Get a Deep Sleep With Tuft & Needle Holiday Discounts of Up to 25% Right Now

This sale saves you up to 25% on a mattress and lets you bundle accessories for some extra savings.

from CNET https://ift.tt/AHFTtqx

Latest Tech News

---

• Nvidia’s Project Denver began as x86 but transitioned to Arm
• Insider reports legal constraints drove Nvidia's pivot
• The Arm-based Project Denver CPU debuted in 2011

---

During a technical session at the recent SC24 event, Dave Ditzel, founder of Esperanto Technologies, offered some fascinating insights into Nvidia’s early server processor efforts.

According to HPCwire, Ditzel, who was previously CEO of Transmeta, revealed that Nvidia’s first server CPU, Project Denver, initially started as an x86 CPU but transitioned to Arm due to legal constraints.

Ditzel says Nvidia’s shift to Arm was influenced by its licensing of Transmeta’s Tokamak technology, which could translate x86 code into a RISC instruction set.

Failed attempt to acquire Arm

IAs he explained, “Nvidia brought out a product called Denver. It was actually that same design. It originally started as an x86 [CPU], but through certain legal issues, had to turn itself into an Arm CPU.”

This decision, he said, laid the foundation for Nvidia’s alignment with Arm architecture. Tokamak, developed by Transmeta, was intended to be its third-generation x86 chip following the Crusoe and Efficeon processors. However, the project was never officially launched and was instead licensed to companies like Intel and Nvidia.

ntel, despite acquiring the design, did not announce a product based on it either. “You can guess as to all the reasons why or buy me a beer sometime,” Ditzel said.

Nvidia officially introduced Project Denver as an Arm-based CPU in 2011, later integrating it into its Tegra lineup. HPCwire reports that while there was initial enthusiasm around Arm servers, adoption was limited by challenges in the software ecosystem. Nvidia has since developed its Grace CPU and abandoned its attempt to acquire Arm after regulatory opposition.

Ditzel founded chip design firm Esperanto about seven years ago and because of his previous bad experiences with licensing x86, he opted for RISC-V because it was cheap and there were no legal concerns to get bogged down by.

“At least we have a playground where we can test some new things out, and some lawyer is not going to be ringing your bell,” Ditzel said.

More from TechRadar Pro

• Nvidia's most powerful chip produced gets tested
• This 192-core RISC-V processor could be AMD and Intel's biggest rival yet
• Google's super powerful Arm-based CPU is now available

from TechRadar - All the latest technology news https://ift.tt/4lvjYra

Healthy Eyes Made Easy: 6 Vitamins and Supplements Worth Considering

Here are the top 6 vitamins and supplements needed to keep your eyes healthy.

from CNET https://ift.tt/sT1gH8w

Latest Tech News

Update: On December 6, 2024, the European Pirate Party reported that the European Council Committee stopped the proposal (yet again) as more governments joined the list of countries against it. We made some edits to the page to reflect this.

The EU proposal to scan all your private communications to halt the spread of child sexual abuse material (CSAM) is back on regulators' agenda – but keeps being rejected.

What's been deemed by critics as Chat Control has seen many twists and turns since the European Commission presented the first version of the draft bill in May 2022. The latest development came in October 2024, when a last-minute decision by the Netherlands to abstain from the vote prompted the Hungarian Council Presidency to remove the matter from the planned discussion.

Now, about two months later, the controversial proposal has returned to the topics the EU Council discussed on December 4, 2024. On Friday, December 6, however, the European Council Committee stopped the proposal (yet again) as more governments joined the list of countries against it.

What is the EU CSAM scan bill?

As mentioned, lawmakers have implemented some changes to the EU CSAM bill amid growing criticism from the privacy, tech, and political benches.

Initially, the plan was to require messaging services and email providers to scan all your messages on the lookout for illegal material – no matter if these were encrypted, like WhatsApp or Signal chats, for example, to ensure that communications remain private between the sender and receiver.

Lawmakers suggested employing what's known as client side-scanning, a technique that experts, including some of the best VPN providers and messaging apps, have long warned against as it cannot be executed without breaking encryption protection. Even the UK halted this requirement under its Online Safety Act until "it's technically feasible to do so."

Fast-forward to June 2024, the second version of the EU proposal aims to target shared photos, videos, and URLs instead of text and audio messages upon users' permission. There's a caveat, though – you must consent to the shared material being scanned before being encrypted to keep using the functionality.

This wording made privacy experts furious, with Meredith Whittaker, President of the Signal Foundation, labeling this so-called 'upload moderation' as a "rhetorical game" instead.

📣Official statement: the new EU chat controls proposal for mass scanning is the same old surveillance with new branding.Whether you call it a backdoor, a front door, or “upload moderation” it undermines encryption & creates significant vulnerabilitieshttps://t.co/g0xNNKqquA pic.twitter.com/3L1hqbBRgqJune 17, 2024

In September, another version was leaked by Politico. Communications providers would be free to decide whether or not to use artificial intelligence to flag images and text chats as suspicious. These companies, however, would be required by law to scan all user chats and report when they found illegal content.

As per the European Pirate Party's data before the December 6 meeting, the greater majority of countries have already expressed their support for the new proposal. Even nations like France, who previously were among the opposed governments, have now joined the in-favor list.

At that time, only a few EU members remained either undecided (Italy, Portugal, and Finland) or against (Austria, Belgium, Czech Republic, Estonia, Germany, Luxemburg, Netherlands, Poland, and Slovenia).

A day before the voting (December 5, Patrick Breyer from the European Pirate Party reported that "unconfirmed rumors" claimed an unnamed critical government could join the countries backing up the proposal on Friday.

This, however, didn't happen as Finland voted against it during today's meeting, despite both Italy and Portugal reportedly passing from undecided to "in favor."

from TechRadar - All the latest technology news https://ift.tt/THEUyrX

Latest Tech News

As 2025 approaches and 'New Year, new you' pressures build for those of us not keen on jogging or hitting the gym, you should know that there’s an awesome alternative to traditional fitness with Quell, one of my favorite workout apps of 2024. And today’s big announcement is that the gamified routines it offers are finally coming to Meta Quest headsets in January 2025.

For those of you that missed Quell at launch, it was originally a PC fitness platform. You’d use the controllers to punch your way through its fantasy adventure game Shardfall while wearing resistance bands that make it tougher than typical shadow boxing. The full-body HIT sessions felt intense, yet Shardfall did an excellent job of keeping me motivated by contextualizing my actions – I wasn’t just doing another rep for the sake of it, I was throwing another punch because I needed to defeat the monster in my way.

VR players can soon try this experience using a Meta Quest headset – sans the resistance bands unless they choose to buy some. That’s because Shardfall is coming to Quest as Shardfall: FitQuest VR for $19.99 / £14.99 at Meta.com, and I’m pretty darn excited about it.

Shardfall will be at home in VR

Quell is great, but it'll be right at home in VR (Image credit: Quell)

The Quell hardware system costs $199 / £189 while the software subscription to access Shardfall (and future game releases) only costs $4.99 / £4.99 a month or $39.99 / £39.99 a year which I feel is a pretty good price for the kit when compared to other fitness gear or a gym membership, but it is an added expense that people might not be willing to pay for a system that can only be used with Quell games.

As a $19.99 / £14.99 add-on to your existing VR headset, Shardfall in VR is a much easier sell, and I strongly recommend everyone give it a try at launch. The combat exercises are an engaging boxing sim that does well to incorporate a full-body routine with ducks and jumps to dodge certain attacks and hazards, and it also includes jogging sections between fights that get your heart rate up.

It'll be a shame to lose the resistance bands, but again that could aid Shardfall’s accessibility as its exercises will be a little less challenging – though I expect the full-body routines will still give you a decent workout without them.

When the VR version launches I’ll certainly be there to give it a go, and if you’re looking for a more interesting way to kickstart your 2025 fitness schedule I recommend you join me.

You might also like

• New Garmin firmware update could break your Forerunner – here's what you need to know
• Oura Ring users get new Symptom Radar feature – an early morning scan that could detect colds and flu
• Want to live longer? Turns out being fit might be more important than your body weight

from TechRadar - All the latest technology news https://ift.tt/tL4Bz5c

Latest Tech News

---

• Azure Integrated HSM boosts security with cryptographic key protection
• Reduces latency and scales better than network-attached HSMs
• Keys stay isolated, ensuring tamper-resistant, in-use protection

---

Microsoft has introduced a new hardware security module designed to boost cloud security by enabling cryptographic key protection directly within server environments. 

Azure Integrated HSM addresses latency and scalability challenges often associated with traditional network-attached HSMs while adhering to FIPS 140-3 Level 3 security requirements.

The new hardware module provides locally attached cryptographic services for encryption, decryption, signing, and verification. Keys remain isolated from software, including guest and host systems, ensuring strong physical and logical tamper protection. Unlike traditional HSMs, which introduce network latency or require key release to local environments, Azure Integrated HSM securely retains keys within the module for continuous in-use protection.

Coming to all new Microsoft data center servers

"As part of our systems approach in optimizing every layer in our infrastructure, security is a key priority, and we are designing our infrastructure hardware with multiple layers of defense with dedicated innovations to ensure robust protection for Microsoft and for our customers," noted Mark Russinovich, Microsoft’s CTO for Azure.

The module is designed to integrate seamlessly with both confidential and general-purpose virtual machines and containers, providing dedicated, secure partitions for each workload. These partitions are hardware-isolated, allowing workloads to access keys only through controlled oracle functions. This design boosts security and reduces latency with node-integrated connections and cryptographic hardware accelerators.

Azure Integrated HSM will be installed in all new servers across Microsoft data centers starting next year, bolstering protection across Azure’s hardware fleet. This deployment is part of the Secure Future Initiative, which also includes Adams Bridge quantum-resilient accelerator and Caliptra 2.0 silicon root of trust.

“By integrating advanced hardware security features such as the silicon root of trust and secure control modules, we are providing the foundation for the trust and security that Azure delivers to our customers,” Russinovich said. “We are committed to continuously enhancing our cloud hardware security capabilities to meet the evolving needs of our customers.”

You might also like

• Microsoft expands Azure AI with two new models for the Phi-3 family
• Microsoft says that it is making progress in its Secure Future Initiative
• You'll no longer need to restart your work laptop after every Windows update

from TechRadar - All the latest technology news https://ift.tt/SXnFyKT

Ionvac SmartClean 2000 Robovac

WiFi robotic vacuum.

from CNET https://ift.tt/vUwHO6G

Latest Tech News

---

• SemiQon announces first CMOS transistor for cryogenic conditions
• Engineered for extreme cold: operates efficiently at 1 Kelvin
• Transistor reduces heat dissipation 1,000x, consumes 0.1% usual power

---

Heat is widely recognized as the enemy of sensitive electronic components, but ultra-low temperatures can also pose serious performance challenges.

Now, SemiQon, a Finland-based company focused on quantum computing hardware, has announced the development of what it describes as the first CMOS transistor fully optimized for cryogenic conditions.

The transistor is engineered to function effectively at temperatures as low as 1 Kelvin (-272.15°C or -457.87°F), just 1 degree above absolute zero, where most quantum computers operate. According to the company, this innovation addresses key challenges in scaling quantum computers while also being compatible with existing CMOS manufacturing processes, requiring no new infrastructure.

Space-borne applications

SemiQon says its transistor reduces heat dissipation by 1,000 times compared to conventional room-temperature transistors and consumes only 0.1% of the power. This allows control and readout electronics to be located inside a cryostat with the processors, eliminating heat dissipation problems that could disrupt the system. SemiQon believes this solution simplifies the growing complexity of managing quantum processors as their scale increases.

“It was clear to us and others in the scientific community, that a transistor which can operate efficiently at ultra-low temperatures would offer substantial value to users in the advanced computing sector and wherever these devices are required to function in cryogenic conditions,” said Himadri Majumdar, CEO and Co-Founder of SemiQon.

“Our company is just 2 years old, and already we’ve delivered something which the world has never seen before. Our cryo-CMOS transistor will provide considerable advantages to users both in terms of CapEx and OpEx, as well as by enhancing the functionality of their hardware. This could potentially accelerate the development of quantum technologies, or even enable a new era of cryogenic electronics.”

The transistor’s potential extends beyond quantum computing to high-performance computing and space-borne applications. SemiQon also highlights its impact on energy efficiency, noting that cooling costs for data centers are projected to grow significantly in the coming years.

SemiQon says it expects to deliver its first cryo-optimized CMOS transistors to customers in 2025. A short technical paper on the new transistor can be found on the arXiv pre-print server.

More from TechRadar Pro

• In 2025 quantum computing will move from physical to logical qubits
• Microsoft and Quantinuum partnership advances Logical Qubit development
• How data centers can use liquid above ground to stay cool

from TechRadar - All the latest technology news https://ift.tt/CDGNqyK

Snag These Last Chance Cyber Monday PS5 Deals While They're Still Available

Time is running out to snag discounts on PlayStation 5 consoles, accessories and games. Don’t miss out -- these deals won’t last much longer.

from CNET https://ift.tt/wuZTmrF

Latest Tech News

---

• Geyser Data's Tape-as-a-Service reaches general availability
• Built with Spectra Logic, integrates seamlessly with AWS S3
• No egress fees, secure air-gapped tapes, energy-efficient solution

---

Back in May 2024, cloud archive provider Geyser Data and data storage and management firm Spectra Logic introduced a new Tape-as-a-Service (TaaS) cloud offering combining the durability and cost-efficiency of traditional tape storage with the flexibility of cloud services.

This TaaS solution addresses the growing demand for secure, cost-effective data storage, supporting large volumes while minimizing environmental impact.

Geyser Data claims up to 97% lower CO2 emissions, 87% less power usage, and 85% less e-waste compared to other cloud services. It offers enhanced security with dedicated tapes and full control over encryption keys, along with the freedom of immediate access.

Subscription service

Following a well-received beta phase, the TaaS offering is now generally available.

Nelson Nahum, CEO of Geyser Data, outlined some of the service's key benefits, saying, "New workloads like AI require cold data to be warmer. One of our customers' biggest challenges today is the unpredictable and skyrocketing costs tied to data retrieval and egress fees in other cloud environments. Our service provides a simple and transparent pricing model that eliminates these burdens while giving businesses the storage capacity they need without investing in new hardware."

The 'enterprise-class' tape archiving solution operates on a subscription basis. It integrates with S3 APIs, allowing businesses to manage and store large volumes of data without the variable costs typical of traditional cloud providers, the need for specialized expertise, or reliance on complex on-premises infrastructure.

“By integrating Spectra Logic's Tape Archive Platform-as-a-Service (TAPAS) with Geyser Data's robust cloud software management platform, we've developed a solution that drives significant cost savings while also addressing critical power consumption challenges faced by data-intensive technologies such as AI and machine learning," noted Mitch Seigle, Chief Marketing Officer of Spectra Logic.

"As organizations grapple with the demands of rapidly expanding data volumes, tape storage provides unmatched security, longevity, sustainability, and operational efficiency - making it a pivotal element in modern data infrastructure strategies.”

Geyser Data offers a straightforward pricing model at $28 per tape per month, with each tape capable of storing up to 18TB of uncompressed data. This equates to an effective cost of $1.56 per terabyte. There are no restrictions on the amount of data customers can archive or back up, as the Spectra Cube library is designed to scale to meet demand.

You might also like

• Spectra Logic proves that data tape still has a place in an AI world
• Tape is probably the cheapest way to store 1000TB of data online securely
• Remember VHS era tape? It’s back with a vengeance and a 50TB capacity

from TechRadar - All the latest technology news https://ift.tt/Gnu7d4F

24 Meal Kit and Food Subscription Deals for Cyber Monday

Score monthly drops of wine, beer, snacks or full-on meals with these discounted subscriptions.

from CNET https://ift.tt/CUXyO6z

Latest Tech News

---

• 01.ai trained an AI model for $3 million using 2000 unnamed GPUS
• “Efficient engineering” allows 01.ai to compete globally, company claims
• 01.ai reduced inference costs to 10 cents per million tokens

---

Tech companies in China face a number of challenges due to the American export ban, which restricts access to advanced hardware from US manufacturers.

This includes cutting-edge GPUs from Nvidia, critical for training large-scale AI models, forcing Chinese firms to rely on older or less efficient alternatives, making it difficult to compete globally in the rapidly evolving AI industry.

However, as we’ve seen time and again, these seemingly insurmountable challenges are increasingly being overcome through innovative solutions and Chinese ingenuity. Kai-Fu Lee, founder and CEO of 01.ai, recently revealed that his team successfully trained its high-performing model, Yi-Lightning, with a budget of just $3 million and 2,000 GPUs. In comparison, OpenAI reportedly spent $80-$100 million to train GPT-4 and is rumored to have allocated up to $1 billion for GPT-5.

Making inference fast too

“The thing that shocks my friends in the Silicon Valley is not just our performance, but that we trained the model with only $3 million," Lee said (via @tsarnick).

"We believe in scaling law, but when you do excellent detailed engineering, it is not the case you have to spend a billion dollars to train a great model. As a company in China, first, we have limited access to GPUs due to the US regulations, and secondly, Chinese companies are not valued what the American companies are. So when we have less money and difficulty to get GPUs, I truly believe that necessity is the mother of invention."

Lee explained the company’s innovations include reducing computational bottlenecks, developing multi-layer caching, and designing a specialized inference engine. These advancements, he claims, result in more efficient memory usage and optimized training processes.

“When we only have 2,000 GPUs, the team has to figure out how to use it,” Kai-Fu Lee said, without disclosing the type of GPUs used. “I, as the CEO, have to figure out how to prioritize it, and then not only do we have to make training fast, we have to make inference fast... The bottom line is our inference cost is 10 cents per million tokens.”

For context, that’s about 1/30th of the typical rate charged by comparable models, highlighting the efficiency of 01.ai's approach.

Some people may be skeptical about the claims that you can train an AI model with limited resources and "excellent engineering", but according to UC Berkeley’s LMSIS, Yi-Lightning is ranked sixth globally in performance, suggesting that however it has done it, 01.ai has indeed found a way to be competitive with a minuscule budget and limited GPU access.

You might also like

• China wants to be the global AI leader in personal computing
• AMD launches a new single-slot GPU for data centers
• This startup may be the answer to running AI training on any GPU

from TechRadar - All the latest technology news https://ift.tt/EYv0eVk

We Found the 33 Best Cyber Monday Deals Under $100 Available Right Now

With Cyber Monday deals now rolling in, there's plenty of opportunity to save big.

from CNET https://ift.tt/92y1uwY

We're Not Running Out of Batteries on Christmas Morning Thanks to This Cyber Monday Deal

Right now, you can grab a 28-pack of Duracell AA batteries for just $19 at Amazon.

from CNET https://ift.tt/CybiPol

Latest Tech News

---

• Micron 6550 Ion SSD comes in 30.72TB and 61.44TB capacities
• PCIe Gen5 G8 TLC NAND improves performance and efficiency
• Sequential read write speeds are over 12GBs and 5GBs respectively

---

Modern data centers require high-capacity, energy-efficient SSD storage solutions to support data-intensive applications like AI and machine learning.

Addressing these needs, Micron has introduced the 6550 Ion SSD, available in 30.72TB and 61.44TB capacities.

The 6550 Ion SSD, which Micron is positioning as the “world’s fastest, most energy-efficient 60TB data center SSD,” is available in various form factors, including E1.L and U.2, but it's the E3.S model which really shines, with its 61.44TB of TLC NAND.

Better performance than expected

Previously, in order to deliver a capacity of 61.44TB, manufacturers needed to use QLC NAND, but by making the switch to Gen5 interface G8 TLC NAND, Micron is able to offer faster write performance, endurance, and the thing that data centers are increasingly craving - energy efficiency.

StorageReview tested the new drive, and came away impressed, as reviewer Brian Beeler notes: “The Micron 6550 Ion offers massive performance and capacity benefits over the 6500 Ion, its predecessor. It adheres to PCIe Gen5 standards and incorporates future-proofing capabilities through OCP 2.5. Micron has prioritized security in the 6550 Ion, ensuring TAA and FIPS 140-3 L2 certifiability and a comprehensive security suite."

In terms of performance, Micron says the 6550 Ion delivers sequential read speeds of 12GB/s and write speeds of 5GB/s while maintaining a maximum power consumption of 20W. StorageReview's test recorded read speeds of 12.7GB/s and writes of 8.2GB/s, better than expected.

The review compared the 6550 Ion against the Solidigm P5336, the only 61.44TB SSD widely shipping today, and while acknowledging it wasn't an even match, said, "We noted many areas where the newer Gen5 interface on the Micron 6550 Ion excelled over the Gen4 interface in the P5336, which is expected. The difference between Gen4 vs. Gen5 drives relates to how much bandwidth each drive can push across the wire. The Gen4 PCIe interface with four lanes of connectivity is capable of around 7GB/s, while the newer Gen5 PCI interface doubles that with 14GB/s. Micron’s use of TLC NAND gives it another design advantage over QLC-based competitors."

Energy efficiency is another key selling point of the 6550 Ion. It uses up to 20% less power than competing SSDs and supports multiple operational power states. For example, Power State 1 (PS1) limits power usage to 20W, compared to the maximum 25W. It also consumes ≤5W when idle.

The 61.44TB capacity of the 6550 Ion is impressive, but it's set to be dwarfed when the PCIe Gen5 Phison Pascari D205V (offered in U.2 and E3.L formats) arrives in Q2 2025. As well as a 61.44TB model, there will also be one with double the capacity, 122.88TB, and it will be interesting to see how that compares in terms of performance.

You might also like

• These are the fastest SSDs you can buy right now
• And these are the largest SSDs and hard drives on the market
• Price of world's largest SSD has nearly doubled since launch

from TechRadar - All the latest technology news https://ift.tt/c69BP47