TSMC has introduced a brand-new manufacturing technology roughly every two years over the past decade. Yet as the complexity of developing new fabrication processes is compounding, it is getting increasingly difficult to maintain such a cadence. The company has previously acknowledged that it will start producing chips using its N3 (3 nm) node about four months later than the industry is used to (i.e., Q2), and in a recent conference call with analysts, TSMC revealed additional details about its latest process technology roadmap, focusing on their N3, N3E, and N2 (2 nm) technologies.

N3 in 2023

TSMC’s N3 technology will provide full node scaling compared to N5, so its adopters will get all performance (10% – 15%), power (-25% ~ -30%), and area (1.7x higher for logic) enhancements that they come to expect from a new node in this day and age. But these advantages will come at a cost. The fabrication process will rely extensively on extreme ultraviolet (EUV) lithography, and while the exact number of EUV layers is unknown, it will be a greater number of layers than the 14 used in N5. The extreme complexity of the technology will further add to the number of process steps – bringing it toto well over 1000 – which will further increase cycle times. 

As a result, while mass production of the first chips using TSMC’s N3 node will begin in the second half of 2022, the company will only be shipping them to an undisclosed client for revenue in the first quarter of 2023. Many observers, however, expected these chips to ship in late 2022.

“N3 risk production is scheduled in 2021, and production will start in second half of 2022,” said C.C. Wei, CEO of TSMC. “So second half of 2022 will be our mass production, but you can expect that revenue will be seen in first quarter of 2023 because it takes long — it takes cycle time to have all those wafer out.”

N3E in 2024

Traditionally, TSMC offers performance-enhanced and application-specific process technologies based on its leading-edge nodes several quarters after their introduction. With N3, the company will be changing their tactics somewhat, and will introduce a node called N3E, which can be considered as an enhanced version of N3. 

This process node will introduce an improved process window with performance, power, and yield enhancements. It is unclear whether N3 meets TSMC’s expectations for PPA and yield, but the very fact that the foundry is talking about improving yields indicates that there is a way to improve it beyond traditional yield boosting methods. 

“We also introduced N3E as an extension of our N3 family,” said Wei. “N3E will feature improved manufacturing process window with better performance, power and yield. Volume production of N3E is scheduled for one year after N3.”

TSMC has not commented on whether N3E will be compatible with N3’s design rules, design infrastructure, and IPs. Meanwhile, since N3E will serve customers a year after N3 (i.e., in 2024), there will be quite some time for chip designers to prepare for the new node.

N2 in 2025

TSMC’s N2 fabrication process has largely been a mystery so far. The company has confirmed that it was considering gate-all-around field-effect transistors (GAAFETs) for this node, but has never said that the decision was final. Furthermore, it has never previously disclosed a schedule for N2. 

But as N2 gets closer, TSMC is slowly locking down some additional details. Particularly, the company is now formally confirming that the N2 node is scheduled for 2025. Though they are not elaborating on whether this means HVM in 2025, or shipments in 2025.

“I can share with you that in our 2-nm technology, the density and performance, will be the most competitive in 2025,” said Wei.

Source: AnandTech – TSMC Roadmap Update: 3nm in Q1 2023, 3nm Enhanced in 2024, 2nm in 2025

Fabs are well-known for being an expensive business to be in, so any time a new fab is slated for construction, it tends to be a big deal – especially amidst the current chip crunch. To that end, TSMC this week has announced plans to build a new, semi-specialized fab in Japan to meet the needs of its local customers. The semiconductor manufacturing facility will focus on mature and specialty fabrication technologies that are used to make chips with long lifecycles for automakers and consumer electronics. The fab will be Japan’s most advanced fab for logic when it becomes operational in late 2024 and if the rumors about planned investments are correct, it could also be Japan’s largest fab for logic chips.

“After conducting due diligence, we announce our intention to build a specialty technology fab in Japan, subject to our board of directors approval,” announced CC Wei, chief executive officer of TSMC, during a conference call with investors and financial analysts. “We have received a strong commitment to support this project from both our customers and the Japanese government.”

Comes Online in Late 2024

TSMC’s fab in Japan will process 300-mm wafers using a variety of specialty and mature nodes, including a number of 28 nm technologies as well as 22ULP process for ultra-low-power devices. These nodes are not used to make leading-edge ASICs and SoCs, but they are widely used by automotive and consumer electronics industries and will continue to be used for years to come not only for existing chips, but for upcoming solutions as well. 

“This fab will utilize 20 nm to 28 nm technology for semiconductor wafer fabrication,” Wei added. “Fab construction is scheduled to begin in 2022 and production is targeted to begin in late 2024, further details will be provided subject to the board approval.”

While TSMC disclosed the specialized nature of the fab, its schedule, and the fact that it gained support from clients and the Japanese government, the company is not revealing anything beyond that. In fact, while it confirmed that the cost of the semiconductor production facility is not included in its $100 billion three-year CapEx plan, it refused to give any estimates about its planned investments in the project.

Meanwhile, there are many things that make this fab special for TSMC, Japan, and the industry.

The Most Advanced Logic Fab in Japan

It was late 2005, AMD and Intel started to ship their first dual-core processors and the CPU frequency battle was officially over. Intel was getting ready to introduce its first 65nm chips in early 2006 and all of a sudden Panasonic said that it had started volume production of the world’s first application processors using a 65 nm technology, which it co-developed with Renesas, putting Panasonic a couple of months ahead of mighty Intel. In mid-2007, Panasonic again beat Intel to punch by several months with its 45 nm fabrication process.

But with their 32 nm node, Panasonic was 9 – 10 months behind Intel. And while the company did a half-node shrink of this process, it ultimately pulled the plug on 22nm following other Japanese conglomerates that opted out from the process technology race even earlier. By now, all Japanese automotive and electronics companies outsource their advanced chips to foundries, who in turn, build the majority of them outside of Japan.

By bringing a 22ULP/28nm-capable fab to Japan, TSMC’s plans will not only brings advanced logic manufacturing back to the country, but it would also amount to the most advanced fab in Japan. TSMC is also constructing an R&D center in Japan and cooperates with the University of Tokyo on various matters, so its presence in the country is growing, which is good news for the local semiconductor industry.

Previously TSMC concentrated its fabs and R&D facilities in Taiwan, but it looks like its rapid growth fueled by surging demand for semiconductors as well as geopolitical matters are compelling the foundry to diversify its production and R&D locations. 

What is particularly interesting is that according to a Nikkei report, the Japanese production facility will be co-funded by TSMC, the Japanese government, and Sony. This marks another major strategy shift for TSMC, which tends to fully own its fabs. In fact, if the Nikkei report is to be believed, the whole project will cost around $7 billion (though it is not said whether this is the cost of first phase of the fab, or a potential multi-year investment).

To put the number into context, SMIC recently announced plans to spend around $8.87 billion on a fab with planned capacity of around 100,000 300-mm wafer starts per month (WSPM). TSMC’s facility will presumably cost less and will be built in a country with higher operating costs, so it may well not be a GigaFab-level facility (which have capacity of ~100K WSPM). But still, we are talking about a sizable fab that could have a capacity of tens of thousands of wafer starts per month, which would make it Japan’s biggest 300-mm logic facility ever. Just for comparison, the former Panasonic fab in Uozo (now controlled by Tower Semiconductor and Nuvoton) has a capacity of around 8,000 WSPM.

TSMC has not formally confirmed any numbers about its Japanese fab, but the company tends to build rather large production facilities that can be expanded if needed. Meanwhile, a fab in Japan that well serve needs of local automotive and electronics conglomerates promises to help them to avoid shortages of chips in the future. This would also leave TSMC free to assign its 28nmTaiwanese and Chinese production lines to other applications, including PCs, which is important for the whole industry.

Source: AnandTech – TSMC to Build Japan’s Most Advanced Semiconductor Fab

With memory manufacturers clamoring over themselves to push out DDR5 in time for the upcoming launch of Intel’s Alder Lake processors, G.Skill has unveiled its latest premium Trident Z5 kits. The latest Trident kits are based on Samsung’s new DDR5 memory chips and range in speed from DDR5-5600 to DDR5-6400, with latencies of either CL36 or CL40. Meanwhile, G.Skill has also opted to use this opportunity to undertake a complete design overhaul from its previous DDR4 memory, with a fresh new look and plenty of integrated RGB, and is .

Looking at performance, the top SKU comes with fast speeds of DDR5-6400, with either a latency of CL 36-36-36-76 or CL 40-40-40-76. Both the lower-rated kits of DDR5-6000 and DDR5-5600 are available with the same latencies, while all of the six combinations will be available in 32 GB kits, with 2 x 16 GB memory modules. The new G.Skill Trident Z5 and Z5 RGB memory kits will also feature the latest Samsung memory ICs, with G.Skill hand screening the memory chips themselves to ensure maximum stability and performance.

At the time of writing, G.Skill hasn’t confirmed the operating voltages of each kit. G.Skill also hasn’t unveiled its pricing at this time, but it did say that the Trident Z5 and Trident Z5 RGB kits will be available from November.

Meanwhile in terms of aesthetics, the G.Skill Trident Z5 DDR5 memory features a new design compared with previous Trident Z series kits. The Trident Z5 comes with a new dual texture heat spreader design and is available either with a black top bar (Z5) or a new translucent RGB light bar (Z5 RGB). It’s also available in black and silver, with a black brushed aluminum insert across both colors, making it stand out.

With the RGB enabled G.Skill Trident Z5 RGB DDR5 memory kits, the lighting can be customized via its Trident Z lighting control software or synced with other third-party software supplied by vendors such as ASRock, ASUS, GIGABYTE, and MSI’s own RGB software.

Gallery: G.Skill

Related Reading

• Samsung Teases 512 GB DDR5-7200 Modules
• Insights into DDR5 Sub-timings and Latencies
• Computex 2021: TeamGroup Announces its First DDR5-4800 Memory Module
• TeamGroup Announces Intentions for DDR5, ELITE DDR5-5200 in 2021

Source: AnandTech – G.Skill Unveils Premium Trident Z5 and Z5 RGB DDR5 Memory, Up To DDR5-6400 CL36

Mesh networking kits / Wi-Fi systems have become quite popular over the last few years. Despite competition from startups such as eero (now part of Amazon) and Plume (with forced subscriptions), as well as big companies like Google (Google Wi-Fi and Nest Wi-Fi), Netgear’s Orbi continues to enjoy popularity in the market. Orbi’s use of a dedicated backhaul provides tangible benefit over other Wi-Fi systems using shared backhauls. However, the costs associated with the additional radio have meant that the Orbi Wi-Fi systems have always carried a premium compared to the average market offerings in the space.

Netgear introduced their first Wi-Fi 6E router – the Nighthawk RAXE500 – at the 2021 CES. Priced at $600, the router utilized a Broadcom platform (BCM4908 network processing SoC + BCM46384 4-stream 802.11an/ac/ax radio). Today, the company is updating the Orbi lineup with a Wi-Fi 6E offering belonging to the AXE11000 class. Based on Qualcomm’s Hawkeye (IPQ8074) / Pine (QCN9074) platform, the company is touting their RBKE960 Orbi series to be the world’s first quad-band Wi-Fi 6E mesh system.

Netgear’s high-end Orbi kits have traditionally been tri-band solutions, with a second 5 GHz channel as a dedicated backhaul. With Wi-Fi 6E, a tri-band solution is mandated – 2.4 GHz, 5 GHz, and 6 GHz support are all needed for certification. The 6 GHz channel, as discussed previously, opens up multiple 160 MHz channels that are free of interference. The RBKE960 series supports the three mandated bands, and also retains a dedicated 5 GHz backhaul, making it a quad-band solution with combined Wi-Fi speeds of up to 10.8 Gbps across all four considered together.

Netgear has opted to retain 5 GHz for the backhaul in order to maximize range. While the 6 GHz band is interference-free, the power restrictions prevent the communication in those channels from having as much range as the existing 5 GHz ones. Having a dedicated backhaul ensures that all the ‘fronthaul’ channels are available for client devices (shared backhauls result in a 50% reduction in speeds available for client devices for each additional node / satellite). The benefits of Wi-Fi 6E and what consumers can expect from the 6GHz band have already been covered in detail in our Nighthawk RAXE500 launch piece. The Orbi RBKE960 series supports up to seven 160 MHz channels, allowing for interference-free operation even in dense apartments with multiple neighbors.

The RBKE960 supports 16 Wi-Fi streams, making for an extremely complex antenna design. Netgear has made improvements based on past experience to the extent that the new Orbi RBKE960 performs better than the Orbi RBK850 even for 5GHz communication (the larger size of the unit also plays a part in this).

In terms of hardware features, the router sports a 10G WAN port, 3x 1GbE, and 1x 2.5GBASE-T ports. The satellite doesn’t have the WAN port, but retains the other three. The 2.5GBASE-T port can be used to create an Ethernet backhaul between the router and the satellite. On the software side, the new Orbi creates four separate Wi-Fi networks for different use-cases.

The reduced range in the 6GHz band means that large homes might require multiple satellites to blanket the whole area with 6GHz coverage.

Installation and management is via the Orbi app. Netgear also includes the NETGEAR Armor cyber-security suite with integrated parental controls – some features in Armor are subscription-based.

Netgear is also introducing an ‘Orbi Black Edition’ available exclusively on Netgear’s own website. With the RAXE500 setting the stage with its $600 price point, it is no surprise that the RBSE960 satellite costs the same (trading the WAN port and other features for an extra 4×4 radio). A kit with a router and a single satellite (RBKE962) is priced at $1100, while the RBKE963 (an additional satellite) bumps up the price tag to $1500. With home Wi-Fi becoming indispensable thanks to the work-from-home trend among other things, Netgear believes consumers will be ready to fork out what is essentially the price of a high-end smartphone or notebook for a reliable and future-proof Wi-Fi solution.

Source: AnandTech – Netgear Updates Orbi Lineup with RBKE960 Wi-Fi 6E Quad-Band Mesh System

Seagate has been maintaining a line of Xbox-certified external SSDs since late 2016. The current Game Drive for Xbox SSD is based on the Seagate Fast SSD’s internals and industrial design. With the USB 3.2 Gen 1 (BarraCuda) Fast SSD reaching EOL (that market segment has since moved on to USB 3.2 Gen 2), the time has now come for Seagate to revamp the internals of the Game Drive for Xbox SSD and give it a new look.

The company is introducing a new Game Drive for Xbox SSD that takes inspiration from the currently available HDD equivalent – sporting a sleek all-black look with a green LED bar. The 96mm x 50mm x 11mm bus-powered portable SSD weighs just 51g, and sports a USB 3.2 Gen 1 micro-Binterface. It is compatible with Xbox Series X, Series S, and all generations of Xbox One. The package includes a 46cm USB 3.0 cable (micro-B to Type-A).

Seagate is planning to launch only a single SKU of the new product – a 1TB version (STLD1000400) for $170. Interestingly, the 3-year warranty is also accompanied by data-loss protection using Seagate’s Rescue Data Recovery Services. Availability is slated for later this month, well in time for the holiday season.

The article will be updated with additional information once we hear back from Seagate regarding details of the SSD’s internals.

Source: AnandTech – Seagate Updates Game Drive SSD for Xbox with New Look and Internals

One of the key semiconductor technologies beyond 3D FinFET transistors are Gate-All-Around transistors, which show promise to help extend the ability to drive processors and components to higher performance and lower power. Samsung has always announced that its first generation GAA technology will align with its ‘3nm’ nodes, with its 3GAE and 3GAP processes. As part of the Samsung Foundry Forum today, some more insight was put into the timeline for the rollout, as well as talk of its 2nm process.

Source: AnandTech – Samsung Foundry: 2nm Silicon in 2025

Despite most discussion about chip manufacturing focusing on the leading edge and blazingly fast and complex side of the industry, the demand for the ‘legacy’ process technologies is also higher than ever, but also by volume a lot bigger than the latest and greatest. These legacy processes form the backbone of most modern electronics, and so being able to offer equivalent technology at lower cost/power is often a win-win for manufacturers and chip designers alike. To that end, Samsung is announcing a new 17nm process node, designed for customers still using a planar 28nm process, but want to take advantage of 14nm FinFET technology.

Source: AnandTech – Samsung Foundry’s New 17nm Node: 17LPV brings FinFET to 28nm

It’s hard not to notice that we’re in the middle of a semiconductor crunch right now. Factories are running at full steam, but pinch points in the supply chain are causing chaos and bottlenecks – whether that means not enough packaging materials, the cost of shipping has increased 10x, or additional tariffs, it’s causing various industries that rely on semiconductors to wait for supply and then pay over expected prices. Nonetheless, everything that is made is being sold, and so all of the big foundries are driving more investment into their supply chain ecosystem as well as raw manufacturing, and Samsung is no different.

Source: AnandTech – Samsung Foundry to Almost Double Output by 2026

One thing Noctua is famed for, other than its high-end design and engineering team delivering top quality air-cooling products, is the brown/beige color scheme. Some users may detest the off-key and non-conventional color, which rarely goes with other colors inside their PC, and so they have to shun Noctua and look elsewhere. Others swear by the design, and ASUS has gone one step further by teaming up with Noctua to create an NVIDIA GeForce RTX 3070 graphics card. The ASUS GeForce RTX 3070 Noctua Edition features two NF-A12x25 PWM cooling fans with a semi-passive design and aims to be one of the coolest and quietest air-cooled RTX 3070 on the market.

ASUS x Noctua: The Start of Something Bigger?

In August, Twitter user @KOMACHI_ENSAKA spotted that an ASUS and Noctua collaboration may have been in the works via a listing on the Eurasian Economic Commission (EEC) website. Putting the rumors to rest, ASUS and Noctua have announced their collaboration today, and they both present the ASUS GeForce RTX 3070 Noctua Edition, with a slightly higher clocked OC version too.

The most striking thing visually with the ASUS GeForce RTX 3070 Noctua Edition models is the dual beige Noctua NF-A12x25 PWM cooling fans attached to a custom heatsink design with multiple fins and heat pipes behind the beige. Noctua and ASUS have opted for an extensive cooling design as the RTX 3070 Noctua Edition and OC version take up 4.3 slots worth of space and measure in at 12.2 inches in length.

Both the regular and OC models operate with a semi-passive design which means that whenever the temperature drops below 50°C, the fans will switch off. Both models have a standard I/O, including dual HDMI 2.1 and three DisplayPort 1.4a video outputs. Providing power to the graphics cards is a pair of 8-pin PCIe power inputs and comes with a recommendation that users install a 750 W power supply or greater.

The ASUS GeForce RTX 3070 Noctua Edition with a dual-branded backplate

Touching on the specifications of the base model, it features a GPU base frequency of 1500 MHz, with a gaming mode boost clock of 1725 MHz and an OC mode boost clock of 1755 MHz. The OC variant features the same base clock of 1500 MHz but with more aggressive boost clocks of 1815 MHz in gaming mode and 1845 MHz in OC mode. Both models share a memory speed of 14 Gbps effective, with 5888 CUDA cores on a 256-bit memory interface and 8 GB of GDDR6 video memory. 

The ASUS GeForce RTX 3070 Noctua and OC version will be available from mid-October, although no pricing has been given at the time of writing.

Source: Noctua

Gallery: ASUS and Noctua Team Up, The ASUS GeForce RTX 3070 Noctua Addition Announced

Related Reading

• NVIDIA Announces RTX A2000 Video Card: Low Profile & Low Power for ProViz
• NVIDIA Announces GeForce RTX 3080 Ti & 3070 Ti: Upgraded Ampere Cards Coming in June
• Intel Video Cards Get a Brand Name: Arc, Starting with “Alchemist” in Q1 2022
• AMD Announces Radeon RX 6600 XT: Mainstream RDNA2 Lands August 11th For $379

Source: AnandTech – ASUS GeForce RTX 3070 Noctua Edition Announced