Industry stub

HW News - Corsair SF PSU Recall, Incognito Mode Data Logging, & Intel Afraid of Benchmarks

Posted on June 6, 2020

As ever, hardware news trudges on unabated. This week we have an interesting product recall from Corsair, pertaining to its SF-series of SFX PSUs. PSU recalls tend to be kind of rare in general, even more so when it’s Corsair doing the recalling. There’s also news of Google facing a massive class action lawsuit, allegedly over deceiving Chrome users into thinking that incognito mode equals Google not collecting user data. We also have coverage of Windows 10 getting hardware-accelerated GPU scheduling, the DOD’s agonizing transition to IPv6, and more. 

GN’s deep dives also continue. So far, we have our heavily tuned Ryzen 5 3600 vs. i5-10400, where we look at RAM timings and overclocks, specifically. Then, we’ve been doing some investigative work on why everything is out of stock, and when we can expect a reprieve. For fun, we took the wraps off of part 1 of our Cyberpunk 2077 PC mod project, and we recently explored i5-10600K GPU bottlenecks. Speaking of the i5-10600K, we recently published our written review. It remains unchanged from our initial video review, and is intended to serve those who would prefer to read the content, rather than watch it. 

Lastly, for those near a MicroCenter, we noticed AMD’s Ryzen 7 3700X is currently priced at $260 for an in-store pick up. MicroCenter will slash an additional $20 if you pair it with an eligible motherboard. Let’s get into the news, with the article and video embed below. 

Stock Availability of New CPUs

We’ve been keeping an eye on the availability of new CPUs from both AMD and Intel. AMD’s Ryzen 3 3100 and 3300X came out first, technically, with a release date of May 21. Intel’s release date was May 20. Both were postured to compete with the other, and neither had inventory ready -- especially AMD, which decided to paper launch the Ryzen 3 CPUs.

One of our video editors, Keegan, has had the Ryzen 3 3300X on pre-order since launch because he plans to upgrade his system, and a benefit of working here was that he knew the test results early enough that a pre-order was the same as waiting for a review. The Ryzen 3 3300X and 3100 were pushed back to late May, then to June 16, then pulled forward. We spoke with a major retailer and were simply told that the retailer is waiting on AMD to release the inventory.

As for Intel’s new CPUs, some of the non-K SKUs can be found right now, but i5 and i9 inventory is scarce.

Intel’s New Hot Take On Benchmarks: Forget Them

In a YouTube video addressing the upcoming virtual Computex 2020, Intel CEO Bob Swan presented this hot take on how the PC community should move away from benchmarks, and instead focus on the “benefits” of technology at large. 

“We should see this moment as an opportunity to shift our focus as an industry from benchmarks to the benefits and impacts of the technology we create. The pandemic has underscored the need for technology to be purpose-built so it can meet these evolving business and consumer needs. This is a moment for our industry to come together. Creating technology that enriches lives, that creates value that supports and accelerates positive business and societal benefits should be our collective goal.” said Swan. Swan also said, “And now more than ever, we'll all have an increasing sense of responsibility not just for the products we make but the role we play for the world and our ability to make a difference.”

In the same breath, Swan also said that “Later this summer, we'll introduce Tiger Lake and cement our position as the undisputed leader in mobile computing and PC innovation." 

So, fewer benchmarks in favor of trusting a multi-billion dollar company at its word when it claims undisputed leadership. Okay. GN would like to give the Wheel of Fortune wheel one more spin and solve this puzzle: ironic juxtapositions. One of the points of benchmarking and testing is to determine just how valid a company’s claim on leadership is. Intel, by the way, is the same company that pushed “real-world benchmarks” when it started losing.

It also seems as if Intel is using the current state of events to perhaps echo its past cries to move away from synthetic benchmarks in favor of its non-quantitative “real world” performance claims. Though, in this case, you can substitute “benefits” for “real world.” It goes without saying, but these types of comments paint Intel in a less-than-favorable light, and it more or less reeks of an attempt to take some shine off of AMD. Intel wants to be Apple: Everything is magical and based purely on the emotion that marketing instills in you. 

It’s true that, in regards to the mobile market, it’s Intel’s market to lose, so to speak. However, Intel is working as hard as it can to snatch defeat from the jaws of victory. AMD keeps making strides in that segment; for instance, for the first time, HP is now offering both Intel and AMD options in its OMEN line of gaming laptops. As for Tiger Lake, time -- and testing, not trusting in “benefits” -- will tell the story of just how far Intel has come, and how much longer it can claim the mobile computing crown.  


AMD Has Shipped 553M GPUs Since 2013

In a new report by JPR, it seems AMD has shipped over half a billion GPUs in its Radeon IP portfolio since 2013, including both dGPUs and APUs.

Starting in 2013 with AMD landing semi-custom silicon deals with Microsoft and Sony (for the Xbox One and PlayStation 4, respectively), AMD’s annual GPU shipments topped 87M for that year, and have only gone up since. 

Moving on to 2014, that number more than doubled to 177M shipments. In 2015, AMD saw another impressive jump to 246M shipments, and by 2016, shipments were at 322M. From 2016 onward, it was more of the same upward trend: 398M by 2017, 474M by 2018, and 553M by 2019. Interestingly, Sony’s PlayStation 4 accounts for 20% of AMD’s cumulative Radeon shipments, while Microsoft's Xbox One comes in at 9%. Additionally, notebook APUs and discrete desktop GPUs make up a significant portion of AMD’s shipments.

As usual, Intel takes the lead for most iGPU shipments due to the fact that nearly all of Intel’s CPUs come with an iGPU. Nvidia, for its part, takes the lead for the most dGPU shipments. However, AMD beats both Intel and Nvidia in terms of raw GPU shipments and demand. JPR credits this to AMD’s channel and platform diversity, and also notes that once AMD gets its RDNA 2 IP in Samsung phones, the GPU landscape will look very different.   


Corsair Begins Recalling SF-Series PSUs

Over at Corsair’s forums, the company has posted a notice detailing its voluntary product replacement for its SF-series of SFX form factor PSUs. On top of selling these units to consumers separately, Corsair also uses them in its Corsair One line of compact gaming PCs.

In its notice, Corsair notes that it has noticed an increase in the RMA rates for SF-series PSUs, which prompted an investigation into the returned units. As it turns out, according to Corsair, certain SF-series PSUs are subject to fail if exposed to “both high temperatures and high humidity.”

The issue can apparently manifest either directly out of the box as the unit is powered on for the first time, or at some point down the road, as the PSU is exposed to certain environmental conditions. The fault seems to be on the primary (AC) side of the PSU, and isolated from the secondary side where the PSU delivers power to components.  

The issue could possibly affect any PSU within the lot code ranges of 194448xx to 201148xx, manufactured between October 2019 and March 2020. Corsair states that any PSU made before October 2019 is unaffected. Corsair also says it will try to replace affected customers’ PSUs in advance, where possible, as having to RMA a PSU is certainly a disruptive process. Lot codes can be found either on the PSU’s packaging or on the PSU itself.

If you find yourself in possession of one such affected Corsair PSU, follow the links below.  


Submit a ticket:

$5B Class Action Lawsuit Targets Google’s Incognito Mode

Google is now the target of a new class action lawsuit, this time attracting the ire of the District Court of Northern California, where the complaint was officially filed.

The lawsuit alleges that Google is tracking users and scraping their data -- unbeknownst to the users -- while using incognito mode in Google Chrome. The crux of the lawsuit seems to be on how Google’s incognito mode is defined; or rather, how users perceive it. Support documents for Chrome make it clear incognito mode isn’t exactly private. However, the lawsuit alleges that nowhere is Google upfront about its data collection practices, and that Google is giving users a false sense of control over their data while browsing in incognito mode. 

“Well aware of consumers’ legitimate and reasonable concerns over privacy, Google assured, and continues to assure, its consumers and users that they, and not Google, are ‘in control of what information [they] share with Google.’ Google further represents that ‘across our services, you can adjust our privacy settings to control what we collect and how your information is used.’ Nothing could be further from the truth,” the lawsuit reads. 

The lawsuit outlines that Google uses its many tentacles to scrape user data, including Google Analytics, Google Ad Manager, Google Applications, and the Google Sign In button for certain websites. 

“When an internet user visits a webpage or opens an app that uses such services(over 70% of all online publishers use sucha service), Google receives detailed, personal information such as the user’s IP address (which may provide geographic information), what the user is viewing, what the user last viewed, and details about the user’s hardware. Google takes the data regardless of whether the user actually clicks on a Google-supported advertisement—or even knows of its existence. This means that billions of times a day, Google causes computers around the world to report the real-time internet communications of hundreds of millions of people to Google.”

Interestingly, the lawsuit seems to lean on the Federal Wiretap Act as a means to uphold its claims. The lawsuit further states that Google has infringed upon users’ privacy (really nothing new here) and has intentionally deceived them. The lawsuit is seeking $5,000 in damages per user, for any user that has used Chrome’s incognito mode since June 1, 2016.


Windows 10 May 2020 Update Adds Hardware-Accelerated GPU Scheduling

Microsoft has begun rolling out its May 2020 update for Windows 10, and one of the premier features is the Windows Display Driver Model (WDDM) version 2.7. With WDDM 2.7, Microsoft will deliver on its promise of hardware-accelerated GPU scheduling, a feature it has been previewing for Windows insiders.

In addition to installing the May 2020 update, users will also need new drivers from their respective GPU vendor (AMD, Intel, Nvidia). AMD has said a future driver will add proper support, while Nvidia technically supports WDDM with its game ready drivers, just not the latest version. Nvidia notes that its driver version 450.99 will allow users to toggle the “Hardware-accelerated GPU Scheduling setting” but it may not work. Nvidia also notes that a future game ready driver package will enable full support.

At any rate, hardware-accelerated GPU scheduling should theoretically make room for higher performance and reduced latency, as it should allow the GPU to manage its VRAM directly and without the overhead of the OS. The feature is supposed to be API-agnostic, meaning it will work with DirectX, Vulcan, or OpenGL. We’ll keep an eye on it and see what happens once support has been enabled at the driver level.


Intel Says Goodbye To Coffee Lake-S

It seems like Coffee Lake’s time in the sun is over, as Intel has released a slew of PCNs signaling the EOL for Coffee Lake-S, in addition to some other product lines.

Coffee Lake, (or 8th-gen, or Skylake... 3? if you’re counting) was more iterative 14nm++ silicon that succeeded Kaby Lake-S. However, Coffee Lake was exciting, if for nothing else than it represented Intel’s first hexa-core processors in the mainstream desktop Core series. Coffee Lake-S chips began entering the market in late 2017, with some models dropping in 2018, and properly marked Intel’s response to AMD’s first Zen-based chips and renewing the core wars.  

Also, don’t confuse Coffee Lake-S with Coffee Lake Refresh, the latter of which is 9th-gen and home to 9xxx-series CPUs. Those aren’t going anywhere. The EOL notices affect nearly the entire 8th-gen line up, including Celerons, Pentiums, and the Core series. Xeons seem to be the only exception at the moment. Intel is also sunsetting certain Compute Sticks and NUCs that use 8th-gen chips. 

The EOL was effective beginning June 1st; Intel is further taking orders until December 18th, 2020. Last orders will ship June 4th, 2021. 

Source: PCN(s): 117617 - 00, 117618 - 00

Secondary source:

The DOD Is Still Struggling To Transition To IPv6 After 17 Years

Since 2003, the Department of Defense has been trying to move over to IPv6, and has botched at least two of those attempts. If recent news is any indication, its third attempt doesn’t appear to instill confidence that it will end any differently.

As summarized by the US Government Accountability Office, the DOD’s most recent attempt at a migration to IPv6 began in April of 2017. The DOD’s past attempts at moving to IPv6 were cut short largely because of security concerns, as well as a lack of personnel trained in IPv6. However, in an audit of the DOD’s current IPv6 plan, the GAO found that the DOD has failed to address three out of four critical steps in developing an IPv6 strategy -- which means this attempt isn’t likely to go any better than the rest of them, assuming nothing changes. 

The Office of Management and Budget (OMB) has previously required all federal agencies to plan for a transition to IPv6, and as such, has outlined a few basic requirements in establishing a plan. The OMB has set forth these planning requirements: 

  • Assign an official to lead and coordinate agency planning 
  • Complete an inventory of existing IP compliant devices and technologies
  • Develop a cost estimate
  • Develop a risk analysis

Out of those four, the DOD has only managed to complete one: assign an official lead to coordinate the IPv6 transition planning. Furthermore, in February of 2019, the DOD released its own plan for migrating to IPv6 that consisted of 35 transition steps -- 18 of which were supposed to be completed by March 2020. To date, the DOD has completed only 6 of those 18 steps. The GAO notes that without an inventory of IP compliant devices, an estimated cost, and an estimated risk analysis, the DOD is significantly reducing the probability of a successful migration schedule. 

Unsurprisingly, after its audit, the GAO is recommending that the DOD pretty much do what it was already supposed to have done per the OMB’s requirements. That is, develop an inventory of IP compliant devices (IPv4 and IPv6), an estimation of the costs to migrate its infrastructure to IPv6, and detail possible risks. However, it seems that the DOD doesn’t agree with the recommendation to complete an inventory of IP compliant devices, but has seemingly agreed to the other two.  

While the internet at large is still transitioning to IPv6, the DOD’s transition seems to be going agonizingly slow. In the meantime, IPv4 and IPv6 operate in parallel, though they won’t forever. And at this rate, the DOD may very well run out of time, because it doesn’t have another 17 years.