Danh mục: Tin tức

The worst rivals for any chip designer and PC maker are not its direct competitors, but rather devices that their potential customers already own. Therefore, to make those customers buy something new, they need to advance their products at a rapid pace so that a new PC would offer radically better experience than a three-years-old computer. This is apparently what Intel and Lenovo are doing in their joint co-engineering lab in Shanghai. 

Intel and Lenovo’s Advanced System Innovation Lab serves as a breeding ground where engineers from both companies combine their skills to build next-generation laptops that provide strong performance, elegance, features, and user experience, reports DigiTimes.

Lenovo and Intel have a long history of working together on multiple innovative products, including ThinkPad X-series as well as ThinkPad X Fold series. To build such systems, Intel and Lenovo not only need to overcome hardware-related challenges such as performance, power management, and thermals, but also software-related issues. In addition, Intel’s dedicated teams work closely with Lenovo in other co-engineering labs located in Zizhu and Pudong

“We share a long and illustrious history of deep engineering collaboration with Lenovo,” said Zheng Jiong (ZJ), senior director of client customer engineering for Intel China’s client computing group (CCG). “We work together very well and are thankful for the innovation support Lenovo has given us through joint labs like these.” 

A notable achievement of the Advanced System Innovation Lab in Shanghai is development of an OLED display driver that can run two OLED screens instead of one, which opens doors to a number of potentially interesting use cases.  

“This work was critical to the development of our platform,” said Zhijian Mo, director of platform design and development in Lenovo’s intelligent devices group. 

Furthermore, both companies joined forces with DRAM makers to enhance LPDDR5 memory data transfer rates.

While Lenovo remains a key partner, Intel also teams up with other global PC OEMs and software vendors. Their collective goal is to break technological barriers, identify core issues, and engineer enhanced PC solutions. Several advancements in CPU, power, thermal management, and other PC parts have emerged from this cooperative approach.

Looking ahead, as work on the Meteor Lake platform nears completion, plans for the Lunar Lake platform are already in motion and expected to be ready in 2024. To make Lunar Lake-based PCs radically better than systems in use today, the two companies are again collaborating across multiple fronts, but this is probably something the two companies would prefer not to discuss in detail for now. 

“It is a very special project that involves detailed co-engineering efforts between both our teams,” said Mo.

As Elon Musk awaits his doctor’s permission to fight Mark Zuckerberg in an on-again / off-again cage match, the two billionaires’ companies continue to spar in the consumer adoption arena. Meta forged ahead with Threads’ aggressive update schedule today as it tries to challenge Musk’s erratic X (rebranded Twitter). Zuckerberg announced today that the young platform now supports sharing posts to Instagram DMs, custom alt text for photos and videos and a new mention button.

The Send on Instagram option lets users publish their Threads posts directly to Instagram DMs via the Send button. The feature could be seen as one of the “retention-driving hooks” Meta cooked up to ensure “people who are on the Instagram app can see important Threads,” as Reuters reported Chief Product Officer Chris Cox said in a recent company meeting. The strategy is allegedly tied to reports of falling engagement after the fledgling platform added over 100 million users in less than a week. Zuckerberg reportedly described the decline as “normal” and expected retention to grow as Meta continues to flesh out the social channel, which launched in early July.

Meta

Meanwhile, the custom alt-text option is an accessibility feature allowing Threads users to add (or edit existing auto-generated) alt text for photos and videos before uploading. The new mention button makes it easier to tag profiles in your Threads posts. Finally, the platform is making it easier to verify your identity on fediverse platforms like Mastodon. “You can now add your Threads profile link on supported platforms to verify your identity,” Instagram head Adam Mosseri posted today.

These are only the latest additions Meta has rolled out in the past few weeks. It announced earlier this month that a much-needed web version is on the way, offering desktop use for the first time, along with a proper search function. Other post-launch additions include a chronological feed and the ability to sort your following list and view your liked posts.

While Wi-Fi 7 is right around the corner, many are still getting acclimated to Wi-Fi 6E. We’ve already tested several Wi-Fi 6E routers, with the latest coming courtesy of MSI.

MSI’s RadiX AXE6600 is a compelling entry into the wireless gaming router field with its array of six antennas and heavy use of MSI’s Mystic Light to liven things up stylistically. MSI also enhances the RadiX AXE6600 with a 2.5Gbps WAN/LAN port (one of the four remaining 1 Gbps ports can also be used for WAN if you desire). 

Available for a reasonable $256, RadiX AXE6600 delivers good looks, a comprehensive web GUI and performed well in our benchmarking suite. That makes it one of the best gaming routers available and also one of the best routers overall.

Looking at the router, the feature that stands out most is the antenna array, with six poseable elements along its sides and back panel. Each antenna features an LED for syncing with MSI’s Mystic Light. There’s one additional LED located in the MSI logo on top of the router. 

Interestingly, while you can enable Mystic Light from the web interface (or using the smartphone app), you cannot adjust the colors using these methods. Instead, you must install the MSI Center app on a Windows PC to change the RGB settings.

A status panel near the top rear of the router also highlights internet connectivity, power status, and activity from the four gigabit ethernet ports. There are also buttons for Wi-Fi and WPS, along with an “LED Off” button which, understandably, turns off the LEDs on the status panel.

Moving around to the rear of the router, you’ll find a 2.5 GbE WAN port, four GbE LAN ports, a USB 3.0 port, and a power button. There’s also a plethora of vents along the front, sides, and bottom of the RadiX AXE6600 to keep the router cool.

MSI RadiX AXE6600 Wi-Fi 6E Gaming Router Specifications

MSI RadiX AXE6600 Setup

Although you can set up the RadiX AXE6600 using a smartphone, I used the tried-and-true web browser method. After navigating to the http://msirouter.login address, I could log in using the credentials printed on the bottom of the router. Then, I was greeted with a vibrant, easy-to-navigate user interface with large buttons and clear graphics.

Thankfully, the router already had the latest firmware installed. However, updating the firmware is a simple affair. You just must navigate to Advanced → Administration → Firmware Upgrade. You can manually upgrade using a firmware package downloaded directly from MSI or an online tool that automatically checks for new updates and installs the firmware.

When you first login to the router, you’re greeted with the Dashboard, which presents information on hardware resources consumed and currently-connected devices. You’ll also find presets to change the router’s QoS (Quality of Service) settings. The four that most users will likely enable at some point include AI Auto, Gaming, Streaming, and WFH.

There’s also a dedicated Game Center subsection where you can enable Game Boost mode. If you select the Traditional QoS mode, you’re also given the option to prioritize MSI devices. Game Center is also where you’ll find controls to enable the Mystic Light function.

Like most Wi-Fi 6E routers, the RadiX AXE6600 ships with three wireless bands enabled (2.4GHz, 5GHz, 6GHz), each with its own SSID. In this case, the router defaulted to MSI_2G_FE, MSI_5G_FE, and MSI_6G_FE as the SSIDs, but you will want to set your own.

However, the RadiX AXE6600 can also use Smart Connect, which sets a single SSID for the router. Whether your device maxes out at Wi-Fi 5, Wi-Fi 6 or Wi-Fi 6E, it will connect to the router using the same SSID, with the client choosing the appropriate band automatically. However, keep in mind that Smart Connect can often diminish network performance.

The Radix AXE6600 offers basic parental controls, including the ability to filter for Adult content, gambling, violence, drugs/firearms, malicious content and games. Each is individually selectable, and you can specify the MAC address to which the settings apply.

MSI RadiX AXE6600 Performance

We conducted multiple tests in a single-family, two-story home with a 500 Mbps connection, using a laptop with an AMD RZ616 160 MHz network adapter as the client. Another PC, attached via Ethernet, functioned as the server to receive traffic. We used iPerf to test throughput and ping to test latency. Four sets of tests were conducted on the 2.4-GHz, 5-GHz and 6-GHz bands:

Near uncongested:  Testing laptop approximately 6 feet away from the router, no substantial traffic being carried across other devices

Far uncongested: Testing laptop approximately 25 feet away from the router, no substantial traffic being carried across other devices

Near congested: Testing laptop approximately 6 feet away from the router with videos streaming on four devices throughout the house

Far congested: Testing laptop approximately 25 feet away from the router with videos streaming on four devices throughout the house

The RadiX AXE6600 put up solid numbers on the 6-GHz band, as witnessed by our iPerf test scores. The RadiX AXE6600 can hit a theoretical 4,804 Mbps on the 6-GHz channel, but real-world iPerf numbers for every router we’ve tested are in the 150 to 550 Mbps range. 

The RadiX AXE6600 managed 532 Mbps in the 6-foot iPerf test, which still far surpassed that of the next-closest competitor, the Netgear Nighthawk RAXE300. The results weren’t nearly as impressive 25 feet away, where the RadiX AXE6600 managed just 283 Mbps. Congested performance at 6-GHz was also strong, reaching 500 Mbps at 6 feet. However, that performance dropped at 25 feet, dropping to just under 200 Mbps, putting it well behind the other Wi-Fi 6E competition.

We witnessed a similar Dr. Jekel and Mr. Hyde performance at 5-GHz, with the 6-foot numbers coming in at a healthy 486 Mbps, well over 100 Mbps faster than the next-closest competitor. The tables turned at 25 feet, with the RadiX AXE6600 coming in at just 149 Mbps, below the average performance of the other assembled routers. With congested traffic, the RadiX AXE6600 pulled up last place at 6 feet and 25 feet.

At 2.4-GHz, the RadiX AXE6600 had some of the weakest throughput numbers of any router we tested. However, it still had very low latency and its speeds were more than good enough for the kinds of devices most people would have on a 2.4-GHz band (smart home devices, old tablets, etc). 

The 2.4-GHz uncongested iPerf numbers reached 128 Mbps at 6 feet and 44.5 Mbps at 25 feet, both of which are quite low. The router slipped into last place among its competitors on the 2.4-GHz band when the network was congested, managing just 73.6 Mbps at 6 feet compared to 101.2 Mbps for the nearest competitor and 151.4 Mbps for the first-place Asus RT-AXE7800.

One thing that was consistent through all the tests, however, was the ping. Across 6-GHz, 5-GHz and 2.4-GHz channels with congested or uncongested traffic, the RadiX AXE6600 maintained speedy pings of 1 to 4 ms. Latency matters more than throughput when you’re playing games because most of the game data is actually on your PC, but your movements and commands need to reach the server quickly.

Bottom Line

The MSI RadiX AXE6600 offers strong performance on the 6-GHz band, a wealth of features, and striking looks at an attractive price point. On the other hand, 5-GHz performance was inconsistent, and 2.4-GHz performance put it at the back of the pack compared to other assembled Wi-Fi 6E routers. The RadiX AXE6600 has an MSRP of $349.99; however, retailers like Amazon offer the RadiX AXE6600 with a $90 discount, taking it to just $256 shipped.

If you have Wi-Fi 6E devices on your network, the RadiX AXE6600 is by far the best-performing router that we’ve tested. The numbers that the router was able to consistently deliver on the 6-GHz band far outpaced the competition at 6 feet and were in the same ballpark at 25 feet. 5-GHz performance was a bit more sporadic (although it took a considerable lead in the 6-foot iPerf test). 2.4-GHz performance, while on the lower side, is more than adequate for the older, slower devices that typically use that band.

We also can’t ignore that MSI has crafted an easy-to-use UI for configuration changes. There are plenty of presets for those that want AI to take over for QoS, or you can get down in the weeds and make granular adjustments on your own. And for those that aren’t enamored with the flashiness of the Mystic Light, it can be turned off.

If you don’t have any Wi-Fi 6E devices on your network or don’t plan on enlisting any soon, there are plenty of cheaper Wi-Fi 6 routers out there to consider, including the budget-minded Asus RT-AX1800S. But if you want blazing-fast Wi-Fi 6E performance, the RadiX AXE6600 is tough to beat.

Just when we thought dual-cores were dead, Intel is rumored to be bringing back the “obsolete” core configuration for another generation. A Tweet by leaker chi11eddog alleges Intel is developing a new 14th Gen entry-level dual-core chip that will feature one of Intel’s three hybrid CPU architectures found in its Best CPUs, Alder Lake, Raptor Lake, or Raptor Lake Refresh.

According to chi11eddog, this new chip will be the spiritual successor of the Pentium Gold G7400. Thanks to Intel’s removal of the Celeron and Pentium sub-brands altogether, the new chip will be dubbed the “Intel 300,” similar to other entry-level chips (like the “Intel Processor N200”).

The chip’s reported specifications include a 3.6GHz base frequency (no turbo clock), a 6MB L3 cache capacity, and a core configuration consisting of two P-cores with four threads. Power consumption is rated at 46W. The Twitter leaker did not reveal any specifications for the iGPU, but for obvious reasons, we expect this CPU to sport one of Intel’s lower-end configurations — like its UHD 730 Graphics unit.

Compared to Intel’s other entry-level CPUs like the Core i3 series, this “Intel 300” will be the cheapest and least powerful entry in Intel’s desktop lineup. The chip features two fewer cores than the Core i3-13100 (Intel’s lowest-end i3 right now), and a much slower clock speed thanks to the lack of Turbo Boost technology. As previously stated the Intel 300 will also effectively be the replacement of Intel’s previous-generation desktop Celeron and Pentium processors which also sported dual-core configurations.

Because of its dual-core design, we don’t expect this chip to be great for anything beyond casual web browsing, office work, and video streaming. Even with the addition of Hyper-Threading, the physical limits of dual-core designs make them sub-optimal for gaming tasks, especially in modern game engines which can take advantage of 6 six CPU cores.

Chi11eddog believes the CPU will be arriving as soon as this quarter, so we shouldn’t have to wait long for this new entry-level chip to arrive on store shelves — assuming it exists. With its 14th Gen moniker, the new dual-core will allegedly arrive in conjunction with Intel’s 14th Gen Raptor Lake Refresh lineup which is also expected to launch very soon.

Users looking to expand their storage without spending more than they have to should take a look at this offer from Amazon on the Intel 670p 1TB SSD. This SSD has been going for around $39 lately but today is marked to one of it’s best prices yet—just $34 — or around 3 cents per GB.

This is not the fastest SSD on the market and is probably not ideal for anyone looking for a high speed gaming drive but it’s plenty fast for anyone looking for some extra storage. We reviewed the Intel 670p SSD when it debuted in 2021 and appreciated its performance which has held up over the years.

This offer applies only to the 1TB model but there are other capacities available in this line as well, including a 500GB and 2TB edition. All of the drives in this line have an M.2 2280 form factor and use a PCIe 3.0 X4 interface along with a Silicon Motion SM2265 controller. The Intel 670p SSD is designed with Intel 144-layer QLC memory. This particular size is able to reach read/write speeds as high as 3500/2500 Mbps. Again, that might not be the fastest on the market but it’s not bad for a drive in this price range.

Users have option 256-bit AES encryption to take advantage of for added security. The drive is supported by a limited 5-year manufacturer’s warranty from Intel that voids should the drive reach 370 TBW. It’s also backed by Amazon’s 30-day return policy.

Visit the Intel 670p 1TB M.2 2280 SSD product page at Amazon for more information and purchase options.

Asus has started to sell some of the best graphics cards under the company’s Tianxuan (Heavenly Chosen) portfolio. Marketed under the TX Gaming moniker, the new trio of graphics cards encompasses the GeForce RTX 4070, GeForce RTX 4060 Ti, and GeForce RTX 4060.

The Tianxuan series may be new to some since it’s one of Asus’ China-exclusive sub-brands. The company has launched many gaming laptops from the Tianxuan series but is extending the product portfolio to graphics cards, specifically the GeForce RTX 40-series parts.

The TX Gaming graphics cards spit images of Asus’ Atlas Shark (Megalodon) series. They’re using the same cooler but with a different color theme. While the Atlas Shark series cooler stick to a black and grey combination, the TX Gaming series arrives with a white and grey theme with teal accents. There aren’t any visible changes to the cooling solution. It’s still a thick shroud with three axial, dual-ball bearing cooling fans. The graphics cards also have a reinforced metal backplate that helps with cooling.

Regardless of the model, the TX Gaming graphics card measures 12.48 x 5.35 x 2.13 inches (317 x 136 x 54mm), so we’re looking at a triple-slot design. The size of the cooler seems a bit overkill for the lower-tier SKUs, such as the GeForce RTX 4060 Ti (160W) or GeForce RTX 4060 (115W). However, it looks like Asus didn’t want to waste resources redesigning the cooler for each SKU. Instead, the company used the same cooler for all three graphics cards.

As their part numbers insinuate, the TX Gaming graphics cards feature a factory overclock. However, Asus didn’t share this information on its product pages or online store. We only know that the graphics cards require one 8-pin PCIe power connector for external power. Many will be glad that Asus didn’t opt for the pesky 16-pin power connector. At any rate, you’d want a power supply with a minimum capacity of 650W for whichever model you pick up. Asus didn’t alter the design of the display outputs. Therefore, the TX Gaming graphics cards provide one HDMI 2.1 port and up to three DisplayPort 1.4a outputs.

Asus sells the TX Gaming GeForce RTX 4070 (TX-RTX4070-O12G-Gaming) for $733.54. Meanwhile, the TX Gaming GeForce RTX 4060 Ti (TX-RTX4060TI-O8G-Gaming) and TX Gaming GeForce RTX 4060 (TX-RTX4060-O8G-Gaming) retail for $526.91 and $386.82, respectively.

There’s seemingly no end to the number of gadgets in the world that are able to run Doom, and thanks to the Raspberry Pi RP2040 microprocessor, this number keeps growing. Today we’re sharing another fascinating Doom port, this one put together by Bob from TheKeebProject who’s managed to successfully run Doom inside of a custom keycap.

The project is built around a tiny custom PCB that features the RP2040 microprocessor. It’s small enough to fit inside of a keycap and is sealed within clear resin. The port is a customized fork of an existing RP2040 Doom port originally created by Graham Sanderson. A few modifications were made to enable support for both SPI and I2C display modules.

According to Bob, the goal of this project was to emulate Doom with as much of its original functionality as possible. Bob wanted to ensure stereo sound, the ability for network multiplayer, save/load functions and any other feature you’d expect from the original game including the original resolution of 320 x 200px.

Is it practical to play Doom on a keycap? We would argue that’s subjective. Is it fascinating and worth creating a keycap that can play Doom? We think so and the end result is certainly worthy of a few bragging rights.

The only way to get the Doom experience he wanted in the form factor of a keycap was to create a PCB from scratch. It’s made just small enough to fit inside the keycap alongside a display module capable of supporting the 320 x 200px resolution. A USB port is included to allow for input as well as power to the Doom key.

Bob was kind enough to make the project open source for anyone interested in checking it out in greater detail. You can find his RP2040 Doom fork source code over at GitHub. There you can also check out the original RP2040 code it was forked from by Graham Sanderson to see how it works, as well.

If you want to see this Raspberry Pi project in action, the best place to see it is over at Instagram where Bob has shared a demo video of the keycap running Doom. You can also follow him for more updates and cool projects.

Gaming monitors are an ever-expanding and exciting category when it comes to computer hardware. Not only are promising panel technologies like Mini-LED and OLED coming to even lower price points, but we’re seeing refresh rates shoot through the roof for the most dedicated enthusiasts. Now, AOC is throwing down the gauntlet with its new Agon Pro AG276QZD, which features a 26.5-inch OLED panel that is certain to be a stunner among the best gaming monitors.

The AG276QZD has a QHD (2560 x 1440) resolution (110.8 pixels per inch) and a heady 240Hz refresh rate. As you might expect from a gaming-centric monitor, that 240Hz refresh rate is backed by Adaptive-Sync and Nvidia G-Sync compatibility. The use of OLED technology ensures incomparable contrast with inky blacks and improved color accuracy versus competing panel technologies (IPS, VA, etc.). AOC says that the AG276QZD can hit 98.5 percent of the DCI-P3 color space and 100 percent of sRGB.

AOC also boasts a 0.03ms GtG response time for the 10-bit panel (1.07 billion colors). For those wondering, the AG276QZD uses an anti-glare coating on the display rather than the glossy finish seen on some OLED monitors.

Regarding connectivity, the AG276QZD features two HDMI 2.0 and two DisplayPort 1.4 ports. There’s also an integrated two-port USB 3.2 (Gen 1) hub and a 3.5mm headphone jack. For those who don’t want to be bothered with adding an external speaker setup, the AG276QZD features two built-in 5-watt speakers. Other features include a Kensington lock and a 100×100 VESA wall mount pattern.

Given that this is a gaming monitor, you won’t be surprised to learn that AOC has integrated Light FX RGB lighting in the AG276QZD. You can sync lighting with other AOC peripherals such as gaming keyboards and mice using AOC’s G-Menu software.

The AOC Pro AG276QZD is currently available, priced at $999 from Amazon. That price gets you a 4-year advance replacement and zero dead pixel warranty.

High-tech giants Amazon, Apple, Intel, Nvidia, and Samsung are looking forward to becoming anchor investors in Arm’s upcoming initial public offering, according to reports from Bloomberg and Nikkei, citing sources with knowledge of the matter. This IPO might amass up to $10 billion and elevate Arm’s valuation to a staggering $60 to $70 billion, according to Bloomberg’s estimates.

Arm, which is owned by SoftBank Group, is preparing for what will potentially be the year’s largest tech IPO, and has engaged with major customers for months regarding their participation. The company has held talks with key global chip designers about becoming anchor investors, and companies like Amazon, Apple, Intel, Samsung, and Nvidia are expected to invest in Arm upon its market listing (note that none of these companies have formally confirmed intention to buy an Arm stake, however). Arm’s strategy involves offering these processor developers medium- to long-term shareholding with the intention of ensuring stock price stability during the listing.

Arm’s global impact on chip designs is undeniable. In the smartphone sector, Arm-based SoCs command close to a 100% market share. They also command a sizeable share in the consumer electronics industry, and, thanks to the efforts of Apple and Qualcomm, Arm’s technology is gaining traction in PCs as well. Throughout the history of Arm, over 250 billion chips based on its architectures have been produced.

Apple, Samsung, Nvidia, and Qualcomm produce some of the world’s best known Arm-based SoCs, which is why Apple (who already has a stake in Arm), Samsung, and Nvidia have reason to invest in the chip IP designer. But Amazon seems to have a particularly good reason to invest in Arm: apparently, Amazon Web Services dominates the Arm server market, with over half of all Arm-based server CPUs globally deployed in AWS datacenters, according to a Bernstein Research report cited by The Register.

The Graviton system-on-chip family, introduced in 2018 and optimized for AWS powered around 20% of AWS CPU instances by mid-2022 — which is a significant portion. These SoCs, optimized for AWS’s requirements, have more cores per socket than competing x86 offerings, reduced power consumption, and significant cost benefits — partly because Amazon does not have to pay premium to AMD or Intel and to a large degree because of lower total cost of ownership and power consumption. The report asserts that in some cases AWS can achieve 20% to 70% lower costs at the same performance compared to competing x86 CPUs, though the details are unclear. 

Given the tangible advantages that AWS has with its Arm-based SoCs, it is evident that the company has invested massively in datacenter software ecosystem for Arm processors. Therefore, the company is likely more than interested in further development of the Arm instruction set architecture (ISA) in general and datacenter-oriented Neoverse CPU cores in particular. 

SoftBank, which acquired Arm in 2016 for $32 billion, tried to sell the company to Nvidia for about $40 billion in 2020 – 2022, but regulatory issues prevented the sale. In the aftermath, SoftBank chose to steer Arm towards an IPO. If the IPO reaches its peak valuation, it would rival the tech sector’s largest IPOs — such as Alibaba in 2014 and Meta Platforms (formerly Facebook) in 2012.

Swiss researchers have found holes in AMD Ryzen processor security. AMD has outlined the newly uncovered “Inception” attack in its official CVE-2023-20569 bulletin. Like some of the most infamous CPU vulnerabilities, Inception is a speculative side channel attack, which can possibly lead to privileged data leakage to unprivileged processes. At the time of writing AMD is not aware of any Inception exploits outside of security research circles.

Unfortunately for AMD and its users, Inception affects the latest AMD Ryzen processor families based on Zen 3 and Zen 4 cores — across data center, desktop, HEDT, and mobile. However, we must be thankful that, as details of Inception go live, mitigations are in the pipeline.

In its security bulletin, AMD says that customers may have a choice between a standalone microcode patch or a BIOS update that incorporates the microcode patch. AMD CPU users may be familiar with the quite frequent AGESA microcode update releases, and some patches will be delivered this way later this month, while others might have to wait until December. Users are asked to check with their OEM, ODM, or MB for a BIOS update specific to their product.

To be clear, AMD says that users of products based on the Zen or Zen 2 CPU architectures don’t need any patching “because these architectures are already designed to flush branch type predictions from the branch predictor.” This is a little different from what the researchers from ETH Zurich say in their Inception paper (PDF), so we hope things will become clearer soon.

So, what is the new Inception vulnerability? We already mentioned it’s a speculative side channel attack like the infamous Spectre. Processor makers thought they had eliminated the ability of attackers to snoop at the look up table used for accessing DRAM in Spectre mitigations. However, the ETH Zurich team discovered that, on Zen architecture processors, they could “make the CPUs manufactured by AMD believe that they had seen certain instructions before, whereas in reality that had never happened,” explained one of the researchers to EE News Europe. Another summary of Inception is provided by COMSEC.

Over the past one and a half years, we have studied two phenomena that enable an unprivileged attacker to leak arbitrary information on all modern AMD CPUs –
▪ Phantom speculation: We can trigger misprediction without any branch at the source of the misprediction.
▪ Training in Transient Execution: We can manipulate future mispredictions through a previous misprediction that we trigger. –
Putting the two together gives rise to a new type of attack called Inception: we can inject future mispredictions through a previous misprediction that we trigger — in the absence of branches.

Ultimately, Inception means that an attacker of an unpatched AMD Ryzen system can obtain leaked data from anywhere in the computer’s memory.

AMD says that it believes the Inception vulnerability is “only potentially exploitable locally, such as via downloaded malware.” Thus, it says good general system security practices should keep Inception dangers away from your PC. However, the researchers reckon Inception could be used by an attacker in the context of cloud computing, where several customers share the same processing hardware resource.

In case you missed it, AMD’s fiercest CPU-making rival, Intel, was also in the news just hours ago for a notable newly discovered security vulnerability — the Intel ‘Downfall’ bug, which affects architectures spanning from Sky Lake to Tiger Lake/Ice Lake.

Memory makers are ramping up production capacity for high-bandwidth memory (HBM) due to rapidly surging orders from cloud service providers (CSPs) and developers of processors for artificial intelligence (AI) and high-performance computing (HPC), such as Nvidia, according to TrendForce. The market research firm estimates that annual bit shipments of HBM will grow 105% by 2024. 

TrendForce claims that leading DRAM producers like Micron, Samsung, and SK Hynix supplied enough HBM memory in 2022 to keep pricing predictable. Yet, a rapid increase in demand for AI servers in 2023 led to clients making early orders, pushing production limits. TrendForce, the market research firm now projects that aggressive supplier expansion will raise the HBM sufficiency ratio from -2.4% in 2022 to 0.6% in 2024.

In addition to surging demand for HBM in general, analysts from TrendForce also note ongoing transition of demand from HBM2e (used on Nvidia’s H100 cards, AMD’s Instinct MI250X, and Intel’s Sapphire Rapids HBM and Ponte Vecchio products) to HBM3 (used on Nvidia’s H100 SXM modules, AMD’s upcoming Instinct MI300-series APUs and GPUs) The demand ratio for these is estimated at around 50% for HBM3 and 39% for HBM2e in 2023 and then HBM3 will represent 60% of shipped HBM in 2024. This burgeoning demand, coupled with a higher average selling price (ASP), is expected to significantly elevate HBM revenue in the coming year.

But to meet demand for HBM, makers of memory need to increase output of HBM. This is not particularly easy as in addition to making more memory devices, DRAM producers need to assemble these devices in 8-Hi or even 12-Hi stacks, which requires specialized equipment. To satisfy demand for HBM2, HBM2E, and HBM3 memory, DRAM makers need to procure additional tools to expand their HBM production lines and delivery and testing time for them is between 9 and 12 months, so tangible increase of HBM output is now expected sometimes in Q2 2024.

SK Hynix leads in HBM3 production with most of the output going to Nvidia, TrendForce claims, whereas Samsung mostly produces HBM2E for other processor developers and CSPs. Micron, which does not make HBM3, can only supply HBM2E (and Intel uses such memory on its Sapphire Rapids HBM processor, according to reports) while it is prepping to ramp up production of HBM3E (which the company calls HBM3 Gen2) in early 2024. Just yesterday Nvidia introduced the industry’s first HBM3E-based AI and HPC platform and given that demand for Nvidia’s products is overwhelming, it is likely that Micron will capitalize on the new GH200 Grace Hopper HBM3E-enabled platform.

Meanwhile, TrendForce anticipates that in 2023 ~ 2024 Samsung and SK Hynix will have nearly equal market shares, collectively comprising approximately 95%. By contrast, share of Micron is projected to be between 3% and 6%.

TrendForce observes a consistent decline of the average selling price for HBM products annually. To boost customer demand and respond to softening demand for previous-generation HBM types, suppliers are reducing HBM2e and HBM2 prices in 2023. Although 2024 pricing strategies remain undecided, there is potential for further price cuts for HBM2 and HBM2e due to increased HBM supply and suppliers’ ambitions to expand market share, according to TrendForce. 

 Nonetheless, HBM3 prices are expected to remain stable, and with its higher average selling price relative to HBM2e and HBM2, it could drive the HBM revenue to an impressive $8.9 billion in 2024, marking a 127% YoY growth.

SK Hynix late on Tuesday demonstrated its 321-layer TLC NAND memory at the Flash Memory Summit 2023. This new type of flash is going to enter mass production in 2025, but the company is already showing it off to demonstrate that it is ready for the future.

The demonstrated memory device has 1Tb capacity (128GB) and 3D TLC architecture. SK Hynix says that the 321-layer memory IC features a 59% improvement in productivity compared to a 512Gb 238-layer 3D TLC device, which indicates a tangible storage density increase. Meanwhile, the company does not disclose how it produces 321-layer NAND memory as well as approximate per bits costs compared to previous-generation nodes.

If SK Hynix’s lowest-capacity 321-layer NAND device can store 128GB of data, then it is reasonable to expect premium devices to feature larger capacity and enable SSDs with capacities way larger than what we have today.

SK Hynix yet has to disclose how it has managed to achieve 321 active layers. While it is more than likely that the company used string stacking technique that places one or more 3D NAND devices on top of each other and connects them, the question is whether it stacked two or more 3D NAND devices on top of each other.

Traditionally, SK Hynix labels its multi-layer NAND memory as ‘4D NAND’ as it uses CMOS under Array (CuA) architecture that places NAND logic under the 3D NAND memory cells array to reduce die size and costs. Meanwhile, similar layout is also used by Micron, which uses the term ‘3D NAND.’

Speaking of NAND logic, it is noteworthy that SK Hynix did not share interface speed supported by its 321-layer NAND memory, possibly because it is too early to set them as the company only plans to start mass production of such flash sometimes in the first half of 2025.

“With another breakthrough to address stacking limitations, SK hynix will open the era of NAND with more than 300 layers and lead the market,” said Jungdal Choi, head of NAND development at SK Hynix, said during a keynote speech. “With timely introduction of the high-performance and high-capacity NAND, we will strive to meet the requirements of the AI era and continue to lead innovation.”

Is LK-99, the new substance that has taken the scientific community by storm, the world’s first room-temperature superconductor, a regular superconductor or a non-discovery? As more time passes since the release of Sukbae Lee et al.’s original paper, more researchers are trying to synthesize their LK-99 samples and test them for diamagnetism and superconductivity. New studies from groups in Mainland China, Taiwan and India claim that there’s no breakthrough and that the material may not even be diamagnetic. 

In summarizing the findings of these three research groups, the University of Maryland’s Condensed Matter Theory Center tweeted, “With a great deal of sadness, we now believe that the game is over. LK99 is NOT a superconductor, not even at room temperatures (or at very low temperatures).”

However, these dispositive studies are far from the final word, particularly since each group made its own LK-99-like substance, which may not have been identical to what Lee et al. created. We also still have studies from other sources who claim they have replicated some of Lee et al.’s claims. 

Kapil Kumar et al., a group of scientists from India’s CSIR National Physical Laboratory, released a paper explaining that an LK-99 sample they created in the lab had no evidence of superconductivity at any temperature and was “highly resistive” as well. This result is exactly the opposite of what you’d want to see. The sample was apparently diamagnetic at 280K (6.85 degrees Celsius), but that doesn’t necessarily correlate with LK-99 being a superconductor.

A separate group from National Taiwan University (NTU) ran LK-99 experiments in a live video broadcast on YouTube but failed to show superconductivity. Wang Limin, a professor of physics at NTU, told Liberty Times that on the fifth evening of testing, the material was found to have diamagnetism, but it was not a superconductor. However, he also said that the research continues. 

Diamagnetism occurs when both magnetic poles repel a substance. In the last week or so, we’ve seen several videos which show a tiny black substance, claimed to be LK-99, floating above or below a magnetic field. Superconductors exhibit diamagnetism, so these videos imply that the magnetic levitation proves LK-99 to be a superconductor. However, as the Condensed Matter Theory Center pointed out in a tweet, “NO, diamagnetism is NOT interesting, many materials (e.g. Pb, Cu, P going into LK99) are diamagnetic. It is a run-of-the-mill property.”

Meanwhile, Kaizhen Guo et al., researchers at Peking University’s International Center for Quantum Materials, published a paper claiming that LK-99 actually has some soft ferromagnetic components which cause it to “half levitate” (levitate while still partially touching supporting materials). However, the group stated that they didn’t detect the Meissner Effect or zero resistance, both required for superconductivity.

These three studies certainly suggest that LK-99 could be more useful for powering a hype machine than superconducting electricity. But it’s important to note that each of these groups synthesized its own substance, meaning the material they tested may not have been identical to the sample that Sukbae Lee et al. used in their experiments.

One solution: get LK-99 samples directly from the group in Korea that’s making the superconductivity claims and test those. Apparently, that is happening. According to the Flat Hat, a student newspaper from the College of William and Mary, Korean researchers are sending some LK-99 samples directly to physics professors at Virginia University. So stay tuned.

A new security vulnerability, called Downfall, was revealed today by Intel and the researcher who discovered it, Daniel Moghimi. The new attack uses Gather Data Sampling to steal data and other sensitive information from other users on a computer with Intel processors from 2015 through 2019 ranging from sixth gen Skylake through eleventh gen Rocket Lake and Tiger Lake.

Intel has posted about the vulnerability in a security advisory, INTEL-SA-00828, and has reserved CVE-2022-40982.

Moghami, a senior research scientist at Google (and formerly of the University of California San Diego posted details on downfall.page.

“The vulnerability is caused by memory optimization features in Intel processors that unintentionally reveal internal hardware registers to software,” Moghami wrote. “This allows untrusted software to access data stored by other programs, which should not be normally be accessible. I discovered that the Gather instruction, meant to speed up accessing scattered data in memory, leaks the content of the internal vector register file during speculative execution.”

On the page, Moghami shows demos stealing 128-bit and 256-bit AES keys from other users, as well as spying on typed characters and taking data from the Linux kernel. He suggests that even if you don’t own an Intel powered-device, Intel’s dominance in the server market means that everyone on the internet is affected, and that “in cloud computing environments, a malicious customer could exploit the Downfall vulnerability to steal data and credentials from other customers who share the same cloud computer.”

Intel is releasing microcode for its affected chips. The company “recommends that users of affected Intel Processors update to the latest version firmware provided by the system manufacturer that addresses these issues.” If you’re not using Intel SGX, a hardware-based memory encryption technology from Intel, you can load it from the operating system.

The overhead, per Moghami and Intel, could be as high as 50% depending on if a workload uses Gather. Notably, Intel will have an “opt-out mechanism” in the microcode that allows the mitigation for Downfall to be disabled in order to “avoid the performance impact on certain vectorization-heavy workloads.”

The researcher recommends against opting out: “This is a bad idea. Even if your workload does not use vector instructions, modern CPUs rely on vector registers to optimize common operations, such as copying memory and switching register content, which leaks data to untrusted code exploiting Gather.”

Moghami will present Downfall at the BlackHat USA conference on August 9 and USENIX Security Symposium on August 11. His technical paper can be found here.

Newer Intel chips, like 12th Gen Alder Lake, 13th Gen Raptor Lake, and Sapphire Rapids server chips aren’t affected. 

X is giving advertisers new ways to have some control over what type of content can appear near their ads. The company formerly known as Twitter new “sensitivity settings” that allow advertisers to choose between different types of content filtering for their ads.

The new controls arrive as X is increasingly desperate to win back advertisers. The company’s ad revenue has dropped since Elon Musk took over as brands cut spending on the platform amid concerns about the rise of hate speech and other unsavory content. Since then, watchdog groups have reported several instances of ads from major brands being placed near accounts, and other users.

With the new tool, X says it “will use machine learning to reduce adjacency to varying levels of content according to a brand’s sensitivity threshold in an upcoming campaign.” For now, the two settings available to advertisers include “conservative” and “standard.” The company notes that all rule-breaking content is meant to be excluded from ads regardless of what advertisers have opted into.

X

Under the most restrictive “conservative” setting, ads would be excluded from appearing near “targeted hate speech, sexual content, gratuitous gore, excessive profanity, obscenity, spam and drugs” in the “for You” timeline. The “standard” option would avoid the same topics, but allow spam and drug-related content, according to an example shared by X. The company also plans to add a “relaxed” setting for advertisers who want to “maximize reach” of their ads with the fewest limits on what can appear nearby.

The update isn’t the first time X has introduced tools to promote brand safety. The company previously added other keyword-based that were also meant to limit ad-buyers’ exposure to problematic content. But those changes, introduced in December, seem to have had little effect on X’s ad business. 

Though the company has repeatedly claimed that it has successfully the reach of hate speech on its site, researchers have said otherwise. Last month, Bloomberg that researchers at the Center for Countering Digital Hate (CCDH) indicating hate speech has surged, both in terms of sheer volume and in engagement, since Musk’s takeover of the company. X the findings and is now CCDH, alleging the group “illegally” scraped data.