Hardware stub

HW News - New CPU Vulnerability, Intel A380 GPU Launches, AMD Ryzen 7000 Release Date

Posted on June 23, 2022

Hardware news this week discusses a new CPU vulnerability (Hertzbleed) that affects all CPUs, alongside other topics, like the Intel A380 GPU launch, AMD's Zen 4 leaked release date for September, and leaks about the Intel Core i9-13900K and 13900KF.

The video embed and show notes follow.

01:21 | Intel A380 GPU Launched in China

  • First Intel Arc dGPU launched in China.
  • Targeting 1080p 60fps gaming.
  • Release in China this month, “summer” for global launch.

Intel has officially launched the A380, its first discrete GPU in the Arc lineup. For now, it’s China only. Availability in other markets will follow later this summer.

Intel says its target market is “mainstream gamers and content creators” with this GPU, which is a slot power GPU positioned low in the stack. It’s above the A310, but below the A580, A750, and A770. Intel claims over 60FPS at medium settings in League of Legends, Moonlight Blade, Naraka: Bladepoint, and PUBG: Battlegrounds, but this metric is kind of meaningless without a comparative and without knowing all the details about the test setup.

The A380 has 8 Xe cores, 6 GB of GDDR6 VRAM, and a 96-bit memory bus. Intel notes support for all of the DirectX 12 Ultimate features, including hardware ray tracing. HEVC and H.264 encoding and decoding are supported in hardware, as well as AV1 encode, which Intel claims is a first.

Availability will be first from system integrators in China, then as individual components from OEMs in China, followed by a full launch in other markets. For now, the only product that has surfaced is a prebuilt PC by MSI available for preorder on JD, a Chinese retailer.

Intel is suggesting a price of 1030 yuan, which is currently about $154. Intel is comparing this to the AMD Radeon RX 6400 in internal benchmarks, and claims up to 25% price-to-performance advantage over it. The RX 6400 is currently going for 160 to 190 USD and, if Intel keeps the price of the A380 equivalent to what it is in China, it could be an interesting entry-level option. This low end of the market isn’t well served right now, so it’ll be nice to see some competition. Higher end cards from Intel should be coming late summer, according to their most recent updates.

We’ll try to get our hands on one of these for testing when we can, and you’ll definitely hear about it when we are able to.

Source: https://www.intel.com/content/www/us/en/newsroom/article/intel-arc-a380-graphics-available-china.html
Source 2: https://www.intel.com/content /www/us/en/products/docs/arc-discrete-graphics/3.html

04:41 | Intel i9-13900KF Preview Leaked by SiSoftware

In leaks this week, SiSoftware posted an Intel i9-13900 set of results and a preview. The 13900 will be part of Intel’s upcoming codename “Raptor Lake” CPUs, following the Alder Lake CPUs in the 12000 series.

The Archive.org store of the SiSoftware preview indicates an 8P/16E configuration, with 32 total threads for the i9-13900K and 13900KF. The listing suggests a maximum 250W Turbo, with 65-125W standard rating for other desktop parts.

The SiSoftware listing suggests 36MB of L3 cache on the high-end parts, which would be an increase of 6MB over the existing Alder Lake CPUs. PCIe Gen5 is included, as expected, and DDR5 support remains present.

SiSoftware mentions that the E-cores move to 16C/16T maximally, with cache the same except for L2 (now at 4MB from 2MB) between 4 cores.

Sources: http://web.archive.org/web/20220613101811/https://www.sisoftware.co.uk/2022/06/10/intel-13th-gen-core-raptorlake-i9-13900-preview-benchmarks/
https://ranker.sisoftware.co.uk/show_run.php?q=c2ffcfe988e9d4ecdaebdfeed8fe8cb181a7c2a79aaa8cffc2fa&l=en

05:54 | Microsoft Cloud Gaming Push & Game Demos

  • Xbox partnering with Samsung on an Xbox app for Smart TVs to allow cloud game streaming.
  • Windowed-mode game improvements being tested on the Windows Insider program coming to release soon.
  • Project Moorcroft program to bring back game demos.
  • Games showcased coming within 12 months.

Xbox shared some plans about the future of gaming in the Xbox ecosystem in a recent blog post, then talked about new games in the live Xbox and Bethesda Games Showcase.

The first item of note from the blog post is Project Moorcroft. This program will be Xbox’s push to bring back the game demo. There isn’t a lot of detail on this yet, but so far we know that Xbox plans to compensate game developers for the demos and provide analytics on how users interact with the demos. Xbox wants to recreate E3, allowing Game Pass users to trial curated segments of upcoming games ahead of release. This hearkens back to a simpler time in gaming when demo discs were commonplace and came with your favorite magazine or were otherwise distributed as freeware.

Xbox also talked about expansions in its cloud-gaming segment. Xbox will be partnering with Samsung to bring an Xbox app to Samsung smart TVs that will allow Game Pass Ultimate users to stream cloud-enabled games straight to the TV. In addition to the paid Game Pass Ultimate games, users will be able to play Fortnite without a subscription to the Game Pass service. This will require logging into a Microsoft account from Samsung Gaming Hub or Media Hub and pairing a Bluetooth controller with the TV. At this point, it seems like the company means any Bluetooth controller, not just Xbox controllers.

Xbox went on to announce that the Cloud Gaming beta is now extended into the Argentina and New Zealand markets, to allow Game Pass Ultimate members there to stream cloud-enabled games to smartphones, tablets, Windows PCs, recent Xbox models, and the just-announced Samsung smart TV app.

Further, Xbox also plans to let Game Pass Ultimate members play games they already own from the cloud that are not part of Game Pass. How exactly that will work is not detailed at this time.

Game streaming in general has a rocky track record, but many companies seem intent on making it a reality. High latency is a core issue making it only practical for slow-paced games that don’t rely on reflexes. 

Moving to non-cloud, Microsoft mentioned that improvements to playing games in windowed mode will be coming to release soon, but no exact date was mentioned. These were originally mentioned in February of this year and are currently being tested in the Windows Insider program. Microsoft claims significant improvements to latency with these changes. This will also enable certain features like Auto HDR and Variable Refresh Rate within windowed mode.

Primary Source: https://news.xbox.com/en-us/2022/06/09/whats-next-for-gaming/

11:02 | Game Updates from Bethesda & Microsoft

Moving on to the Xbox and Bethesda Games Showcase live event, Xbox talked about notable games on the horizon. All the games mentioned are targeted to launch within the next 12 months.

Todd Howard of Bethesda showed off gameplay for the highly-anticipated Starfield, or as Todd likes to call it, “Skyrim in Space.” This is the first new IP from Bethesda in 25 years and features a “NASA punk” art style. Starfield has been in development since late 2015, but was in the  developers’ minds much earlier, with the name being trademarked in 2013.

Xbox has partnered with Kojima Productions to create a game “like no one has ever experienced or seen before leveraging the power of the cloud.” There’s no real detail here, but Kojima usually brings unique ideas into his games.

Other points of interest include:

  • Developer Riot has also teamed up with Xbox to bring several of their games into Game Pass, including League of Legends, Valorant, and more.
  • Forza Motorsport coming Spring 2023.
  • Persona 5 Royal, Persona 4 Golden and Persona 3 Portable being ported to Xbox and Windows.
  • Halo Pelican ship being added to Microsoft Flight Simulator.

Primary Source: https://news.xbox.com/en-us/2022/06/12/xbox-and-bethesda-games-showcase-recap/

12:01 | Hertzbleed CPU Vulnerability

  • New family of side-channel attacks called “frequency side channels.”
  • Worst case can allow extraction of cryptographic keys from remote computers.
  • True mitigation falls to developers to use constant-time best practices.

Intel posted information about a new CPU vulnerability disclosure this week. The vulnerability has been dubbed “Hertzbleed” and it affects nearly all modern x86 CPUs, similar in that regard to Spectre. AMD and Intel are both affected. Both companies have published security bulletins for the vulnerability.

A joint research team from UTA, UIUC, and UW published a paper describing the new family of “frequency side-channel” attacks. The group reported that the worst case scenario for this vulnerability would allow attackers to use Hertzbleed to extract cryptographic keys from remote computers. 

The root cause is the frequency scaling behavior of modern processors with features like Intel’s Turbo Boost and AMD’s Precision Boost. So it’s not a bug, it’s a feature. The researchers’ website states:

“Hertzbleed takes advantage of our experiments showing that, under certain circumstances, the dynamic frequency scaling of modern x86 processors depends on the data being processed. This means that, on modern processors, the same program can run at a different CPU frequency (and therefore take a different wall time)...”

Another type of side-channel attack, “power side-channel,” is similar but requires a power measurement interface. Hertzbleed does not require such a device and can be thought of as turning power side-channel attacks into remote timing attacks. Timing attacks rely on complex analysis of the time it takes for code to execute against differing data sets. The Hertzbleed site goes on to say:

“The cause is that, under certain circumstances, periodic CPU frequency adjustments depend on the current CPU power consumption, and these adjustments directly translate to execution time differences (as 1 hertz = 1 cycle per second).”

Typically, ensuring that cryptographic programs take the same amount of actual wall time to execute against both correct and incorrect keys would guard against this kind of attack; this concept is referred to as “constant-time.” However, the researchers claim that current best practices on how to write constant-time code, like Intel’s, are not sufficient to guarantee constant-time execution on modern processors.

Speaking of Intel, the company has since published an article titled: “Frequency Throttling Side Channel Software Guidance for Cryptography Implementations,” abbreviated as FTSCSGFCI, which is coincidentally the name for Intel’s next mobile CPU.

Intel said this about the behavior:

“[...] there is an inverse correlation between the average throttling frequency and the power consumption before frequency throttling: a workload with higher power consumption before throttling tends to run at lower average throttled frequency, and vice versa. Furthermore, since the power consumption of a workload may be correlated with the data being processed, the throttling frequency may also be correlated with the data, which becomes a frequency side channel. The CPU frequency change also causes a difference in the execution time of the workload and results in a timing side channel.”

This diagram contained within Intel’s article illustrates the situation.

Intel recommends using instructions whose execution time is data-independent in order to mitigate timing side-channels, especially for developers of cryptographic algorithms. Intel gives substantial detail for developers in its articles.

This vulnerability was explicitly tested against Coffee Lake, Coffee Lake Refresh, Comet Lake, Rocket Lake, Ice Lake, and Tiger Lake, with the researchers claiming it also worked against Ryzen processors. AMD confirmed the affected CPUs in its own bulletin, including multiple Ryzen and Threadripper lines, but excluding Ryzen 5000. We don’t know if that exclusion means those CPUs aren’t affected, but AMD chose not to mention them.

Source 1: https://www.hertzbleed.com/
Source 2: https://www.hertzbleed.com/hertzbleed.pdf
Source 3: https://www.intel.com/content/www/us/en/developer/articles/technical/software-security-guidance/technical-documentation/frequency-throttling-side-channel-guidance.html

16:26 | Plastic Processors

  • The first high-yield plastic processors have been manufactured.
  • Architecture is called FlexiCore and has 4 and 8-bit designs.
  • Contains 2104 “semiconductor devices,” similar to Intel 4004.

If you thought teaching sand to think was a mistake, get ready. A team of engineers at the University of Illinois Urbana-Champaign have designed the world’s first high-yield processors out of plastic, instead of silicon. The engineers worked with British company PragmatIC Semiconductor for the manufacture of this new design, which has been dubbed FlexiCore.

This isn’t the first time a processor has been made of plastic, with the very first being created in 2011 in Europe by researchers at the IMEC and TNO organizations. Their processor was 8-bit, contained 4000 transistors, and ran at 6 Hz. More recently in 2021, Arm reproduced its M0 32-bit microcontroller in plastic. Neither of these earlier designs, however, had good yields or hit the low cost goal of FlexiCore. No plastic processors have been manufactured at any scale, until now.

For FlexiCore, the team designed both 4 and 8-bit designs to help minimize size and maximize yields. 81% of the 4-bit processors came out functional, which is good enough to lower the cost per processor down below 1 cent each. The 8-bit version did not yield as well but we don’t know what that yield number actually was.

The processors are made using a flexible thin-film semiconductor called indium gallium zinc oxide (IGZO) and can function even when bent around a radius as tight as a few millimeters. In order to achieve this, the engineers couldn’t just copy a design intended for silicon, but had to make their design from scratch with plastic in mind. The low 4 and 8-bit logic, separation of memory with instructions from memory with data, and cutting down the instruction set itself all helped. Additionally, instructions are executed in a single clock cycle, rather than a multistep pipeline like all modern CPUs have. One of the engineers is quoted as saying,

“In general, we were able to simplify the design of FlexiCores by tailoring them to the needs of flexible applications, which tend to be computationally simple.”

The 4-bit design comes in at 5.6 square millimeters with 2104 “semiconductor devices.” This is comparable to the Intel 4004 from 1971 which had 2300 transistors.

Cheap and flexible processors like this could have many applications, from something useful like a flexible gas sensor wrapped around a pipe to detect leaks, to notably less useful things like interactive packaging on food. Get ready for smart everything; it’s coming. We can’t wait for our forks to start telling us when it’s time to stop eating.

Source: https://spectrum.ieee.org/plastic-microprocessor

19:49 | Leak: Alleged Ryzen 7000 Launch Date Leaked

  • Photo posted on Twitter by wxnod shows an AMD slide being presented regarding socket AM5.
  • Date on the slide is 9/15 and the text allegedly translates to “time to sell.”

Next up is a quick one. Leaker wxnod on Twitter shared the alleged launch date of AMD’s socket AM5 and Ryzen 7000 processors. wxnod claims the photo was taken at an AMD China event and shows a slide displaying socket AM5 with Chinese characters. The slide reads “jiu3yue4 shi2wu3 hao4,” or September 15th, then reads “kai1shou4,” or “on sale.” Easy enough: The AM5 product launch, at least in China, is expected for September 15th on-shelf availability.

We read some of the other text around the borders of the image, including the reflection, and there wasn’t much of further value there. It’s mostly talking about prior launches.

This lines-up with AMD’s official launch window of “Fall 2022,” which is publicly known. Nothing else of note is revealed here, but when we have official word from AMD on launch we’ll let you know.

Source: https://twitter.com/wxnod/status/1537824618613833728/photo/1

20:49 | Rumor: AMD CPU Lineup Named

Relatedly, rumor propagator Greymon55 tweeted an alleged set of CPU names for the AMD Ryzen 7000 series.

There’s not much to this one: The suggestion is an initial launch containing the 7950X, 7900X, 7800X, and 7600X.

Honestly, whether or not that rumor is founded in anything, it’s probably true. We could have guessed that AMD would use its existing product names and then put a 7 in front of them. Not a big surprise.

Source: https://twitter.com/greymon55/status/1537852496080961537

21:30 | AMD Mobile APU Engineering Sample Spotted

  • Upcoming Zen4 and RDNA3 mobile APU surfaced in MilkyWay@home BOINC project database.
  • Twitter bot BenchLeaks picked it up.
  • 16 “processors” reported, which are likely threads.

Twitter bot BenchLeaks picked up an unknown AMD CPU this week in the database for BOINC project MilkyWay@home. The database entry reports “Family 25 Model 112 Stepping 0” and 16 processors, which is likely the number of threads.

This information lines up with the upcoming “Phoenix Point” mobile APUs revealed in AMD’s notebook roadmap. Phoenix Point will be 4nm chips containing Zen 4 CPU cores, RDNA 3 graphics, and support for DDR5.

Source: https://milkyway.cs.rpi.edu/milkyway/show_host_detail.php?hostid=931791
Source 2: https://twitter.com/BenchLeaks/status/1537117447484059648

22:15 | Lian Li Launches Infinity Mirror Fans

In some smaller product news, Lian Li launched its new UNI FAN SL-INF120 this past week, which uses an infinity mirror hub at 0.5mm depth, but reflecting -- as one would expect -- “infinitely.” At least, to the extent that 7 rings is approximately infinite.

The fans carry this pattern to the sides of the housing and also use 40 LEDs distributed along the hub and frame.

The UNI FAN series uses a pin-to-pad contact system that allows the fans to lock to each other and eliminate cabling when used sequentially. It’s been out for a while now, but this is the illuminated upfit.

Source: https://www.youtube.com/watch?v=wZNFLLGaW08

23:30 | Alphacool Launches Apex Thermal Paste

Finally, Alphacool launched some thermal paste this week. It’s not the most exciting news, but we have a thermal paste addiction here.

The new paste is branded “Apex,” and claims a 17W/mK thermal conductivity. As a note, thermal conductivity is measured in under different conditions between brands, so it’s not typically comparable cross-brand. It often is comparable intrabrand, though.

The paste ships in 4g syringes and is about $16.

Source: https://www.aquatuning.us/thermal-pads-paste/thermal-compounds/29598/alphacool-apex-17w/mk-thermal-grease-4g