Here’s a histrionic quote for you: “AMD must cease the sale of Ryzen and EPYC chips in the interest of public safety.”
That’s a real quote from Viceroy Research’s deranged, apoplectic report on CTS Labs’ security allegations against AMD’s Ryzen architecture. The big story today seemed to mirror Meltdown, except for AMD: CTS Labs, a research company supposedly started in 2017, has launched a report declaring glaring security flaws for AMD’s processors. By and large, the biggest flaw revolves around the user installing bad microcode.
There are roots in legitimacy here, but as we dug deep into the origins of the companies involved in this new hit piece on AMD, we found peculiar financial connections that make us question the motive behind the reportage.
The goal here is to research whether the hysterical whitepapers -- hysterical as in “crazy,” not “funny” -- have any weight to them, and where these previously unknown companies come from.
The past week of hardware news has been peculiarly busy for this time of year, with a deluge of news posting toward the latter half of last week. For major stories, [H]ardOCP’s coverage of nVidia’s GPP agreements has undoubtedly garnered among the most attention in the news cycle, with additional stories of interest covering hacks to get Coffee Lake CPUs functional in Z170 and Z270 motherboards.
We’ve got a couple of minor news items – new liquid coolers, a mini-review of a chair – and a couple of game industry items, like Valve’s return to game development.
Find the written and filmed recaps below:
“Intel & NVIDIA working together” will surely raise eyebrows, but this isn’t similar to how AMD and Intel recently worked to make Hades Canyon. Rather than work together on a product, the two companies sent high-ranking researchers and engineers to meet with the US Government, alongside numerous other AI and machine learning organizations. Intel and nVidia are the most relevant to our line of work, and the representatives at each organization worked to educate government officials on the needs of AI and machine learning development.
In addition to this news, Intel also is working on a $5B expansion for 10nm production, covered further down, and the company’s new Coffee Lake Pentium CPUs have been rumored and assigned specs.
Show notes below the video, if you prefer to read.
Samsung recently officially confirmed that they are producing ASICs (Application-Specific Integrated Circuits) intended for cryptocurrency mining, being sold to unnamed clients for ASIC mining machines. These machines are different from GPU miners, and do not meaningfully affect desktop GPU supply.
As the name implies, ASICs are chips designed for a single purpose. There’s nothing unusual about producing ASICs, but mining-specific ones have been the domain of TSMC until now, primarily with client Bitmain. Samsung won’t be doing the mining themselves, just supplying the hardware: TechPowerUp suggests the order was placed by “Chinese clients” which were mentioned in a recent earnings report. Our understanding is that the varieties of cryptocurrency which ASICs can effectively mine are ones that are now beyond the capabilities of home mining operations, like Bitcoin, so they’re used by massive currency farms. SHA-256 algorithms are best mined with ASIC miners.
While researching GPU prices and learning that GDDR5 memory price has increased by $20-$30 on the bill of materials lately, we started looking into the rising system memory prices. RAM pricing has proven somewhat cyclic over the past few years. We’ve reported on memory price increases dating back to 2012, and have done so seemingly every 2 years since that time. This research piece pulls five years of trend data, working in collaboration with PCPartPicker, to investigate why memory prices might be increasing, when we can expect a decrease, and more.
DRAM prices are crazy right now. We’ve driven that point into the ground over the past few years, but pinpointing a “when” and a “why” is a difficult proposition. With the help of PCPartPicker, we’ve identified some general trends that seem almost cyclic, and provide some relief in pointing toward an eventual downturn.
We’re revisiting a topic from July 2017, initially published in the middle of one of last year’s cryptocurrency booms. That topic was our discussion with GPU add-in board partners and PSU makers, where we collected anonymized, aggregate thoughts on cryptomining and its impact on the consumer GPU market. Given the tremendous growth of the cryptocurrency community in the time since, and the recent explosion of GPU prices up to 3-5x their MSRP (depending on if it’s a primary or secondary seller), we decided it was time to revisit the topic once more.
This information is anonymized and aggregated for a few reasons: One, no one would be able to share their thoughts otherwise, as this isn’t a topic that can be officially approached; two, it allows folks to speak more freely, as if there were an official response, you can be assured it’d tread the line of neutrality to a point of being bereft of insight. We spoke to most of the major GPU board partners and some PSU maker representatives, including the original group of folks we spoke with in mid-2017, now back to re-evaluate their positions from six months ago.
UPDATE: We worked with Google's local Fiber team directly -- who responded quickly to this post -- and got Fiber installed and working. After a month of settling in, everything seems good now. We haven't had any additional issues with Google Fiber, and can now recommend the service over the competition (easily). As long as Google doesn't lose other customers in the system, like it did with us, we can strongly recommend the service.
Original Article: Google Fiber isn’t all that it’s cracked-up to be.
The company has routinely demonstrated impressive bouts of incompetence as we’ve tried to subscribe to the service, and today was the latest artistic expression of that ineptitude. Thus far, Google hasn’t been any better than the old TWC or AT&T ISPs, with regard to support, and has been significantly worse in installation and setup. Once fiber is setup, we hope that the speeds will account for the tremendous pain that Google and its contractors have been; we imagine it’ll all be worth it, as it’s still gigabit speeds, and it’s still going to help on our uploads – it’s just a matter of getting everything working.
For this, we’re ignoring that it took a few years for the crew to embed the lines in the roads. That’s expected, and not something we’re complaining about. This complaint is more about the post-payment service.
We signed-up for Google Fiber in August of 2017, or 4-5 months ago. Our first appointment for Google Fiber installation was scheduled for November 6, 2017. 73 days later, we still do not have Fiber installed. It is presently January 18, 2018. We have also been charged for the service, despite having no service. Our “free month” credit, bordering on a scam, has been consumed, and we’ve been billed for the second month of no service.
Adding a day to our CES 2018 trip meant that we were able to regroup with about half of the NA-based "YouTubers" and technical media outlets, allowing for a lot of discussion on content creation, quality of content, and products shown at the event. One of those meetings was with Luke Lafreniere, formerly of Linus Media Group and presently of Floatplane (under the same roof, technically), who joined us to "review" CES 2018. During this on-camera-only content, we covered major product launches at the show, best and worst products, improvement to the case industry, and lacking product elsewhere. We also venture into the topic of virtual reality gaming and its waning marketing, alongside discussion of blockchain branding attached to nearly every major vendor.
It's a fun, looser video, and allows us to decompress with other media. For our video audience -- and even those who prefer the articles -- we think this one will be enjoyable for all of you, if only for the candid approach to the PC industry in early 2018.
Intel has released its own internal testing of architectures dated from Skylake to Coffee Lake, using Windows 10 and Windows 7, in A/B testing between the Meltdown kernel patch. We’ve done some of our own testing (but need to do more), but not with the applications Intel has tested. As usual, exercise grain-of-salt-mining for first-party numbers, but it’s a starting point.
Intel claims that it’s found its CPUs largely retain 95-100% of their original performance (from pre-patch, with some worst-case scenarios showing 79% of original performance – Skylake in SYSMark 2014 SE Responsiveness, namely. On average, it would appear that Intel is retaining roughly 96% of its performance, based on its own internal, first-party data.
Here’s the full chart from the company:
There’s been a lot of talk of an “Intel bug” lately, to which we paid close attention upon the explosion of our Twitter, email, and YouTube accounts. The “bug” that has been discussed most commonly refers to a new attack vector that can break the bounding boxes of virtual environments, including virtual machines and virtual memory, that has been named “Meltdown.” This attack is known primarily to affect Intel at this time, with indeterminate effect on AMD and ARM. Another attack, “Spectre,” attacks through side channels in speculative execution and branch prediction, and is capable of fetching sensitive user information that is stored in physical memory. Both attacks are severe, and between the two of them, nearly every CPU on the market is affected in at least some capacity. The severity of the impact remains to be seen, and will be largely unveiled upon embargo lift, January 9th, at which time the companies will all be discussing solutions and shortcomings.
For this content piece, we’re focusing on coverage from a strict journalism and reporting perspective, as security and low-level processor exploits are far outside of our area of expertise. That said, a lot of you wanted to know our opinions or thoughts on the matter, so we decided to compile a report of research from around the web. Note that we are not providing opinion here, just facts, as we are not knowledgeable enough in the subject matter to hold strong opinions (well, outside of “this is bad”).
We moderate comments on a ~24~48 hour cycle. There will be some delay after submitting a comment.