Processors - Page 7

Google has some big upgrades going on under its custom Tensor G5 processor inside of its newly-announced Pixel 10 family of smartphones, including the new flagship Pixel 10 Pro Fold foldable smartphone.

The company has posted a new blog explaining its new Google Tensor G5 processor being the "biggest upgrade yet" and delivering a major performance boost, with up to 34% more CPU performance. There's up to 60% more TPU performance. Mixed with the upgraded CPU performance, this makes the new Pixel smartphones more responsive with everyday use like web browsing, and the latest AI features (which are packed into Pixel 10).

Google's new Tensor G5 processor is fabbed on TSMC's leading 3nm process node, allowing the company to cram in more transistors onto the chip, making it more powerful and efficient. There's an all-new Image Sensor Processor (ISP) inside of the new Tensor G5 chip, allowing higher-quality videos on a Pixel 10 smartphone, even in low lighting, with 10-bit video by default for 1080p, and 4K30.

AMD has apparently got a new Zen 5 desktop processor on the boil which is aimed at the budget end of the market, namely the Ryzen 5 9500F.

VideoCardz spotted the processor (which was also flagged by BenchLeaks on X) and it's a 6-core (12-thread) effort running at 5.05GHz as shown in the leaked GeekBench result which has surfaced for the CPU.

Obviously take this with some salt - it's a single leaked benchmark at the moment, so it's far from certain this chip is actually inbound - but it would appear to be a more affordable take on the AMD Ryzen 5 9600. That's a 6-core processor, and the compromise with the Ryzen 5 9500F would be slightly slower clocks, alongside a lack of integrated graphics (which is what the F suffix indicates).

AMD's next-generation Zen 7 processors will be coming to the current AM5 socket according to new leaks, with 32 cores and 64 threads being on offer with flagship Zen 7 chips in the future... all on AM5.

In a new video from Moore's Law is Dead, the leaker says that over the last couple of weeks he's been hearing more information and seeing more documentation regarding AMD's next-gen Zen 7 CPUs on the current AM5 socket. This has been a "recent decision" over at AMD, with plans changing recently that would see 32C/64T chips based on the future-gen Zen 7 architecture on AM5.

AMD committed to the AM5 socket lasting until 2025 and beyond back in December 2023, with Zen 3, Zen 4, and the current-gen Zen 5 processors all launched on AM5. AMD's next-gen Zen 6 processors will be released in 2026, also on AM5, while Zen 7 should be released in 2028, with these leaks suggesting Zen 7 is also on AM5.

Intel's next-gen Core Ultra 300 series "Arrow Lake Refresh" CPUs are coming later this year, with leaks on the flagship Core Ultra 9 385K processor being around 5-10% faster in gaming.

The performance leaks are from Moore's Law is Dead, where in his latest video he said that Intel's new Arrow Lake Refresh processors won't be Zen 6 competitors for AMD, but will beat the current fleet of Zen 5 chips, but won't beat the Zen 5 + X3D processors like the Ryzen 7 9800X3D and 9950X3D.

Intel will reportedly be increasing clock speeds with aims for 6.0GHz (up from 5.7GHz on Arrow Lake) while there will be "quite a bit" of an increase to the ring bus clock speeds, and plans to increase the die-to-die clocks to eliminate bottlenecks that were seen in Arrow Lake.

AMD's new Ryzen Threadripper PRO 9995WX processor has been tested running an insane 100 copies of the original Doom at once, and 8 copies of Crysis.

PC enthusiast YouTuber Level1Techs was using the Falcon Northwest Talon workstation PC powered by the new AMD Ryzen Threadripper PRO 9995WX processor with 96 cores and 192 threads installed into an ASUS WRX90 SAGE SE motherboard, a huge 256GB of RAM, and dual NVIDIA RTX Pro 6000 workstation GPUs with 96GB of VRAM each for a total of 192GB VRAM.

The YouTuber was running 400 copies of Doom at the same time, but then Level1Techs host Wendell accidentally left 100 copies of Doom running on the new system, only noticing they were running when his Cinebench R24 benchmark score was lower than he expected. But, that's when he revealed just how powerful AMD's new 96-core, 192-thread CPU was.

AMD's Ryzen 7 5700X3D processor has reached the end of the line, at least according to a report from a Dutch tech site.

This comes from Tweakers (as highlighted by VideoCardz) which claims that 'several' online stores have confirmed that the Ryzen 5700X3D isn't being shipped to retail outlets by AMD any longer.

So, any remaining stock is all that's left - in theory - and that explains why prices have been going up of late (in the Netherlands, anyway). Tweakers observes that the Ryzen 5700X3D has risen by around 15% in recent weeks, presumably as the inventory of the CPU started to dwindle.

Intel's next-generation Nova Lake-S desktop CPUs won't be here until 2026, but we've got some new leaks to share on Nova Lake-S with its up to 52-core configuration.

In a new video posted by leaker Moore's Law is Dead, we're hearing that the new Intel Nova Lake-S processors will feature 52 cores and 52 threads in total, something we've heard in previous leaks, with "bLLC" or "Big Last Level Cache" that is Intel's current answer to AMD's innovative (and kick ass) 3D V-Cache technology. The bLLC will be on-die L3 in the middle of the ringbus.

There are reportedly five Tiles on Nova Lake-S, with 2 x bLLC dies = 8+16 Compute Tiles, with MLID noting that we should expect games to only really utilize one bLLC Tile. We should expect the flagship Nova Lake-S processor to be made on TSMC's bleeding-edge N2P process node, with 52 cores + 144MB of bLLC per CCD for a huge 288MB of L3 cache in total.

Earlier this week, we heard some chatter about a new high-end (200W) 16-core AMD Ryzen 9000 3D V-Cache processor with two banks of cache, one per chiplet - and apparently this rumor is a fake.

This comes from another leaker - so obviously, we need to take a debunking like this with appropriate seasoning, as we did with the original rumor - but 'wjm47196' on the Chiphell forums (in China) seems pretty certain this is fakery.

As flagged by HXL on X (see the above post, noticed by Wccftech), the Chiphell leaker (who is well-known) wrote: "Fake, there is no such thing." They then elaborated that there may be a new X3D chip for the Zen 5 generation in the works, though.

AMD's next-generation AM6 socket will house next-generation Zen 7 processors, and it'll feature 2100 pins with a higher pin density than the AM5 socket, but it won't be any bigger.

We're hearing about AMD's next-gen AM6 socket from the folks over at Bits and Chips, reporting that AM6 has a higher pin density than AM5, something that was pointed out in an AMD patent labeled "US20250149248". The new AM6 socket will have up to 2100 pins, up from 1718 pins on AM5, which is a 22% increase in pin density, all while retaining the same socket size as AM5.

This is great news for AM5-compatible coolers, as they should work, on the next-gen AM6 socket as well. It also technically means that AM4-compatible coolers should work with AM6, but that'll be harder as the next-gen Zen 7 family of CPUs could see cooling manufacturers needing to release new AM6-ready cooling designs depending on the chiplet layout of the next-gen Zen 7 chips.

Intel is reportedly struggling with its next-gen 18A semiconductor process that was meant to be one of the saving graces of the company but is delivering huge headaches for its next-gen processors.

In a new report from Reuters, the production process that Intel is using to work its way back into securing chip-making deals for high-end, high-margin chips is "facing a big hurdle on quality as it puts newer technologies to the test". Intel has been promising its investors for months now that it would be increasing manufacturing using its new Intel 18A process.

Intel has spent billions of dollars developing its new 18A process node, including the construction and upgrades of multiple semiconductor factories, where it wanted to take the fight to semiconductor dominator, TSMC. Intel wants to shift from solely designing chips that TSMC produces for them, to owning a competitive contract manufacturing business capable of rivaling TSMC.

AMD is reportedly cooking up two new Ryzen 9000 series "Zen 5" CPUs with higher TDPs and massive injection of L3 cache, according to new leaks.

In a new post on X by leaker @chi11eddog, we're hearing that AMD is working on new Granite Ridge (Zen 5-based Ryzen 9000 desktop CPUs) with a 16-core, 32-thread CPU with 200W TDP and a huge 192MB of L3 cache, meaning we might be seeing dual X3D cache on a new flagship CPU.

AMD's current flagship Ryzen 9 9950X3D and Ryzen 9 9950X processors both have 16C/32T but TDPs of 170W, so this new 16C/32T + 200W TDP + 192MB cache chip has 30W more available to it, and 64MB more L3 cache. The Ryzen 9 9950X3D features 128MB of cache in total, while this new CPU has a huge 192MB of cache.

AMD has a cheaper spin on the Ryzen 7 9700X that drops the integrated graphics, and ASUS has spilled the beans on the processor (at least in theory).

This was spotted by Italian tech site Bits and Chips (as highlighted by VideoCardz), which found the Ryzen 7 9700F - note the 'F' suffix meaning it lacks the GPU - on the ASUS website.

The Ryzen 7 9700F (which has been rumored previously) pops up in the CPU support details page for ASUS motherboards on the official site if you type in the model name, confirming its existence - and presumably the fact that the 9700F is not far off launching. (Either that, or someone has made a silly mistake here).

Intel will be reintroducing SMT or Hyper-Threading to future-generation CPUs says its new CEO Lip-Bu Tan, in the middle of laying off 15% of its overall staff and 50% of its management layers.

We don't know if the return of SMT will be for data center CPUs, but it would be nice to see the reintroduction of Hyper-Threading to consumer gaming processors to increase its competitiveness against CPU rival AMD. Intel began moving away from Hyper-Threading (or SMT) with its 12th Gen Core CPUs, as it started moving toward asymmetrical CPU designs.

Intel first enabled SMT on P-Cores and then after a while, completely eliminated Hyper-Threading on its 15th Gen Core CPUs. Intel's next-gen Nova Lake CPUs will NOT feature SMT, or even the generation after that, with rumors that we'll see "Unified Cores" with its future-gen Titan Lake CPUs.

Samsung Electronics has announced that it has signed a new $16.5 billion contract to supply semiconductors to a "major global corporation" which is rumored to be Tesla for FSD chips.

Tesla needs as many FSD (Full Self-Driving) chips that it can get, with reports that the Elon Musk-led company secured orders for its latest Tesla FSD chips with TSMC on its 4nm/5nm process nodes in March 2025. The new $16.5 billion contract started on July 24, 2025 and runs through to December 31, 2033.

The "major global corporation" hasn't been disclosed, but insider @Jukanrosleve on X says it's "presumed to be with Tesla" which makes sense. Samsung is also fabbing NVIDIA's new custom Tegra T239 SoC for Nintendo and its new Switch 2 gaming handheld, but a huge Tesla contract would be a big win for Samsung Foundry which has been struggling in the semiconductor business against TSMC for many years now.

Chinese server CPU manufacturer, Zhaoxin, has just unveiled its next-gen KH-50000 chiplet CPUs for servers boasting up to 96 cores.

The new Zhaoxin KH-50000 chiplet CPUs are a serious upgrade over the current KH-40000 offerings, with KH-40000 offerings featuring up to 32 cores, 32 threads, 64MB cache, 8-channel DDR4 memory support, and 128 PCIe 4.0 lanes. The chips feature a chiplet CPU architecture, meaning they come in various SKUs with 16-core and 12-core offerings with CPU clocks of up to 2.5GHz.

Zhaoxin's next-gen KH-50000 chiplet CPUs boast a considerable upgrade in CPU power, unveiled at the Shanghai World Expo 2025 event, based on a chiplet architecture and feature support for an LGA socketed platform. The new Zhaoxin KH-50000 chips feature up to 96 cores (no confirmation on SMT, but we could see 96C/192T like AMD's new EPYC processors).

Intel Foundry is in serious trouble if it can't find big customers to use its Intel 14A process, and, wouldn't you believe it, there are rumors flying that Apple might use Intel as its semiconductor partner for upcoming M-series processors.

In a new report, GF Securities analyst Jeff Pu, who has the inside track at Apple, claims Intel is sampling early versions of Intel 14A PDK to its customers, and that both Apple and NVIDIA are interested in using Intel Foundry as a chipmaking partner.

Pu said: "The next focus Intel 14A process will incorporate second-gen RibbonFET and PowerDirect, marking a technological evolution built upon the foundation of the PowerVia introduced in Intel's 18A. Targeting both AI and edge applications, Intel has already provided early versions of the 14A PDK to key customers, with several expressing interest in producing test chips. We anticipate NVIDIA's gaming GPU (low-end version) and Apple's M series to emerge as adopters of Intel 14A".

NVIDIA is still cooking its upcoming N1X AI PC processor, based on Arm CPU cores and Blackwell GPU cores. But now we've got some Geekbench scores that show it's as fast as a GeForce RTX 5070, and faster than all other integrated GPUs on the market.

In some new Geekbench OpenCL browser benchmarks, we have the upcoming NVIDIA N1X processor scoring 46,361 points. We've also got some details on the Blackwell-based GPU which rocks 6144 CUDA cores and 48 SM units, the same core count as the GeForce RTX 5070 which is based on the GB205 "Blackwell" GPU.

NVIDIA's GeForce RTX 5070 features the GB205 GPU, which contains 192 TMUs, 80 ROPs, 48 RT Cores, and 192 Tensor Cores. There's also 12GB of GDDR7 memory with up to 672GB/sec of memory bandwidth, a GPU boost clock of up to 2512MHz and a 250W TDP.

AMD's brand new Ryzen Threadripper PRO 9995WX processor has been overclocked to 4.9GHz on all CPU cores using an interesting method of cooling: BMW's M4 radiator (yeah from the car) and a Toyota Highlander radiator fan (also from the car).

The new Zen 5-based Ryzen Threadripper PRO 9995WX processor features 96 cores and 192 threads of processing power, and is capable of consuming over 1000W of power with manual overclocking and tweaking, leading a tech YouTuber to use the BMW M4 radiator and Toyota Highlander radiator fan to keep things much cooler than a high-end AIO cooler.

Car radiators are much bigger than any traditional AIO radiator, where they're not limited to the constraints of 360mm or even a bigger 420mm in size, with their pumps capable of pumping far more liquid per second which increases the cooling process. Geekerwan used the M4 radiator from his personal BMW, which is one of the best-performing radiators on the market.

SIXUNITED recently unveiled its new STHT1 Thin Mini-ITX motherboard, ushering in AMD's new Ryzen AI Max 300 series "Strix Halo" APU to the PC enthusiast market. Check it out:

AMD's new Strix Halo APUs have only been released inside of high-end laptops and Mini-PC systems so far, with the new Strix Halo APU featuring 16 cores and 32 threads of Zen 5 processing power, enhanced RDNA 3.5-based integrated graphics with the Radeon 8060S GPU with up to 40 GPU cores, and support for up to 128GB of RAM.

SIXUNITED is a Chinese ODM that has just unveiled its new Mini-ITX motherboard with the Strix Halo APU ready with a non-socketed design that allows builders to choose between three different Ryzen AI Max series processors, with TDPs of between 45W and 120W.

NVIDIA and MediaTek have reportedly delayed their collaborative Arm-based AI PC processor into Q1 2026, after rumors that there were issues with the silicon in the last few days.

In a new report from DigiTimes, the outlet says their supply chain sources say that a combination of delays in Microsoft's operating system roadmap, ongoing chip revisions at NVIDIA, and weakening demand in the overall notebook market have seen the Arm-based AI PC processor delayed until the first quarter of 2026.

The two companies originally aimed to launch the "N1X" processor in Q3 2025, aiming at both the premium consumer and commercial sectors, but after the unveiling event not taking place at Computex 2025 in late May, and now the new issues with the silicon and Windows on Arm, we're looking at a possible CES 2026 reveal and Q1 2026 release.