Latest News - Page 532

The power required to run complex, massive clusters of high-performance AI GPUs continues to skyrocket with the power of the AI chips themselves, but new research has a reduction in energy consumption required by AI processing by at least 1000x.

In a new peer-reviewed paper, a group of engineering researchers at the University of Minnesota Twin Cities have showed an AI efficiency-boosting technology, which is in lamens terms a shortcut in the regular practice of AI computations that massively reduces energy consumption for those workloads.

AI computing sees data transferred between components processing it (logic) and where data is stored (memory and storage). The moving around of this data back and forth is the main factor in power consumption being 200x higher than the energy used in the computation, according to this research.

AMD Fluid Motion Frames of AFMF is the company's driver-based frame generation solution, which means that with a compatible Radeon GPU and the latest AMD Software: Adrenalin Edition, you can enable the technology in thousands of games via this one-click solution.

As a driver-based solution, AFMF differs from AMD's FSR 3 and NVIDIA DLSS 3 Frame Generation, which require per-game implementation. AFMF debuted not that long ago, and the consensus is that its impressive tech has the potential to become a game changer for boosting performance in a wide range of machines, including mobile devices like the Steam Deck.

In a surprise announcement, AMD has launched AMD Fluid Motion Frames 2 into 'Technical Preview,' promising lower latency and higher performance. Thanks to new AI-optimized enhancements, AMD calls it a "significant upgrade" from the initial release.

According to reports, one of the most popular GeForce RTX 40 Series GPUs, the GeForce RTX 4070, could be in short supply as of August 2024 due to issues regarding the supply of GDDR6X memory.

According to the Board Channels forum (via @harukaze5719 on X), an unforeseen issue with a new batch of GDDR6X memory modules will directly impact the GeForce RTX 40 Series GPU supply. GDDR6X memory can be found in the GeForce RTX 4070 and above, so the shortage will also apply to the newer RTX 4070 SUPER, the RTX 4070 Ti SUPER, the RTX 4080 SUPER, and the flagship GeForce RTX 4090.

According to the latest Steam Hardware & Software survey, the GeForce RTX 4070 is one of the most popular new GPUs among PC gamers. It hits a 1440p sweet spot, making it an attractive option. If supplies become constrained, it will be interesting to see if gamers opt for the AMD alternative or wait to purchase a new GPU.

AMD's new Ryzen AI 300 series "Strix Point" APUs have a new integrated RDNA 3.5 GPU, which has arrived in the form of the Radeon 890M and 880M integrated GPUs.

The new RDNA 3.5-powered Radeon 890M and 880M have been tested, which show us we can expect up to 41% more performance over previous-gen RDNA 3 integrated GPUs. AMD's new Ryzen AI 300 series "Strix Point" APUs have new Zen 5 CPU cores, new RDNA 3.5 GPU cores, and a new XDNA 2 NPU for AI workloads.

AMD's new RDNA 3.5 GPU acts as a refresh and optimization of RDNA 3, with the RDNA 3.5 GPU powering the Radeon 890M integrated GPU inside of the Ryzen AI 9 HX 370, and the Radeon 880M inside of the Ryzen AI 9 365.

AMD's new Zen 5-based Ryzen 9 9950X processor is nearly here, with Bilibili tech influencer "Ordinary Uncle Tony" using LN2 cooling and overclocking the 9950X to 6.5GHz, breaking the Cinebench R23 world record.

The new Ryzen 9 9950X processor was pushed to 6.5GHz using LN2 cooling, with Ordinary Uncle Tony hitting a multi-core score in Cinebench R23 of 55,327 points during a live event held in China called China Joy. As it stands, the current world record Cinebench R23 (16-core) score on HWBot's official charts is 50,843 by the Ryzen 9 7950X overclocked by Safedisk.

Tony ran his Ryzen 9 9950X processor at stock, scoring 42,689 points at its stock 5.0GHz. When aiming for his 6.0GHz target, Tony reached a score of 51,204 points which beat out the current 16-core Cinebench R23 world record by Safedisk and the 7950X.

SK hynix has just announced it has introduced the best-performing GDDR7, a next-generation graphics memory product at 32Gbps speeds.

The South Korean company said that the development of GDDR7 in March comes amid "growing interest by global customers in the AI space in the DRAM product that meets both specialized performance for graphics processing and fast speed". SK hynix promised volume production of its new GDDR7 memory in Q3 2024.

SK hynix's new GDDR7 memory comes at an operating speed of 32Gbps, which is a whopping 60% improvement from the previous-gen GDDR6 standard, and GDDR7 can scale up to 40Gbps depending on the product. When used on a high-end graphics card, SK hynix's new GDDR7 memory is capable of over 1.5TB/sec of memory bandwidth, enough speed to transfer 300 x 1080p movies (at 5GB each) per second.

TSMC is currently scurrying to get fabs built on US soil right now, with Intel and its own semiconductor business with Intel Foundry, poaching TSMC senior engineers to help its long-term growth in the semiconductor industry... against TSMC.

Intel has been poaching senior engineers away from TSMC as they are coming into the US to help the Taiwan semiconductor giant set up its foundries in Arizona, USA. Intel is recruiting the TSMC staffers for its own Intel Foundry business, which is kind of dog as TSMC is helping Intel, and Intel is poaching staff out from under them.

On top of that, Intel has outsourced the production of its next-gen Core Ultra 200V "Lunar Lake" CPUs to TSMC using its 3nm process nodes for the CPU, GPU, NPU, and on-package memory on a single chiplet on a compute tile on 3nm. The SoC and I/O dies will are the only part of the chiplet made on TSMC's 6nm process node.

Sabrent has up some amazing deals on Amazon right now, with up to 26% discount on select products, check them out:

First up we've got the Sabrent Universal Docking Station with Stand for Tablets and Notebooks, which is discounted 20% from its original price of $99.99 down to $79.95. There's also the Sabrent USB-C Universal Laptop Docking Station, with tons of USB-C connectivity, dual HDMI 2.0 and dual DisplayPort 1.4 display connectors, discounted 21% down from $99.95 to $79.95.

Sabrent also has the USB-C to 2.5GbE ethernet adapter, which is made for the Apple MacBook Pro, MacBook Air, and iPad Pro 2018 and later. From its name, you should work out it's a USB-C to 2.5GbE ethernet adapter, converting USB-C to wired 2.5GbE (super-fast) ethernet. This is discounted 26% from $26.99 to $19.99 right now.

Fallout series co-creator Tim Cain shares some interesting tidbits on Interplay's original plans for the franchise's infamous vaults.

The Fallout series has little over 100 vaults, all-told...but this number was supposed to be much, much higher. Interplay, the devs of the first Fallout game, had planned for about 1,000 vaults at one point in time.

In a recent YouTube video, Fallout producer, co-creator, lead programmer and designer Tim Cain explains why there were so many vaults in planning and, more importantly, why there aren't that many vaults in the actual games.

Windows 11 just saw the addition of another advert - or 'recommendation' as Microsoft likes to call them - in what will come as a surprise to pretty much no one.

The newest ad to join the growing collection of various nags and prods to use mostly Microsoft services is for OneDrive.

What's particularly annoying here is that it's a full-screen pop-up pushing the cloud storage service, which as Windows Latest reports, appeared after the tech site had installed a recent Windows 11 update.

Sony has revealed two new expensive special edition PlayStation 5 controllers for purchase in August to commemorate the launch of upcoming first-party games, signalling an emphasis on the "highly profitable" accessories market.

Console-makers are no stranger to accessories. Since the earliest Gen 1 console days, we've seen a multitude of extra controllers roll out, many of which like the NES' turbo controller added a new level of capability to gameplay. The advent of modern consoles has facilitated the expensive "pro" enthusiast line of accessories, ranging from the $199 DualSense Edge to the $150 Xbox Elite controllers.

Given these prices, it should be no surprise that accessories are a highly profitable market, especially for Sony, who's aforementioned DualSense Edge costs half as much as a digital-only PS5. And Sony has said as much--during a 2023 business segment briefing, SIE exec Eric Lempel affirmed that accessories like special edition DualSense controllers are a "highly important and highly profitable category."

Intel's next-generation Core Ultra 9 288V flagship "Lunar Lake" CPU has appeared on Geekbench, the only Lunar Lake CPU with a default PL1 power setting of 30W.

The new Core Ultra 9 288V "Lunar Lake" CPU with its higher PL1 set to 30W is 13W higher than other members of the upcoming Intel Core Ultra 200 series CPUs. OEMs can play around with the PL1 value for their respective systems, but with the Core Ultra 9 288V processor, Intel sets it to 30W.

The Core Ultra 9 288V will have 32GB of on-package LPDDR5-8533 memory, P-Core boosts of up to 5.1GHz, E-Core boosts of up to 3.7GHz, and Xe2-GPU "Battlemage" integrated graphics through Arc 140V at up to 2.05GHz.

ASUS has just launched its new AMD Ryzen AI 300 series "Strix Point" APU-powered laptops, for both gamers and creators, starting from $1399.

ASUS launched the new ProArt P16, ProArt PX13, and ZenBook S16 laptops, all powered by the Strix Point APUs, which feature new Zen 5 CPU cores, RDNA 3.5 GPU cores, and an upgraded XDNA 2 NPU for AI workloads with up to 50 TOPS of AI performance.

All three of the new ASUS laptops feature the highest-end Ryzen AI 9 HX 370 chip, with the ProArt P16 featuring a 16-inch 4K+ OLED panel with touchscreen capabilities, true 10-bit color, 500 nits of HDR peak brightness, and more.

TSMC has just completed its first wave of purchases plans for ASML's bleeding-edge High-NA EUV lithography machines, which TSMC will use for its next-gen A14 process in Q3 2027, and A14P process node expected in 2028.

The news is coming from DigiTimes which reports that with the large-scale procurement plan established, TSMC continues to be ASML's largest customer. TSMC is also "the key" to ASML's future operational growth momentum as the development of the Chiense market continues to be suppressed.

Intel announced in April 2024 that its R&D team at its R&D base in Hillsboro, Oregon, USA, that it had completed the industry's first High-NA EUV assembly. If you didn't already know, Intel is TSMC's number one main rival... and Intel had ASML's first High-NA EUV machines installed.

NVIDIA is gearing up into the official release of its next-gen Blackwell AI GPU architecture, with reports suggesting the supply of high-end RTX 40 series GPUs has been "significantly reduced" by up to 50% and that supply shortages and price hikes could be expected.

In a new report from UDN, we hear that if the price of NVIDIA graphics cards increases, AIB manufacturers like ASUS, GIGABYTE, MSI, and ASRock are expected to seize the business opportunities as prices increase for consumers.

These companies are key partners of NVIDIA and its AI servers, and as the AI server boom continues (it's really just getting started), board quotations are rising simultaneously, which makes their momentum even more bullish for these Taiwanese companies, adds UDN.

NVIDIA is the company powering the tech industry's massive push into artificial intelligence-powered systems, as the green team is making the incredible hardware that makes it possible for these impressive tools and features to exist.

NVIDIA's dominance in this market was achieved by providing the best hardware for training AI systems, which briefly made the green team the most valuable company on the planet, taking the crown of long-standing tech giants such as Amazon and Microsoft. NVIDIA has since moved down to third place but remains the dependent player in the world of AI-focused microprocessors. With the push into AI many developers want to continue training their creations but don't necessarily want to rely on or keep fueling the massive beast that is NVIDIA.

A new report from The Information provides an example: Microsoft and OpenAI are in talks with several chip designs to create a new AI chip to rival NVIDIA. One of those companies was Broadcom, which is ranked the 13th most valuable company in the world for its solutions in semiconductors and software infrastructure.

X, formerly known as Twitter and the digital echo chamber for Elon Musk and his politics, also has a chatbot and powerful AI called Grok. Created by xAI, alongside hundreds of thousands of high-powered NVIDIA GPUs, Grok is described as an AI with a "rebellious streak" that will deliver candid, unfiltered responses.

X recently updated its terms and settings for all users. By default, it uses all X data for training. Grok now has access to everybody's posts, including yours, if you're on X. This move follows Meta and is understandable, given that massive amounts of raw data are a key ingredient for training and creating complex AI models like Grok.

Several AI companies and models have been under fire lately, with reports indicating that some have been scraping YouTube and other public forums to train AI. According to a Microsoft executive, if it's online, it's free to scrape. So, yes, X, Elon, and Zuckerberg are not alone in looking to social media platforms for AI training. The good news is that you can opt-out.

Call of Duty's multiplayer matchmaking uses a custom SBMM, or skill-based matchmaking, which has been praised and criticized in equal measure. Critics of SBMM in a game like Call of Duty claim that skill-based matchmaking leads to similar-feeling rounds where competition is high but unpredictability is low. SBMM, in a nutshell, is all about pairing players of a certain skill level with other players of a similar skill.

As part of a new in-depth paper and fascinating report surrounding the "complex and multifaceted domain" that is Call of Duty matchmaking, Activision confirmed that in early 2024, it ran a secret 'Deprioritize Skill Test in Call of Duty: Modern Warfare III' in North America. Activision put around 50% of the population into random matches without the usual skill check and grouping.

"For the treatment group, we loosened the skill constraints," Activision writes. "The other half of the population was left with the standard configuration." The test was run for a month, and the results were fascinating. Outside of the top 10% of players, the returning player rate metric was down by a staggering 90%.

Two weeks ago, Apple rolled out the public beta for iOS 18, the operating system update that was slated to add Apple's highly anticipated AI features. Unfortunately, the public beta didn't contain any Apple Intelligence features, leaving users somewhat confused as to where they are.

A new report from Bloomberg's Mark Gurman, a known Apple insider, reveals the Cupertino company will be rolling out Apple Intelligence software updates several weeks after the launch of iOS 18, iPadOS 18, and macOS Sequia, which means the touted AI-powered features won't be present at the launch of the new iPhone 16 line-up.

However, software developers will be getting the new Apple Intelligence features in the first betas for iOS 18.1 and iPadOS 18.1 as soon as next week. Why is this happening? Gurman writes that the new AI features will be rolled out in a software update in October, and according to people who spoke to Gurman under the agreement of anonymity, the delay is to give Apple developers more time to squash as many bugs as possible.

AMD delayed the launch of its Zen 5-based Ryzen 9000 series CPUs, but it could've all been over something as simple as a typo: with a Ryzen 9 9700X processor spotted, which is arriving as the Ryzen 7 9700X.

Maybe the company had some very last-minute naming changes, but a batch of Ryzen 7 9700X processors was shipped with the wrong label of "Ryzen 9 9700X". This could've been the reason for the Zen 5 desktop CPU launch delay, as AMD only explained that "during final checks, we found the initial production units that were shipped to our channel partners did not meet our full quality expectations".

AMD continued: "Out of an abundance of caution and to maintain the highest quality experiences for every Ryzen user, we are working with our channel partners to replace the initial production units with fresh units".