Watch our full video on our GTC coverage: https://www.youtube.com/watch?v=v99ZOjd90XU
Also our full coverage of the next generation of AMD APUs: https://www.youtube.com/watch?v=6gok9J_kbA8&t
02:17 | MSI, Great Leaker, Leaks 3080 Ti
If MSI existed in some ancient mythology, it’d be the revered “MSI, Great Leaker.” You’d offer up a sacrifice in exchange for information. It could be a Magic card, at that point: Legendary Creature - Dragon. Sacrifice one Ethics token and scry 2. Use this ability any time you could accidentally leak your supplier’s plans for the next quarter.
MSI shipped a box, as it often does, emblazoned with the name of the product within: A crate of RTX 3080 Ti Ventus 12GB cards was spotted by user “Lok Lok” on Facebook, and this confirms a rumor we’d been told by a board partner (and previously published) about the memory capacity.
NVIDIA was originally planning a 20GB model of the RTX 3080 Ti, but internal debate eventually led to a more reasonable 12GB SKU. The 20GB model was a pure marketing battle against AMD’s 16GB SKUs and would have offered no benefit to gamers, but would have potentially undercut RTX 3090 24GB sales that target 3D artists. That’s where we land on the 12GB model that MSI accidentally showed.
A post by Hot Hardware suggests a $1000 price-point for the RTX 3080 Ti, which would place it equal to the 6900 XT and about $300 over MSRP for the RTX 3080 (or $500 below the RTX 3090).
04:20 | Atari Likes the Stock: New Crypto Business
Don’t look now, but the desecrated Atari corpse that has been shambling about for the last few decades is now about to be halved -- or duplicated, rather. The executive necromancers atop Atari’s corporate ladder, not satisfied with how much of Atari’s history and legacy has already been stained, have decided to create a new business unit solely focused on blockchain.
Atari announced that, in addition to some top of the food chain leadership changes, it will also split its business into two separate divisions: Atari Gaming and Atari Blockchain. Regarding the change in leadership, Atari announced that effective April 6, 2021, Wade J. Rosen will become CEO of the entire Atari Group, while Frédéric Chesnais, former Atari CEO, will focus on Atari Blockchain and licensing.
Atari Gaming will continue to extract -- and exploit -- any meaningful value left in Atari’s brands and IPs, for better or worse. Atari will also continue to develop free to play mobile games and expand upon its licensing model -- both in sectors gaming and non-gaming. For instance, Atari is still pushing its idea for Atari branded hotels and casinos. Atari also claims to be committed to the Atari VCS (which was actually a lot better than we thought), and developing an entire ecosystem around it, consisting of third-party and first-party content.
Atari Blockchain, as you may have guessed, will focus on all things blockchain and cryptocurrency related -- such as Atari’s cringey Atari Token. Additionally, Atari seems to be foaming at the mouth to get further into the NFT (Non-fungible Token) business, and it seems this division will have strong emphasis on that. Atari is like that friend that constantly jumps from one get-rich-quick scheme to the next. Once Atari is interested in something, the ride is probably already over.
Earlier this year, we took a look at the embarrassingly delayed Atari VCS console and noted that, at least under the hood, the unit exceeded our expectations. We also looked closer at the software and OS environment, as well as gaming performance in some further coverage.
07:31 | GPU Warranties Tanking
PC Gamer ran a story this past week about GPU warranties, catching that MSI and Gigabyte have both reduced warranty periods for GPUs in the current mining-laden market. Gigabyte’s new $600 mining GPU has a warranty of just 90 days, although our understanding is that the EU will require a 2-year period. We’ve typically seen warranties in the 2-3 year range, often at least a 2 year parts warranty.
It is feasible that a future could develop where on-card chips are used to track power load over the lifespan of the device. We spoke with a manufacturer about this a few years ago (and won’t name them), when the company expressed interest in using a sort of “Car Facts” approach to GPUs, allowing them to selectively ignore the warranty if the card was deemed to be used out of reason. Of course, this quickly deteriorates into a slippery slope argument of what’s defined as “reasonable.”
10:27 | Recap: Nvidia GTC 2021, AMD 5000G APUs
Both Nvidia and AMD had some headlining news this week, which we’ve already covered over at our YouTube channel.
Nvidia’s GTC 2021 was all-digital this year, unsurprisingly, and featured the traditional keynote from Nvidia CEO and immaculate kitchen owner Jensen Huang. Nvidia’s GTC usually highlights all of Nvidia’s business segments, which have certainly grown over the years. This year, Nvidia discussed graphics, AI, networking, data center and servers, and automotive. Dotted throughout the keynote were various product announcements, with key emphasis on Nvidia’s first data center CPU design. We had our hopes up that Jensen would pull out an Ampere powered, 64-core cake whisk from out of his oven; alas, it was not meant to be. Maybe next year.
That said, Nvidia did announce its first data center, Arm-based CPU, dubbed Grace. Also announced was the Alps supercomputer, a new DGX SuperPOD, the new Bluefield 3 DPU, and a slew of new GPUs aimed at the data center and workstations. These GPUs include the A5000 and A4000 line, as well as the A10 and A16 GPUs. Check out our video for more details.
AMD also trotted out its Ryzen 5000G line of APUs, a line of Zen 3 CPUs with on-board Vega 8 graphics. The new APUs are aimed at addressing entry-level to mid-range systems, and will land at OEMs for pre-built systems first, before becoming widely available to the DIY market later this year.
The line consists of the Ryzen 7 5700G, Ryzen 7 5700GE, Ryzen 5 5600G, Ryzen 5 5600GE, Ryzen 3 5300G, and the Ryzen 3 5300GE. All parts are underpinned by the Zen 3 architecture, and built on TSMC’s N7 7nm process. The GE suffixes are to identify the lower frequency/TDP variants.
Source: Nvidia -- https://www.youtube.com/watch?v=v99ZOjd90XU
AMD -- https://www.youtube.com/watch?v=6gok9J_kbA8
13:40 | Intel CEO Pat Gelsinger’s Comments on Nvidia
Following the summit held at the White House regarding the global chip shortage, Fortune was able to talk with Intel CEO Pat Gelsinger about a few different things. The overall interview is rather short, but worth a read. The topics included the US Government’s proposed $50B plan to bolster the US semiconductor industry, as well as a discussion on the current state of US-based silicon manufacturing.
Gelsinger also offered a bit of commentary on Nvidia’s product announcements at GTC 2021, with a particular emphasis on Nvidia’s Grace CPU, which is aimed at data center applications. Fortune inquired about Gelsinger’s thoughts on Intel’s competitive position with Nvidia, as well as accelerating the use of AI within CPUs.
“We announced our Ice Lake last week with an extraordinarily positive response. And in Ice Lake, we have extraordinary expansions in the A.I. capabilities. [Nvidia is] responding to us. It's not us responding to them. Clearly this idea of CPUs that are A.I.-enhanced is the domain where Intel is a dramatic leader,” Gelsinger told Fortune.
Gelsinger doubled down, reminding us that Intel acquired Habana Labs, a silicon designer specializing in AI, back in 2019 with the express intent of accelerating AI in the data center. Intel also announced a massive partnership with AWS last year, with the goal of accelerating deep learning, AI, and HPC in cloud computing.
“So clearly, I'd say the idea of CPUs is Intel's provenance. We're now building A.I. into that and we expect this to be an area where we are on the offense, not the defense going forward,” Gelsinger said towards the end of the interview.
15:57 | Intel Looking to Build Automotive Silicon
Over the last few weeks, Intel has announced a number of moves that are aimed at getting the company back on track, and on the surface, a lot of these moves seem to be steps in the right direction. One of the most critical and biggest changes Intel announced was the creation of Intel Foundry Services, whereby Intel will manufacture silicon for external customers.
We’ve noted previously that Intel’s IFS could be a watershed moment for the company, and it’s in a unique position relative to other foundries. Intel not only possesses the means to manufacture silicon, but obviously is capable of designing silicon, as well as licensing its existing catalog of x86 IP. This puts Intel in a position where it can not only leverage the foundry model, but could also emulate some degree of Arm’s licensing model success.
It’s obviously a bit early to know where exactly Intel is going to steer IFS, but it’s not hard to imagine the possibilities with IFS. That said, Intel recently told Reuters that it has held discussions on building specialized silicon for the automotive industry. Additionally, Intel has expressed an interest in stepping in to try and mitigate some of the symptoms caused by the semiconductor shortage, which the automotive industry has been feeling acutely.
“We're hoping that some of these things can be alleviated, not requiring a three- or four-year factory build, but maybe six months of new products being certified on some of our existing processes. We've begun those engagements already with some of the key components suppliers,” Intel’s Pat Gelsigner told Reuters.
Automotive silicon absolutely makes sense for Intel. Intel’s IFS needs customers, and the automotive industry needs another source for its silicon needs. Moreover, as Intel has more granular control over its supply chains and logistics, it hasn’t been affected to the same degree as others have in regards to component shortages, such as substrates.
Gelsinger told Reuters that its discussing the matter with a number of companies that design automotive silicon, and hopes to produce chips within six to nine months.
18:00 | Epic’s Battle Against Steam is Expensive
Epic Games and its respective online storefront, Epic Games Store, has long been on a crusade to claw away market share from market incumbent Valve. Epic Games has implemented a number of somewhat atypical business practices for its online store, all while unabashedly pointing a critical finger at Steam.
Such practices include taking a 12% revenue cut on third-party games distributed through Epic Games Store (Steam and other stores have historically taken 30%), waiving the 5% royalty fee for Unreal Engine for monetized games until a game accrues $1M in revenue, and advance payouts of “minimum guarantees.”
On top of trying to crack Steam’s grip on PC gaming, Epic is also deeply entangled in a bitter lawsuit with Apple over Apple’s iOS store. That’s an entirely different story, which we've discussed before, but the ongoing lawsuit has produced no shortage of financial filings from both companies. It’s from these filings that we are able to get a picture of just how much money Epic is dumping into its Epic Games Store.
According to court documents and testimonies, Epic has shown year-over-year losses for the Epic Games Store since its inception in 2018. For 2019, Epic noted a $181M loss, while it noted a loss of $273M in 2020 and a projected loss of $139M in 2021. Furthermore, According to Apple’s legal math, Epic has spent $444M on minimum guarantees in 2020 alone. According to Epic’s own year in review for Epic Games Store, users spent a reported $700M on the platform in 2020, but only $265M of that figure represented third-party game sales. In that same report, Epic also noted that there were over 749 million copies of free games claimed in 2020.
Epic maintains that its 12% cut has allowed it to cover the costs of distribution, as well as inject more investments into the store. Also, Epic still claims that at some point, that revenue cut will more than sustain the storefront. It seems most of Epic’s losses are coming from advance payouts to developers and the cost of securing free games. Epic has spent an obscene amount of money in order to attract developers and secure timed exclusives for the Epic Games Store. For instance, it’s well known at this point that Epic paid more than $10M in upfront payments for the exclusive distribution of Control.
For any other company trying to take on Steam, this wouldn’t be feasible or sustainable. However, to put it simply: Epic can afford it. Epic could likely continue pumping money into its games store just off its Fortnite profits, although the company isn’t hurting for investment money, either. Recently, Epic Games completed a round of funding that saw the company raise $1 billion in new investments -- Sony accounted for a cool $200M of that.
21:50 | Nvidia Comments on Supply Issues
Nvidia had some disheartening, if unsurprising comments on the GPU supply issues permeating the market. The comments were brief, and came at its Annual Investor Day, which also coincided with the kick off of Nvidia’s GTC 2021 conferences. Nvidia’s CFO Colette Kress addressed investors and outlined how Nvidia’s Q1’2022 revenue is tracking above its previous outlook.
Nvidia had previously projected a Q2’2022 fiscal revenue of $5.30B, plus or minus 2%. According to Nvidia, the company is on track to punch above that, thanks to strong demand from all of its business sectors, and in no small part to the insatiable demand for its graphics products.
Regarding supply constraints, Nvidia doesn't expect much to change this year. “Overall demand remains very strong and continues to exceed supply while our channel inventories remain quite lean. We expect demand to continue to exceed supply for much of this year. We believe we will have sufficient supply to support sequential growth beyond Q1,” said Kress.
22:45 | AMD RX 580 RMA Scam
If smuggling wasn’t enough, now it seems consumers have to deal with scams. Recent reports are warning about scammers who are posing as AMD China or XFX and sending out a fake letter claiming a recall for the RX 580. The bogus recall lists PCB and manufacturing defects as the reason for the recall, and offers users a GTX 1050 Ti or GTX 1060 3GB as replacement for trading in their RX 580.
Of course, AMD and XFX had something to say about this. AMD China released an official statement via Weibo, and XFX put out a statement on Bilibili -- both popular Chinese social media platforms. The statements warn users to disregard any letters or emails claiming a recall, and state that the brands will be taking legal action against the scammers.
AMD’s Radeon RX 580 has proven to be quite popular among Ethereum miners, and as the card has long since been discontinued, would-be scalpers have resorted to fraud.
23:49 | Gigabyte Makes CMP 30HX Mining Card Official
Also in mining news is one of the first officially official Nvidia CMP 30HX SKUs. We’ve seen some SKUs pop up overseas, and not long ago we saw leaked images of Gigabyte’s CMP 30HX card. Now, Gigabyte has made the CMP 30HX D6 6G card official, and according to listings found by -- wait for it -- @momomo_us, could go for over $700.
On the surface, the D6 6G doesn’t appear to be drastically different from Gigabyte’s GeForce GTX 1660 Super Gaming SKU, sans the lack of display outputs. Both are powered by Nvidia's TU116 silicon and use similar components, as well as Gigabyte’s WindForce cooler. The biggest difference, pricing aside, is the CMP 30HX D6 6G will only come with a three-month warranty. That’s opposed to the more standard warranty of three years for graphics cards.
Gigabyte is the first vendor to officially list one of its Nvidia CMP SKUs, and we expect more vendors to follow suit before too long. Pricing is still a big question mark, and if these low-end CMP SKUs aren't priced right, they’ll do nothing to detract miners from gaming-oriented SKUs.
24:55 | TSMC Power Outage
Reuters and Digitimes reported that TSMC suffered a power outage at one of its facilities, specifically Fab-14A P7. According to reports, the power outage seemed to be brief, and the fab was able to run on generator power for a time. However, it seems TSMC still had a disruption in production, having to pause manufacturing on older process nodes.
Digitimes notes that the disruption impacted up to 40,000 wafers, and could cost TSMC as much as $35 million, all told. At the time of writing, TSMC is still assessing the impact the outage has had, and we’re sure the company will release a statement once its content with its findings.
According to the reports, the outage was the result of an underground transmission line being severed during construction work in the surrounding area. TSMC’s Fab 14 is one of the company’s eight Gigafabs producing 12” wafers, and the fab mostly deals with processes nodes between 40nm and 90nm.
GeForce Experience 3.22, Game Ready Driver Update
Nvidia released its latest GeForce Experience and game ready driver updates, which is one of its more interesting and feature-heavy releases in recent memory.
The GeForce Experience 3.22 update adds a number of features, such as one click GPU overclocking in the new, freshly out of beta performance panel. For RTX 20 and 30-series GPUs, the performance panel offers a few different ways to tune and monitor the GPU, including the one-click overclock.
According to Nvidia, “there’s a new automatic tuner which finds the best overclock settings using an advanced scanning algorithm, and manages your GPU tuning profile for you.”
The GeForce Experience 3.22 update also lists changes and enhancements to Nvidia’s Reflex Latency Analyzer, including the ability to log all performance data to a CSV file. There’s also the usual bug fixes and stability improvements.
Additionally, the latest GeForce Game Ready 466.11 WHQL driver is highlighting that Open Broadcaster Software (OBS) is now natively supporting Nvidia’s Broadcast Noise Removal, starting with OBS Studio 27.
Additionally, Nvidia notes that with this driver release it is introducing improvements to how Reflex works in CPU-bound scenarios.
“In each NVIDIA Reflex title, users can optionally enable “Boost Mode” to improve Reflex’s effectiveness in CPU-bound scenarios, increasing GPU clock speeds, allowing the GPU to submit frames to the display faster and further reduce latency,” says Nvidia.
Editorial: Eric Hamilton
Host, Additional Writing: Steve Burke
Video: Keegan Gallick, Andrew Coleman