Processors - Page 39

DIY gadget YouTuber "Bitluni" has built his own personal RISC-V megacluster, inside of a compact design, check it out:

Overall, the megacluster features 256 x RISC-V-based microcontrollers, operating at a mind-boggling 14.7GHz single-core frequency. Impressive. Remember, these are RISC-V chips, not x86-based processors inside of a regular PC.

The idea behind this is to fit 16 of the superclusters into a single interface, but that created issues down the line, so he designed his own "cluster blade" that featured two microcontrollers mounted on the board, each of them managing the supercluster on top of it. The cluster blade allowed Bitluni to give each supercluster its bus interface.

AMD's next-generation high-end Ryzen AI "Strix Halo" APUs have turned up in new shipping manifestos, featuring up to 64GB capacities and up to 120W TDPs.

The shipping manifestos were noticed by "nbd.ltd" and @harukaze5719, with multiple shipments of the same APU that would've been sent out for evaluation and testing purposes. These manifestos show the Maple DAP reference evaluation platform, which is available in both 32GB and 64GB memory capacities.

The Maple DAP reference systems feature 32GB or 64GB of pre-soldered memory, set by AMD, unlike Intel which is using on-package memory with its next-generation Lunar Lake CPUs.

Intel has just splashed out on buying ASML's bleeding-edge High-NA EUV lithography machines, placing orders for the rest of the machines that the Netherlands-based company will make this year.

In a new report from The Elec, we're hearing that Intel is buying ASML's latest High-NA EUV lithography machines by placing reserve orders for the 5-6 kits that the company will make this year. Intel will be using ASML's new High-NA EUV lithography machines for its upcoming Intel 18A (1.8nm) and Intel 14A (1.4nm) process nodes.

Intel will have the most High-NA EUV lithography machines in 2025, with competitors including SK hynix and Samsung set to get their first batches of High-NA EUV lithography machines in the second half of 2025. Intel will have a rather large advantage over their competitors, especially if production goes well, yield rates are healthy, and the chips that Intel is pumping out are performance leaders.

AMD's next-generation Zen 5-based "Ryzen AI 100 series" processors will be unveiled and launched at Computex 2024 next month, with the latest Ryzen AI 165 processor teased in new leaks.

In a new leak from Chinese reviewer and leaker Golden Pig Upgrade, we can expect the AMD Ryzen AI 165 "Strix Point" APU to feature 10 cores and 20 threads, with an XDNA2-based NPU for AI workloads, offering 45 TOPS of AI performance from the NPU alone.

We've already heard rumbles of a Strix Point APU being officially called the Ryzen AI 9 HX 170, with 12 cores and 24 threads of Zen 5-based CPU power boosting at up to 5.1GHz with the same XDNA2-based NPU for AI workloads. But the new 10-core, 20-thread Ryzen AI 165 "Strix Point" APU could feature a 2+8 or 4+6 configuration (AMD will have Zen 5 and Zen 5c CCXs for Strix Point).

Intel's Panther Lake is still a good way off - if the rumor-mongers are right, Arrow Lake might even be further off than we hoped - but that hasn't stopped the leak machine churning out speculation about Panther Lake CPUs.

Mind you, this latest Panther Lake leak, which comes from X (formerly Twitter) as flagged up by VideoCardz, should be treated with more skepticism than normal, as we're not familiar with the source - a certain Prakhar Verma.

So, fistfuls of salt in hand, the contention is that Panther Lake will be mobile chips only - with no desktop offerings. Theoretically we will get Panther Lake U series CPUs, efficient variants to follow Lunar Lake (which should debut late this year), and Panther Lake H chips to succeed Arrow Lake high-end mobile processors.

AMD's Zen 5 desktop processors - which should be Ryzen 9000, based on previous leaks - might end up with a generational increase in performance that'll disappoint many folks, or that's the conclusion some are drawing given fresh chatter from the grapevine.

Wccftech flagged up this rumor which comes from Weibo - add extra seasoning because of that, but then again, we are told that this is a post from Lenovo China's manager.

According to that manager, we can expect Ryzen 9000 desktop to bring forth an IPC increase of 10% compared to Ryzen 7000 chips - and that's a much lower figure than we've heard from other sources. Although it's worth noting that the source does observe that gains are higher than this in Cinebench single-threaded benchmarks (without mentioning any exact figure).

Intel has been in a spot lately regarding its flagship high-end 13th and 14th Gen Core i9 desktop processors, with widespread reports of instability. The extent and detailed causes of the crashes are still unknown. Still, many believe it to be due to the increased power requirements of high-end Intel CPUs and motherboard profiles and official specs and guidelines bypassing proper voltage regulation.

Various motherboard makers have released BIOS updates to address the issue, implementing new 'Intel Baseline Profiles' to improve stability at the cost of performance. The band-aid approach was simple: reign everything in, and the issues will disappear.

In a new statement for its customers and partners, Intel states that the new 'Intel Default Settings' represent a range of power delivery profiles based on a motherboard's capabilities. In fact, Intel wants customers to continue pushing the new CPUs as it "recommends customers implement the highest power delivery profile compatible with each individual motherboard design."

AMD has shared some data on its CPU shipments courtesy of Mercury Research's first quarter (Q1 2024) report. And it's good news for Ryzen and AMD's growing server market, with Unit Share and Revenue Share increasing year-over-year - with Desktop and Server products leading the charge.

AMD cites growing demand for 4th Gen EPYC and Ryzen 8000 Series processors for the growth in the Client sector. Its server share (compared primarily to Intel) is now 23.6% compared to 18% a year ago. For Desktop processors, aka Ryzen, AMD's share has increased to 23.9% compared to 19.2% a year ago.

Mobile CPUs, which are Ryzen chips for laptops, mini PCs, and PC gaming handhelds, also saw their share increase in Q1 2024 to 19.3% compared to 16.2% a year ago. However, this figure presents a slight dip from the previous quarter - chalked up to a decline in overall shipments for low-end PCs.

Intel's next-gen flagship CPU could be dialing back clock speeds considerably - by a whole lot, in fact, at least if this leak turns out to be correct.

Upfront, let's arm ourselves with the usual helping of seasoning, and indeed a heavier dose than normal here, as this is a rumor that comes from Weibo, not exactly our favored source for reliability (though it does get some things on the money).

At any rate, Wccftech noticed that user 'MebiuW' posted about how Intel's Arrow Lake range - which could be revealed in Q3 - will witness clock speeds being reined in considerably.

The global client-based CPU market showed impressive growth in Q1 2024, with John Peddie Research reporting that shipments hit 62 million units - a 33% increase compared to Q1 2023. However, compared to the previous quarter, this figure represents a 9.4% decrease.

Even though the dip compared to the previous quarter bucks the recent growth trend, the report notes that the Q1 dip, when viewed with the context of year-over-year growth, "is an indication that the market has stabilized and has returned to its traditional cyclic behavior."

Looking into the numbers more closely, Notebooks and Laptops are driving the growth as desktop market share has declined. A year ago, the split between Notebooks and Desktops was 68% and 32%, with this figure changing to 73% Notebooks and 27% Desktops in Q1 2024. With integrated graphics being a key part of the mobile CPU space, iGPU shipments increased by 30% year-over-year.

Apple has announced the latest iteration of its custom processor, the new M4 chip, which will power the new iPad Pro line-up. The M4 chip is the second generation of Apple's 3-nanometer technology (28 billion transistors), and the company is touting a brand-new display engine that will utilize the Ultra Retina XDR OLED display on iPad Pro.

The new M4 CPU features up to 10 cores (4 performance cores and 6 efficiency cores), with a next-gen 10-core GPU that "builds on the next-generation GPU architecture introduced in M3" with new caching tech called Dynamic Caching, hardware-accelerated ray tracing, and hardware-accelerated mesh shading. It's the most powerful GPU ever to hit the iPad, so it will be interesting to see how it fares with modern games.

Of course, AI is a vital component of the new M4 chip - it's "Apple's fastest Neural Engine ever," sporting an AI performance of 38 TOPS, higher than Intel's current Core Ultra and AMD's current Ryzen offerings. The M4 also features faster memory bandwidth to help push AI performance even further.

It looks like we could be in for a huge naming scheme change for AMD's upcoming "Strix Point" APUs, with the new ASUS VivoBook S16 laptops featuring an AMD Ryzen AI 9 HX 170 processor.

We've never heard of the AMD Ryzen AI 9 HX 170 processor, but ASUS lists it as a 12-core, 24-thread CPU that clocks at up to 5.1GHz with 36MB of cache and Ryzen AI performance of up to 77 TOPS. Until now, no "AI" wording has made it into any processor from AMD like this, so it looks like we could see the Ryzen AI 5, Ryzen AI 7, and Ryzen AI 9 series CPUs in the near future.

ASUS will unveil the new VivoBook S16 laptops at Computex 2024 next month, and I'm sure we're going to get the official naming from AMD at their Computex 2024 press conference. The new Strix Point CPUs aka Ryzen AI 9 HX 170 processor will deliver up to 77 TOPS of AI performance from its beefed-up XDNA2-based Ryzen AI NPU and newer RDNA 3.5 GPU. The NPU on its own will provide 45 TOPS of AI workload performance.

Last month, Intel launched an investigation into the recent reports of game crashes for owners of the company's high-end CPU.

A report from BenchLife via VideoCardz, claims that Intel has sent guidance to motherboard partners to roll out the Intel Default Settings on Z789 motherboards. These new default settings will restrict a PL2 of 188 watts, which is the maximum boost power that the CPU can draw sparingly during heavy loads. For context, PL1 is the base power the CPU can sustain for longer periods of time.

The report claims Intel's guidance is to restrict the performance of its high-end CPUs, particularly the Core i9-14900K, which appeared to be the CPU getting the most reports of instability issues. Notably, GIGABYTE has already rolled out its update for PL2, being restricted to 188W on the Core i9-13900K and Core i9-14900K, and according to testing conducted by Hardware Unboxed, it resulted in an overall 20% performance reduction for both CPUs.

Someone has purchased an AMD Ryzen 9 7950X3D processor with 3D V-Cache stacks on both Zen4 core dies, which is 64MB more L3 cache than a retail 7950X3D chip.

A user posted on Bilibili, with their purported Ryzen 9 7950X3D featuring 192MB of L3 cache, which is 64MB more L3 cache than retail Ryzen 9 7950X3D. The obvious red flags: this could be an error with CPU-Z which lists 192MB, it could be a modification, or something else.

This could be an engineering sample Ryzen 9 7950X3D with 3D V-Cache stacks on both Zen 4 core dies, as the Ryzen 9 7950X3D (and Ryzen 9 7900X3D) feature just one 3D V-Cache stack. Maybe this is a bigger discovery, an engineering sample 7950X3D with 192MB of L3 cache that never hit the market... very cool. Or, to spoil the party, this could just be a simple error.

Intel's next-gen Core Ultra 9 285K "Arrow Lake" CPU has a rumored 5.5GHz boost clock, compared to up to 6.2GHz on the current-gen flagship Core i9-14900KS processor.

The news on the Core Ultra 9 285K processor is coming from some posts by MebiuW on Weibo, where the big architectural changes that Intel will be bringing with Arrow Lake -- as well as the new Core Ultra 200 series branding -- will see a hefty CPU frequency reduction.

Intel's current-gen Raptor Lake Refresh CPUs can boost at up to 6.2GHz out of the box with the Core i9-14900KS, while even the Raptor Lake CPUs before it with the Core i9-13900KS we saw 6.0GHz. The new Arrow Lake CPUs however, seem to be topping out at 5.5GHz, but will feature a new tile-based design with an integrated NPU for AI workloads... something desktop Core CPUs didn't have until Meteor Lake hit laptops late last year.

Intel's Arrow Lake desktop processors might pitch up earlier than expected, or at least ahead of what the rumor mill seems to have been floating as the most likely arrival timeframe recently.

If you've been following the chatter around Arrow Lake (Core Ultra 200), you won't have missed a certain amount of pessimism creeping in of late.

In the early days of leaks, there was talk of massive performance gains, and while that's still true to an extent, enthusiasm about Arrow Lake seems to have been waning somewhat with suggestions that Intel is struggling to finish the next-gen desktop silicon, and that it might even get pushed to 2025. (Or at least that volume production may not happen until next year, anyway).

Intel has just announced that it has surpassed 500 AI models running optimized on its Core Ultra "Meteor Lake" CPUs.

The company issued a press release today, saying that the industry's leading AI PC processor is available right now enabling new AI experiences, immersive graphics and optimal battery life. This significant milestone is a result of the investments by Intel in client AI, the AI PC transformation, framework optimizations and AI tools including OpenVINO toolkit.

Intel said that the 500 AI models that were trained can be deployed across the CPU, GPU, or NPU, with models drawing from categories of local AI inferencing, including large language, diffusion, super resolution, object detection, image classification/segmentation, computer vision, and more.

Intel's next-generation Core Ultra 200 "Arrow Lake-S" desktop processors will launch later this year, with 6 different SKUs and a flagship 24-core CPU based on next-gen P-Core and E-Core architectures.

Now we've got some new information regarding upcoming SKUs in the Arrow Lake-S desktop CPU family, with the Core Ultra 9 285K, Core Ultra 7 265K, and Core Ultra 5 245K. The new Core Ultra 9 285K will be the successor to the Core i9-14900K, the Core Ultra 7 265K is the successor to the Core i7-14700K, and finally, the Core Ultra 5 245K is the successor to the Core i5-14600K.

We'll also see non-K variants of these processors, with the Core Ultra 9 275, Core Ultra 7 255, and Core Ultra 5 240 processors. The leaks are coming from "OneRaichi" on X, who said that "the rules of names look like a little change".

AMD is preparing a huge wave of next-gen Zen 5 processors in the coming months, with its Strix Halo APUs shaping up to be beasts.

In a new video from Moore's Law is Dead, we're hearing whispers of AMD's new Strix Halo-LP APU, which would be a low-power version of Strix Halo with 6 to 8 cores, and 20 CUs of GPU power that would provide RTX 3050 performance-per-watt-style power.

MLID says that AMD's new Zen 5-powered Strix Halo APUs will come in four different SKUs: Best, Better, Good, and the new LP model we're talking about today. The Strix Halo "Best" APU will feature 8, 12, or 16 cores based on the new Zen 5 architecture, with 40 CUs of RDNA 3.5-based GPU power that provides RTX 4070-class performance. We'll see a 256-bit memory bus on the Strix Halo "Best" APU with around 270GB/sec of memory bandwidth.

Ampere Computing already has its first-gen AmpereOne CPU on the market, but its next-generation AmpereOne-2 and AmpereOne-3 processors have been teased: AmpereOne-3 will reportedly be offered with up to 256 cores ready to go.

AmpereOne features 192 physical cores, a 350W TDP, and 8-channel DDR5 memory, meaning it competes against Intel Xeon and AMD EPYC processors, but now we're hearing about AmpereOne-2 and AmpereOne-3. The Next Platform has the scoop, with Ampere gearing up for its second-generation AmpereOne processor known as AmpereOne-2.

The new AmpereOne-2 processors will feature an upgraded "A2" core architecture, and support for 12-channel DDR5 memory for even more performance. Ampere Computing says that its next-generation AmpereOne-2 processor will have a 33% increase in DDR5 memory controllers (12-channel versus 8-channel) with 50% more memory bandwidth, which will provide some very nice performance improvements.