HW News - Intel Comeback, TSMC 3nm Silicon, Phone Radeon GPUs, NVIDIA Mining GPUs
Posted on January 25, 2021
This round of HW News comes as an all digital CES 2021 comes to a close, so we’re focusing mostly on non-CES news here. We say mostly, because a couple of stories do have a bit of overlap as it relates to what was announced at CES 2021.
At any rate, most notably we have more confirmed price adjustments regarding graphics cards, with EVGA, Zotac, and MSI all joining Asus in raising prices due to expired tariff exemptions. There’s interesting news in Qualcomm acquiring Nuvia, and what it could mean long-term for Qualcomm’s CPU design ambitions. We also have news on TSMC possibly already receiving orders from Intel for non-CPU products, and how TSMC may be making some of Intel’s Core i3-series chips this year.
There’s more of course, so hit the article and video embed below.
01:48 | Parts Shortages Extend to Car Manufacturers
Supply issues in the PC enthusiast space absolutely dominated the news cycle for the entire second half of 2020 -- it was bad enough to be the centerpiece of our Disappointment Build for the year. New GPUs, CPUs, and consoles aren’t the only things soaking up the world’s supply of silicon, and other industries are starting to feel the effects in a big way. Semiconductor shortages have now begun to shut down automotive factories around the world. Ford has closed factories in Louisville, Kentucky and Chennai, India, citing the microchip shortages in each instance. Ford closed the factories on Jan. 14th and said they’ll remain closed at least until the 24th of January. The problem extends well beyond Ford and seems to affect most of the industry, including Volkswagen, Honda, Toyota, Subaru, Honda, Mercedes-Benz, and Nissan are all reporting either temporary shutdowns or reduced production in factories around the world. CNN talked to Kristin Dziczek, who is the vice president of research at a Michigan think tank -- the Center for Automotive Research. Dziczek said "it's temporary, but it's not going to be short-term," CNN continued: “The problems are likely to last throughout the first three months or so of the year. Supplies should hopefully be back to normal in the second quarter.” Anyone looking to spend their stimulus check on a new truck may find themselves restricted to last year’s models, which, much like GPUs, is sure to be a devastating and life-shattering loss that ruins 2021 for friend of the site Gordon Mah Ung. At least we can look forward to the chat sections of automotive videos also being filled with “OUT OF STOCK” comments.
Source (Ford): https://timesofindia.indiatimes.com/business/india-business/ford-shuts-plant-on-chip-shortage/articleshow/80317762.cms
https://www.cnn.com/2021/01/11/tech/ford-plant-shutdown-computer-chip-shortage/index.html
Source (Audi): https://www.bbc.com/news/technology-55704936
Secondary source: https://www.caranddriver.com/news/a35228204/microchip-shortage-car-production-disruption/
06:03 | TSMC Working on 3nm Silicon
Even while multiple industries struggle with the supply of larger process chips, TSMC continues to forge ahead with smaller process nodes. TSMC announced that it plans to enter risk production of their 3nm process this year and that they’re already working on 2nm as well. Risk production of silicon is defined as the period during which the technology is ready, but hasn’t yet been tested by any customers.
Risk production will see TSMC refine the process before mass production begins in 2022. Tseng Guan-wei, an analyst at the market research firm Trendforce, said that “the 3-nm process allows 250 million transistors per square nanometre while 2-nm technology will enable more than 310 million transistors to be packed into the same area.” This, Tseng said, would increase the performance to power efficiency from one generation to the next.
Mario Morales, an analyst for market research firm IDC, notes that Apple, Qualcomm, and Nvidia will be the first to utilize TSMC’s 3nm process. TSMC has not yet announced any firm schedule for 2nm development or production.
Source: https://www.digitimes.com/news/a20210115PD208.html
07:45 | New Intel CEO: Prepared to Come Back Strong
As part of our CES-replacement coverage of Intel, we covered the surprising news that Intel CEO Bob Swan would be stepping down. Intel veteran and former VMware CEO Pat Gelsinger will be stepping into the role starting in February. Ahead of this transition, Gelsinger held a company-wide meeting that was covered by an Oregon-local newspaper, the Oregonian, which is unfortunately paywalled. Since we’re not otherwise all that interested in local Oregon news -- sorry -- we relied on secondary sources for the contents of Gelsinger’s speech.
Gelsinger is entering Intel at a pivotal moment for the company, with 2020 delivering massive blows to the company that had already been showing signs of weakness for a few years. Apple launched laptops that used its own M1 chip, having moved away from the Intel chips that they had used previously. To make matters worse, AMD was finally able to take the gaming performance crown from Intel after years of Ryzen chewing away at Intel’s CPU stronghold. These blows all follow Intel’s struggle to deliver on either 7 or 10nm architectures, having stretched 14nm technology further than anyone could have expected them to.
Despite these setbacks, Gelsinger seems unbowed. Saying to Intel employees in the meeting: “We have to deliver better products to the PC ecosystem than any possible thing that a lifestyle company in Cupertino [makes]”, and that “We have to be that good, in the future.” This likely reference to Apple isn’t necessarily a dig, either: Apple, unlike Intel, doesn’t do the core of its business in silicon design and manufacturing, but rather in product design and marketing. For Apple to best Intel at anything CPU-related is embarrassing for incumbent Intel.
Source: https://www.theverge.com/2021/1/15/22232554/intel-ceo-apple-lifestyle-company-cpus-comment
11:13 | EVGA, Zotac, and MSI Raise GPU Prices
Not long after news broke that Asus would be initiating some price hikes on their graphics cards, we now have word that EVGA, Zotac, and MSI will be doing the same. As was previously reported, the MSRP increases are a direct symptom of US tariff exemptions that expired at the end of last year.
Visiting EVGA’s website, one will be greeted with a notification regarding price adjustments for Nvidia’s RTX 30-series cards.
“Due to ongoing events, EVGA has made price adjustments on the GeForce RTX 30 Series products. This change was necessary due to several factors and will be effective January 11, 2021. EVGA has worked to reduce and minimize these costs as much as possible. For those who are currently in the EVGA.com Notify Queue system or Step-Up Queue, EVGA will honor the original MSRP pricing through April 16th, 2021 if your purchase position is processed before this date,” says EVGA.
The price increases in most cases seem to be less than $100, for what that’s worth. Zotac, on the other hand, seems to be implementing much steeper prices. As evidenced by the Internet Archive’s Wayback Machine, you can see past and present GPU prices for Zotac (via The Verge). Since December, the cost of the Zotac Gaming GeForce RTX 3060 Ti has gone up by nearly $100, while the RTX 3090 has gone up by ~$350.
MSI told Tom’s Hardware that it will also be making adjustments to its current and future graphics cards -- including the ones it announced at CES, with prices TBD.
"I can confirm that MSI will be raising the prices on its graphics cards due to the reimplementation, to the reactivation of tariffs. It’s for that reason, you know, we’re still working on the prices,” MSI’s David Yee told Tom’s Hardware.
Adding insult to injury is the fact that AMD and Nvidia are still struggling to get stock into retail channels, cryptocurrency is on the rise, and there’s no indication that stock will be any better until we get well out of the first quarter. We’ll see, but things remain grim for the time being.
Source: https://www.theverge.com/2021/1/13/22228470/nvidia-evga-zotac-raise-prices-rtx-3080-3070-3060-3090
https://www.tomshardware.com/news/msi-gpu-price-increase-tariffs
14:44 | Qualcomm Is Acquiring Nuvia
Qualcomm is making some waves this week in the CPU design landscape, as it announced that it plans to acquire Nuvia, a silicon startup focused on HPC silicon for the datacenter. Nuvia was formed by notable chip designers John Bruno, Manu Gulati, and Gerard Williams III, all of whom have contributed to Apple’s extremely successful line of A-series SoCs.
According to Qualcomm, it will acquire Nuvia for $1.4B, before “working capital and other adjustments.” As ever, this deal will be subject to the usual regulatory approvals before closing, but there’s no red flags here to speak of; it’s not like Nvidia acquiring Arm.
Qualcomm states that Nuvia’s technology and IP will be integrated across Qualcomm’s portfolio, with what will likely be an emphasis on its flagship Snapdragon SoC. Qualcomm reports that Nuvia’s founders -- and all of its employees -- will now be joining Qualcomm.
Unpacking a few things, it’s not hard to see why Qualcomm would look to absorb Nuvia. Nuvia had ambitions of doing for the server space what Arm did for the mobile space. By acquiring Nuvia, Qualcomm is positioning itself to reenter the server market, after its previously unsuccessful attempt at breaking into the market. Secondly, as Nvidia is set (pending regulatory approval) to acquire Arm, Nuvia’s IP and talent could secure the company’s future roadmap, should Nvidia dare to disrupt Arm’s neutral licensing model.
And then, there’s Apple Silicon, which is set to change the complexion of the CPU landscape in big ways, as it relates to x86. Qualcomm could position itself as a premier chip designer and offer an alternative to Arm and Apple silicon, depending on the success of Nuvia designs.
It will take time for this acquisition to bear fruit, but it could mark another interesting shift in the CPU landscape.
Source: https://www.qualcomm.com/news/releases/2021/01/13/qualcomm-acquire-nuvia
16:39 | Samsung Exynos SoCs With Radeon Silicon Are on the Way
It was almost two years ago at this point, but back in 2019 Samsung announced a joint partnership with AMD that would allow it to license AMD’s Radeon/RDNA IP for its Exynos SoCs. At the time, there was no timeline or time frame given for the partnership, just that it would be a multi-year collaboration.
And while we still don’t have a solid indication of when we will see the first Radeon endowed Exynos chip hit the market, Samsung offered something of an update, the first we’ve heard since the initial announcement.
In announcing the new Exynos 2100, Samsung said that it was working with AMD to have a next-gen mobile GPU inside of its next flagship product, which we assume will be the successor to the Exynos 2100.
As we mentioned nearly two years ago, this could prove to be an important avenue for AMD, as it will be the only chip designer to have graphics IP in x86, consoles, and now Arm-based SoCs.
Source: https://www.tomshardware.com/news/samsung-announces-radeon-exynos-soc-is-coming
17:47 | TSMC May Be Manufacturing Intel’s Core CPUs in 2H21
As we’ve mentioned numerous times now, Intel is expected to make some decisions and announcements this year regarding how it will approach its CPU design and manufacturing business in the wake of years of 10nm chip delays, and more recent delays to the 7nm roadmap. Ratcheting up that pressure is the recent call for Intel to explore a new strategy in decoupling its CPU design and manufacturing processes from one another.
It’s that same pressure that likely led to the resignation of current CEO Bob Swan, as Intel looks to rectify its ongoing manufacturing crisis and get itself back on the rails, so to speak.
It has already committed its Xe HPG GPUs and certain Atom and Xeon SKUs to third-party fabs -- TSMC, most likely, although Intel hasn’t officially stated that. A (paywalled) report out of Bloomberg states that Intel is in continued talks to both TSMC and Samsung as it further explores its manufacturing options, and Intel itself hasn’t ruled out licensing other fabs’ process technology for use in its own facilities.
While there’s still a lot left to know, a report from TrendForce claims that TSMC has already received orders for Intel’s non-CPU products (i.e., chipsets) and that it expects orders in the future for Intel Core i3 CPUs.
“Intel has outsourced the production of about 15-20% of its non-CPU chips, with most of the wafer starts for these products assigned to TSMC and UMC, according to TrendForce’s latest investigations. While the company is planning to kick off mass production of Core i3 CPUs at TSMC’s 5nm node in 2H21, Intel’s mid-range and high-end CPUs are projected to enter mass production using TSMC’s 3nm node in 2H22,” says TrendForce.
TrendForce also suggests that Intel plans to outsource some of its more entry-level products, while using its own fabs for more complex or higher-margin products, and likely anything built on Intel’s new 10nm SuperFin.
“TrendForce believes that increased outsourcing of its product lines will allow Intel to not only continue its existence as a major IDM, but also maintain in-house production lines for chips with high margins, while more effectively spending CAPEX on advanced R&D. In addition, TSMC offers a diverse range of solutions that Intel can use during product development (e.g., chiplets, CoWoS, InFO, and SoIC).”
Source: http://www.trendforce.com/presscenter/news/20210113-10651.html
21:23 | DisplayPort 2.0 Monitors, GPUs Delayed
If you were expecting any news regarding DisplayPort 2.0 hardware at CES 2021, then you were sorely disappointed like the rest of us. The DisplayPort 2.0 specification was made official back in 2019, with the first DisplayPort 2.0 supported monitors set to come in 2020. Like everything else last year, it seems DisplayPort 2.0 devices became awash in a year lost to health and economic crisis.
Both The Verge and Tom’s Hardware were able to confirm that DisplayPort 2.0 devices have been delayed, primarily due to the fact that various vendors have not been able to come together and test hardware against one another for testing and interoperability purposes.
“In 2020 VESA had no PlugTests, which has slowed the deployment of DisplayPort 2.0. VESA is now planning our next PlugTest for this Spring in Taiwan, so we expect to get this process rolling again,” a VESA spokesperson told The Verge.
These PlugTests are where multiple hardware companies can gather and test various hardware and systems for compatibility and debugging purposes. VESA usually hosts a few every year, but there were no such events in 2020. These PlugTests also play an important role in how VESA certifies hardware/products for compliance with its standards.
“That also kind of proves out the compliance testing so we can certify equipment. So without being able to test a lot of different things and seeing how things work together, it's kind of hard to finish that whole compliance program. So that's been part of the problem,” VESA’s Craig Wiley told Tom’s Hardware.
According to VESA, DisplayPort 2.0 monitors and products are now expected in the second half of 2021.
Source: https://www.theverge.com/2021/1/14/22230654/displayport-2-monitors-delay-vesa-statement
https://www.tomshardware.com/news/where-is-displayport-20
22:47 | Nvidia May Restart CMP Line of GPUs
In response to soaring cryptocurrency rates, many consumers and enthusiasts are concerned that the GPU market is headed for a repeat of the cryptomining bubble and crash of 2018. It’s presently hard enough to get a graphics card at MSRP -- or at all -- without miner demand.
In a conference call with J.P. Morgan, analyst Harlan Sur volleyed several questions at Nvidia's CFO, Colette Kress. One such question was how cryptocurrency mining contributed to the overall strength of the gaming market, and how it might affect Nvidia’s guidance.
“We don’t have visibility on how much of the RTX 30 Series end demand comes from mining. So, we don’t believe it’s a big part of our business today. Gaming demand is very strong, and we think that’s larger than our current supply,” says Kress.
Kress also highlighted that, in relation to “last time,” inventory has already been very lean for several months now, and Nvidia also has a better view into its inventory channels and where that inventory is going. Kress also mentioned that during the last cryptocurrency boom, Nvidia was transitioning from Pascal to Turing, which made it hard to manage availability for both generations of products.
Also, Kress stated that if Nvidia sees enough demand, it could bring back its CMP line of mining-specific cards. These are cards that don’t have any video ports, and may also use salvaged silicon from other wafers.
“So, in summary, if crypto demand begins or if we see a meaningful amount, we can also use that opportunity to restart the CMP product line to address ongoing mining demand,” says Kress.
Editorial: Eric Hamilton & Keegan Gallick