JEDEC and the JC-42.6 Subcommittee for Low Power Memories has announced the publication of the new JESD209-5B standard which now includes improvements to LPDDR5, as well as an extension for the standard in the form of new LPDDR5X.

Source: AnandTech – JEDEC Publishes LPDDR5X Standard at up to 8533 Mbps

Continuing our Q2 earnings coverage this month, AMD is next out the gate in reporting their earnings. And, has been the story now for most of the last year, AMD is enjoying explosive revenue growth across the company. CPU, GPU, and semi-custom sales are all up, pushing the limits of what AMD can do amidst the current chip crunch, and pushing the company to new levels of profitability in the process.

For the second quarter of 2021, AMD reported $3.85B in revenue, making for yet another massive jump over a year-ago quarter for AMD, when the company made just $1.93B in a then-record quarter. Now, half-way through 2021, AMD’s financial trajectory is all about setting (and beating) records for the company, as evidenced by the 99% leap in year-over-year revenue – falling just millions short of outright doubling their revenue.

AMD’s big run-up in revenue is also reflected in the company’s other metrics; along with that revenue AMD’s net income has grown by 352% year-over-year, now reaching $710M. And if not for an unusual, one-off tax benefit for AMD’s Q4’2020, this would have been AMD’s most profitable quarter ever – and indeed is on a non-GAAP basis. Meanwhile, as you might expect from such high net income figures, AMD’s gross margin has risen even further and now sits at 48%, up 4 percentage points from the year-ago quarter and 2 points from last quarter.

Breaking down AMD’s results by segment, we start with Computing and Graphics, which encompasses their desktop and notebook CPU sales, as well as their GPU sales. That division booked $2.25B in revenue for the quarter, $883M (65%) more than Q2 2020. Accordingly, the segment’s operating income is (once more) up significantly as well, going from $200M a year ago to $526M this year.

As always, AMD doesn’t provide a detailed breakout of information from this segment, but they have provided some selective information on revenue and average selling prices (ASPs). Overall, client CPU sales have remained strong; client CPU ASPs are up on both a quarterly and yearly basis, indicating that AMD has been selling a larger share of high-end (high-margin) parts. According to AMD this is the case for both desktop and laptop sales, and making this the fifth straight quarter of revenue share gains.

Meanwhile the company is reporting similarly good news from their GPU business. As with CPUs, ASPs for AMD’s GPU business as up on both a yearly and quarterly basis. According to the company this is being driven by demand for high-end Radeon 6000 video cards, as well as AMD Instinct (data center) sales. AMD began initial shipments of their first CDNA 2 architecture-based Instinct accelerators in Q2, opening the spigot there for data center GPU revenue going into Q3.

Moving on, AMD’s Enterprise, Embedded, and Semi-Custom segment has once again experienced a quarter of rapid growth, thanks to the success of AMD’s EPYC processors and demand for the 9th generation consoles. This segment of the company booked $1.6B in revenue, $1035M (183%) more than what they pulled in for Q2’20, and 19% ahead of an already impressive Q1’21. The big jump in revenue also means that the segment is even further into the black on an operating income basis, continuing to close the gap with the Computing and Graphics segment even with the all-around growth.

Overall, both the enterprise and semi-custom sides of this segment are up on a yearly basis. AMD set another record for server processor revenue this quarter on the strength of EPYC processor sales. Meanwhile semi-custom revenue was up on both a yearly and a quarterly basis, reflecting the continued demand for the latest generation of consoles.

Looking forward, AMD’s expectations for the quarter and for the rest of the year have been bumped up once again. For Q3 the company expects to book $4.1B (+/- $100M) in revenue, which if it comes to pass will be 46% growth over Q3’20. Meanwhile AMD’s full year 2021 projection now stands at a 60% year-over-year increase in revenue versus their $9.8B FY2020, which is 10 percentage points higher than their forecast from the end of Q1.

Finally, while AMD doesn’t have any major updates on the ongoing Xilinx acquisition, the company has reiterated that it remains on-track. Which means that if all goes according to plan, it will close by the end of the year.

Source: AnandTech – AMD Reports Q2 2021 Earnings: Company-wide Growth Drives Doubled Revenue

Alongside Intel’s sizable announcement today regarding their manufacturing roadmap over the next half-decade, the company is also announcing their first major customer for their third-party foundry service, IFS. And in an example of how Intel’s entry into the contract fab business is going to make for some strange bedfellows, it turns out that major customer is Qualcomm.

Per Intel’s announcement, Intel and Qualcomm are partnering up to get Qualcomm products on Intel’s 20A process, one of the company’s most advanced (and farthest-out) process node. The first of Intel’s “Ångström” process nodes, 20A is due in 2024 and will be where Intel first implements Gate-All-Around (GAA) transistors, one of the major manufacturing technology milestones on Intel’s new roadmap.

Given that 20A isn’t due out for another three years, neither company is saying much more about the partnership at this point – we’re talking about chip designs that are still in their earliest stages – but even being able to name a major customer like Qualcomm is a big deal for Intel. Not only does it show that another major industry player has a degree of faith in what Intel is trying to accomplish with its silicon lithography technology, but it helps to validate Intel’s efforts to open up into the contract fab business.

Meanwhile, an announcement like this opens the door to all kinds of speculation over just what Qualcomm will be building over at Intel. Qualcomm is best known for their mobile SoCs, and the company already has significant experience using multiple fabs as a customer of both TSMC and Samsung. So it may be that Qualcomm is looking to build a mainstream mobile SoC or two at Intel as a way to get experience working with Intel and prove that Intel’s fabs will meet their needs. Alternatively, Qualcomm may be looking to take advantage of Intel’s PC-tuned manufacturing lines to produce Nuvia-infused laptop SoCs – which would mean Intel would be directly producing competing chips.

There are a lot of possibilities here over the long-run, though in the short-run it’s likely that Qualcomm is going to play things conservatively. So suffice it to say, it will be interesting to see just what Qualcomm is using their rival’s fabs for in a few years.

Qualcomm is excited about the breakthrough RibbonFET and PowerVia technologies coming in Intel 20A. We’re also pleased to have another leading-edge foundry partner enabled by IFS that will help the U.S. fabless industry to bring its products to an onshore manufacturing site.

-Cristiano Amon, President and CEO, Qualcomm

Source: AnandTech – Intel’s First High-Profile IFS Fab Customer: Qualcomm Jumps on Board For 20A Process

Kicking off another earnings season for the tech industry, we as always start things off with Intel, who is the first big chipmaker out of the gate. Over a year into the coronavirus pandemic – and slowly heading out of it – Intel has seen its ups and downs as product demands have shifted and the company’s ability to execute over the long term has been challenged by fab delays. Following a particularly painful (by Intel standards) first quarter, the company is hoping to put those problems behind them with a stronger second quarter.

For the second quarter of 2021, Intel reported $19.6B in revenue, a decline of less than $100M versus Q2’20, and what Intel is calling a flat difference overall. More importantly, perhaps, is that Intel’s profitability has also held quite steady (and significantly improved over Q1), with Intel booking $5.1B in net income for the quarter, a YoY decline of 1%. Overall, with a lone caveat, Intel’s Q2 performance has exceeded their earlier projections.

Intel’s famed gross margin has also recovered on both a quarterly and yearly basis. At 57.1% it’s up almost 2 percentage points higher than Q1, and almost 4 percentage points higher than Q2’20. Intel’s gross margin has been subject to greater than usual fluctuations as of late – typically dropping whenever a major new product is ramping – but at least for Q2 it is on the rise as Intel enjoys a very profitable quarter.

Breaking things down on a group basis, there are a couple of major points to immediately take away. The first is that, while still recorded per Generally Accepted Accounting Principles (GAAP) rules, Intel is all but ready to cut loose its NAND memory business, which it’s in the process of selling to SK hynix. That said, the deal has not been approved and a closing date has not been set, so while Intel is opting to exclude it from their non-GAAP results (and future business projections), they aren’t free of it quite yet.

Second, this is the first full quarter that can be used for year-over-year comparisons with the coronavirus pandemic. While Intel’s manufacturing side has long since stabilized there, year-over-year numbers are sometimes in odd places as the demand mix a year ago was very unusual, to put it mildly.

For Q2, Intel’s Client Computing Group was once again the biggest winner; that division pulled in $10.1B in revenue, and is up 6% YoY. According to Intel, both laptop and desktop revenue is up, as Intel has surpassed shipping 50 million Tiger Lake processors. That said, these revenue gains are largely volume-driven; ASPs for both desktop and mobile are down, due in part to what Intel is noting to be increased sales of low core count processors. Intel’s improving fab situation has also played a part here – according to the company, 10nm production costs have dropped, helping to improve the division’s operating income.

Meanwhile Intel’s Data Center Group is really starting to feel the impact of the increasingly competitive server environment. While revenue was up 16% versus Intel’s tough Q1, it’s still down 6% on a yearly basis. Breaking that down further, both cloud server provider and communication sales are down, buoyed somewhat by an uptick in enterprise and government sales. But with chip volume and ASPs both down a bit, group revenue can only fall. Intel has long wanted to get away from CCG leading the company’s earnings – in big part by growing its server revenues – but for the moment it looks like that won’t be in the cards. Though as Intel continues to ramp up 10nm production (and thus Ice Lake Xeon production) there is some opportunity to recover in future quarters.

As previously mentioned, Intel is looking to cut loose its NAND business, which is now the only part of the company’s Non-volatile Solutions Group (Optane is DCG). None the less, for the moment Intel still has to account for the group’s revenue, which is down 34% on a yearly basis. Because Intel is downplaying the group so much, they aren’t offering any meaningful written commentary on why revenues are down, but it would seem to be pandemic-related. NSG recorded a very big bump in revenue a year ago, and things have since fallen back towards the industry baseline.

Rounding out Intel’s remaining divisions, both the ioT and Mobileye groups are up significantly on a yearly basis, with revenue jumping 124% in the case of Mobileye. Both groups are benefitting from the post-pandemic recovery, as well as additional design wins in the case of Mobileye. Otherwise, Intel’s final group, Programmable Solutions, saw revenues decline a few percent due to what Intel calls “inventory digestion and supply constraints.”

Overall, Q2’2021 marks a better quarter for Intel on both a quarterly basis and a yearly basis. Along with recovering from their tough Q1 three months ago, Intel this year isn’t facing another fab delay, as Intel’s big 7nm delay was first announced as part of their Q2’20 earnings. To be sure, the company is essentially flat on revenue and profit on a yearly basis – while its competition has generally grown – but as Intel gets further into CEO Pat Gelsinger’s tenure, there is increasing confidence that Intel will hit its goals (and a decreasing window to miss them).

Looking forward, based on their stronger-than-expected Q2 performance and optimism about Q3 and beyond, Intel is increasing its full-year guidance. The company is now projecting revenue to grow on a yearly basis, while gross margin projections remain unchanged at 56.5%. With that said, Intel’s projections all exclude their NAND business, with the company assuming that the deal will be approved by regulators this year as originally planned.

Finally, the next big business update from Intel will come on Monday, when Intel hosts its Intel Accelerated event. That webcast will be focused on Intel’s process and packaging roadmaps, and is a big step in Intel’s efforts to establish their IDM 2.0 strategy. We’re hoping to see more on Intel’s 10nm and 7nm roadmaps there, as well as more on Intel’s next-gen packaging technologies. In the meantime, Intel’s efforts seem to have at least caught the collective ear of Silicon Valley, as the company has revealed today that they’re in talks with 100 potential foundry customers.

Gallery: Intel Q2 2021 Earnings Presentation

Source: AnandTech – Intel Reports Q2 2021 Earnings: Client Computing Leads the Way

Today OnePlus is announcing the 2021 successor to the OnePlus Nord – the company mid-range line-up phone. The new phone takes the hardware platform from the OPPO Reno 5 Pro in that it is now powered by the new MediaTek Dimensity 1200. 

Source: AnandTech – OnePlus Announces Nord 2 with Dimensity 1200 SoC

The design of most leading edge processors and ASICs rely on steps of optimization, with the three key optimization points being Performance, Power, and Area (and sometimes Cost). Once the architecture of a chip is planned, it comes down to designing the silicon of that chip for a given process node technology, however there are many different ways to lay the design out. Normally this can take a team of engineers several months, even with algorithmic tools and simulation to get a good result, however that role is gradually being taken over with Machine Learning methods. Cadence today is announcing its new Cerebrus integrated ML design tool to assist with PPA optimization – production level silicon is already being made with key partners as the tool directly integrates into Cadence workflows.

Source: AnandTech – Cadence Cerebrus to Enable Chip Design with ML: PPA Optimization in Hours, not Months

Known for its core design IP that ends up in everything from IoT to smartphones to servers, Arm is now presenting that it has enabled one of its key microcontrollers in a new form factor: rather than using silicon as a base, the company has enabled a processor core in plastic. The technology has been in the works for almost a decade, but Arm has been waiting on the fabrication methods to create a fully working core. Now the company has something working in a tangible medium and the research has been published in Nature.

Source: AnandTech – PlasticArm: Get Your Next CPU, Made Without Silicon

Today at a private GlobalFoundies event, CEO Tom Caulfield accompanied by Senate Majority Leader Chuck Schumer, announced that the company is set on expansion. At the heart of this new initiative is a doubling of Fab 8, GF’s leading manufacturing facility, at the cost of around $1B. Accompanying this is the disclosure that GF is going to build another manufacturing facility close to Fab 8, in Malta NY, as part of a Private-Public partnership. Details of the new facility were not given.

CEO Tom Caulfield

GlobalFoundries is a contract manufacturer of microprocessors, focusing on adjacency technologies from 12nm and larger geometries. While most column inches are spent discussing the leading-edge manufacturing at GF’s competitors, in a discussion with the CEO we were told that GF addresses around 70% of the semiconductor market and in the current climate is currently running all of its facilities at maximum production.

GF has three main fabs in Malta NY, Dresden Germany, and Singapore – all three are running at maximum output, and GF recently announced a new plant in Singapore capable of 450K wafers per year. Tom Caulfield told us prior to that announcement that the Malta fab is around two-thirds full of equipment, Dresden is at about half, but Singapore is full, hence the new Singapore fab. In March GF announced a $1.4B expansion divided equally between the three sites, with Production capacity is expected to increase by 13% this year and by 20% next year as a result of the increased funding. Today’s announcement commits to adding additional machines at Malta to scale out to the space already there, for another 150k wafers per year, at a cost of $1B.

New website with the new branding. They should have put ‘Moore’

The other element of the announcement is the new fab in Malta. The deployment of a new facility, especially at scale, costs billions. GlobalFoundries today acknowledges that it will take billions, citing the US government’s desire to increase national manufacturing in light of the global scale and building more on American soil. Exactly how GF will implement a new facility has not been disclosed – no timeline, no costs, no information about where the funding is coming from, or what process nodes will be manufactured on-site. It was announced that it would be a private-public partnership, developing chips for high-growth areas such as automotive, 5G, and IoT. The fab is set to create 1000 technical jobs and another few thousand in ancillary positions in the local area to support it. Discussions were also made in light of the semiconductor supply chain, and the need to invest and evolve that part of the business alongside manufacturing improvements. Senator Schumer spoke about the need to pass grow semiconductors, holding up a bag of chips alongside a wafer.

These announcements are part of a train of recent disclosures and talk about GlobalFoundries. Last week it was rumored that Intel was seeking to acquire GF for $30 billion, however today GF announced a complete logo change and rebranding of the business, which doesn’t tend to occur if a company is in the process of acquisition talks. Alongside this, GF is expected to bring forward its Initial Public Offering (IPO) from 2022 to late 2021. The company is currently owned wholly by the Emerati state holding company Mubadala, and the IPO is on the back of some growth of GF in light of the high semiconductor demand environment. GlobalFoundries expects 2021 revenue to be around $6.2 billion, a +9% growth over 2020.

Official Press Release from GlobalFoundries

Related Reading

• GlobalFoundries To Build New 450K Wafer-per-Year Fab in Singapore
• AMD and GlobalFoundries Update Wafer Supply Agreement: Orders Through 2024, Now Non-Exclusive
• GlobalFoundries Upgrades for Silicon Photonics in Quantum Computers
• GlobalFoundries to Invest $1.4B in Expansion, Potential Earlier IPO
• GlobalFoundries’ 22FDX with MRAM is Ready
• GlobalFoundries and SiFive to Design HBM2E Implementation on 12LP/12LP+
• GlobalFoundries and TSMC Sign Broad Cross-Licensing Agreement, Dismiss Lawsuits

Source: AnandTech – GlobalFoundries To Spend Billions: Doubling Fab 8, Creating New Fab in NY

Russian outlet Vedomosti.ru today is reporting that the conglomerate Rostec, a Russian state-backed corporation specializing in investment in technology, has penned a deal with server company Yadro and silicon design company Sintakor to develop RISC-V processors for computers, laptops, and servers. Initial reports are suggesting that Sintakor will develop a powerful enough RISC-V design to power government and education systems by 2025.

The cost of the project is reported to be around 30 billion rubles ($400m), with that the organizers of the project plan to sell 60,000 systems based around new processors containing RISC-V cores as the main processing cores. The reports state that the goal is to build an 8-core processor, running at 2 GHz, using a 12-nanometer process, which presumably means GlobalFoundries but at this point it is unclear. Out of the project funding, two-thirds will be provided by ‘anchor customers’ (such as Rostec and subsidiaries), while the final third will come from the federal budget. The systems these processors will go into will operate initially at Russia’s Ministry of Education and Science, as well as the Ministry of Health.

Sintakor (or Syntacor) already develops its own core with the RISC-V architecture, rather than licensing a design. There have been questions as to whether any current RISC-V design is powerful enough to be used in a day-to-day work machine suitable for administrative services, however with the recent news that Canonical is enabling Ubuntu/Linux on some of SiFive’s RISC-V designs, chances are that by 2025 there will be a sufficient number of software options to choose from should the Russian processor adhere to any specifications required. That being said, it is not uncommon for non-standard processors in places like Russia or China to use older customized forks of Linux to suit the needs of the businesses using the hardware. Sintakor’s documentation states that their highest performance 64-bit core already supports Linux.

Sintakor’s latest design

This news is an interesting development given that Russia has multiple home-grown CPU prospects in the works already, such as the Elbrus 2000 family of processors that run a custom VLIW instruction set with binary translation for Intel x86 and x86-64; these processors already offer 8-core and multi-socket systems running on Linux. Development on Elbrus is still ongoing with Rostec in the mix, and the project seems focused on high-powered implementations in desktop to server use. In contrast, the new RISC-V development seems to be targeting low-powered implementations for desktop and laptop use. Russia also has Baikal processors using the MIPS32 ISA, built by a Russian supercomputer company.

It will be interesting to see how this story develops: $400m should be sufficient to build a processor and instruct system design at this level, which puts the question on how well the project will execute.

Sources: @torgeek, Vedomosti.ru

Related Reading

• Russia’s Elbrus 8CB Microarchitecture: 8-core VLIW on TSMC 28nm
• Intel to Create RISC-V Development Platform with SiFive P550 Cores on 7nm in 2022
• Intel Licenses SiFive’s Portfolio for Intel Foundry Services on 7nm
• SiFive Announces First RISC-V OoO CPU Core: The U8-Series Processor IP
• SiFive Unveils Freedom Platforms for RISC-V-Based Semi-Custom Chips
• Western Digital Rolls-Out Two New SweRV RISC-V Cores For Microcontrollers

Source: AnandTech – Russia To Build RISC-V Processors for Laptops: 8-core, 2 GHz, 12nm, 2025

Legendary overclocker Vince “KINGPIN” Lucido, part of EVGA’s in-house design team, posted to his Facebook page an image that has caused a bit of a stir. The post is captioned ‘The red pill’, with an image of the appears to be the rear of an EVGA Dark series motherboard using AMD’s famous Zen logo. This points to a potential AMD Ryzen version of its board, and this would be the first AMD-based EVGA motherboard since back in days of AMD’s Athlon 64 processors. Most recently, EVGA has made hardware parts for two primary vendors: motherboards for Intel and graphics cards for NVIDIA. So making something AMD again is quite a shock.

EVGA’s Dark series of motherboards typically cater towards enthusiasts and overclockers, with premium controller sets and a wide variety of features designed for pushing silicon to the limit. Dissecting the teased image posted by Vince Lucido on Facebook, we can see the EVGA X570 or X570S Dark series motherboard will feature one 24-pin 12 V ATX motherboard power input two 12 V EPS ATX CPU power inputs, and two full-length PCIe 4.0 slots.

Looking at what memory support the board might have, the rear of the board looks to only feature two memory slots, which is common on Dark series models, with two or more PCIe 4.0 M.2 slots likely to be a feature. Glancing the rear of the power delivery, there does look to be at least 14 power stages, with given the pedigree and status of EVGA’s Dark series models, it could easily be a 16-phase monster.

At the moment, EVGA hasn’t officially said anything about the launch of the EVGA X570/X570S Dark, but we expect an announcement in due course, hopefully with pricing and expected availability.

Related Reading

• The ASUS ROG Strix X570-E Gaming Motherboard Review
• Biostar Announces B550T-Silver Mini-ITX Motherboard for AMD’s Ryzen 5000 Family
• Investigating Performance of Multi-Threading on Zen 3 and AMD Ryzen 5000
• Best AMD Motherboards: June 2021

Source: AnandTech – EVGA Teases an AMD X570 Dark Edition: For Ryzen 5000 Enthusiasts

Earlier this year, new Intel CEO Pat Gelsinger outlined his new ‘IDM 2.0’ vision for Intel. This vision was a three pronged strategy based on improving its own process node technology, mixing in other foundry technology where needed, and also realigning its manufacturing for a new foundry service offering allowing other semiconductor companies to use Intel’s manufacturing expertise. As part of the journey towards Gelsinger’s IDM 2.0 vision, we were told to expect updates at a more regular cadence, and the announcement of ‘Intel Accelerated’ in a couple of weeks is the next event on the calendar.

Source: AnandTech – Intel Accelerated Webcast on July 26th: Update on Process Technology and Roadmaps