CPU, APU & Chipsets News - Page 27

A new leak about AMD Ryzen 8000 (Granite Ridge) processors fills in a good deal more info about the next-gen chips (bring your own salt), but it may disappoint because it puts forward a later timeframe for arrival than previously expected.

Let's get that nugget out of the way upfront, with a new report from Hilbert Hagedoorn at Guru3D claiming that the Zen 5 release date will be the second half of 2024.

Previously, we'd heard chatter from the rumor mill - courtesy of YouTube leaker Moore's Law is Dead - that Zen 5 desktop chips might be out in the first half of next year, and quite possibly Q1 2024.

Rapidus Corp. is a newly-formed Japanese semiconductor manufacturing company with pretty lofty goals, making it a significant player in the chip space in just four years. This would put it in direct competition with TSMC and Samsung, with Rapidus producing its 2-nanometer chips by 2027.

The company is already investing billions in the new plant, with strategic investments from the likes of Sony and Toyota and even the Japanese government. According to Bloomberg, the government has already budgeted over USD 2.4 billion (or ¥330 billion) annually for the new semiconductor manufacturer.

"Huge opportunities are ahead of us if we can become first in the market and focus on making chips for specific areas, such as AI," Rapidus chairman Tetsuro Higashi tells Bloomberg.

Intel's Arrow Lake processors, which in theory are set to follow Meteor Lake (and Raptor Lake refresh), could arrive late next year with a massive performance boost.

That's according to YouTube leaker Moore's Law is Dead (MLID), whose latest video concerns Intel's upcoming processor ranges, encompassing all the aforementioned Core families.

For us, the most significant piece of news here is that 15th-gen Arrow Lake's performance cores will be hugely faster than Meteor Lake. MLID asserts that Arrow Lake will offer an IPC (Instructions per Clock) boost of 22% to 34%.

Recently we reported on a new and mandatory AMD AGESA 1.0.0.7 BIOS Firmware for all AM5 motherboards built for the new Ryzen 7000 CPUs. This a direct response to the burnout issues found on some Ryzen 7000X3D chips that were damaging both the chip and motherboard.

AMD's root cause for the issue was around voltage. The new firmware limits the SoC voltage to 1.3V. It adds other measures/fixes to ensure processors don't exceed their temperature specifications - while not affecting memory overclocking via AMD EXPO profiles.

With potentially damaged hardware on the line, no doubt there's a rush to get this out to as many customers (and board partners) as possible, but with ASUS using a BETA release of the new AGESA 1.0.0.7 BIOS firmware, it looks like there are still issues to work out, which could delay the proper and final rollout of firmware to fix the burnout issue by weeks.

Regarding non-memory chip manufacturing, Samsung has been trailing Taiwan Semiconductor Manufacturing Company (TSMC) in terms of cutting-edge process nodes and yields for some time. Even so, the company believes that when it comes to rolling out a 2-nanometer semiconductor manufacturing process, it will catch up to TSMC and even surpass it.

Samsung's president and manager of the company's foundry business, Dr. Siyoung Choi, made these remarks during a keynote at the Korea Advanced Institute of Science & Technology in South Korea (via Wccftech) where he put a timeline of five years for Samsung to catch up to TSMC.

Dr. Siyoung Choi did note that TSMC is currently ahead of Samsung when it comes to chip production. However, he cites Samsung's use of GAA technologies for current 3nm production as a step ahead of TSMC, which still utilizes the tried and true FinFET.

AMD's next-gen Zen 5 processors are once again the subject of chatter on the grapevine, with a YouTube leaker toning down expectations (slightly) on how fast Ryzen 8000 chips could be.

The latest video from RedGamingTech (RGT) still sticks to the line that Zen 5 might offer an IPC (Instructions per Clock) uplift of between 20% and 25%, as the YouTuber said in their last leak.

This time around, though, RGT notes that while multiple sources in both cases are pegging IPC gains at about 20% and 25% respectively, the leaker believes that 20% is the more likely figure.

Quarterly reports and financial documents for large tech companies are often a mix of numbers and figures that require digging to find some of the most interesting takeaways. And that's still certainly the case with AMD's latest Q1 2023 financial results, but with the current state of the PC market, it's hard not to see the historic operating loss of USD 145 million as a cause for concern.

Especially when you compare that to the last year's operating income of USD 951 million, making it a 115% decrease and the first overall loss for AMD in years. As AMD's business encompasses several sectors, the Client Segment covering AMD Ryzen products looks to have seen the most notable decline, with revenue down 65% year on year, from over USD 2 billion to USD 700 million.

This segment alone reported an operating loss of USD 172 million, a stark contrast to the rapid growth seen in the Embedded Segment, which saw massive growth with revenue up 163%. This is reflected in the following statement from AMD Chair and CEO Dr. Lisa Su, who notes that AMD will prioritize its data center and embedded portfolio.

It looks like Intel might be preparing to change the name of its Core family of processors, departing from the terminology used for a long time now, namely (pun not intended) the Core i3, i5, i7 and i9 tiers of CPU.

This originates from a leak of an Intel processor via the Ashes of the Singularity benchmark database, an 18-thread CPU with the name 'Core Ultra 5 1003H', apparently switching the Core i5 naming scheme with Core Ultra 5.

At the time, it wasn't clear whether this might just be a different codename used for an engineering sample chip, but now, as VideoCardz points out, Bernard Fernandes, who is Intel's Director of Global Communications, has tweeted on the subject.

A new tweet from @OreXda simply states that 'AMD signed to use Samsung Foundry 4nm,' which points to AMD using Samsung's 4nm process for future products. It's unclear what these products will be, but looking back at Samsung's plans in 2021, there was word that AMD was eyeing the Samsung Foundry to produce 4nm CPUs designed for Chromebooks.

All of this comes down to TSMC's limited capacity, which for 4nm is apparently at its absolute peak, and with recent reports pointing to Samsung's fabrication process yielding better results as time goes on - a switch for certain low-power products makes sense for AMD.

Sourcing chips from multiple sources, as long as the quality is there, is a sound strategy and will allow AMD to get products out on time instead of delaying things ala the Ryzen 7040 Phoenix APUs. However, it's worth stating that product delays can happen for several reasons beyond manufacturing.

Ryzen CPUs burning up has been one of the most talked about things in the PC space in recent weeks, with AMD releasing an official statement noting that excessively high voltages on AM5 motherboards with the new Ryzen 7000X3D chips were damaging both the chip and motherboard.

A catastrophic outcome that has led to several outlets conducting their own tests alongside AMD - with a deep dive from Gamers Nexus dropping on the weekend showcasing that it was able to get an AMD Ryzen 7 7800X3D to effectively "explode" and melt the motherboard-side connections.

Even though the issues and voltage spikes are seemingly linked with overclocked memory using EXPO or XMP kits, AMD's root cause does note that the new firmware limiting the SoC voltage to 1.3C won't affect how memory performs.