Nanya Technology's Board of Directors today has approved a strategic partnership with PieceMakers Technology, Inc. ("PieceMakers") to jointly develop customized ultra-high-bandwidth memory solutions. As part of the collaboration, Nanya Technology will subscribe to a cash capital increase of up to NT$ 660 million, purchasing up to 22 million common shares at NT$ 30 per share in PieceMakers. Upon completion of the capital increase, Nanya Technology is expected to acquire up to approximately 38% stakes of PieceMakers.

To meet the growing demand for high-performance memory driven by AI and edge computing, this collaboration will combine Nanya Technology's 10 nm-class DRAM innovation with PieceMakers' expertise in customized DRAM design to develop high-value, high-performance, and low-power customized ultra-high-bandwidth memory solutions, unlocking new opportunities in AI and high-performance computing markets.

A high-stakes trial between technology giants Arm and Qualcomm has revealed deeper tensions in the semiconductor industry, as Arm seeks the destruction of chip designs from Qualcomm's $1.4 billion Nuvia acquisition. The case, being heard in Delaware federal court, centers on a licensing dispute that could impact the future of AI-powered Windows PCs. Arm CEO Rene Haas took the stand Monday, adding allegations that Qualcomm violated licensing agreements following its 2021 acquisition of chip startup Nuvia. The issue is whether Qualcomm should pay Nuvia's higher royalty rates for using Arm's intellectual property rather than its own lower rates. Internal documents revealed Nuvia's rates were "many multiples" higher than Qualcomm's, with the acquisition potentially reducing Arm's revenue by $50 million.

During cross-examination, Qualcomm's legal team challenged Arm's motives, suggesting the dispute is part of a broader strategy to confront a customer increasingly viewed as a competitor. When presented with documents outlining potential plans for Arm to design its own chips, Haas downplayed these ambitions, emphasizing that Arm has never entered chip manufacturing. Allegedly, Arm sent letters to Qualcomm's customers, including Samsung, warning about possible disruption if Nuvia's IP design before acquisition in 2021 must be destroyed. Haas defended these communications, citing frequent inquiries from industry partners.

Marvell Technology, Inc. (NASDAQ: MRVL), a leader in data infrastructure semiconductor solutions, today announced the development of FBNIC, a custom 5 nm network interface controller (NIC) ASIC in collaboration with Meta to meet the company's infrastructure and use case requirements. The FBNIC board design will also be contributed by Marvell to the Open Compute Project (OCP) community. FBNIC combines a customized network controller designed by Marvell and Meta, a co-designed board, and Meta's ASIC, firmware and software. This custom design delivers innovative capabilities, optimizes performance, increases efficiencies, and reduces the average time needed to resolve potential network and server issues.

"The future of large-scale, data center computing will increasingly revolve around optimizing semiconductors and other components for specific applications and cloud infrastructure architectures," said Raghib Hussain, President of Products and Technologies at Marvell. "It's been exciting to partner with Meta on developing their custom FBNIC on our industry-leading 5 nm accelerated infrastructure silicon platform. We look forward to the OCP community leveraging the board design for future innovations."

Huawei has reportedly started shipping its Ascend 910C accelerator—the company's domestic alternative to NVIDIA's H100 accelerator for AI training and inference. As the report from China South Morning Post notes, Huawei is shipping samples of its accelerator to large NVIDIA customers. This includes companies like Alibaba, Baidu, and Tencent, which have ordered massive amounts of NVIDIA accelerators. However, Huawei is on track to deliver 70,000 chips, potentially worth $2 billion. With NVIDIA working on a B20 accelerator SKU that complies with US government export regulations, the Huawei Ascend 910C accelerator could potentially outperform NVIDIA's B20 processor, per some analyst expectations.

If the Ascend 910C receives positive results from Chinese tech giants, it could be the start of Huawei's expansion into data center accelerators, once hindered by the company's ability to manufacture advanced chips. Now, with foundries like SMIC printing 7 nm designs and possibly 5 nm coming soon, Huawei will leverage this technology to satisfy the domestic demand for more AI processing power. Competing on a global scale, though, remains a challenge. Companies like NVIDIA, AMD, and Intel have access to advanced nodes, which gives their AI accelerators more efficiency and performance.

Corsair, a world leader in high-performance gear for gamers and content creators, today announced a fantastical new lineup for the ever-growing Corsair Custom Lab: the Artist Series Collection. Three stunning new peripheral sets are available exclusively on the Corsair webstore for a limited time, specially crafted by talented artists. Players can show off their passion for gaming - and their favorite creators - as they deck out their desktops in an explosion of imagination, creativity, and self-expression.

INTRODUCING THE ARTIST SERIES: WHERE ART MEETS GAMING
Corsair Custom Lab is an all-encompassing custom builder that empowers users to fashion their setups with unique designs that fit their personal style. Its gallery of gaming gear includes the K65 PLUS WIRELESS Keyboard, the M75 WIRELESS RGB Mouse, and the MM300 Mouse pad line. Corsair has collaborated with illustrative powerhouses Nacho Customz, Elina Clevergull, and Yunzhen Ho to create a limited-edition Artist Series drop that inspires all gamers to creatively express themselves. In the Corsair Custom Lab, every gamer can sign their deskscape with a distinctive style; and now, Corsair's three featured artists present even more showstopping designs to choose from.

Intel Corp. and Amazon Web Services. Inc., an Amazon.com company, today announced a co-investment in custom chip designs under a multi-year, multi-billion-dollar framework covering product and wafers from Intel. This is a significant expansion of the two companies' longstanding strategic collaboration to help customers power virtually any workload and accelerate the performance of artificial intelligence (AI) applications.

As part of the expanded collaboration, Intel will produce an AI fabric chip for AWS on Intel 18A, the company's most advanced process node. Intel will also produce a custom Xeon 6 chip on Intel 3, building on the existing partnership under which Intel produces Xeon Scalable processors for AWS.

This week at Gamescom 2024, HP Inc. unveiled its latest gaming gear from OMEN and HyperX. The new offerings take customization to the next level, allowing gamers to fine-tune their aesthetics and ensure their desktops are future-ready with OMEN components.

"HP is committed to pushing the boundaries of technology to deliver unparalleled, personalized experiences for every gamer," said Josephine Tan, Senior Vice President & Division President of Personal Systems Gaming Solutions. "With our latest innovations in OMEN and HyperX, being showcased at Gamescom 2024, we are excited to elevate the gaming experience. Our goal is to provide users with the tools they need to perform at their best and ensure their systems are upgradable, allowing them to fully immerse themselves in their gaming worlds for years to come."

Intel has officially started shipping the "0x129" microcode update for its 13th and 14th generation "Raptor Lake" and "Raptor Lake Refresh" processors. This critical update is currently being pushed to all OEM/ODM partners to address the stability issues that Intel's processors have been facing. According to Intel, this microcode update fixes "incorrect voltage requests to the processor that are causing elevated operating voltage." Intel's analysis shows that the root cause of stability problems is caused by too high voltage during operation of the processor. These increases to voltage cause degradation that increases the minimum voltage required for stable operation. Intel calls this "Vmin"—it's a theoretical construct, not an actual voltage, think "speed for an airplane required to fly". The latest 0x129 microcode patch will limit the processor's voltage to no higher than 1.55 V, which should avoid further degradation. Overclocking is still supported, enthusiasts will have to disable the eTVB setting in their BIOS to push the processor beyond the 1.55 V threshold. The company's internal testing shows that the new default settings with limited voltages with standard run-to-run variations show minimal performance impact, with only a single game (Hitman 3: Dartmoor) showing degradation. For a full statement from Intel, see the quote below.

A fleet of recent reports have highlighted stability issues affecting Intel's 13th and 14th-generation desktop processors, raising concerns among consumers and industry professionals. The problem, which has gained significant attention over the past few months, is related to the processors' physical degradation over time. Custom PC builder Puget Systems has shared insights from its experience with these processors, revealing a nuanced perspective on the issue. While it has observed an increase in CPU failures, particularly with the 14th-generation chips, its failure rates remain notably lower than those reported by some game development studios and cloud gaming providers, who have cited failure rates as high as 50%. An interesting observation is that Puget Systems recorded a higher failure rate with AMD Ryzen 5000 and Ryzen 7000 series than Intel's 13/14th generation, with most failures happening at Puget's shop rather than the "field" in customers' hands.

Puget Systems attributes their more modest failure rates of Intel processors to their conservative approach to power management settings. By adhering strictly to Intel's specifications and developing their own power settings that don't hurt performance, they've managed to mitigate some of the stability issues plaguing other users. Intel has acknowledged the problem and announced plans to release a microcode patch by mid-August, with extended warranty program. This update is expected to prevent further degradation but may not reverse existing damage. Despite the elevated failure rates, Puget Systems' data shows that the issue, while concerning, still needs to be at critical levels for their operations. The company reports that failure rates for 13th and 14th gen Intel processors, while higher than ideal, are still lower than those they experienced with Intel's 11th gen chips and some AMD Ryzen processors. In response to the situation, Puget Systems is taking several steps, including maintaining its current power management practices, promptly validating Intel's upcoming microcode update, and extending warranties for affected customers. Below, you can see failure rates by month, by Intel's Core generation, as well as by "shop" vs "field" testing.

MCHOSE, a highly recommended peripheral brand known among keyboard enthusiasts in China for its premium, high-performance products, unique style, and competitive pricing, is making its debut in the international market. Their first full CNC aluminium custom mechanical keyboard, the GX87, will debut on Indiegogo starting at $59 beginning August 6th. As the world's first mass-produced full aluminium quick disassembly TKL (Tenkeyless) custom mechanical keyboard, the GX87 is poised to drive a revolutionary shift in the pricing of the high-end mechanical keyboard market.

The MCHOSE GX87 stands out as the first-ever CNC aluminium prebuilt quick disassembly TKL mechanical keyboard. As a pioneering brand in the mass production of full aluminium custom mechanical keyboards, MCHOSE has partnered with top-tier supply chain resources in China to bring forth practical and cost-effective keyboard solutions. The introductory pricing of the GX87 sets a remarkably competitive stance against other brands, truly reflecting a sincere approach to enthusiasts worldwide.

According to The Information, OpenAI is reportedly in talks with Broadcom about developing a custom AI accelerator to power OpenAI's growing demand for high-performance solutions. Broadcom is a fabless chip designer known for a wide range of silicon solutions spanning from networking, PCIe, SSD controllers, and PHYs all the way up to custom ASICs. The latter part is what OpenAI wants to focus on, but all the aforementioned IP developed by Broadcom is of use in a data center. Suppose OpenAI decides to use Broadcom's solutions. In that case, the fabless silicon designer offers a complete vertical stack of products for inter-system communication using various protocols such as PCIe, system-to-system communication using Ethernet networking with Broadcom Tomahawk 6 and future revisions, alongside storage solutions and many other complimentary elements of a data center.

As a company skilled in making various IPs, it also makes ASIC solutions for other companies and has assisted Google in the making of its Tensor Processing Unit (TPU), which is now in its sixth generation. Google TPUs are massively successful as Google deploys millions of them and provides AI solutions to billions of users across the globe. Now, OpenAI wants to be part of the AI chip game, and Broadcom could come to the rescue with its already-established AI success and various other data center componentry to help make a custom AI accelerator to power OpenAI's infrastructure needed for the next generation of AI models. With each new AI model released by OpenAI, compute demand spikes by several orders of magnitude, and having an AI accelerator that exactly matches their need will help the company move faster and run even bigger AI models.

Apple is reportedly working on creating in-house chips designed explicitly for its data centers. This news comes from a recent report by the Wall Street Journal, which highlights the company's efforts to enhance its data processing capabilities and reduce dependency on third parties to supply the infrastructure. In the internal project called Apple Chips in Data Center (ACDC), which started in 2018, Apple wanted to design data center processors to handle the massive user base and increase the company's service offerings. The most recent advancement in AI means that Apple will probably serve an LLM processed in Apple's data center. The chip will most likely focus on inference of AI models rather than training.

The AI chips are expected to play a crucial role in improving the efficiency and speed of Apple's data centers, which handle vast amounts of data generated by the company's various services and products. By developing these custom chips, Apple aims to optimize its data processing and storage capabilities, ultimately leading to better user experiences across its ecosystem. The move by Apple to develop AI-enhanced chips for data centers is seen as a strategic step in the company's efforts to stay ahead in the competitive tech landscape. Almost all major tech companies, famously called the big seven, have products that use AI in silicon and in software processing. However, Apple is the one that seemingly lacked that. Now, the company is integrating AI across the entire vertical, from the upcoming iPhone integration to M4 chips for Mac devices and ACDC chips for data centers.

CHERRY XTRFY, the specialist for high-quality gaming peripherals, announces the availability of the new K5V2 keyboard - the first gaming keyboard with CHERRY's new mechanical MX2A switches.

CHERRY XTRFY redefines the 65% keyboard class with the K5V2 Compact. The renowned company combines well-known CHERRY quality with a modern design and durable mechanical MX2A Red switches. The keyboard's advanced per-key customizable RGB illumination creates impressive effects with innovative strip LED technology, which can be further enhanced with the USB 3 mode.

Blink, an Amazon company, today announced the next-generation Blink Mini 2. The new Blink camera packs a punch in a compact, weather-resistant design that can now be used indoors or outdoors with the purchase of the new Blink Weather Resistant Power Adapter (sold as part of a bundle or separately). Blink Mini 2 offers enhanced image quality with improved low light performance, a wider field of view, and a built-in LED spotlight for night view in color. Powered by the company's custom-built chip, Blink Mini 2 utilizes on-device computer vision (CV) to support smart notifications, including person detection, which is available with a Blink Subscription Plan (sold separately).

"It is clear customers love Blink—in fact, the Blink business has grown 5x over the last four years," said Liz Hamren, chief executive officer at Blink. "We are building on this momentum with the addition of Mini 2 to Blink's affordable and easy-to-use suite of devices. Mini 2 was rebuilt from the inside out, keeping everything customers expect from Blink while adding even more utility through features like person detection, all at an incredible price point."

Discover the MOZA TSW Truck Wheel, engineered for an unparalleled truck simulation experience. With its classic four-spoke design and standard 400 mm diameter, this wheel brings the authenticity of a truck's control into your hands. The rim, wrapped in hand-stitched microfiber leather, offers a blend of comfort and grip, while the aerospace-grade aluminium frame guarantees long-term durability. Equipped with customizable RGB buttons and LED lights, it allows for extensive personalization.

Compatibility with all MOZA bases, as well as third-party systems through a universal adapter, ensures a seamless fit into any setup. Enhanced by its high-strength steel construction and versatile adjustment options, the MOZA TSW Truck Wheel is the definitive choice for simulation enthusiasts seeking depth, functionality, and an immersive driving experience. Expected to start shipping on March 9th.

An Ebay Store—customgpu_official—is selling memory modified GeForce RTX 2080 Ti graphics cards. The outfit (located in Palo Alto, California) has a large inventory of MSI GeForce RTX 2080 Ti AERO cards—judging from their listing's photo gallery. Workers in China are reportedly upgrading these (possibly refurbished) units with extra lashings of GDDR6 VRAM—going from the original 11 GB specification up to 22 GB. We have observed smaller scale GeForce RTX 2080 Ti modification projects and a very ambitious user-modified example in the past, but customgpu's latest endeavor targets a growth industry—the item description states: "Why do you need a 22 GB 2080 Ti? Large VRAM is essential to cool AIGC apps such as stable diffusion fine tuning, LLAMA, LLM." At the time of writing three cards are available to purchase, and interested customers have already acquired four memory modded units.

They advertise their upgraded "Turbo Edition" card as a great "budget alternative" to more modern GeForce RTX 3090 and 4090 models—"more information and videos" can be accessed via 2080ti22g.com. The MSI GeForce RTX 2080 Ti AERO 11 GB model is not documented within TPU's GPU database, but its dual-slot custom cooling solution is also sported by the MSI RTX 2080 SUPER AERO 8 GB graphics card. The AERO's blower fan system creates a "mini-wind tunnel, pulling fresh air from inside the case and blowing it out the IO panel, and out of the system." The seller's asking price is $499 per unit—perhaps a little bit steep for used cards (potentially involved in mining activities), but customgpu_official seems to be well versed in repairs. Other Ebay listings show non-upgraded MSI GeForce RTX 2080 Ti AERO cards selling in the region of $300 to $400. Custom GPU Upgrade and Repair's hype video proposes that their modified card offers great value, given that it sells for a third of the cost of a GeForce RTX 3090—their Ebay item description contradicts this claim: "only half price compared with GeForce RTX 3090 with almost the same GPU memory."

GALAX is known to go overboard with its top flight graphic card models—the introduction of GeForce RTX 40 SUPER models has further lengthened the manufacturer's naming conventions. A Hall of Fame (HOF) OC LAB Master Edition card based on NVIDIA's freshly launched GeForce RTX 4070 SUPER GPU has been introduced via a Galaxy BBS blog post. As befits such a fancily named card, GALAX has rolled out the red carpet with a very special cooling solution that is designed to temper a substantial (211 MHz) overclock over Team Green's reference settings. Its 2685 MHz spec sits atop the customized GeForce RTX 4070 SUPER GPU pile—VideoCardz notes that another high-end option—GIGABYTE's AORUS RTX 4070 SUPER MASTER model—trails by a 30 MHz margin.

The GALAX GeForce RTX 4070 SUPER Hall of Fame OC LAB Master Edition graphics card sports an almost all-white design, even its PCB is outfitted in a pale hue. This Ada Lovelace AD104-350-A1-driven flagship is specced with a 250 W TGP as standard, user-adjustable up to 320 W. A three-pin to sixteen-pin adapter is supplied by default; ensuring that more than enough juice is supplied. GALAX states that the card features a 12+3 phase design, coupled with a power section is controlled by an XDPE10281 PWM. The Chinese manufacturer hints that more white Master Edition SUPER models are incoming (see below). We hope to see further announcements, and full product pages uploaded to their web site(s), but GALAX is unlikely to sell these top flight cards outside of their native market.

According to new data from Chinese Customs, China's imports of integrated circuits suffered their steepest annual drop on record in 2023, falling 15.4% to $349.4 billion. The decline marks the second straight year of falling chip imports and can be attributed to economic uncertainty and US export controls on advanced semiconductors. Shipment volumes of imported chips also saw a substantial 10.8% year-over-year decrease as demand within China stagnated. The country's important tech manufacturing sector has struggled under strict zero-Covid policies and a lackluster recovery post-pandemic. Flagship manufacturing companies like TSMC recorded modest declines in 2023 sales, though TSMC still forecasts overall growth this year.

Sentiment plunged further when the Biden administration heightened restrictions on China's access to cutting-edge AI-capable chips from NVIDIA and other top American suppliers. The escalating US export controls have choked off China's pipeline to advanced semiconductors needed for AI and supercomputing applications. However, early positive signs for global semiconductor demand have emerged, with worldwide chip sales rising for the first time in over a year this past November. The increase was driven by growing demand for AI and other emerging technologies that rely on advanced computing chips. While the US seeks to limit China's progress in this key strategic area, an inflection point for the battered global chip sector may be nearing.

NVIDIA revealed basic price guides for GeForce RTX-40 SUPER graphics cards at the recently concluded CES 2024 trade show, but their board partners largely stayed coy about figures for customized options (review embargoes will be lifted soon). ZOTAC broke the mold later on in the week, with press material updated to reflect that non-overclocked models will adhere to Team Green's basic MSRP. However, premium charges for overclocked SUPER Twin Edge OC, AMP HOLO, Trinity OC and AMP Extreme AIRO cards remain a mystery. VideoCardz decided to conduct some weekend detective work, and fiddled around on Newegg and Best Buy online stores—although the focus shifted to other brands/manufacturers.

Workaround methods were implemented in order to prematurely extract card prices, before NVIDIA's staggered schedule of reveals for customized versions of the GeForce RTX 4070 SUPER, RTX 4070 Ti SUPER and RTX 4080 SUPER GPUs (throughout January). The leaked results show that GIGABYTE and PNY have custom overclocked GeForce RTX 4070 SUPER models targeting base MSRP at $599, while MSI has several options exceeding that base level—ranging from $10 to $50 premiums. GIGABYTE's GAMING OC card also tops the table at $649. Jumping up to the GeForce RTX 4070 Ti SUPER tier, we see a GIGABYTE Gaming OC model sitting at $849.99 and an MSI VENTUS 3X OC going for $899.99. The sole custom GeForce RTX 4080 SUPER within the VideoCardz article appears to be an MSI VENTUS 3X OC; we are witnessing a $100 extra tacked on for this design.

GIGABYTE revealed its GeForce RTX 40 SUPER graphics card range earlier this week, and TPU staff in attendance at CES 2024 were most pleased to see some of these models on display at the manufacturer's Las Vegas Convention Center booth. The existing EAGLE OC shroud design was the first to attract our photographer's attention—it has been reused to house and temper NVIDIA's latest GeForce RTX 4070 Ti SUPER 16 GB GPU.

The futuristic aesthetic—comprised of cosmic-themed graphics, symbols, and geometric shapes—and protective metal backplate encase GIGABYTE's WINDFORCE cooling system. We see three (alternate spinning) 80 mm unique blade fans, along with eight composite copper heat pipes, and a copper plate directly touching the GPU. The GeForce RTX 4070 Ti SUPER EAGLE OC 16G has an integrated RGB lighting system.

Today at the "AI Everywhere" event, Intel launched its 5th Gen Intel Xeon processors (code-named Emerald Rapids) that deliver increased performance per watt and lower total cost of ownership (TCO) across critical workloads for artificial intelligence, high performance computing (HPC), networking, storage, database and security. This launch marks the second Xeon family upgrade in less than a year, offering customers more compute and faster memory at the same power envelope as the previous generation. The processors are software- and platform-compatible with 4th Gen Intel Xeon processors, allowing customers to upgrade and maximize the longevity of infrastructure investments while reducing costs and carbon emissions.

"Designed for AI, our 5th Gen Intel Xeon processors provide greater performance to customers deploying AI capabilities across cloud, network and edge use cases. As a result of our long-standing work with customers, partners and the developer ecosystem, we're launching 5th Gen Intel Xeon on a proven foundation that will enable rapid adoption and scale at lower TCO." -Sandra Rivera, Intel executive vice president and general manager of Data Center and AI Group.

Apple has announced a partnership deal with Amkor, one of the leading chip packaging and testing manufacturers, which will build a two billion US Dollar silicon packaging facility in Peoria, Arizona. Being the only US-based OSAT (outsourced semiconductor assembly and test) provider, Amkor has decided to invest its funds and apply for the CHIPS Act, hoping to get a part of the funding from the US government's grant budget. The state-of-the-art facility in Arizona will feature over 500,000 square feet (46,452 square meters) of cleanroom space for packaging and testing chips. Using Amkor's latest technologies, the plant will support advanced computing, automotive, and communications chip packaging. It is tailored to meet the capacity needs of major customer Apple starting in 2025-2026. Apple will be the largest customer, with the Amkor facility packaging Apple-designed chips produced at the nearby TSMC wafer fabrication plant.

Building a chip packaging facility in the US with advanced packaging types means that the domestic manufacturing of advanced silicon is now possible across almost the entire supply chain, with OSAT now being present on US soil as well. In the initial phase, this partnership will enable domestic advanced packaging capabilities for leading-edge chips down to 3 nm nodes, which Apple plans to utilize for its A and M series of processors. Along with the creation of an estimated 2,000 local jobs, the investment serves as a boost to the local economy as well. Additionally, Amkor is TSMC's strategic partner, meaning future designs and packaging will cooperate without any delays.

MAINGEAR, the leader in premium-quality, high-performance, custom PCs, today announced the launch of its latest lineup of Pro Series Workstation PCs, meticulously engineered and configurable with the industry's most powerful components, to cater to the diverse needs of professionals across multiple industries.

Ideal for game developers, photo editors, graphics designers, videographers, 3D rendering artists, music producers, CAD engineers, data scientists, and AI/Machine Learning developers, the MAINGEAR ProWS Series introduces a range of desktop workstations crafted to crush the most intensive tasks, elevate productivity and streamline workflow.

Rapidus Corporation, a company involved in the research, development, design, manufacture, and sales of advanced logic semiconductors, today announced an agreement with Tenstorrent Inc., a next-generation computing company building computers for AI, to jointly develop semiconductor IP (design assets) in the field of AI edge devices based on 2 nm logic semiconductors.

In addition to its AI processors and servers, Tenstorrent built and owns the world's most performant RISC-V CPU IP and licenses that technology to its customers around the world. Through this technological partnership with Rapidus, Tenstorrent will accelerate the development of cutting-edge devices to meet the needs of the ever-evolving digital society.

During its Ignite conference, Microsoft introduced a duo of custom-designed silicon made to accelerate AI and excel in cloud workloads. First of the two is Microsoft's Azure Cobalt 100 CPU, a 128-core design that features a 64-bit Armv9 instruction set, implemented in a cloud-native design that is set to become a part of Microsoft's offerings. While there aren't many details regarding the configuration, the company claims that the performance target is up to 40% when compared to the current generation of Arm servers running on Azure cloud. The SoC has used Arm's Neoverse CSS platform customized for Microsoft, with presumably Arm Neoverse N2 cores.

The next and hottest topic in the server space is AI acceleration, which is needed for running today's large language models. Microsoft hosts OpenAI's ChatGPT, Microsoft's Copilot, and many other AI services. To help make them run as fast as possible, Microsoft's project Athena now has the name of Maia 100 AI accelerator, which is manufactured on TSMC's 5 nm process. It features 105 billion transistors and supports various MX data formats, even those smaller than 8-bit bit, for maximum performance. Currently tested on GPT 3.5 Turbo, we have yet to see performance figures and comparisons with competing hardware from NVIDIA, like H100/H200 and AMD, with MI300X. The Maia 100 has an aggregate bandwidth of 4.8 Terabits per accelerator, which uses a custom Ethernet-based networking protocol for scaling. These chips are expected to appear in Microsoft data centers early next year, and we hope to get some performance numbers soon.