Intel officially unveiled the final member of their Alder Lake-based NUC12 family earlier this week. The NUC12 Enthusiast (like the three previous Enthusiast NUCs) caters to the gaming / creators market looking for a small form-factor machine with a discrete GPU. As a refresher, Intel created the NUC Enthusiast category back in 2016 with the introduction of the Skull Canyon NUC (NUC6i7KYK). With a 4″ x 5″ motherboard, it had a slightly larger footprint compared to the traditional NUCs. However, the increased size allowed the incorporation of a 45W TDP processor with increased graphics flex. The second generation Hades Canyon moved to a slightly larger board (5.5″ x 8″), while retaining the industrial design of the Skull Canyon NUC. It used the Kaby Lake-G processors with a Kaby Lake processor and an AMD GPU packaged together (with a total TDP budget between 65W and 100W). The NUC11 Enthusiast (Phantom Canyon) went for a more traditional gaming notebook-type architecture with a Tiger Lake-U Core i7-1165G7 and a NVIDIA RTX 2060 laptop GPU. The NUC12 Enthusiast retains a similar architecture. The key difference lies in the fact that this is first NUC to utilize Intel’s Arc discrete GPU. The specifications of the GPU are much more powerful than the NVIDIA RTX2060, and this has resulted in a redesign of the cooling solution as well as the chassis dimensions compared to the NUC11 Enthusiast.

Similar to the Phantom Canyon family, Serpent Canyon will also come in two varieties – a barebones version, and another with a 1TB SSD / 16GB of DDR4-3200 RAM / Windows 11 Home pre-installed. The SKUs utilize the Intel Core i7-12700H notebook processor and the Intel Arc A770M discrete GPU with 16GB of VRAM.

The NUC12 Enthusiast sports a rich set of I/Os. There are two Thunderbolt 4 ports (one in the front and one in the rear) that also carry the display output from the Intel Iris Xe Graphics in the Core i7-12700H. Two USB 3.2 Gen 2 Type-A ports and a SDXC UHS-II slot, along with an audio jack and a quad-microphone array round out the front panel. On the rear, we have an audio output jack (supporting TOSLINK), a single 2.5 Gbps LAN port, four USB 3.2 Gen 2 Type-A ports (with a hub chip behind), and the display outputs (1x HDMI 2.1 4Kp60 and 2x Display Port 2.0 (1.4 certified)) from the Intel Arc A770M.

The table below compares the specifications of the flagships in the last three generations of Enthusiast NUCs.

The block diagram below gives some insights into the design of the system in relation to the I/O capabilities.

Despite the Arc A770M supporting a PCIe 4.0 x16 link to the host processor, the Serpent Canyon configuration keeps the connection at x8. Both PCIe 4.0 M.2 2280 slots are CPU-attached ones. The SD card slot is connected via a PCIe lane instead of USB – this should allow maximum possible performance for different SD cards. While the official specifications indicate that the slot is UHS-II, the technical product specifications document also indicates SD Express support. This depends on the exact SD controller being used in the board, and we have reached out to Intel for clarification. Three of the four Type-A ports in the rear are enabled by a 1:4 Gen 2 hub, which is not ideal in terms of bandwidth sharing. However, the availability of additional ports is always welcome. On the display front, the front Thunderbolt 4 port can support a display bandwidth of around 17 Gbps, while the rear port can support up to 35 Gbps.  With multi-stream support on the Type-C port, the system can drive a total of six diifferent displays – five at 4Kp144 (DP/Alt-DP) and one at 4Kp60 (HDMI). Two 8Kp60 displays can also be driven using a multi-cable / -port solution.

Intel also provided a complete teardown picture along with the press release. The combined cooling solution for the CPU and dGPU with the thermal shroud and heat pipes is clearly seen. Whether this solution aids in / enables performance tuning via the Intel Deep Link Dynamic Power Share feature remains to be seen in hands-on evaluation.

Overall, the Serpent Canyon NUC is a huge step-up for Intel. Moving to a fully in-house solution for both the CPU and dGPU in a small form-factor portable machine will enable the company to gain a larger share of the gaming / creator systems / eSports total addressable market. Based on paper specifications, the level of integration and gaming prowess in the NUC12 Enthusiast should be well beyond what has traditionally been possible in this form factor. On the pricing front, the Mini-PC version with pre-installed OS will come in at $1350, while the barebones version can be purchased for $1180 later this month. These numbers roughly track the introductory pricing for previous-generation Enthusiast NUCs.

Gallery: Intel Announces Serpent Canyon: Alder Lake and Arc Amalgamate in NUC12 Enthusiast

Source: AnandTech – Intel Announces Serpent Canyon: Alder Lake and Arc Amalgamate in NUC12 Enthusiast

Even though the main focus this month has been on AMD’s upcoming launch of its Ryzen 7000 desktop series processors based on the Zen 4 architecture, AMD is also in the process of launching its 7020 series of processors designed for entry-level mobile, codenamed Mendocino. It has launched four new SKUs for mobile, including two Ryzen series models and two Athlon variants, all based on its Zen 2 architecture.

As we learned in May, AMD’s Zen 2 based ‘Mendocino’ APUs are designed as part of its lower-end mobile processor stack for entry-level thin notebook and laptop solutions. The affordable APUs incorporate up to 4 Zen 2 CPU cores, as well as AMD’s integrated Radeon 610M graphics, all with a maximum TDP of 15W.

Focusing specifically on the AMD 7020 series APUs announced today, the top SKU is the Ryzen 5 7520U. Using AMD’s new Ryzen Mobile CPU numbering system, the 7520U is a model year 2023 chip under the Ryzen 5 series banner using its Zen 2 core architecture. The U suffix technically means 15-28W, but in this case AMD has confirmed that all of the current chips have a max TDP of 15W.

The AMD Ryzen 5 7520U benefits from a 2.8 GHz base frequency across its four cores, with a 1T boost frequency of up to 4.3 GHz. It also includes a total of 6MB of cache, split between 4MB of L3 and 2MB of L2 (512KB per core).

Moving down the stack is the Ryzen 3 7320U, which has four cores and eight threads but a base core clock speed of 2.4 GHz and a single core boost frequency of up to 4.1 GHz. Like the Ryzen 5 7520U, it also benefits from a combined cache of 6 MB across its L2/L3 cache structure.

Looking at the Athlon-branded 7020 series chips, the Athlon Gold 7220U offers two cores and four threads, with a base frequency of 2.4 GHz, a single core boost frequency of up to 3.7 GHz, and 5 MB of shared L2/L3 cache. The Athlon Silver 7120U is pretty much the Athlon Gold 7220U, but it has a slightly lower 1T boost frequency of 3.5 GHz and half as much L3 cache (for a total of 3MB instead of 5MB)

AMD Radeon 610M with two graphics cores for performance at the entry-level

All of today’s announced AMD’s Zen 2 Ryzen and Athlon 7020 series will support up to 32 GB of LPDDR5 memory in a dual-channel (64-bit) configuration, and feature two graphics cores/CUs based on its RNDA 2 technology, which AMD is branding the Radeon 610M integrated graphics chip. All four mobile APUs will also include a TDP of up to 15 W. 

At the time of writing, AMD hasn’t revealed specific pricing aside from a total laptop price range of between $399 and $699. However, it has announced that its AMD Ryzen 7020 Series Ecosystem partners are Acer, HP, Lenovo, and Microsoft, with notebooks featuring these chips expected to start appearing on retail shelves sometime in Q4 2022. 

Source: AMD

Source: AnandTech – AMD Launches Mendocino APUs: Zen 2-based Ryzen and Athlon 7020 Series with RDNA 2 Graphics

With NVIDIA’s fall GTC event in full swing, the company touched upon the bulk of its core business in one way or another in this morning’s keynote. On the enterprise side of matters, one of the longest-awaited updates was the shipment status of NVIDIA’s H100 “Hopper” accelerator, which at introduction was slated to land in Q3 of this year. As it turns out, with Q3 already nearly over H100 is not going to make its Q3 availability date. But, according to NVIDIA the accelerator is in full production, and the first systems will be shipping from OEMs in October.

Source: AnandTech – NVIDIA: H100 Hopper Accelerator Now in Full Production, DGX Shipping In Q1’23

Among the spate of announcements from NVIDIA today as part of their fall GTC 2022 event, the company is delivering a surprising shake-up to their DRIVE automotive SoC plans. Effective immediately, NVIDIA has cancelled Atlan, their planned post-Orin SoC for 2025 automobiles. In its place, NVIDIA is announcing Thor, an even more powerful SoC set to launch in the same 2025 timeframe.

Source: AnandTech – NVIDIA Drops DRIVE Atlan SoC, Introduces 2 PFLOPS DRIVE Thor for 2025 Autos

While we’re still over a quarter out from the end of 2022, Intel already has its eyes aimed at 2023 and its eventual refresh of its mobile processors. To that end, today the company has announced that they are making some branding changes for the low-end.

Starting in 2023, Intel will be retiring the Pentium and Celeron brands for laptop processors. In its place, Intel will have a singular “Intel Processor” brand for the low end of the market, while the Core branding (with its multiple tiers) will remain in place for the rest of Intel’s mobile product stack.

Notably, this change only applies to future laptop parts. At this point Intel is not announcing a change for desktop parts or embedded parts. But with that said, I would not be the least bit surprised if these change ultimately came to desktops as well, as mobile is effectively Intel’s leading consumer market segment these days. So technology and names tend to percolate up to the desktop segment, keeping the two in sync.

Intel’s current generation Pentium and Celeron offerings are both based on Alder Lake-U processors with a single performance core and four (one block of) efficiency cores. The only differences between these SKUs, besides price, is clockspeeds – specifically, that the Celeron parts lack turbo. So if Intel is going to pursue a similar strategy in future generations, then it’s not outlandish to fold two similar products under a single brand. Though the decision to forgo any kind of specific branding is an unusual one for Intel.

With that said, there’s also been a notable absence of “pure” Atom parts in this segment in this generation. Intel has yet to produce a true entry-level part using its Gracemont Atom cores; so everything below the Alder Lake Pentiums/Celerons has been the last-generation Tremont Atoms. So larger changes may be afoot for Intel’s cheapest laptop product segment.

Source: AnandTech – Intel to Drop Celeron and Pentium Branding From Laptop Parts In 2023

Micron this week broke ground on its leading-edge memory production facility near Boise, Idaho. The company will invest $15 billion in its new fab as a part of its ambitious plan to invest $40 billion in its U.S.-based manufacturing capacities by the end of this decade as well as spend $150 billion on new fabs by 2030 globally.

Micron’s upcoming leading-edge fab will produce DRAM and will be a rather colossal manufacturing facility. At build-out, when the fab is fully equipped with tools, its cleanroom space will reach 600,000 feet² (55,700 meters²), which is about two times larger compared to cleanroom space at GlobalFoundries’ Fab 8 and which is comparable to cleanroom space at giant fabs operated by Micron’s rivals Samsung and SK Hynix is South Korea. Essentially, Micron will operate one of the largest semiconductor production facilities in the U.S.

The new fab will be located adjacent to Micron’s R&D center and headquarters near Boise, Idaho, which will bring together scientists, process technology developers, and manufacturing engineers in one location, something that promises to speed up time-to-yield and time-to-market for advanced DRAMs. 

“With this facility, Micron will closely couple R&D and manufacturing, providing synergies that will enable us to accelerate the production ramp of advanced memory technology,” said Sanjay Mehrotra, president and CEO of Micron.

Micron is currently prepping the site for the new fab and plans to begin construction in early 2023 in a bid to start bringing cleanroom space online gradually starting in 2025. The facility will be equipped with modern deep ultraviolet (DUV) and extreme ultraviolet (EUV) lithography tools and will make memory using one of Micron’s advanced EUV-enabled production nodes. The company intends to start fabbing DRAMs at its new facility sometimes in 2025 and then ramp up production to the full capacity in the following years. 

At present it is hard to guess which fabrication process will be adopted at the new fab. Considering the fact that Micron is expected to start DRAM production using its first EUV-enabled manufacturing technology (1γ) sometimes in mid-2023 – early-2024, it is likely that the fab near Boise, Idaho, will adopt the company’s second EUV-enabled process (1δ). Yet, this is an educated guess (based on habitual introduction of new DRAM nodes every 18 months or so) at this point.

Micron intends to make 40% of its global DRAM output in America in 2030s, a rebalance that the company has not done in decades. Therefore, in addition to Micron’s new fab near Boise, Idaho, the company plans to build a yet another DRAM facility in the U.S. Currently the company is in the final stages of its selection process for another site in America. 

Micron will invest $15 billion in the new manufacturing facility near Boise, Idaho. The company also plans to get incentives from the local and state authorities and support from the federal government enabled by the CHIPS and Science act signed into law last month.

“The investment, made possible by the anticipated grants and credits provided by the CHIPS and Science Act, also enhances Micron’s supply chain resilience and will establish a new strategic capability for the U.S.,” said the head of Micron.

It is noteworthy that Micron is not the only DRAM maker to start building new memory fabs when demand for both 3D NAND and DRAM is down. Last week SK Hynix began to expand its M15 site with its new M15X building as the company is preparing for increased DRAM demand starting in 2025.

Source: Micron

Source: AnandTech – Micron Breaks Ground on Its Billion EUV DRAM Fab in the U.S.

In preparation for AMD’s Ryzen 7000 series processors, which will launch on September 27th, ASRock has announced that it has developed a new BIOS for its AM5 motherboards. ASRock states that its new firmware has been built to decrease booting times on its motherboards.

With every new platform, chipset, and processor launch, firmware is one of the focal points surrounding motherboards. Over the last couple of years, leading vendors such as ASUS, GIGABYTE, and MSI have kept their firmware consistent, regardless of whether the board is designed for AMD or Intel. ASRock has announced that it has developed a new BIOS specifically for its AM5 motherboards, which are designed to support the upcoming AMD Ryzen 7000 series Zen 4 processors based on TSMC’s 5 nm process node.

ASRock X570S PG Riptide motherboard firmware

ASRock claims that their new BIOS for AM5 will provide better compatibility, although it doesn’t expressly state what, as well as quicker booting or POST times into Windows. As it stands, ASRock has announced five X670E models currently, including the X670E Taichi and Taichi Carrara models, the X670E Steel Legend, the X670E Pro RS, and the X670E PG Lightning. ASRock also hasn’t stated whether or not the layout or core GUI has changed compared with previous generations such as X570 or B550.

Another thing ASRock has announced is that all of its X670E and X670 models will support BIOS Flashback (a feature now native to the AM5 platform), with users able to update their motherboard to the latest firmware with just a USB flash drive and 24-pin power connected to the power supply. 

ASRock’s new firmware for its X670E and X670 motherboards will be available to download from the appropriate product pages on the ASRock website after September 27th. At the time of writing, it hasn’t confirmed any information regarding the pricing of its new AM5 motherboards.

Source: ASRock

Source: AnandTech – ASRock Announces New AM5 BIOS Design, Promises Faster Booting Times

Ahead of the launch of AMD’s latest Ryzen 7000 processors, which will hit retail shelves on September 27th, MSI unveiled pricing on four of its X670/X670E motherboards. Given the world’s current financial climate and features such as PCIe 5.0 connectivity to M.2 storage drives and at least one PCIe x16 slot, the writing has been on the wall for a while regarding pricing.

Currently listed in their US store, MSI has two premium (MEG), one mid-range (MPG), and one entry-level regular X670 (Pro) model, with prices ranging from $290 (Pro X670-P WIFI) up to $1300 (MEG X670E Godlike). 

MSI MPG X670E Carbon WIFI ($480) ATX motherboard

As we’ve seen with previous generations on both Intel and AMD platforms, MSI’s flagship for Ryzen 7000 is the MEG X670E Godlike ($1300). MSI is advertising a 24+2+1 power delivery, with up to six M.2 slots, 10 GbE, 2.5 GbE, Wi-Fi 6E, and a touchscreen 4.5″ M-Vision dashboard panel.

Sitting just behind the Godlike is the MSI MEG X670E Ace ($700), with support for up to six M.2 drives, an advertised 22+2+1 power delivery, and 10 GbE/Wi-Fi 6E networking to sweeten the deal. Both models benefit from PCIe 5.0 PEG slots and one PCIe 5.0 x4 M.2 slot for the latest PCIe Gen 5 SSDs due sometime in November.

X670E Mid-Range Model for $480, X670 Entry-Level at $290

The MPG X670E Carbon ($480) represents MSI’s mid-range MPG offerings with advertised 18+2+1-phase power delivery, an 8-layer PCB, and Wi-Fi 6E connectivity. Despite costing nearly $500, MSI offers a 2.5 GbE NIC considering this has been offering this on entry-level models for the last couple of generations. Importantly, however, this is the cheapest MSI X670E motherboard – meaning it’s the cheapest board MSI will be offering at launch with PCIe 5.0 slots.

Coincidentally, we just reviewed MSI’s MPG Z690 Carbon WIFI, which is MSI’s equivalent motherboard for the Intel LGA1700/Z690 market.That board has an MSRP of $400 and a current selling price of $350 on Amazon. So there is a significant premium right now for the AM5 board, on top of what’s already a relatively high price for a mid-range Intel motherboard.

Finally, the entry-level MSI Pro X670-P WIFI (non-E) model has 2.5 GbE, an advertised 14+2+1-phase power delivery, and supports up to four M.2 slots. Despite dropping PCIe 5.0 in favor of PCIe 4.0, the MSRP is sitting at $290, which is very surreal pricing for an entry-level model.

Gallery: MSI Unveils X670E Pricing Ahead of AMD Ryzen 7000 Launch, Starting at $290

Despite not agreeing with MSI’s current X670E/X670 MSRP pricing, things could change over the coming weeks and months as more vendors announce its offerings to the market. The introduction of B650/B650E boards should also offer a cheaper alternative, though those boards won’t be arriving for at least another month. As it stands, the MSI MEG X670E Godlike will cost $1300, the MEG X670E Ace will cost $700, the MPG X670E Carbon WIFI has a price tag of $480, and the Pro X670-P WIFI will cost $290.

All four models are expected to launch on September 27th, along with AMD’s Ryzen 7000 series processors.

Source: MSI

Source: AnandTech – MSI Unveils X670 Pricing Ahead of AMD Ryzen 7000 Launch, Starting at 0

When a major South Korean memory firm invests over $11 billion in a fab, that raises a couple of eyebrows. But when it comes within a major $100+ billion capital expenditure (CapEx) package, it certainly warrants some attention.

Semiconductor business in general and memory business in particular are very cyclical in their nature. Just a year ago almost all chips were in short supply and prices of commodities like memory or display drivers were high, but now that sales of PCs are declining, memory prices are declining too. But several years down the road demand for PCs and other client devices will rise once again, and so will demand for memory. Which is why SK Hynix is already preparing for this with its Fab M15X expansion plan.

SK Hynix’s Fab M15X will be a two-story building occupying 60,000 m² of land and will be located adjacent to existing Fab M15 in the Cheongju Technopolis industrial complex. The fab will produce 3D NAND (which means more chemical vapor deposition and etching tools in the cleanroom) and/or DRAM (which is more lithography-intensive, so more DUV and EUV equipment in the cleanroom) memory chips, depending on demand by the time it comes online sometimes in 2025. Since at present it is unclear/undecided what the fab will produce, the company cannot disclose planned production capacity of the upcoming manufacturing facility.

At a planned size roughly equal to combined space of SK Hynix’s existing Fab M11 and Fab M12, the company’s Fab M15X expansion looks more like an entirely new fab built adjacent to existing Fab M15 than an expansion project of the existing fab. Yet since the two manufacturing facilities share infrastructure and various facilities, the maker prefers to call it Fab M15X. 

“Looking back on the past 10 years, SK hynix could grow into a global company as it boldly carried out investment during crisis,” said Park Jung-ho, vice chairman and co-chief executive of SK Hynix. “As we look to prepare for the next 10 years now, I believe starting the M15X will be a first step to lay foundation for a solid future growth.”

Source: SK Hynix

Source: AnandTech – SK Hynix Starts Prepping for Next Semiconductor Boom with Billion Memory Fab

While all eyes are on the impending launch of AMD’s new Ryzen 7000 desktop processors, the chipmaker also has its wheels in motion for the future of its mobile product lineup. And while we’re still quite a bit away from the first Zen 4 mobile parts, more immediate on AMD’s roadmap is their Zen 2-based Mendocino SoC, which is aimed at mainstream laptops. Mendocino APUs are set to launch next quarter, and to prepare for that launch, AMD today is updating their mobile processor numbering scheme to accommodate those future products.

The short version of matters is that while the new numbering system is quite similar to AMD’s previous system (e.g. Ryzen Mobile 6000), the company is now dedicating a digit to represent the version of the Zen architecture used. With AMD set to have Zen 2 (Mendocino), Zen 3 (Rembrandt), and eventually Zen 4 (Phoenix) mobile APUs all on the market at the same time, AMD has decided that they need to better disclose the architecture used underneath – a “necessary evil”, as former AnandTech CPU editor Dr. Ian Cutress put it, to avoid any improprieties (perceived or otherwise) that AMD is misleading customers by offering multiple versions of the Zen architecture.

Source: AnandTech – AMD Updates Ryzen Mobile CPU Numbering System Ahead of Mendocino APU Launch

Qualcomm this morning is taking the wraps off a pair of new SoCs for the mid-range and entry-level smartphone markets. Refreshing the company’s longstanding 600 and 400 series of chips, Qualcomm is announcing the Snapdragon 6 Gen 1 and Snapdragon 4 Gen 1. Both SoCs are receiving similar spec bumps, incorporating newer and faster IP blocks from Qualcomm – such as Arm Cortex-A78 derived CPU cores – as well as moving to newer, more contemporary manufacturing processes.

The Snapdragon 600/400 lineups were last updated in mid and early 2021 respectively, so as Qualcomm is already preparing for 2023, the time has finally come to update the bottom half of their product stack. Following Qualcomm’s broad cascading IP strategy, this generation of parts sees both SoC lineups migrate to Cortax-A78 CPUs for their main CPU cores, and in the case of the 6 Gen 1, doubling the number of high-performance CPU cores. Both SoCs also come with faster Adreno GPUs, though in traditional Qualcomm fashion, the company isn’t offering much in the way of details on the underlying hardware there.

Notably, however, Qualcomm’s 2023 mid-range/low-end parts aren’t making the jump to the Armv9 architecture. Unlike the 8 Gen 1 and 7 Gen 1, which incorporated Arm’s new Armv9 cores, Qualcomm’s cascading development strategy means that the 6 and 4 series will remain a bit farther behind the curve. For end users this should have little significance for the moment, but for smartphone vendors and software developers, it does mean Qualcomm won’t complete the Armv9 transition for at least another generation.

Meanwhile, coming up on nearly a year since Qualcomm announced their initial Snapdragon 8 Gen 1 SoC, today’s announcement from Qualcomm brings their remaining smartphone SoC families in alignment with their new product branding strategy. The 6 and 4 series pick up from where the 600 and 400 series left off, respectively, resetting the counted with the inaugural Gen 1 parts. Like the rest of simplified “Gen” series, this also means that Qualcomm is doing away with individual model numbers for its Kyro/Hexagon/Adreno/Spectra blocks, obfuscating a bit what generation of IP Qualcomm is using there.

Source: AnandTech – Qualcomm Unveils Snapdragon 6 Gen 1 and 4 Gen 1 SoCs: Updating Mid-Range and Entry-Level Phones