a 2.4GHz quad-core 64-bit Arm Cortex-A76 CPU with 2GB, 4GB or 8GB of LPDDR4X-4267 SDRAM – Upton suggested that a 16GB variant could become available in 2025. It’s also configurable with 16GB, 32GB or 64GB of MLC eMMC non-volatile memory, and includes Gigabit Ethernet and PCIe 2.0 support.
The CEO also confirmed that CM5 is mechanically compatible with its predecessor, CM4, which means that users can upgrade their single-board computers without having to upgrade other parts of their infrastructure. It fits M2.5 screws in each corner of the board for mounting the device.
CM5 is configurable thanks to a new range of accessories, including an upgraded USB-C powered I/O board with a standard 40-pin GPIO connector, two HDMI 2.0 connectors, two USB 3.0 connectors, a Gigabit Ethernet jack with PoE+ support, a microSD card socket and more.
The I/O board is also available with an enclosed case, which positions the ports all on one side of the neat box. It’s sold with an integrated fan, which connects to the board’s four-pin fan connector.
Other equipment includes a cooler, an antenna kit and a development kit, which bundles together a pre-configured CM5, the I/O board, the case, the cooler, the antenna and more.
Priced alone, the Raspberry Pi Compute Module 5 is available to buy from $45.
Microsoft introduces Azure Boost DPU to handle data-centric workloads
Specifically for Azure, integrates compute, storage, and networking
Custom hardware-software design boosts efficiency, lowers power needs
Traditional CPUs excel at general-purpose tasks but struggle to manage multiplexed streams from millions of network connections. GPUs, while effective for AI computations, are similarly not well-suited to these specialized workloads. To address this, the Data Processing Unit (DPU) was developed as a new class of silicon to complement CPUs and GPUs in large-scale data centers and efficiently handle high-volume processes.
At its recent Ignite event, Microsoft announced the Azure Boost DPU, a new SoC created in-house specifically for Azure.
The company’s new DPU is designed to tackle challenges in cloud and AI environments, improve security, and efficiently manage large-scale workloads across networks. Its introduction marks a key step toward the disaggregation of the data center, a growing trend that separates compute, storage, and networking functions into specialized hardware to improve efficiency and scalability.
Hardware and software co-design
“As part of our comprehensive approach to optimizing Azure infrastructure, we are working hard to remove infrastructure constraints that stand in the way of delivering customer value - whether from performance bottlenecks or resource limitations,” said Pradeep Sindhu, Corporate VP Silicon at Microsoft.
"We are innovating at every layer of the stack, from silicon to systems to software, to enhance the security, efficiency, performance, agility, and scale of our infrastructure."
Azure Boost DPU integrates high-speed Ethernet, PCIe interfaces, storage engines, and security features into a single, fully programmable SoC. Its custom hardware-software co-design delivers higher performance and lower power consumption, enabling the DPU to handle cloud storage workloads with three times less power and four times the performance of traditional CPUs, according to Microsoft.
By addressing key bottlenecks, Azure Boost DPU can meet the evolving demands of cloud and AI applications, while improving security through its tightly integrated data protection, and cryptography engines.
“Azure Boost DPU is the newest addition to our family of infrastructure enhancements through hardware innovation. As we continue to push the boundaries of what's possible, we remain committed to delivering innovations that make our infrastructure more capable, efficient, and scalable to meet the evolving needs of our customers,” Sindhu said.
Meta and Marvell Technology design new chip hardware together
FBNIC looks to boost internal networking capabilities at Meta
The chip will likely never go on commercial sale though
A collaboration between Meta and Marvell Technology has resulted in the development of a powerful new network chip.
The chip, called FBNIC, was designed with the aim of enhancing Meta’s internal networking capabilities, reducing downtime and increasing overall efficiency.
Announced at the 2024 OCP Global Summit, FBNIC claims to offer significant performance improvements - but it won’t be available for commercial sale anytime soon.
A custom approach to networking
The partnership created a custom Network Interface Controller (NIC) developed using a 5nm process, combining custom firmware, software, and hardware to deliver optimized network performance aimed at helping Meta streamline its networking operations.
The FBNIC chip supports a wide range of Ethernet network interfaces, ensuring Meta’s infrastructure can handle large amounts of data with high-speed connections, making it ideal for the massive scale of the company's operations - especially when factoring in AI.
It also comes with a multi-host PCIe interface, which supports four independent Gen5 x4 ports allowing the chip to communicate efficiently across multiple servers. Furthermore, the custom firmware control provides access to all hardware internals, giving Meta the ability to tailor the chip’s performance to its specific needs.
In addition to developing FBNIC for its use, Meta has chosen to contribute the board design to the Open Compute Project (OCP), an initiative aimed at promoting open-source hardware designs for data centers.
“The future of large-scale, data center computing will increasingly revolve around optimizing semiconductors and other components for specific applications and cloud infrastructure architectures,” said Raghib Hussain, President of Products and Technologies at Marvell.
“It’s been exciting to partner with Meta on developing their custom FBNIC on our industry-leading 5nm accelerated infrastructure silicon platform. We look forward to the OCP community leveraging the board design for future innovations."
Security researchers from Check Point Research discover new malware loader written in Godot's programming language
Godot is a popular open source game development platform
At least 17,000 devices were infected with infostealers and cryptojackers so far
Hackers are abusing a popular gaming engine to infect people’s computers with malware used to steal private data and cryptocurrency.
Researchers from Check Point Research have detailed a previously undetected hacking technique targeting users of the Godot Gaming Engine, an open source game development platform used to build both 2D and 3D games across Windows, macOS, Linux, Android, iOS, HTML5, and others, with a community of more than 2,700 developers.
Check Point says since late June 2024, crooks have been building malicious code written in GDScript (Godot’s Python-like scripting language) calling on some 200 GitHub repositories and more than 220 Stargazer Ghost accounts, which were hosting a piece of malware called GodLoader.
Infostealers and cryptojackers
In typical malware loader fashion, GodLoader would drop different malware to the infected devices, with the researchers spotting mostly RedLine stealer, and XMRig, a popular cryptojacker.
RedLine is an infamous infostealer capable of grabbing passwords, crypto wallet details, and other data stored in browsers, sensitive data, session cookies, and more. XMRig turns the infected device into a cryptocurrency miner, generating tokens for the attacker (while rendering the computer useless for pretty much anything else).
GodLoader, the researchers further explained, was downloaded at least 17,000 times, which is a rough estimate on the number of infected devices. However, the attack surface is much, much larger.
Check Point argues that in theory, crooks could hide malware in cheats, cracks, or modes, for different Godot-built games. Looking at the number of popular games developed with Godot, that would put the attack surface at approximately 1.2 million people.
Since GodLoader is yet to be flagged by most antivirus programs, it is essential to remain vigilant at this time, and careful when dealing with Godot-related content.
Elon Musk seeks $6 billion for xAI’s Memphis data center
Funding to purchase 100,000 Nvidia chips for AI projects
Data center will power Tesla’s Full Self-Driving capabilities
Elon Musk knows that staying competitive in the AI space requires deep pockets - back in February 2024, the world’s richest man said Tesla would spend "over a billion dollars" on Nvidia's H100 and AMD's Instinct MI300 hardware, adding, “The table stakes for being competitive in AI are at least several billion dollars per year at this point.”
More recently, the controversial billionaire gave us a tour inside the massive AI supercluster, dubbed Cortex, being used by X (formerly Twitter) and followed that up with news that Colossus, a new 100k H100 training cluster for Musk’s artificial intelligence company xAI, is now up and running.
According to CNBC’s David Faber, Elon Musk has raised up to $6 billion (at a $50 billion valuation) to purchase 100,000 Nvidia chips for xAI’s Memphis data center, which will - among other things - power Tesla’s full self-driving (FSD) capabilities.
Musk's government role
Sources familiar with Musk’s plans told Faber that the $6 billion funding comes from $5 billion of sovereign funds in the Middle East and $1 billion from other investors who will have the option to increase their investment.
Following Donald Trump’s election victory and the announcement that Elon Musk will lead the new Department of Government Efficiency within the president-elect’s administration, we can expect to see sizable government changes regarding AI regulation.
Musk’s future role, despite potential conflicts of interest, is to "dismantle government bureaucracy, slash excess regulations, cut wasteful expenditures, and restructure federal agencies."
However, no one believes for a second that this will be the limit of his involvement in wider policy decisions.
As CNBC reports, “Trump plans to repeal President Joe Biden’s executive order on AI, according to his campaign platform, stating that it 'hinders AI innovation and imposes radical left-wing ideas on the development of this technology,' and that 'in its place, Republicans support AI development rooted in free speech and human flourishing.'”
Anthropic is rolling out new ways for you to change how your Claude AI chatbot communicates with a new range of pre-set and custom writing styles. Instead of explaining in every prompt how you want the AI to respond, you'll be able to tailor its conversation to your needs. That means Claude will be as formal or casual, verbose or concise as you wish. It’s a major step to making talking to an AI feel more natural and more like talking to a person, in this case, a person who speaks in a way you prefer.
There are three preset options you can pick from right away. The names are pretty much on the nose. Formal style is for professional, precise responses; Concise is for shorter and direct answers; and Explanatory style is more about explaining and teaching, with extra context and details.
However, the new custom style option is the real eye-catcher. You can create your own personal communication styles for Claude to mimic by uploading samples in the tone and style you want, along with your own words describing the way you want Claude to write. Over time, you can fine-tune and improve the descriptions to make the AI chatbot a perfect mimic. You can see how it works below.
(Image credit: Anthropic)
AI at home
"Whether you're a developer writing technical documentation, a marketer crafting specific brand guidelines, or a product team planning extensive project requirements, Claude can adapt to your preferred way of writing," Anthropic explained in a blog post. "With styles, Claude adapts to your unique context and communication choices, helping you achieve more while working in a way that feels natural to you."
For those keen on consistent style, like in business or professional communications, the new feature has obvious appeal. Anthropic pointed to customers like GitHub, which is using the style options to improve internal operations and create marketing copy. It's worth raising questions about how customizable AI might reshape communication in professional and creative spaces. Could this blur the line between human-authored and AI-generated content? And in scenarios where tone and voice are crucial—like legal documents or sensitive communications—how much responsibility should fall on the AI versus the user to ensure accuracy and appropriateness?
Anthropic's approach is somewhat unique, but customizing AI responses isn’t entirely new. ChatGPT, Google Gemini, and Microsoft Copilot all have features for customizing tone and voice. Still, Anthropic's option to infer your preferred style from uploaded content could make it a lot easier to teach the AI how you want it to write.
Dash Pro Flash Drive delivers up to 4TB storage with high speeds
Achieves 1050MBps on USB 3.2; supports USB-C and Thunderbolt
Includes adapters, and features durable aluminum housing
Despite the rise of cloud storage, a USB flash drive remains a practical and reliable tool for data storage and transfer.
Their portability, ease of use, and ability to operate without an internet connection make them invaluable for creatives and professionals, whether sharing large files, backing up critical data, or accessing information on the go. If there’s a drawback to flash drives, it might be capacity, but the Dash Pro from Oyen Digital solves this by offering sizes up to 4TB.
Measuring just 3.1 x 0.97 x 0.33 inches and weighing 2 ounces, the drive supports USB-A, USB-C, and Thunderbolt 3 & 4 connections and the internal NVMe PCIe 4.0 x4 interface, powered by the Phison E21 controller, offers speed and reliability.
Heat protection
Pre-formatted with exFAT, the drive is ready to use with Windows, macOS, and Linux operating systems.
By using 3D TLC NAND, the USB 3.2 Gen2 Dash Pro achieves speeds of up to 1050MBps when connected via USB-C, USB 3.2, or Thunderbolt 3 & 4 (it is not compatible with Thunderbolt 1 or 2), and up to 525MBps with USB 3.0.
The Dash Pro has a durable aluminum housing that effectively absorbs and dissipates heat, preventing the internal components from overheating. With an operating temperature range of 32°F to 158°F, the drive promises dependable performance even in harsh conditions.
Included with the Dash Pro are a USB-A to USB-C adapter, a USB-A extender, and a lanyard, which should hopefully prevent you from losing the drive. Certified by CE and FCC, Dash Pro is backed by a three-year warranty.
The Dash Pro is available for purchase from B&H in 1TB, 2TB, and 4TB capacities, with the largest drive priced at $439. Whether you need fast transfers, secure data storage, or wide compatibility, the Dash Pro Flash Drive is a versatile and high-performance option.
MyHeritage gained a lot of attention for turning old photos into videos with its Deep Nostalgia technology in 2024, and they're also the company behind DeepStory, which makes images speak by creating talking portraits from photos or paintings. Now, the company has upgraded the tech with its new LiveMemory tool.
Like Deep Nostalgia, LiveMemory uses AI to make short animated videos from still images that theoretically show what might have happened right after the photo was taken. It's an upgraded version of the same feature, capable of making a kid on a bike in a picture ride away or a couple on their wedding day turn and kiss. Or at least that's the pitch.
I decided to try it out myself, as it's easy enough to use if you have an account and the MyHeritage mobile app. However, you only get a few tries with the free trial, and you need to pay up to remove the watermark.
To make a LiveMemory, you upload whatever picture you want to see transformed. A few minutes later, you get an email from MyHeritage with the video. To spare my friends and family, I started with a picture of myself and a much-missed pet dog named Malfoy. You can see the resulting video below.
Uncanny Nostalgia
I was mostly impressed with how much the tech has improved from Deep Nostalgia. The movements of my head and Malfoy's head and body are quite realistic, and his tail, unseen in the photo, does look exactly like his actual tail. Even my wry expression is well observed, considering the AI still had to go on for the video.
That said, I don't consider myself that wall-eyed, and while Malfoy's tail looked right, it also looked like it was growing out of the side of his body. We used to joke that Malfoy was a dog built by a committee in the dark, but even he had his tail in the right location.
I decided to go simpler and just upload a straightforward portrait of myself at about a year old. You can see how that went below.
Boneless boy
Again, the expressions are great; it looks like I'm really enjoying a joke, and my head and neck are moving like an actual human being. On the other hand, the AI doesn't seem to realize that the young child in the image would definitely not have that many teeth gleaming in his mouth. Those teeth apparently stole all of the digital bones from my hands. Watching my clay-like fingers squish each other and occasionally pass through each other like monstrous tentacles is more likely to induce nausea than nostalgia.
Compare that to the official launch video from MyHeritage, seen below. You can tell that even if there will be a lot of videos that people don't like, the ones that hit the mark will be very popular, just like Deep Nostalgia. Maybe just be careful not to use photos where hands are clasped together like mine.
Robots struggle to learn from each other, and rely on human instruction
New research from UC Berkeley shows that the process could be automated
This would eliminate the struggles of manually training robots
Despite robots being increasingly integrated into real-world environments, one of the major challenges in robotics research is ensuring the devices can adapt to new tasks and environments efficiently.
Traditionally, training to master specific skills requires large amounts of data and specialized training for each robot model - but to overcome these limitations, researchers are now focusing on creating computational frameworks that enable the transfer of skills across different robots.
A new development in robotics comes from researchers at UC Berkeley, who have introduced RoVi-Aug - a framework designed to augment robotic data and facilitate skill transfer.
The challenge of skill transfer between robots
To ease the training process in robotics, there is a need to be able to transfer learned skills from one robot to another even if these robots have different hardware and design. This capability would make it easier to deploy robots in a wide range of applications without having to retrain each one from scratch.
However, in many current robotics datasets there is an uneven distribution of scenes and demonstrations. Some robots, such as the Franka and xArm manipulators, dominate these datasets, making it harder to generalize learned skills to other robots.
To address the limitations of existing datasets and models, the UC Berkeley team developed the RoVi-Aug framework which uses state-of-the-art diffusion models to augment robotic data. The framework works by producing synthetic visual demonstrations that vary in both robot type and camera angles. This allows researchers to train robots on a wider range of demonstrations, enabling more efficient skill transfer.
The framework consists of two key components: the robot augmentation (Ro-Aug) module and the viewpoint augmentation (Vi-Aug) module.
The Ro-Aug module generates demonstrations involving different robotic systems, while the Vi-Aug module creates demonstrations captured from various camera angles. Together, these modules provide a richer and more diverse dataset for training robots, helping to bridge the gap between different models and tasks.
"The success of modern machine learning systems, particularly generative models, demonstrates impressive generalizability and motivated robotics researchers to explore how to achieve similar generalizability in robotics," Lawrence Chen (Ph.D. Candidate, AUTOLab, EECS & IEOR, BAIR, UC Berkeley) and Chenfeng Xu (Ph.D. Candidate, Pallas Lab & MSC Lab, EECS & ME, BAIR, UC Berkeley), told Tech Xplore.
PureVPN has been delivering incredible VPN services since 2007. As a result of their products, users can enjoy freedom on the internet and access region-specific sites.
One of PureVPN's unique features is that it lets users choose from four modes: Stream, Internet Freedom, Security/Privacy, and File Sharing. Each has its own dedicated servers, which guarantees faster speeds and smoother experiences online.
This Black Friday, PureVPN is offering a whopping 89% off their 5-year plans, starting at just $1.99/month. Choose from three plans: Standard (VPN + Tracker Blocker), Plus (adds Password Manager & File Encryption), or Max (adds Remove My Data & Dark Web Monitoring). To make use of the offer, you'll need to head over to the PureVPN website.
If you've missed out on this incredible deal or would like to see the latest PureVPN deals, head over to our PureVPN coupon codes hub.
PureVPN.comget 89% off 5-year plans Get 89% off on PureVPN's 5-year plan, starting at just $1.69/month, and stand a chance to win up to $1,000 in prizes. Plus, earn an additional 25% discount by using the coupon code: FUTURE25. Don’t miss this limited-time offer to secure your online privacy and save big! US only: Available during Black FridayView Deal
Our top picks from PureVPN
Standard VPN
A package that provides online anonymity, is fully featured, and is super fast. You'll also get a One-tap Tracker Blocker.
A superb VPN package that delivers fully-featured and fast access as well as a strong password manager and a file encryption tool. You'll also get a One-tap Tracker Blocker.
With over 6,500 servers across more than 70 countries worldwide, PureVPN is well placed to keep you and your data secure. Use your VPN account with as many devices as you like, although only 10 can be connected simultaneously.
Use PureVPN on PCs and laptops running Mac, Windows, and Linux, as well as iPhones and Android phones, Android, Apple and Firestick TV devices, and even your browser, with support for Google Chrome, Mozilla Firefox, and Microsoft Edge. In short, use it everywhere!
PureVPN also offers a student discount program that enables users to take advantage of an additional 15% off their chosen PureVPN plan.
from TechRadar - All the latest technology news https://ift.tt/S07EgpO
New Chinese mini PC supports Intel 12th-14th gen CPUs
Can take high-performance GPUs including the Nvidia RTX 4090M
Apple Mac Studio-inspired chassis has multiple ports
FEVM isn’t a brand many outside of China will have heard of - established by Hefengyi Electronics, it manufactures and sells a range of high-performance mini PCs, the latest being the FN60G WE, an upgrade of the previously released FN60G.
As you can see from the photo above, the device bears more than a passing resemblance to the Apple Mac Studio - but runs either Windows or Linux.
The compact mini PC supports Intel’s 12th to 14th generation desktop processors, including the high-performance Core i9-14900KF, which boasts 24 cores and 32 threads. This makes it well-suited for professional workloads and content creation.
Drives up to five screens
The upgraded FN60G WE features an improved design with a 3.8-liter chassis that measures 168 x 168 x 137mm (6.6 x 6.6 x 5.4 inches) and can accommodate high-performance GPUs, such as the top-tier Nvidia RTX 4090M. The system supports up to 96GB of DDR5-5200 memory and includes dual M.2 PCIe 4.0 SSD slots for fast data transfer speeds and ample storage.
The FN60G WE offers a welcome selection of ports, including 2 x HDMI (4K/60Hz), 2 x DisplayPort (8K/60Hz), 1 x USB Type-C with DisplayPort Alt Mode (8K/60Hz) and 15W USB Power Delivery, 1 x USB Type-C, 2 x USB 3.2 Gen 2 Type-A, 1 x USB 3.2 Gen 1 Type-A, 3 x USB 2.0 Type-A, 2 x 2.5 Gb Ethernet, a 3.5mm audio jack, and an SD card reader.
The mini PC can drive up to five screens, making it ideal for multitasking or complex workflows.
If you're worried about it overheating, don't be - the FN60G WE features an advanced cooling system with independent fans for the RAM and storage, coupled with a Thermalright AIO water cooler for the GPU, which should keep things running smoothly during intensive tasks.
Currently, the FN60G WE is not available in the United States, with its release limited for now to China, with pricing starting at $380 for a barebones model. There’s a Chinese-language video showing off the mini PC, which you can watch below - but remember to turn on the English subtitles.
Game Assist mode is part of the Game Bar and is now in testing
It pops up a compact Edge panel right in your gaming session
You can look at online tips and guides without Alt-Tabbing out of the game
Microsoft is adding a new feature to the Edge browser that could give it a distinct edge over Chrome (and other rivals), at least for PC gamers.
This is the Game Assist mode, which is now in preview (testing) with Microsoft Edge beta version 132, and it’s part of the functionality integrated into the Game Bar on Windows 11.
The basic idea is to give you access to, say, game guides or tips while you’re playing a game. So, instead of having to Alt-Tab constantly in and out of the game to look at said guide in the browser – or have the guide open on your phone, staring at a small screen – you can have it right there in your game session.
Game Assist is essentially a mini version of Edge running in the Game Bar overlay, so you can look at those hints and tips – or whatever you want – and still see the game world, and when you’re about to be attacked by a monster or something that might be similarly disastrous if you’d Alt-Tabbed out.
This gaming-centric Edge panel shares the same data as the main Edge browser on your PC, so you get access to your favorites, cookies, and so on. It’s also trimmed down and streamlined so it doesn’t sap resources, Microsoft notes – which obviously wouldn’t be helpful for gaming.
Furthermore, Game Assist is aware of what you’re playing and can surface tips or walkthroughs appropriate to that game.
Analysis: Game on!
All in all, this sounds like a great idea – providing what Microsoft says about the Game Assist feature not using too much in the way of resources holds true. Of course, as it’s in beta right now, if you do try the capability out, it’s possible that Game Assist could be wonky in any number of ways.
To begin with, the test version only supports the English language and a ‘selection’ of popular games. The initially supported PC games include Baldur’s Gate 3, Diablo IV, Fortnite, and League of Legends, among others, but no doubt more will be added as we go along – assuming Game Assist makes it out of testing.
We don’t see why it wouldn’t, as this is a pretty neat idea for gamers who don’t have a multiple-monitor setup that allows them to peruse guides or tips on a separate large screen (and not a fiddly little smartphone screen). It also avoids clumsy Alt-Tabbing, as mentioned.
If it works well, this is a feature that we could see persuading a number of PC gamers to give the browser a whirl and perhaps drive more adoption of Microsoft Edge (which is, in fact, already rated as our top pick out of the best web browsers).
Report claims Snapdragon X laptops only capture 0.008% of the global PC market
Microsoft Surface leads Snapdragon X adoption in niche segment
Qualcomm faces stiff competition from AMD, Intel, and Apple
The PC market is continuing to be a challenge for Snapdragon X laptops, with recent reports showing limited market penetration in its first full quarter since launch, as total sales reached under 720,000 units.
This figure represents less than 0.008% of the total PCs shipped globally over the same period, equivalent to fewer than one out of every 125 devices.
Despite a notable sequential growth of 180% in Q3 2024 compared to Q2, the Snapdragon X Series remains a small segment of the broader Windows market, capturing less than 1.5% of the ecosystem.
AI-capable PCs are gaining momentum
Microsoft and other popular brands have already transitioned some devices to the Snapdragon X platform, however, the market for these devices remains niche, even as Qualcomm pushes forward with integration into more consumer and business devices.
Canalys told TechRadar Pro, “As this was the first full quarter of shipments for Snapdragon X Series PCs, we saw sequential growth of around 180% compared to Q2 2024. However, as a proportion of the total Windows market, the products remain very niche, at less than 1.5% share. The top shipping vendor was Microsoft, which has transitioned most of their Surface line to the platform. Behind them was Dell who has embraced the new platform quite strongly in terms of SKU count, followed by HP, Lenovo, Acer and Asus (all four with similar volumes).”
While Snapdragon X laptops struggle to gain traction, the broader category of AI-capable PCs is rapidly expanding. In Q3 2024, shipments of these devices reached 13.3 million units, accounting for 20% of all PC shipments. This category includes desktops and notebooks equipped with chipsets dedicated to AI workloads, such as AMD’s XDNA, Intel’s AI Boost, and Qualcomm’s Hexagon.
Windows devices led the AI-capable PC market for the first time, capturing 53% of the segment. The surge in demand was supported by the Windows 11 refresh cycle and processor advancements. Sequential growth for AI-capable PCs stood at 49%, underscoring the market’s increasing appetite for AI-driven computing capabilities.
Nevertheless, AI-capable PCs face significant hurdles. Canalys' data suggests that consumers and channel partners remain cautious about adopting these premium offerings. For instance, Microsoft’s Copilot+ PCs, requiring at least 40 NPU TOPS and other high-performance specifications, have yet to fully convince buyers of their value.
A November 2024 poll of channel partners revealed that 31% do not plan to sell Copilot+ PCs in 2025, while 34% expect these devices to account for less than 10% of their sales. With Windows 10’s end-of-support deadline approaching, PC vendors are under pressure to drive upgrades among users still relying on ageing devices.
To stand out in the competitive AI-capable PC market, vendors are exploring unique strategies. HP has emphasized collaborations with independent software vendors (ISVs) to enhance on-device AI experiences. Lenovo has invested in proprietary AI tools embedded within its PCs, such as Creator Zone and Lenovo AI Now.
Dell and Lenovo are leveraging on-device AI to complement their broader AI service ecosystems. Meanwhile, Apple has taken a distinct approach, focusing on its vertically integrated ecosystem.
AMD's 192-core EPYC 9965 CPU now a third cheaper than at launch
5th-gen Turin chip has seen stellar reviews from the tech press
Half the price of Intel's 128-core Granite Rapids Xeon 6980P
AMD’s new 192-core CPU monster, the EPYC 9965, launched last month, four months after its initial announcement. The CPU, which uses the Zen 5c core architecture, boasts 192 cores and 384 threads and competes (almost) directly with Intel’s recently launched 128-core Granite Rapids Xeon 6980P.
In addition to being AMD’s most complex microprocessor to date, the EPYC 9965 was also its most expensive, at nearly $15,000.
Fast forward to today however, andWiredzone is currently selling the chip for a more modest $10,034 - a significant discount. Unlike some online deals that disappoint with limited stock or long delivery times, Wiredzone has the processors in and ready to ship, with volume discounts also available.
What's behind the price cut?
The reason for this surprise price drop isn’t entirely clear. It could be that AMD has been able to produce more processors than first anticipated, or perhaps demand hasn’t been as high as expected (which seems very unlikely).
By comparison, Intel’s Xeon 6980P launched at $17,800, but its price has since increased. Newegg is currently selling it for $18,900, and Connection lists it for $20,527 - though neither retailer has the product in stock at the moment. That could suggest another reason for the sudden price drop.
At the lower cost, AMD might see an opportunity to appeal to budget-conscious enterprises or organizations seeking high-performance processors without paying a premium.
By undercutting Intel’s prices and ensuring ample stock availability, AMD could strengthen its position in the competitive server CPU market while attracting new customers who might otherwise consider Intel’s offerings.
Reviewers previously declared that with the Xeon 6900P family, Intel was finally a contender again in the server CPU arena.
However, the arrival shortly afterwards of AMD’s 5th generation Turin products quickly rained on that parade, comfortably putting AMD back on top. As Phoronix noted in its gushing review, “The new top-end AMD EPYC Turin processor performance can obliterate the competition in most workloads and delivers a great generational leap in performance and power efficiency."
Quordle was one of the original Wordle alternatives and is still going strong now more than 1,000 games later. It offers a genuine challenge, though, so read on if you need some Quordle hints today – or scroll down further for the answers.
What letters do today's Quordle answers start with?
• N
• S
• S
• D
Right, the answers are below, so DO NOT SCROLL ANY FURTHER IF YOU DON'T WANT TO SEE THEM.
Quordle today (game #1036) - the answers
(Image credit: Merriam-Webster)
The answers to today's Quordle, game #1036, are…
NUDGE
SNOOP
SHACK
DRYLY
DRYLY is the kind of word that crops up occasionally in Quordle or Wordle with the potential to cause all kinds of trouble, due to the fact that it contains no vowels. Or does it? Debate rages (in some, admittedly niche, circles) as to whether Y counts as a vowel or not. The answer is that it sometimes does and it sometimes doesn't. Here, it most definitely does, standing in for the 'I' sound in the middle and the 'E' sound at the end. However, I don't count Y as a vowel in my Quordle hints, because it would be far too confusing to do so.
DRYLY aside, today's game is not too difficult, but I did run it close having guessed SPOON when the answer was SNOOP. That aside, it was pretty approachable as Quordle goes.
Huawei develops SSD-tape hybrid for warm and cold data storage
MED system offers efficient storage with NAND speeds
First-gen holds 72TB, uses 10% of the power required by disk drives
It’s no secret that Chinese companies are becoming increasingly inventive when it comes to circumventing the challenges caused by the US export tech restrictions that prevent access to advanced semiconductor technology and critical components.
Huawei, in particular, is leading the way here as it looks to become China’s answer to Nvidia, and was recently reported courting Chinese hyperscalers to use its new Ascend AI chip instead of Nvidia's H100.
Blocks & Files reports that Huawei has developed an SSD-tape hybrid which combines Magneto-Electric Disk (MED) archive storage with a Huawei-developed tape drive. This hybrid will enable the delivery of warm and cold data storage in a single solution.
(Image credit: Blocks & Files/Huawei)
Expected to arrive next year
This development was first reported back in March 2024, but Blocks & Files now has a presentation image and more details about the MED.
The site writes: “The MED is a sealed unit presenting a disk-like, block storage interface to the outside world, not a streaming tape interface. Inside the enclosure, there are two separate storage media devices: a solid-state drive with NAND, and a tape system, including a tape motor for moving the tape ribbon, a read-write head, and tape spools.”
Unlike traditional tape cartridges with a single long reel, the MED features a compact design with a tape reel half the length of an LTO tape and an additional empty reel for used tape. This setup allows the MED to handle data storage more efficiently while integrating both solid-state and tape storage technologies.
The system functions as a dual-purpose solution, acting as an archive for cold data and nearline storage for warm data. Warm data is accessed rapidly through the SSD at NAND speed, making it ideal for frequently used information. Cold data, stored on the tape, requires more time to retrieve (up to two minutes), as the system must locate the data and adjust the tape ribbon to the correct position for reading.
The first generation of the MED, expected to arrive next year, reportedly holds 72TB of data (it’s unclear whether this refers to raw or compressed capacity) and consumes just 10% of the power required by disk drives. According to Huawei, a MED rack will deliver 8GBps, hold more than 10PB, and require less than 2kW of electricity. A second-generation model could follow in 2026 or 2027.
MSI launches two business laptops powered by Intel’s Core i7-13620H CPU
Includes Microsoft Office 2024 and Microsoft 365 Basic software
Both models have limited battery life
MSI has released two new business-oriented notebooks in Japan, the Modern 15 H B13M 5039JP and Modern 14 H D13MG 5029JP. These laptops, designed with productivity in mind, are powered by Intel’s Core i7-13620H.
Both models come pre-installed with Microsoft Office Home & Business 2024 and a year's subscription to Microsoft 365 Basic, giving business users access to Word, Excel, PowerPoint, and Outlook, along with 100GB of OneDrive storage.
This offer is currently only available in Japan, and MSI isn’t the only company providing it; Dynabook has included the same software package with two of its new laptops, suggesting a possible trend. It’s certainly something we’d like to see rolled out globally.
Why Windows 11 Home?
The Modern 15 H B13M 5039JP offers a 15.6-inch full HD (1920 x 1080) non-glare LCD, 32GB of memory, a 512GB NVMe SSD, and runs Windows 11 Home.
Connectivity options include USB 3.2 Gen 2 Type-C with video output and USB PD compatibility, three USB 3.2 Gen 2 Type-A ports, HDMI, a 920,000-pixel webcam, an audio combo jack, Wi-Fi 6E, and Bluetooth 5.3. It measures 359 x 241 x 19.9 mm, weighs about 1.9 kg, and its 53.8Whr battery lasts just four hours during video playback (up to eight hours when idle), so you won’t want to stray too far from a power outlet.
The Modern 14 H D13MG 5029JP is the smaller counterpart, featuring a 14-inch WUXGA (1920 x 1200) non-glare LCD, with the same memory and storage configuration as the 15.6-inch model.
It includes Thunderbolt 4, three USB 3.2 Gen 1 ports, a 920,000-pixel webcam, HDMI, an audio combo jack, Wi-Fi 6E, and Bluetooth 5.3. It measures 313.7 x 236 x 18.6 mm, weighs approximately 1.6 kg, and has the same battery but with a longer life - up to six hours during video playback and 14 hours when idle.
Interestingly, although Microsoft’s productivity software is included for free with these two laptops, they ship with Windows 11 Home. MSI recommends Windows 11 Pro for business, so it’s surprising the firm didn’t upgrade the default OS to suit the productivity package. Both laptops come In any color you like, so long as it’s Classic Black. No word on pricing or availability yet.
Stanford scientists are looking to combine SRAM and DRAM
The new memory type would help solve issues with AI computing
Gain Cell memory looks to bridge the gap between the two types
The development of more energy-efficient hardware for artificial intelligence (AI) systems is receiving increased support, with a focus on improving memory technology.
A hybrid type of memory that blends the high density of DRAM (Dynamic Random-Access Memory) with the speed of SRAM (Static Random-Access Memory) is at the forefront of this effort.
The project is being led by electrical engineers at Stanford University, with the team’s goal being to create faster, more efficient memory hardware for AI applications that addresses the current limitations in processing power and energy consumption.
Memory, a key AI bottleneck – hybrid gain cell memory to the rescue
This research is being funded under the CHIPS and Science Act, with a recent boost of $16.3 million in US Department of Defense funding to the California-Pacific-Northwest AI Hardware Hub.
AI systems are heavily reliant on hardware that can efficiently move and process large volumes of data. However, moving data between memory and logic unites takes time, which slows down GPUs and leads to increased energy consumption.
As AI models become larger and more complex, these memory bottlenecks become more pronounced. Therefore, faster and denser memory located directly on chips is seen as a potential solution to this problem.
Stanford University’s H.-S. Philip Wong, an electrical engineer and chair of the AI Hardware Hub, emphasizes the importance of memory in making AI hardware more energy efficient.
Wong’s team has turned to a new type of memory design called Gain Cell memory, which combines the advantages of both DRAM and SRAM. The hybrid gain cell offers a middle ground which has the small footprint of DRAM, but it also provides the faster readout speeds characteristic of SRAM.
The key difference in this new design is the use of two transistors—one for writing data and one for reading rather than the capacitor found in traditional DRAM. This allows the gain cell to retain data more reliably and to boost the signal strength when data is read.
Gain Cell memory has faced limitations such as rapid data leakage in silicon-based designs and slower readout speeds in oxide-based designs. However the Stanford team combined a silicon transistor with an indium tin oxide transistor, significantly enhancing the device's performance, offering faster readouts while maintaining a compact footprint.
The new design can hold data for over 5,000 seconds, far longer than traditional DRAM, which needs refreshing every 64 milliseconds. Additionally, the hybrid memory is around 50 times faster than oxide-oxide gain cells.
Wong likens this advancement to transitioning from a basic 3-gear bicycle to a sophisticated 20-gear bicycle, emphasizing that this evolution of memory technology will extend beyond traditional options like DRAM, SRAM, and flash memory. “We want to provide better options so designers can optimize better…it’s an opportunity to rearchitect computers,” Wong said.
Using dual screens is an effective, and proven way to boost productivity for many of us - and now, Espresso Displays, an Australian company specializing in portable monitors, has introduced a new product that simplifies adding a second screen to any laptop.
The bundle features a magnetic stand that will allow you to position the display alongside your laptop, or above it.
The espresso Display 15 features a 15.6-inch LCD screen with a Full HD (1920 x 1080) resolution with 16.8 million colors and 99% sRGB color coverage. With a brightness of 300 nits and a contrast ratio of 1000:1, the screen should provide consistent image quality, suitable for work tasks and content viewing. The display is super-slim, measuring only 0.2 inches (5.3 mm), and weighs 865 grams, making it highly portable. It comes with two USB-C ports.
Portrait and landscape modes
Stand+ is built from durable aerospace-grade aluminum and includes rubber pads to ensure stability on various surfaces.
The magnetic stand (which is compatible with other espresso displays) offers variable height adjustment and will allow the screen to be used in both portrait and landscape modes, providing flexibility for different workspace setups. Stand+ is also designed for easy attachment and detachment, which espresso says will help users transition seamlessly between tasks.
In addition to the new hardware, espresso Displays has introduced a software tool called Glide, which is part of its multi-screen management system, espresso Flow.
Glide offers automatic digital screen arrangement by detecting the physical location of the espresso Display in relation to a laptop. This removes the need for manual adjustments, as users can simply move their cursor across screens.
espresso Display 15 and Stand+ are available for $299 and currently come with a free case and protector. If you want to buy just the Stand+ that will set you back $69. We reviewed an earlier version of espresso Displays' portable monitor, and you can see what we thought of it here.
