00:58 | Disappointment PC Behind-the-Scenes
Check the news video for this discussion!
05:09 | NVIDIA is Embarrassed of 3050 & 3090 Ti(e)
AMD, Intel, and NVIDIA all ran keynotes coinciding with CES 2022, finally putting out the roadmap for new silicon launches. We’ve already separately covered Intel and AMD’s roadmaps, but we weren’t prebriefed for NVIDIA’s, so we watched the stream with everyone else.
Our main takeaway is simply that NVIDIA is embarrassed of its new RTX 3050 and RTX 3090 Ti. Collectively, the two GPUs got under one minute of total airtime in the 30-40 minute presentation. Instead, NVIDIA breezed past them and spent the rest of its time acting less like a manufacturer and more like a broadcaster: It sold itself to as many partners as it could, not going more than a minute or two before mentioning a game publisher, hardware manufacturer, or boring third-party that is using their GPUs that definitely exist.
NVIDIA spent a lot of time coping right at the beginning, injecting copium into the veins of its executives as they talked about record-breaking GPU sales and all-time high sell-through rate of inventory. That’s all accurate -- JPR and other third parties have shown as much -- but attempting to slyly tell people that a problem doesn’t exist isn’t going to get a good response.
Anyway, the presentation was a waste of our time, so we’ve sacrificed so that you don’t have to. Don’t even bother opening it. The key information is this:
NVIDIA announced an RTX 3050, due out on January 27th (following the RX 6500 XT that’s due on January 19th) and for $250. It’s an 8GB GDDR6 card. NVIDIA did not detail information beyond some mostly useless “TFLOP” numbers.
As for the RTX 3090 Ti, NVIDIA literally spent under 15 seconds on the card. The card will have 24GB of 24Gbps GDDR6X memory. It’s likely that the increased density rumors are true and that NVIDIA will be able to reduce down to a one-sided PCB for memory once again. That’s all we know about the RTX 3090 Ti.
Source: NVIDIA Presentation
09:56 | Rumor: EVGA RTX 3090 Ti Kingpin
Specifically regarding EVGA’s rumored Kingpin variant, Twitter leaker @harukaze5719 claims EVGA will be using an entirely custom PCB and a redesigned cooling shroud. However, it seems the Asetek-based 360mm AIO will still ship with the card, as it did with the RTX 3090 Kingpin. Also, it seems the RTX 3090 Kingpin may enter EOL status very soon, especially if there’s a new SKU in the works. Additionally, due to the PCB and shroud changes, it seems EVGA’s Hydro Copper water blocks will not fit the new RTX 3090 Ti Kingpin.
Furthermore, EVGA would eschew the PCIe 8-pin connectors and opt for two (2x) of Nvidia’s 12-pin (or 16-pin, if equipped with the four signal contacts) connectors. According to the rumor, EVGA is currently tweaking the vBIOS, and may be looking at a March 2022 launch. Separately, Nvidia has been rumored to be looking at a January launch for the RTX 3090 Ti, following a rumored launch for the RTX 3080 12GB and RTX 3070 Ti variants. These announcements might happen while this video is getting edited.
11:11 | Asus Shows Off DDR5 to DDR4 Converter Card
It’ll be a while before DDR5 displaces DDR4 or reaches parity with the fastest DDR4 kits on the market. It’ll also take time for DDR5 pricing to level-out and make sense for consumers.
Intel is the only CPU maker currently supporting DDR5 with its latest Alder Lake line and the Z690 chipset-based motherboards, but AMD just formally announced support on AM5 platforms with Zen4 and is also supporting LPDDR5 on Ryzen 6000 notebooks. Those looking to build an Alder Lake system are then faced with the choice of buying a DDR5-based board and likely spending a small fortune for DDR5, or pairing an Alder Lake CPU with a DDR4 board and being stuck on a DDR4 only system.
Asus is working on a solution to this problem; although, depending on the cost of the solution, it may just end up being a demo.
As spotted by AnandTech, a YouTuber named Bing (presumably an engineer for Asus) has shown off a DDR5 to DDR4 converter card prototype. The concept is simple enough: The card would allow for DDR4 modules to be slotted into the DDR5 DIMM slots via the converter card, and the card – with the help of a special BIOS revision from Asus – would allow the modules to run in DDR4 mode.
There’s plenty of challenges associated with such a product. Aside from a different notching and keying on the DIMMs themselves, DDR5 modules also make use of their own on-board PMICs for power management and voltage regulation, whereas DDR4 modules rely on the motherboard for power management. The converter card and the Asus BIOS also have to reroute the power signal and basically mimic motherboard power management to allow the modules to operate in DDR4 mode.
The most glaring problem otherwise is the height of the card: In the prototype photos shown in Bing’s video, the combined card and DIMM height ends up being the same as a tower cooler on its own. You’d basically be relegated to using an AIO with these.
They don’t look pretty. ASUS is trying to determine interest before proceeding further. This seems like it would be a useful tool, but maybe not a practical consumer product - especially if 2-4 of them would cost enough to push price too close to that of DDR5 anyway.
14:28 | Alienware is Out of Touch
Alienware published a press release this week that is titled: “Alienware Smashes Space-Time Boundaries with Concept Nyx.”
The entire press release doesn’t need any help to make it seem ridiculous. Here’s an excerpt:
“It’s one of the most common and frustrating problems – having multiple gamers under one roof competing for access to their favorite titles and totally crushing Wi-Fi bandwidth. Then there’s the medium: if you’re playing on a PC and want to seamlessly switch to the TV, the very idea is a ridiculous fantasy.”
Alienware next said:
“We’re always thinking ahead to the future – near and far – and thinking about the ways we can provide the best gaming experiences out there. Which is why we’re incredibly stoked about potentially solving --”
They’re incredibly stoked about potentially (?) solving the problem.
“one of the biggest challenges in PC gaming and changing the game completely: how to simultaneously allow players in a household to easily access their full game library and play on any device, even if they want to change screens during gameplay. A tall order, but Alienware is hurtling towards a solution.”
“An R&D project from our consortium of wizards also known as our Experience Innovation Group (EIG),”
This is some of the most Alienware-esque, out-of-touch writing we've seen in a while.
“Imagine you’re on your desktop in your bedroom exploring Night City in CyberPunk 2077. Your roommates are on their laptops and tablets in the living room, battling head-to-head in Rocket League. And your cousin is also over, casually building a new world in Minecraft on her cell phone. Now let’s say it’s time to prepare dinner so you head downstairs and pass the controller to one of your roommates – you can quickly switch to your CyberPunk 2077 experience on the 65-inch TV in the living room and let them takeover exactly where you left off, advancing your game while you cook.
That’s the future of gaming! And it’s starting to take shape today.”
“How would it be done, you ask?”
No one asked that.
“Concept Nyx spotlights the role edge computing could play – where high performance game processing could be done in the home and shared across the local network. Because the processing happens locally (versus needing to travel to and from distant servers), it could offer lower latency, greater bandwidth and more responsiveness - ultimately solving some of the performance shortfalls of current alternatives like cloud gaming. This means multiplayer latency could reach single-digit milliseconds to dramatically decrease any lag of modern-day cloud gaming systems. Pretty cool, huh?”
Alienware basically described Steam in-home streaming or any other in-home streaming solution. Not very interesting, in other words.
19:36 | Rumor: B660 & H610 Board Prices
A leak by TechnoMega Store, based in Ecuador, recently exposed the pricing table for the country’s B660 and H610 ASUS listings. The table had a floor price of $121 for an H610M-E D4 Prime motherboard, with the most expensive B660 leaked listing at nearly $200 USD. How these numbers translate to the US market remains to be determined.
The H610M-E D4 board photo shows 2 DIMM slots, no VRM heatsink whatsoever, a single PCIe slot (x16, at least), 2x M.2 SSD slots, and a whopping 4 USB ports in total. You’d normally have to buy a Dell pre-built to find a motherboard this barren and soul-less. We’re assuming it has about as many layers as a wafer cookie, because it looks cheap.
The leak, as identified by haruzake5719 on twitter, also lists the ASUS Prime B660M-A D4 at $166. That one has half of a VRM heatsink -- apparently a mark of royalty -- and 3 total PCIe full-length physical slots. It runs 4 DIMM slots instead of the 2 on the previous board.
Finally, the ASUS Prime B660M-A WiFi D4 has the other half of the heatsink and a wireless module. That board is priced at $191 USD.
21:19 | Intel To Possibly Disable AVX-512 on Alder Lake
According to a report from the German publication Igor’s Lab, Intel is looking to get rid of AVX-512 support on Alder Lake CPUs. AVX-512, for the uninitiated, is an instruction set that accelerates specific workloads. Intel has built in support for AVX-512 in various CPU platforms, dating back to 2016 with its Knight’s Landing platform. AVX-512 support has even shown up in consumer oriented architectures such as Tiger Lake and Rocket Lake, and to a certain extent, Alder Lake. Though, assuming the veracity of Igor’s report, this could be ending soon.
Despite AVX-512 technically being available with Alder Lake, it’s something of a mixed bag, due to the hybrid core topology of Alder Lake; AVX-512 only works with the P cores, meaning the E cores will need to be disabled for best possible performance with AVX-512. Additionally, there has always been a debate regarding how useful AVX-512 support is for desktop users, regardless of the CPU or architecture in question.
According to Igor, who is citing an unnamed source, Intel is going to push motherboard vendors to release firmware updates that would disable AVX-512 support entirely. Intel maintains that AVX-512 was never actually “supported” and that motherboard BIOS features inadvertently exposed AVX-512 support. As ever, this is likely a move to allow Intel to segment its product stack and price server and workstation class CPUs higher, as AVX-512 tends to be targeted at those markets anyway. We also expect this may be the end of Intel’s AVX-512 journey on client CPUs, at least in terms of out of the box support for the feature.
22:47 | HPE to Blame for 77TB Data Lose at Kyoto Uni.
The Kyoto University in Japan has suffered a massive loss of 77TB of research data, and it looks like a botched software update from HPE is to blame. According to reports, the university has lost some 34 million files from 14 research groups – and some of those files appear to be gone for good. According to The Stack, the university says that at least one third of the researchers will not get their data back, and the school is looking directly at HPE.
HPE has since issued a letter that was published by Kyoto University where the company takes full responsibility and claims that the update pushed out to the supercomputer caused a script to malfunction and delete critical backups.
According to HPE (via The Stack):
“The backup script includes a find command to delete log files older than 10 days. In addition to functional improvement of the script, the variable name passed to the find command for deletion was changed to improve visibility and readability. However, there was a lack of consideration in the release procedure of this modified script. We were not aware of the side effects of this behavior and released the [updated] script, overwriting [a bash script] while it was still running,” HPE admitted. “This resulted in the reloading of the modified shell script in the middle of the execution, resulting in undefined variables. As a result, the original log files in /LARGE0 [backup disc storage] were deleted instead of the original process of deleting files saved in the log directory.”
It seems for the time being, Kyoto University has suspended backup operations on the supercomputer and is investigating the issues.
24:36 | Asus Begins Recalling Faulty Z690 Hero MOBOs
Asus announced via a new webpage (linked below) that it has begun the recall process for batches of its faulty Z690 Hero motherboards. Recently, news and user reports surfaced of some Z690 Hero motherboards having a series of error codes, melted MOSFETs, and even signs of smoke coming from the boards.
After an investigation on Asus' part, and independent investigations done by those like Buildzoid, it was determined that certain Z690 Hero boards have a capacitor installed backwards. A capacitor installed improperly (backwards) would reverse the polarity and damage anything near it, such as MOSFETs.
According to Asus:
"To our valued ASUS Customers, ASUS is committed to producing the highest quality products and we take every incident report from our valued customers very seriously. We have recently received incident reports regarding the ROG Maximus Z690 Hero motherboard. In our ongoing investigation, we have preliminarily identified a potential reversed memory capacitor issue in the production process from one of the production lines that may cause debug error code 53, no post, or motherboard components damage. The issue potentially affects units manufactured in 2021 with the part number 90MB18E0-MVAAY0 and serial number starting with MA, MB, or MC.”
If you’re affected, you can follow the link below to verify your serial number and part number and start the RMA process.
You can check our serial number here: https://www.asus.com/support/rog-maximus-z690-hero-checking
27:05 | Intel Demos PCIe 5.0 SSD with Alder Lake
Ahead of CES 2022, Intel is showing off Samsung’s recently announced PM1743 PCIe 5.0 SSD paired with an Alder Lake i9-12900K-based system. Intel has been working with Samsung to test and evaluate Samsung’s new crop of PCIe 5.0 SSDs, and Intel and Samsung seem more than ready to show off what the two companies have been able to achieve.
Ryan Shrout, Chief Performance Strategist at Intel and former editor at PC Perspective, offered a demo of a pair of Samsung’s PM1743 SSDs. Shrout showed off the demo system, which consisted of an Alder Lake i9-12900K socketed into an Asus Z690 motherboard (not of the on fire variety, thankfully) paired with an RTX 3080.
Shrout began the demo by showing a “top of the line” PCIe 4.0 SSD in IOMeter, where the SSD was topping out at ~7GB/s. He then ran the Samsung PM1743 through IOMeter, where the drive came in at around ~13GB/s, which seems consistent with Samsung’s performance claims. Furthermore, in response to requests on Twitter, Shrout then showed off two drives being addressed independently via Raid 0 in IOMeter for a combined bandwidth of 28GB/s.
Samsung states the PM1743 is on track for mass production in the first quarter of 2022. To be clear, Samsung’s PM1743 is an enterprise drive, though in time, a consumer variant will trickle down.
Writing: Eric Hamilton, Steve Burke
Video: Keegan Gallick