South Korean startup Panmnesia has unveiled an interesting solution to address the memory limitations of modern GPUs. The company has developed a low-latency Compute Express Link (CXL) IP that could help expand GPU memory with external add-in card. Current GPU-accelerated applications in AI and HPC are constrained by the set amount of memory built into GPUs. With data sizes growing by 3x yearly, GPU networks must keep getting larger just to fit the application in the local memory, benefiting latency and token generation. Panmnesia's proposed approach to fix this leverages the CXL protocol to expand GPU memory capacity using PCIe-connected DRAM or even SSDs. The company has overcome significant technical hurdles, including the absence of CXL logic fabric in GPUs and the limitations of existing unified virtual memory (UVM) systems.

At the heart of Panmnesia's solution is a CXL 3.1-compliant root complex with multiple root ports and a host bridge featuring a host-managed device memory (HDM) decoder. This sophisticated system effectively tricks the GPU's memory subsystem into treating PCIe-connected memory as native system memory. Extensive testing has demonstrated impressive results. Panmnesia's CXL solution, CXL-Opt, achieved two-digit nanosecond round-trip latency, significantly outperforming both UVM and earlier CXL prototypes. In GPU kernel execution tests, CXL-Opt showed execution times up to 3.22 times faster than UVM. Older CXL memory extenders recorded around 250 nanoseconds round trip latency, with CXL-Opt potentially achieving less than 80 nanoseconds. As with CXL, the problem is usually that the memory pools add up latency and performance degrades, while these CXL extenders tend to add to the cost model as well. However, the Panmnesia CXL-Opt could find a use case, and we are waiting to see if anyone adopts this in their infrastructure.Below are some benchmarks by Panmnesia, as well as the architecture of the CXL-Opt.

TrendForce reports that a recovery in demand for general servers—coupled with an increased production share of HBM by DRAM suppliers—has led suppliers to maintain their stance on hiking prices. As a result, the ASP of DRAM in the third quarter is expected to continue rising, with an anticipated increase of 8-13%. The price of conventional DRAM is expected to rise by 5-10%, showing a slight contraction compared to the increase in the second quarter.

TrendForce notes that buyers were more conservative about restocking in the second, and inventory levels on both the supplier and buyer sides did not show significant changes. Looking ahead to the third quarter, there is still room for inventory replenishment for smartphones and CSPs, and the peak season for production is soon to commence. Consequently, it is expected that smartphones and servers will drive an increase in memory shipments in the third quarter.

A new startup emerged out of stealth mode today to power the next generation of generative AI. Etched is a company that makes an application-specific integrated circuit (ASIC) to process "Transformers." The transformer is an architecture for designing deep learning models developed by Google and is now the powerhouse behind models like OpenAI's GPT-4o in ChatGPT, Anthropic Claude, Google Gemini, and Meta's Llama family. Etched wanted to create an ASIC for processing only the transformer models, making a chip called Sohu. The claim is Sohu outperforms NVIDIA's latest and greatest by an entire order of magnitude. Where a server configuration with eight NVIDIA H100 GPU clusters pushes Llama-3 70B models at 25,000 tokens per second, and the latest eight B200 "Blackwell" GPU cluster pushes 43,000 tokens/s, the eight Sohu clusters manage to output 500,000 tokens per second.

Why is this important? Not only does the ASIC outperform Hopper by 20x and Blackwell by 10x, but it also serves so many tokens per second that it enables an entirely new fleet of AI applications requiring real-time output. The Sohu architecture is so efficient that 90% of the FLOPS can be used, while traditional GPUs boast a 30-40% FLOP utilization rate. This translates into inefficiency and waste of power, which Etched hopes to solve by building an accelerator dedicated to power transformers (the "T" in GPT) at massive scales. Given that the frontier model development costs more than one billion US dollars, and hardware costs are measured in tens of billions of US Dollars, having an accelerator dedicated to powering a specific application can help advance AI faster. AI researchers often say that "scale is all you need" (resembling the legendary "attention is all you need" paper), and Etched wants to build on that.

As Intel gears for the launch of Lunar Lake and Arrow Lake processors, the company appears to be preparing a new line of high-performance processors for gaming laptops. Recent developments suggest that the company is adapting its desktop-grade Arrow Lake-S chips for use in ultra-high-performance notebooks. The buzz began when X user @InstLatX64 spotted Intel testing a peculiar motherboard labeled "Arrow Lake Client Platform/ARL-S BGA SODIMM 2DPC." This discovery hints at the possibility of Intel packing up to 24 cores into laptop processors, eight more cores compared to the 16 cores expected in standard Arrow Lake-H mobile chips. By utilizing the full potential of Arrow Lake-S silicon in a mobile form factor, Intel aims to deliver desktop-class performance to high-end notebooks in a BGA laptop CPU.

The leaked chip would likely feature eight high-performance Lion Cove P-cores and 16 energy-efficient Skymont E-cores, along with an integrated Xe2 GPU. This configuration could provide the raw power needed for demanding games and professional applications in a portable package. However, implementing such powerful hardware in laptops presents challenges. The processors are expected to have a TDP of 45 W or 55 W, with actual power consumption potentially exceeding these figures to maintain high clock speeds. Success will depend not only on Intel's chip design but also on the cooling solutions and power delivery systems developed by laptop manufacturers. As of now, specific details about clock speeds and performance metrics remain under wraps. The test chip that surfaced showed a base frequency of 3.0 GHz, notably without AVX-512 support.

Legendary overclocker Vince Lucido, aka KINGPIN, has reportedly partnered with PNY to develop next-generation GPUs for extreme overclocking. KINGPIN, known for his unparalleled expertise in pushing hardware to its limits, revealed the partnership during a recent interview with Gamers Nexus at Computex 2024. The move comes as welcome news to enthusiasts who have been eagerly awaiting KINGPIN's next venture since EVGA's departure left a noticeable gap in the high-end GPU segment. Previously, he was the leading engineer of EVGA's high-end KINGPIN designs aimed at pushing the GPU to its limits. However, since EVGA decided to leave the GPU business, KINGPIN was looking for a new company to work on the next-generation GPU designs.

This time, the company of choice for KINGPIN is now PNY. While he has been in contact with many companies like GALAX and ASUS, he claims that it would be very crowded to work there as there are "too many cooks in the kitchen" with these companies already having in-house overclockers. He has also been talking with MSI, but the company wasn't interested in making GPUs for extreme overclocking. However, PNY has been very interested in shaking up the high-end GPU market. KINGPIN claims that there is a massive hole in the high-end GPU market, and he hopes to fill it with a collaboration with PNY. Next-generation GPU designs assisted by KINGPIN will reportedly arrive for the upcoming NVIDIA GeForce RTX 50 series of GPUs when we hope to see the legacy EVGA left to continue at PNY.

GIGABYTE TECHNOLOGY Co. Ltd, a leading manufacturer of premium gaming hardware, today launched the cutting-edge AMD Radeon PRO W7000 series workstation graphics cards, including the flagship GIGABYTE Radeon PRO W7900 Dual Slot AI TOP 48G as well as the GIGABYTE Radeon PRO W7800 AI TOP 32G. Powered by AMD RDNA 3 architecture, these graphics cards offer a massive 48 GB and 32 GB of GDDR6 memory, respectively, delivering cutting-edge performance and exceptional experiences for workstation professionals, creators and AI developers.⁠⁠

GIGABYTE stands as the AMD professional graphics partner in the market, with a proven ability to design and manufacture the entire Radeon PRO series. Our dedication to quality products, unwavering business commitment, and comprehensive customer service empower us to deliver professional-grade GPU solutions, expanding user's choices in workstation and AI computing.

According to the report from Phoronix, the upcoming Linux 6.11 kernel will introduce initial display support for the highly anticipated Intel Battlemage graphics processors. Battlemage, built on Intel's Xe2 architecture, represents the company's latest effort to challenge established players in the graphics arena. This new line of GPUs is designed to succeed the current DG2/Alchemist hardware, promising enhanced performance and improved power efficiency. The Linux 6.11 kernel will provide the fundamental capability to drive displays connected to Battlemage GPUs. While this initial support is a crucial first step, it lays the groundwork for more comprehensive functionality in future updates. Linux users and developers can look forward to testing and providing feedback on these new graphics solutions.

Intel's focus on power efficiency is evident in the features accompanying Battlemage support. The kernel will introduce Panel Replay functionality, a technology aimed at reducing display power consumption. This aligns with the growing demand for energy-efficient computing solutions, particularly in mobile and laptop segments. However, the work is far from complete. Intel's Linux graphics driver team continues to refine and optimize Battlemage support, with the goal of delivering a seamless experience by the time these GPUs hit the market later this year. The introduction of Battlemage support not only expands options for Linux users but also intensifies competition in the GPU market, potentially driving innovation across the industry. With promises of up to 1.5x over the previous generation Arc GPUs, we are in for a decent upgrade this year.

PC enthusiasts who crave powerful graphics in compact systems have a new option from Zephyr. The Chinese manufacturer has launched the world's first ITX-sized GeForce RTX 4070 graphics card. Dubbed the Sakura Blizzard, this GPU packs NVIDIA's AD104-250 chip and 12 GB of GDDR6X memory into a footprint of just 172 x 123 x 42 mm. While slightly taller than standard PCIe brackets, the two-slot cooler should fit most Mini-ITX cases. The card's cute pink shroud and solitary cooling fan give it a unique aesthetic. But don't let the pink looks fool you - Zephyr claims this compact powerhouse can keep the GPU and memory up to 10°C cooler than some dual-fan RTX 4070 designs, which needs to be taken with a grain of salt. Thermal testing videos show the fan spinning at 2,400 RPM to maintain GPU temperatures around 73°C under load in a 25°C room. Meanwhile, synthetic benchmarks reportedly demonstrate no performance compromises versus full-sized RTX 4070 implementations.

Zephyr's initial production run has already sold out in China. However, a second batch is slated for mid-July availability to meet the apparent higher demand for small form factor RTX 40-series GPUs. The launch comes just weeks after NVIDIA unveiled new "SFF-ready" design guidelines at Computex 2024. As the power-hungry RTX 40 lineup hit the market, many voiced concerns over the cards' ever-growing dimensions. NVIDIA's renewed SFF PC focus signal options like the Sakura Blizzard could become more common. For space-constrained enthusiasts, having top-tier GPU muscle in a properly-cooled Mini-ITX card is a big win. Zephyr's ITX-sized RTX 4070 shows powerful things can come in small packages, and we hope more manufacturers follow this philosophy.

Stability AI, a maker of various generative AI models and the company behind text-to-image Stable Diffusion models, has released its latest Stable Diffusion 3 (SD3) Medium AI model. Running on two billion dense parameters, the SD3 Medium is the company's most advanced text-to-image model to date. It boasts features like generating highly realistic and detailed images across a wide range of styles and compositions. It demonstrates capabilities in handling intricate prompts that involve spatial reasoning, actions, and diverse artistic directions. The model's innovative architecture, including the 16-channel variational autoencoder (VAE), allows it to overcome common challenges faced by other models, such as accurately rendering realistic human faces and hands.

Additionally, it achieves exceptional text quality, with precise letter formation, kerning, and spacing, thanks to the Diffusion Transformer architecture. Notably, the model is resource-efficient, capable of running smoothly on consumer-grade GPUs without compromising performance due to its low VRAM footprint. Furthermore, it exhibits impressive fine-tuning abilities, allowing it to absorb and replicate nuanced details from small datasets, making it highly customizable for specific use cases that users may have. Being an open-weight model, it is available for download on HuggingFace, and it has libraries optimized for both NVIDIA's TensorRT (all modern NVIDIA GPUs) and AMD Radeon/Instinct GPUs.

The race is on for memory manufacturers to bring the next generation GDDR7 graphics memory into mass production. While rivals Samsung and Micron are aiming to have GDDR7 chips available in Q4 of 2024, South Korean semiconductor giant SK Hynix revealed at Computex 2024 that it won't kick off mass production until the first quarter of 2025. GDDR7 is the upcoming JEDEC standard for high-performance graphics memory, succeeding the current GDDR6 and GDDR6X specifications. The new tech promises significantly increased bandwidth and capacities to feed the appetites of next-wave GPUs and AI accelerators. At its Computex booth, SK Hynix showed off engineering samples of its forthcoming GDDR7 chips, with plans for both 16 Gb and 24 Gb densities.

The company is targeting blazing-fast 40 Gbps data transfer rates with its GDDR7 offerings, outpacing the 32 Gbps rates its competitors are starting with on 16 Gb parts. If realized, higher speeds could give SK Hynix an edge, at least initially. While trailing a quarter or two behind Micron and Samsung isn't ideal, SK Hynix claims having working samples now validates its design and allows partners to begin testing and qualification. Mass production timing for standardized memories also doesn't necessarily indicate a company is "late" - it simply means another vendor secured an earlier production window with a specific customer. The GDDR7 transition is critical for SK Hynix and others, given the insatiable demand for high-bandwidth memory to power AI, graphics, and other data-intensive workloads. Hitting its stated Q1 2025 mass production target could ensure SK Hynix doesn't fall too far behind in the high-stakes GDDR7 race, with faster and higher-density chips to potentially follow shortly after volume ramp.

At Computex 2024, Colorful has followed the success of their popular COLORFIRE MEOW Series gaming desktops. The company has unleashed two fierce new gaming laptops - the COLORFIRE MEOW R15 and R16. These portable laptops are adorned with designs inspired by Bobi, COLORFIRE's beloved office cat. From paw print patterns to a sleek orange tabby color scheme, the laptops are littered with feline flair. Under the cat-inspired hood, the MEOW laptops pack a serious punch. They are equipped with the latest AMD Ryzen 8000 Series processors tailored for gaming, content creation, and entertainment. AMD Ryzen AI technology provides a boost to artificial intelligence capabilities. Both run on the AMD Ryzen 7 8845HS processor with eight cores, 16 threads, and a 5.1 GHz maximum boost clock. The Ryzen AI engine delivers up to 39 TOPS for accelerated AI performance. For graphics, the laptops feature up to an NVIDIA GeForce RTX 4070 GPU with 140 W of power.

The R15 has a 15.6-inch 2560x1440 IPS panel with 100% sRGB color and 165 Hz refresh rate. The R16 goes even bigger with a 16-inch 2560x1600 IPS display, 100% sRGB, and 240 Hz refresh. AMD FreeSync is there as well to ensure a tear-free experience. Colorful has outfitted the laptops with dual turbo fans, a 6-heatpipe solution, and performance control software to tame the heat from the powerful components. When it's time to pounce on the latest games, these MEOW laptops have the fierce cooling to keep their claws out.

At Computex 2024, Colorful has introduced a new PC series of premium componentry for games, called iGame Ultra Series. Carrying a premium white aesthetics, the Ultra Series serves as clean and elegant component for PC builds. At the heart of the Ultra Series is the GeForce RTX 4080 SUPER graphics card. The card has a triple-fan cooling solution with vapor chamber technology that keeps temperatures under control. Powering this GPU is the iGame Z790D5 Ultra motherboard. Supporting the latest Intel 14th Gen CPUs and up to 192 GB of DDR5-7800 RAM. An array of ultra-fast PCIe 5.0 and 4.0 M.2 slots allow multiple high-speed SSDs for ridiculous storage performance. Keeping all of these elite components running is the 850 W iGame P850G Ultra W 80 Plus Gold power supply. Its modular design with dedicated PCIe 5.0 connectors ensures clean cable management and stable power delivery. For cooling the CPU, there is the iGame LQ360 Ultra W all-in-one liquid cooler, which sports a massive 360 mm radiator and a trio of high-static pressure fans. All of these premium components are housed in the iGame C23A Ultra W full-tower chassis. With a spacious dual-compartment layout, front USB-C port, and preinstalled vertical GPU bracket, it offers both incredible expandability and sleek aesthetics.

Micron Technology, Inc., today announced the sampling of its next-generation GDDR7 graphics memory with the industry's highest bit density. Leveraging Micron's 1β (1-beta) DRAM technology and innovative architecture, Micron GDDR7 delivers 32 Gb/s high-performance memory in a power-optimized design. With over 1.5 TB/s of system bandwidth, which is up to 60% higher bandwidth than GDDR6, and four independent channels to optimize workloads, Micron GDDR7 memory enables faster response times, smoother gameplay and reduced processing times.

GDDR7 also provides a greater than 50% power-efficiency improvement compared to GDDR6 to better thermals and lengthen battery life, while the new sleep mode reduces standby power by up to 70%. Advanced reliability, availability and serviceability (RAS) features on Micron GDDR7 enhance device dependability and data integrity without compromising performance, broadening the spectrum of applications for Micron GDDR7 to AI, gaming and high-performance computing workloads.

In addition to new PC cases, Phanteks brought several new cooling products to the Computex 2024 show, including a whole new NEXLINQ cooling system, Glacier EZ-Fit series of liquid cooling components and kits, as well as a rather interesting fan prototype that will hopefully find its way to the market.

The big part of the Phanteks showcase was the NEXLINQ, a set of cooling components, or an ecosystem as Phanteks likes to describe it, which includes the M25 G2 fans, Glacier One M25 G2 AIO, and a NEXLINQ Hub. The NEXLINQ and components will feature support for the whole new LINQ-6 connector that will be able to provide both power and D-RGB lighting to the fans.

At Computex 2024, ASUS unveiled major updates to its popular laptop lineups, designed for the "Copilot+" era of AI computing. The first is the Zenbook S16 is a premium 16-inch laptop series powered by AMD's latest Ryzen AI 9 HX 370 processors with dedicated AI acceleration. Remarkably, ASUS has managed to pack this high-performance silicon into an ultra-portable 1.1 cm thin chassis weighing just 1.5 kg. The Zenbook S16 integrates AMD's new NPU capable of a 50 TOPS of AI compute for accelerating AI/ML workloads. The centerpiece is the laptop's stunning 16-inch 3K OLED display made with ASUS Lumina technology. It offers 100% vibrant DCI-P3 color gamut coverage, a blazing-fast 120 Hz refresh rate with 0.2 ms response time, and up to 600 nits brightness. ASUS paired this premium visual experience with a six-speaker audio system for an immersive multimedia experience.

GIGABYTE debuted its LCD Edge View display for showcasing real-time system information. Designed as a small and compact LCD display, it acts as a desktop-standing box capable of displaying critical system information like CPU and GPU temperatures, fan speeds of GPUs and CPU coolers, PC fans, as well as everything else happening inside of a PC. Interestingly, it isn't just a fixed-function display. It can show custom images/videos and display custom text, serving as a small display companion to the main monitor for users to look at occasionally or just monitor system health. The design of the LCD Edge View is to be a plug-and-play white box that could easily be positioned on any desktop, blending seamlessly with the user's setup.

It requires external power connection, and can be driven by any GPU/video output. It also features a USB port.

During Computex 2024, GIGABYTE unveiled its new "AI TOP" series designed to empower users to develop and run AI applications locally on their systems. The AI TOP lineup includes AI-optimized motherboards, graphics cards, and complete system solutions. The flagship motherboard is called TRX50 AI TOP, which boasts support for AMD's Ryzen Threadripper 7000 PRO and a regular series of CPUs. The flagship TRX50 AI TOP motherboard features a special VRM design with beefy heatsinks, four PCIe 5.0 x16 slots for quad-GPU setup, eight-channel DDR5 memory with room for eight DIMMs, and a few M.2 PCIe Gen 5 slots. Next in the AI TOP line is the GIGABYTE GeForce RTX 4070 Ti SUPER AI TOP edition. Formed as a blower-style cooler, the RTX 4070 Ti SUPER AI TOP is envisioned for tight spaces and parallel installation with other GPUs to accelerate local AI development and inference.

Supposedly, there will be more AI TOP motherboards and GPUs than what is showcased. GIGABYTE may have an AI TOP makeover for Intel's upcoming Z890 AORUS Xtreme, and there could be more GPUs in the future with the blower-style AI TOP design. The goal of AI TOP series is optimization for AI workloads, which require lots of GPUs and lots of memory, like the TRX50 AI TOP motherboard shows.

According to a report by KED Global, Korean chipmaking giant Samsung is ramping up its efforts to compete with global giants like TSMC and Intel. The latest partnership on the horizon is AMD's collaboration with Samsung. AMD is planning to utilize Samsung's cutting-edge 3 nm technology for its future chips. More specifically, AMD wants to utilize Samsung's gate-all-around FETs (GAAFETs). During ITF World 2024, AMD CEO Lisa Su noted that the company intends to use 3 nm GAA transistors for its future products. The only company offering GAAFETs on a 3 nm process is Samsung. Hence, this report from KED gains more credibility.

While we don't have any official information, AMD's utilization of a second foundry as a manufacturing partner would be a first for the company in years. This strategic move signifies a shift towards dual-sourcing, aiming to diversify its supply chain and reduce dependency on a single manufacturer, previously TSMC. We still don't know what specific AMD products will use GAAFETs. AMD could use them for CPUs, GPUs, DPUs, FPGAs, and even data center accelerators like Instinct MI series.

AMD, Broadcom, Cisco, Google, Hewlett Packard Enterprise (HPE), Intel, Meta and Microsoft today announced they have aligned to develop a new industry standard dedicated to advancing high-speed and low latency communication for scale-up AI systems linking in Data Centers.

Called the Ultra Accelerator Link (UALink), this initial group will define and establish an open industry standard that will enable AI accelerators to communicate more effectively. By creating an interconnect based upon open standards, UALink will enable system OEMs, IT professionals and system integrators to create a pathway for easier integration, greater flexibility and scalability of their AI-connected data centers.

In addition to its two new CPU cores, Arm has announced three new GPU cores, namely the Immortalis-G925, Mali-G725 and Mali-G625. Starting from the top, the Immortalis-G925 is said to bring up to 37 percent better performance at 30 percent lower power usage compared to last year's Immortalis-G720 GPU core, whilst having two additional GPU cores in the test scenario. It's also said to bring up to 52 percent better ray tracing performance and up to 36 percent improved inference in AI/ML workloads. It's also been given a big overhaul when it comes to ray tracing—due to it being aimed towards gaming phones—and Arm claims that it can either offer up to 52 percent increased performance by reducing the accuracy in scenes with intricate objects, or 27 percent more performance with maintained accuracy.

The Immortalis-G925 supports 50 percent more shader cores and it supports configurations of up to 24 cores, compared to 16 cores for the Immortalis-G720. The Mali-G725 will be available with between six and nine cores, whereas the Mali-G625 will sport between one and five cores. The Mali-G625 is intended for smartwatches and entry-level mobile devices where a more complex GPU might not be suitable due to power draw. The Mali-G725 on the other hand is targeting upper mid-range devices and the Immortalis-G925 is aimed towards flagship devices or gaming phones as mentioned above. In related news, Arm said it's working with Epic Games to get its Unreal Engine 5 desktop renderer up and running on Android, which could lead to more complex games on mobile devices.

According to Jon Peddie Research latest report, the global GPU and CPU markets ended Q1 2024 on growth. GPU shipments reached 70 million units, while PC CPU shipments increased by 33% year-over-year, the second consecutive yearly increase in over two decades. Over the next five years, discrete GPUs are projected to achieve 22% penetration in the PC market as the overall GPU installed approaches 3 billion units by 2026. Among major GPU vendors, AMD's market share dipped 0.7% from last quarter, while Intel gained 0.3% and Nvidia rose 0.4%. However, overall GPU shipments declined 9.9% quarter-over-quarter. The total GPU attach rate for PCs was 113%, slightly down from the prior quarter. Desktop graphics add-in board shipments also decreased 14.8%.

While Q1 is typically flat or down versus Q4, Jon Peddie Research's president suggests this quarter's performance could signal a return to normal seasonality. With Microsoft, AMD and Intel promoting AI PCs, and forecasts pointing to growth in Q2, there are optimistic signs - although semiconductor suppliers are guiding 7.9% down on average for next quarter.

ZOTAC Technology, a global manufacturer focused on innovative and high-performance hardware solutions, will return to COMPUTEX 2024 to showcase its biggest push yet into brand-new product categories. At this year's exhibition, ZOTAC will unveil its first attempt at creating a unique Handheld Gaming PC with advanced controls and features, allowing gamers to enjoy their favorite games on the go like never before with maximum competitive advantage.

Also in ZOTAC's extensive lineup is a full-fledged selection of AI-focused computational hardware, including a new workstation-grade External GPU Box series for hassle-free GPU compute and AI acceleration, ZBOX mini PCs powered by Intel Core Ultra CPUs equipped with integrated neural processing units (NPU), as well as other enterprise-grade solutions, such as GPU Servers and Arm-based NVIDIA Jetson systems, offering users a broad selection of AI accelerators in applications big and small.

USB4 hasn't exactly taken off as yet, largely due to a shortage of devices in the market and with a limited number of device controllers, the competition in the market is also slow. With Thunderbolt compatibility, there are plenty of USB4 compatible hosts out there though, even though you only get 32 Gbps rather than 40 Gbps of data throughput when a USB4 device is connected to Thunderbolt 3 or 4, although Thunderbolt 5 will allow for the full 40 Gbps. A company called ADT-Link has launched a product it calls the UT3G which is pretty much a retail ready bridge solution that allows techies and tinkerers to try out whatever PCIe devices they have over USB4. The adapter has a full-length PCIe x16 slot, but it's limited to a four lane PCIe interface which is compatible with PCIe 4.0, albeit not at the full 64 Gbps that such an interface can deliver.

In addition to the USB4 Type-C port on the PCB there's a standard 24-pin ATX power supply connector for power, which means that this isn't really a portable solution. However, the company has tested the UT3G with various graphics cards from both AMD and NVIDIA to make sure you can use it as an external graphics card dock. The UT3G has also been tested to be compatible with Windows, Linux and macOS, although macOS is limited to AMD GPUs. The board is built around the ASMedia ASM2464PD USB4 to PCIe 4.0 bridge and this shouldn't come as a surprise, as so far this is the only such device controller. ADT-Link doesn't appear to be selling the UT3G to consumers, but it can be picked up online from DFRobot starting at US$129 for a single unit.

The US Trade Representative (USTR) under Biden administration is preparing to reinstate tariffs on certain technology products imported from China, including GPUs and motherboards. The 25% duties, initially imposed by the Trump administration in 2019 but later suspended, are being revived as part of broader efforts to address concerns over China's economic and trade practices. The tariffs are intended to protect American companies from what the administration describes as unfair Chinese trade actions like intellectual property theft and forced technology transfers that undermine U.S. competitiveness. While no specific effective date was provided, the reinstated tariffs are expected to impact major Chinese computing component suppliers significantly. The revival of the Trump-era tariffs marks a reversal from the previous administration's move to temporarily suspend the duties in 2020 as a goodwill gesture during broader trade negotiations with Beijing.

However, those talks ultimately stalled amid the COVID-19 pandemic and rising bilateral tensions over economic and national security issues. Industry groups have expressed concerns that reviving the tariffs could disrupt tech supply chains, increase costs for U.S. companies and consumers, and potentially invite further Chinese retaliation. The tariffs would apply to GPUs, motherboards and other computing components assembled in China regardless of whether the raw components themselves originated from the country. With tensions already elevated over issues like Taiwan and advanced semiconductor production, the tariff announcement could set the stage for further economic friction between the world's two largest economies absent a negotiated resolution on tech trade.

AMD has quietly added some interesting codenames to its ROCm hardware support list. The biggest surprise is the appearance of "RDNA 4" and "Navi 44" codenames, hinting at a potential successor to the current RDNA 3 GPU architecture powering AMD's Radeon RX 7000 series graphics cards. The upcoming Radeon RX 8000 series could see Navi 44 SKU with a codename "gfx1200". While details are scarce, the inclusion of RDNA 4 and Navi 44 in the ROCm list suggests AMD is working on a new GPU microarchitecture that could bring significant performance and efficiency gains. While RDNA 4 may be destined for future Radeon gaming GPUs, in the data center GPU compute market, AMD is preparing a CDNA 4 based successors to the MI300 series. However, it appears that we haven't seen all the MI300 variants first. Equally intriguing is the "MI300X1" codename, which appears to reference an upcoming AI-focused accelerator from AMD.

While we wait for more information, we can't decipher whether the Navi 44 GPU SKU is for the high-end or low-end segment. If previous generations are for reference, then the Navi 44 SKU would target the low end of the GPU performance spectrum. The previous generation RDNA 3 had Navi 33 as an entry-level model, whereas the RDNA 2 had a Navi 24 SKU for entry-level GPUs. We have reported on RDNA 4 merely being a "bug correction" generation to fix the perf/Watt curve and offer better efficiency overall. What happens finally, we have to wait and see. AMD could announce more details in its upcoming Computex keynote.