02:40 | Comet Lake-S Availability Update
A quick update on Comet Lake-S availability. We spoke with Intel regarding the lack of general availability with specific Comet Lake-S SKUs at launch. The statement we received is as follows:
“The unlocked 10th Gen processors are coming to market first, including the Core i9-10900K, Core i7-10700K, and Core i5-10600K. The 10th Gen Core i3-10100 and 10th Gen Core i3-10300 are expected to be available shortly after the 10th Gen unlocked K SKUs.”
Furthermore, Intel told us it would keep GN updated on current i5-10600K availability, as it relates to viewer/reader interest.
Source: Intel to GamersNexus
04:16 | Asetek Doesn’t Understand Math for Rad Card
Asetek, the company behind both desktop and data center cooling solutions, has trotted out a new cooling design aimed at deploying liquid cooled GPUs in restricted cases. The new design is known as the Rad Card.
As is probably obvious given the name, the Rad Card is a radiator constructed like an add-in PCIe card, shroud and all, that slots into a PCIe slot. The Rad Card is being deployed initially as an OEM only option, being first used by Alienware in the Aurora R11 PC. Within the Aurora R11, you can see two populated PCIe slots. One slot is filled with an RTX 2080 Super, and the one just beneath it with the Asetek Rad Card, with a short run of tubing between them. The Rad Card appears to have one integrated radial fan.
Asetek lays claim to some bold performance numbers:
“With Alienware's liquid cooled 2080 SUPER GPU you can reduce noise by up to 69% and GPU temperatures by up to 20%”
You can’t really do this, mathematically. Temperature is an arbitrary scale that doesn’t start at 0. For example, 50 degrees Celsius to 100 degrees Celsius would appear, strictly numerically and ignoring the scale and units, a “100%” increase -- but if you convert the values to Kelvin, it’s 323.15K to 373.15K, so that’d be a “15%” increase. It’s misleading at best, and actually wrong at worst, to claim a “20% reduction” in GPU temperatures. The scale is arbitrary. That means nothing. If they want to use percentages, they should convert to Kelvin -- but that wouldn’ produce a percentage as high as 20.
Further, noise reduction of “69%” means absolutely nothing to anyone. We don’t know what unit of measurement they’re using for noise, and if it’s decibels, that’s a logarithmic scale and can’t be compared linearly with a percentage.
While the Rad Card is currently OEM only, Asetek teased on Twitter that it may not stay that way.
08:48 | ASRock BCLK Frequency Boost for Non-K Intel CPUs
We’ve been hearing rumors for some time about the possibility of overclocking non-K SKU Intel chips on not only new Z490 ASRock motherboards, but select B and H-series boards as well. While the rumors were a bit vague, they pointed to the concept of increasing the base clock (BCLK), rather than the CPU multiplier.
BCLK tuning isn’t anything new, mind you, but the idea of being able to do it with 10th-gen, locked Intel CPUs -- and on non-Z motherboards -- is certainly promising. We still aren’t in full possession of the facts, and we don’t currently have any non-K Comet Lake-S chips to verify the process with. However, at this point, we do know that ASRock does indeed have a base frequency boost feature (BFB Technology, as ASRock calls it), built into some of its 300 and 400-series motherboards via new BIOS updates.
What exactly ASRock is doing to make BFB work is a bit nebulous. However, if we were to venture a guess, it likely has something to do with tweaking the PL1 state on Comet Lake-S CPUs. In a leaked image, it’s clear that ASRock is able to force a 70W CPU to operate as if it were a 125W CPU. This would obviously have implications for your cooling solution. It’s also subject to provoke the unwanted ire and gaze of Intel, a point that even ASRock has apparently accounted for.
“Due to future hardware/firmware updates or other reasons, the availability of BFB is subject to change without notice in advance,” reads the fine print, denoted with an asterisk over at ASRock’s BFB page.
So, for now, it seems BCLK overclocking non-K Intel chips on Z/H/B ASRock motherboards is a go -- until it’s not. We’ll see what happens. I’m a bit skeptical that Intel won’t nuke this with a firmware update; that would’ve been all but guaranteed pre Ryzen. Intel has long held overclocking as a luxury worth paying extra for and limited the feature to the most expensive letters in the alphabet: K, X and Z. Again, we’ll see if the Ryzen effect has since changed Intel’s stance on the topic.
ASRock currently has a list of supported Zxxx/Hxxx/Bxxx motherboards that support BFB through the latest BIOSes available.
12:32 | Lawsuit: NV Allegedly Disguised Crypto Revenue as Gaming
The suit against Nvidia that dates back to 2017, with a class action complaint filed early last year, seems to be gaining steam. Now, the class action complaint has been amended to suggest that Nvidia obfuscated as much as $1B in cryptocurrency revenue by disguising it as gaming revenue.
The reason for this, allegedly, was because Nvidia knew the crypto-mining boom wouldn’t last. This would've painted the false picture that Nvidia’s massive revenues at the time were insulated from the unpredictable volatility of cryptomining, thus misleading investors and shareholders. This also would’ve suggested that Nvidia’s gaming segment was more profitable and bigger than it actually was (at the time), again misleading investors.
The complaint specifically names Nvidia’s senior management -- CEO Jensen Huang, CFO Collette Kress, and SVP and Jeff Fisher -- as the culprits behind the alleged cover-up.
Shareholders are now seeking damages for what they maintain are violations of the federal securities laws.
13:55 | TSMC Cuts Off Huawei Amidst Stricter US Regulations
The US Department of Commerce recently announced that it has an amended export rule, and it’s aimed directly at Huawei and its wholly owned chip design arm, HiSilicon. The amended export rule dictates that any foreign chipmakers using American IP, technology, or equipment will have to apply for a license before shipping to Huawei.
Last year, waves were made when the US added Huawei to the Entity List, essentially blacklisting them from doing business with US-based companies. Key suppliers, like ARM and Google, quickly retracted their business with Huawei. Despite this, The Department of Commerce asserts that Huawei has continually undermined the Entity List.
“Despite the Entity List actions the Department took last year, Huawei and its foreign affiliates have stepped-up efforts to undermine these national security-based restrictions through an indigenization effort. However, that effort is still dependent on U.S. technologies,” said Secretary of Commerce Wilbur Ross.
Ever since the Entity List actions, TSMC has been caught in the middle of the geopolitical spat between the US and China, with pressure on the chipmaker over its relation with Huawei. Huawei is TSMC’s second largest customer behind Apple. So, it quickly becomes apparent as to why TSMC has been reluctant to sever ties. However, in response to the amended export rule, TSMC has now apparently halted all orders from Huawei, per Nikkei Asian Review.
"TSMC has stopped taking new orders from Huawei after the new rule change was announced to fully comply with the latest export control regulation," a person familiar with the situation told Nikkei. This news comes as TSMC formally announced plans to build a 5nm megafab in the state of Arizona.
15:42 | Microsoft Concedes “Wrong Side of History” for Open-Source
Microsoft has officially admitted that it was wrong about open-source. Mind you, this is the same Microsoft whose former CEO, Steve Ballmer, infamously called Linux “a cancer that attaches itself in an intellectual property sense to everything it touches.” Well, that didn’t age well.
In a webinar hosted by MIT, Microsoft president and chief legal officer Brad Smith said that “Microsoft was on the wrong side of history when open source exploded at the beginning of the century and I can say that about me personally." Smith went on, saying that “The good news is that, if life is long enough, you can learn... that you need to change."
True enough. And it’s not really fair to say this is the same Steve Ballmer-led Microsoft of the early aughts. It’s hard to forget Microsoft’s sketchy history, especially in regards to its war with Linux, but Microsoft has since turned some corners in regards to open-source projects. Most recently, it announced its Fluid Framework, a system seemingly aimed at taking the collaborative cloud software crown from Google Docs, is going open-source.
There’s also Power Toys, a suite of tools for power users resurrected from Windows 95 and updated and developed for Windows 10 under MIT License. It’s free, open-source, and hosted on GitHub. Oh, there’s also GitHub, the Git-based collaborative code repository service that Microsoft acquired, which it uses for its own open-source projects. Smith even mentions Microsoft’s GitHub acquisition as part of its commitment to open source. “Today, Microsoft is the single largest contributor to open-source projects in the world when it comes to businesses. When we look at GitHub, we see it as the home for open-source development, and we see our responsibility as its steward to make it a secure, productive home,” says Smith.
Perhaps most notably, there’s also the Windows Subsystem for Linux (WSL) that allows for Linux/GNU distros to be run in Windows 10 without the overhead of a VM, and WSL 2.0 is on the way. So, while Microsoft was on the wrong side of history in respect to Linux, it seems determined to not let history repeat itself. Whether you attribute this stance to a paradigm shift in Microsoft’s leadership or a paradigm shift in computing -- or both -- is up to you.
18:17 | Rumor: Early Clock Speeds for Ryzen 4000
A report coming out of Igor’s Lab seems to detail some early clock speeds for Ryzen “Vermeer,” or Ryzen 4000 as it will come to be.
According to friend of the site Igor, there are already Zen 3 CPUs in the wild with an A0 stepping, which indicate an engineering sample. The samples in question are 8C/16T and 16C/32T chips, respectively. According to OPNs (Ordering Part Numbers) ascertained through BIOS entries, it seems early clock speeds for the 8C/16T variant are 4.6GHz boost and 4.0GHz base. The 16C/32T chip seems to have a base and boost of 3.7GHz and 4.6GHz, respectively. As this is no doubt very early silicon, these frequencies will almost certainly rise with the A1 stepping revision.
At any rate, treat this rumor with the appropriate amount of salt.
20:10 | Intel Acquiring Maker of Killer NICs
Intel has announced that it’s acquiring Rivet Networks, the company that produces the Killer line of NICs, for an undisclosed sum. A little known fact is that Rivet Networks actually started life as Bigfoot Networks. It wasn’t until Qualcomm acquired the company and subsequently spun it off a few years later that it became Rivet Networks.
Intel reports that Rivet Networks’ team will be lumped into Intel’s Wireless Solutions Group, which is itself a part of Intel’s broader Client Computing Group. Rivet Networks’ products and IP, such as the Killer brand, will roll into Intel’s PC Wi-Fi portfolio. The driving factor behind Rivet Networks being spun off from Qualcomm was due to the Killer NIC brand stagnating under that acquisition. Since becoming Rivet Networks, the Killer brand has flourished and seen wide adoption with OEMs and motherboard makers. Ian Cutress with AnandTech had a chance to talk with Intel and Rivet Networks, and expressed concern that Rivet may suffer the same fate under Intel. Rivet's CEO Mike Cubbage, who Cutress says will become Intel's Senior Director of Connectivity Innovations, seemed to assuage those concerns.
According to Cubbage, Rivet Networks’ strengths -- both then and now -- lie within PC and gaming, and at least in that arena, both Intel’s and Rivet Networks' goals are mutually congruent. One could likely take the recent AX201 and Killer AC-1535 chips as evidence of this, as Rivet partnered with Intel on both solutions.
21:53 | NVIDIA Reproduces Pac-Man with AI
Just in time for Pac-Man’s 40th anniversary in Japan, researchers at Nvidia have faithfully recreated the retro classic with GameGAN.
GameGAN is Nvidia’s neural network model that emulates games through GANs, or generative adversarial networks. Based on two separate, but competing neural networks (a generator and discriminator), GameGAN can be trained to recreate games without an underlying game engine, and such is the case with Pac-Man.
The recreation is the result of training GameGAN with some 50,000 episodes (a few million frames, apparently) of Pac-Man. “This is the first research to emulate a game engine using GAN-based neural networks,” said Seung-Wook Kim, an NVIDIA researcher and lead author on the project. “We wanted to see whether the AI could learn the rules of an environment just by looking at the screenplay of an agent moving through the game. And it did.”
The GameGAN rendition relies solely on AI to recreate all elements and rules of the game, and it can be done through training the AI on screen recordings and keystrokes from past gameplay. The GameGAN version of Pac-Man faithfully honors both simple and not so simple physics and rules of the original. For instance, if Pac-Man runs into a Ghost, the screen flashes and it’s game over. Alternatively, when Pac-Man successfully exits the maze on one side, he’s transported back to the opposite end. Just like the orginal, Pac-Man can’t go through maze walls, and he eats dots and power pellets, replete with the appropriate effects.
The research team used Nvidia DGX systems to train GameGAN, and the Pac-Man demo will be available over at AI Playground, as well as future GameGAN projects.
Furthermore, Nvidia says that GameGAN isn’t just about games, but has larger implications for how AI could be used in the future. “We could eventually have an AI that can learn to mimic the rules of driving, the laws of physics, just by watching videos and seeing agents take actions in an environment,” said Sanja Fidler, director of NVIDIA’s Toronto research lab. “GameGAN is the first step toward that.”
Source: NVIDIA to GamersNexus
Editorial: Eric Hamilton
Host: Steve Burke
Video: Keegan Gallick, Andrew Coleman