HW News - ASUS + Noctua GPU, Dual-GPU AMD Card, Linux Grows Among Gamers
Posted on August 13, 2021
This week, we have news on AMD and Valve striking a partnership that hopefully leads to better CPU drivers for Linux, as both companies will be working together to optimize the playing experience on the Steam Deck. Also in AMD news is the company's updated Radeon Pro W6000X-Series of GPUs, which probably aren’t that exciting if you’re not a Mac Pro user.
Moving on, we also have updates on SK Hynix’s purchase of Intel’s NAND flash business, as well as developments regarding the Nvidia-Arm deal. There’s also some Windows 11 news as it relates to TPM 2.0 supported motherboards, a new Steam survey to go over, Sony’s PS5 becoming profitable, and more.
News article and video embed below, as usual.
The first section in the video talks about our Gigabyte exploding power supply testing -- we'll leave that to the video.
AMD and Valve Optimizing Steam Deck for Linux
According to Phoronix, AMD and Valve are working together to further optimize the AMD APU inside the Steam Deck for Linux. And while this is good news to hear, it isn’t really a surprise, given that both AMD and Valve have expressed more interest in taking Linux more seriously. Valve has been pouring resources into its compatibility layer, Proton, while AMD has been on a hiring spree for Linux engineers this summer.
And of course, both parties also have a vested interest in making the Steam Deck as successful as possible. One key area of focus seems to be improving the CPU frequency and power scaling while using Steam Play (Proton). AMD and Valve’s combined efforts could lead to improved CPUFreq driver code based on the ACPI CPPC (Collaborative Processor Performance Control) specification.
With AMD also beefing up development around its Linux scheduler, it’s also possible that AMD could be overhauling the Schedutil governor to better target CPU frequency scaling. As Phoronix notes, AMD will have a presence at this year’s X.Org Developer Conference (XDC), so we can expect to know more then.
Source: https://www.phoronix.com/scan.php?page=news_item&px=AMD-Valve-New-CPU-Freq
ASUS + Noctua File Plans for Collaboration
Fan renderings of a GPU with a distinctive brown-and-beige cooler have been appearing in the news -- despite pre-dating this story by a couple of years. That’s because of an ECC filing spotted by Twitter sleuth KOMACHI_ENSAKA. The filing was listed by Asus and contains a list of card names, including “RTX3070-8G-NOCTUA,” with the assumption being that we could see an Asus GPU featuring Noctua fans in the future.
This could be an interesting move on Asus’ part to play on Noctuas reputation in the community. It is also worth noting that Asus is capable of making (or sourcing) perfectly good fans for their GPUs, when we looked at some partner 30-series cards in 2020, the Asus 3080 TUF and 3070 Strix were among the top performers in noise normalized testing. Further, while Noctua makes very good fans, they aren’t magic and still function like, well, a fan. Having said that, if it comes to fruition, this is a partner card we’d definitely be interested in reviewing.
AMD Announces Dual-GPU Card for $5,000
AMD took the wraps off of three new workstation class GPUs that are built exclusively for Apple’s Mac Pro line. The three new cards make up AMD’s Radeon Pro W6000X-series and consist of the Radeon Pro W6900X, Radeon Pro W6800X, and Radeon Pro W6800X Duo. The new lineup will be a successor to AMD’s current W5000X cards, which are based on RDNA 1.0 and Navi 10 silicon.
AMD’s Radeon Pro W6000X-series will be based on Navi 21 silicon and RDNA 2.0, and will also include AMD’s newer Infinity Cache and Infinity Fabric Link. As these cards are built for Apple and the Mac Pro line, they’ll come in the form of Apple’s MPX modules and are also built with Apple’s Metal API in mind.
The W6900X is built with 80 compute units, 5120 stream processors, a 256-bit memory bus, 128MB of Infinity Cache, 32GB of GDDR6, and a TGP of 300W. Meanwhile, the W6800X is built with 60 compute units, 3840 stream processors, a 256-bit memory bus, 128MB of Infinity Cache, 32GB of GDDR6, and a TGP of 300W.
The W6800X Duo is a dual-GPU (non-MCM) configuration, built with two W6800X GPUs on the same PCB. The two GPUs are linked together through AMD’s Infinity Fabric interconnect, and so the card’s compute units, stream processors, Infinity Cache and VRAM are all doubled, and the card has a TGP of 400W; however, this isn’t a true doubling and will behave differently than a theoretical monolithic GPU of similar CU count (although this is not possible). Apple’s Mac Pro supports up to two MPX modules, meaning users could opt for a quad-GPU array via two W6800X Duo modules -- should they want to spend $10,000 for the privilege.
Steam: RX 6000 Doesn’t Exist, Linux Growing
Another month has come and gone, which means we can pore over another round of Steam’s Hardware and Software survey, this time for July 2021.
There are a couple of highlights worth bringing up, the most interesting of which is that Linux has finally reached a 1% share in the OS market -- within the vacuum of Steam’s collected data on its user base, anyway. Keep in mind that Linux’s server marketshare is domineering.
Steam’s numbers show Linux up by +0.11 points over last month, with both Ubuntu and Manjaro Linux seeing a very slight (+0.01 point) uptick in interest. This is the highest share Linux has had in years, and Valve’s Steam Deck and Proton are likely playing some part in the renewed interest.
Also of note is that this is yet another month where AMD’s Radeon RX 6000 cards have yet to surface on the charts. AMD’s most popular card on the charts is the RX 580, although its share continues to shrink at 1.73% (-0.02 points). Nvidia’s GTX 1060 is still the most popular card among Steam users, but again, its share has also been decreasing at 9.17% (-0.71 points month-over-month). Meanwhile, the fastest moving cards up the charts are the RTX 3060 (+0.13 points) and RTX 3060 laptop GPU (+0.15 points).
Another trend we’ve been watching is the decline in share for quad-core CPUs, which lost another -0.25 points in share last month, while octa-core CPUs went up by +0.51. Elsewhere, 16GB is still the preferred amount of system RAM at 45.53% (-1.55), though configurations beyond 16GB have grown by +0.40 since last month.
Source: https://store.steampowered.com/hwsurvey/Steam-Hardware-Software-Survey-Welcome-to-Steam
The UK is Still Debating Blocking The Nvidia-Arm Deal
As Nvidia’s $40B acquisition of Arm continues to wind its way through global regulatory processes, the UK is becoming an increasingly higher hurdle to clear. Last April, the UK officially stepped in to intervene on the deal, citing national security concerns. These concerns were of course on top of the antitrust and monopoly concerns.
At the time, The UK's Competition and Markets Authority (CMA) was ordered to build a report detailing the possible national security implications that an Nvidia takeover would have, and that report was completed and submitted as of July 20, 2021, according to an updated merger inquiry. And while the details of the report haven’t been disclosed, they don’t appear to be in Nvidia’s favor.
According to Bloomberg, sources familiar with the matter have stated that the implications are worrying, and that the UK is inclined to reject the deal. However, Bloomberg also stated that the UK government was likely to delve deeper into the merger and further review the findings of the report before coming to a final decision.
Nvidia continues to remain confident -- perhaps arrogantly so -- that the deal will go through, which is the tone the company has maintained since the deal was announced. Both Nvidia and Arm’s CEOs have largely been tonedeaf to criticism surrounding the deal, and at least on the surface, don’t appear to be overly worried about these kinds of snags.
“We continue to work through the regulatory process with the U.K. government,” said an Nvidia spokesperson in a statement to Bloomberg. “We look forward to their questions and expect to resolve any issues they may have.”
Primary Source: https://www.bloomberg.com/news/articles/2021-08-03/u-k-considers-blocking-nvidia-s-takeover-of-arm-over-security
Secondary Source: https://www.reuters.com/technology/uk-considers-blocking-nvidias-40-bln-takeover-deal-arm-bloomberg-news-2021-08-03/
Updated merger inquiry: https://www.gov.uk/cma-cases/nvidia-slash-arm-merger-inquiry#history
Sony Reports $500 PS5 Is Now Profitable
In its latest earnings report, Sony’s CFO Hiroki Totoki revealed the disc-based, $500 PS5 model has become profitable. As we’ve said before, consoles and hardware have been historically used as a loss leader, being sold at a loss for the sake of market share and growing the install base while collecting publisher fees for games sold and service fees (like online).
However, the machines usually cross a threshold where they become profitable, for Sony at least, and it seems Sony has crossed that line with the more expensive $500 PS5. For reference, Sony’s PS4 didn’t become profitable until around 6 months after launch, so the timeline here is similar. While the cheaper All Digital PS5 isn’t yet profitable, Sony notes that those losses continue to be offset by software sales and steady sales.
Sony also stated that the PS5 is now the fastest selling PlayStation in history, having already sold more than 10M units since it launched last November. Sony also stated that it remains committed to selling more than 14.8M PS5 consoles for its full fiscal year, which would match the amount of PS4 machines sold in its first year.
Source: https://www.theverge.com/2021/8/4/22609150/sony-playstation-5-ps5-loss-profit
https://www.sony.com/en/SonyInfo/IR/library/presen/er/
Intel Accidently Teases Thunderbolt 5 With 80 Gbps Bandwidth and PAM-3 Modulation
Gregory Bryant, Intel’s EVP and GM for its Client Computing Group, has been on a tour of sorts, visiting Intel’s Israel site, where a lot of its most important research and development happen. That in itself isn’t news. However, as spotted by AnandTech’s Dr. Ian Cutress, some of the photos that Bryant tweeted -- and has since deleted -- are news.
The deleted photo showed off what appear to be early Thunderbolt 5 specifications. The photo can be seen over at AnandTech, where a poster in the background shows “80G PHY Technology.” That in itself suggests that Intel is working on a physical layer capable of supporting 80 Gbps connections, which would be double the bandwidth of Thunderbolt 4.
Elsewhere in the photo, AnandTech noticed the lines “USB 80G is targeted to support the existing USB-C ecosystem” and “The PHY will be based on novel PAM-3 modulation technology.” This suggests that Intel is intent on keeping the USB-C connector, and that PAM-3 will play a significant part in the increased bandwidth of Thunderbolt 5.
SK Hynix Will Spin Intel’s NAND Business Acquisition Into a Separate Company
Late last year, Intel announced that it would be selling off its NAND flash business to SK Hynix, for a total of $9B. The deal came with some caveats, like Intel retaining its Optane technology and IP, as well as a staggered timeline for the acquisition. Like we said at the time, the deal indicated that Intel wasn’t necessarily exiting the storage business (because it’s not), but it is exiting the NAND flash manufacturing business.
The deal included Intel’s NAND flash SSDs, flash and wafer business, as well as its NAND flash fab in China. At the time, SK Hynix didn’t disclose its immediate plans for the acquisition, but now, it’s been revealed that it will spin Intel’s NAND business into a separate company. The unnamed company will be headed up by Robert Crooke, senior vice president and general manager of Intel’s NAND Products and Solutions Group. Crooke took to LinkIn to announce that he would become the CEO of the newly formed company.
“On October 19, 2020, SK hynix announced the acquisition of Intel’s NAND memory and storage business. I am excited to share that we are currently in business development and once all the jurisdictions approve the deal, we will stand up a new global company, headquartered in the US, and owned by SK hynix,” Crooke wrote. “I am honored to be the CEO of this company. Stay tuned for our new company name, I’ll share it here!”
We already knew that SK Hynix would likely use Intel’s NAND business to give itself more of a foothold in the SSD market, and according to CRN’s report, the acquisition will also give SK Hynix a bigger presence in American markets and with SIs. SK Hynix notes that the deal has already been approved by the U.K., the U.S., and the European Union. SK Hynix expects to get approval from China later this year.
Source: https://www.crn.com/news/storage/sk-hynix-to-turn-intel-nand-business-into-stand-alone-us-company
Asus Preps Motherboards For Windows 11
Asus seems to be among the first motherboard vendors getting ready to push firmware out the door that will prep certain motherboards for Windows 11 support. Asus currently has a Windows 11 microsite where it lists compatible motherboards with TPM 2.0 support and their respective BIOS updates, some of which are still under testing.
Updating to the latest BIOS should enable TPM 2.0 support, without users having to trek through ASUS’ UEFI BIOS utility. It seems for now, ASUS boards with Intel 300, 400, and 500-series chipsets, as well as X299, are good to go. For AMD, boards based on 300, 400, and 500-series chipsets are supported. AMD’s TRX40 and WRX80 motherboards should also be ready, according to the list.
Source: https://hothardware.com/news/asus-pushes-first-wave-of-windows-11-bios-updates
https://www.asus.com/microsite/motherboard/ASUS-motherboards-Win11-ready/
Corsair Expects DDR5 Will Require Improved Cooling
In a brief video hosted on Corsair’s YouTube channel, Corsair’s Memory Product Manager, Matt Woithe and DIY Marketing Director, George Makris, discussed DRR5 memory a bit. Nothing overly technical, mind you, but perhaps a glimpse into what Corsair’s engineering department has in mind for DDR5.
In the video, Woithe and Makris discuss a few different talking points regarding DDR5, and DDR memory/DRAM in general. In a way, they sort of reveal that Corsair expects heat spreaders to play an increasingly important role with DDR5, especially as DDR5 DIMMs will make use of local power management by way of an on-die PMIC, which could theoretically lead to some heat waste being expended into the memory.
"DDR5 could conceivably run much hotter than DDR4. They've moved voltage regulation off the motherboard itself, and now it's on the chip. So you actually could be pumping a lot more heat into DDR5 modules," says Makris during the video. The video goes on to discuss Corsair's DHX (Dual-Heat Path Exchange) cooling solution, and how it might relate to Corsair’s DDR5 products.
To be fair, this likely is a non-issue for DDR5 DIMMs running at the JEDEC specified 1.1V. However, some vendors are already talking about DIMMs capable of 126000 MT/s with voltage ranges between 1.1V and 1.6V.
Though DDR5 has officially landed, and memory manufacturers are ramping into volume production, it’ll still be quite a while before consumers have any real need for it. Intel’s Alder Lake will have the honors of claiming “first” when it comes to DDR5 support for client machines when the platform arrives late this year, while AMD will get onboard next year with Zen 4.
Source: https://www.youtube.com/watch?v=DsTu_7iF0JI
https://www.tomshardware.com/news/corsair-says-ddr5-modules-to-run-hotter-than-ddr4
Editorial: Eric Hamilton
Host, Additional Writing: Steve Burke
Video: Keegan Gallick