Tech News
Back to home

Articles tagged with "chips"

Intel details 18A-P process node, touts higher performance, lower power, and better thermals — 9% more performance, thermal conductivity improved by 50%

Intel details 18A-P process node, touts higher performance, lower power, and better thermals — 9% more performance, thermal conductivity improved by 50%

Intel is progressing with its 18A-P process node technology, boasting higher performance, lower power consumption, and improved thermals compared to its baseline 18A technology. The 18A-P process introduces new transistors, tighter process variability control, and enhanced thermals, potentially attracting interest from companies like Apple. This enhanced technology offers a 9% performance boost or an 18% power reduction compared to the base 18A process. Additionally, improvements in skew corner tightening, threshold voltage options, and metal stack resistance and capacitance contribute to better performance efficiency and yield. The 18A-P process also focuses on enhancing thermals, reliability, and voltage behavior critical for both client and data center applications.

Tom's Hardware•
Newegg finally has a worthwhile Intel combo that includes memory, saving almost $240 — snag an Intel Core Ultra 7 270K Plus, ASRock Z890 motherboard, and 32GB of GSkill Trident Z5 DDR5 RAM for only $768.34

Newegg finally has a worthwhile Intel combo that includes memory, saving almost $240 — snag an Intel Core Ultra 7 270K Plus, ASRock Z890 motherboard, and 32GB of GSkill Trident Z5 DDR5 RAM for only $768.34

Newegg is offering a compelling combo deal including an Intel Core Ultra 7 270K Plus processor, ASRock Z890 Pro RS motherboard, and 32GB of G.Skill Trident Z5 RGB RAM for $768.34, saving buyers almost $240. The Intel Core Ultra 270K Plus processor, despite facing tough competition from AMD, offers excellent productivity performance at a lower price point of $328.34. The GSkill Trident Z5 RGB RAM, paired with the ASRock motherboard, provides good performance and overclocking potential. This combo deal on Newegg is a rare find and a solid option for those looking to upgrade their PC amidst high component prices.

Tom's Hardware•
Get 32GB of RAM and 2TB SSD for $406 when paired with AMD's new 9950X3D2 — Asus ROG X870E Apex gets you a monster overclocking AM5 bundle for $2,048, $400 off

Get 32GB of RAM and 2TB SSD for $406 when paired with AMD's new 9950X3D2 — Asus ROG X870E Apex gets you a monster overclocking AM5 bundle for $2,048, $400 off

Newegg offers a premium combo deal featuring the Ryzen 9 9950X3D2 processor, Asus ROG Crosshair X870E Apex motherboard, 32GB of Corsair Vengeance RGB RAM, and a 2TB Western Digital Black M.2 SSD for $2,048.99, saving you $398.98. The Ryzen 9 9950X3D2 processor boasts improved multi-threaded performance and specialized workload capabilities, while the Crosshair X870E Apex motherboard is designed for overclocking. The Corsair Vengeance RAM kit and Western Digital Black SN7100 SSD complete the high-performance bundle. This deal provides a top-tier AM5 system suitable for gaming, productivity, and more.

Tom's Hardware•
Framework's new RTX 5070 12GB graphics module costs a whopping $1,199 — 72% more expensive than $699 8GB version, says pricing is beyond its control

Framework's new RTX 5070 12GB graphics module costs a whopping $1,199 — 72% more expensive than $699 8GB version, says pricing is beyond its control

Nvidia released a new 12GB version of the RTX 5070 mobile GPU with upgraded memory chips, increasing memory throughput. Framework introduced a new graphics module for its Framework Laptop 16 featuring this GPU, priced at $1,199, a significant increase from the $699 8GB version. The high cost is attributed to the expensive GDDR7 memory and the ongoing global memory shortage. Framework clarified that the pricing is influenced by external factors and not within its control, highlighting the challenges faced by consumers due to the current market conditions.

Tom's Hardware•
Palit Group says Galax GPU brand will continue to operate following restructure — Galax management centralized under Palit Group in 'pre-planned' shakeup

Palit Group says Galax GPU brand will continue to operate following restructure — Galax management centralized under Palit Group in 'pre-planned' shakeup

Palit Group has confirmed that the Galax GPU brand will continue to operate and release hardware despite recent reports suggesting otherwise due to an internal restructuring. The management of Galax will now be centralized under Palit Group, aiming to enhance operational efficiency and synergy between the brands. Galax reassured customers that they remain committed to developing and supporting high-performance hardware. The move involves centralizing brand management at Palit Group headquarters, with Galax now being managed directly from Taiwan. Despite the changes, Galax emphasizes that the goal is to strengthen its global presence.

Tom's Hardware•
More details emerge about how Intel now earns more revenue from each wafer by looking to the edges — analyst reports say reduced yield variability across each wafer leads to more sellable CPUs

More details emerge about how Intel now earns more revenue from each wafer by looking to the edges — analyst reports say reduced yield variability across each wafer leads to more sellable CPUs

Intel has seen improved revenue per wafer by reducing yield variability across each wafer, resulting in more sellable CPUs. The company's focus on tightening yield distribution across the wafer edges has led to increased margins and productivity. By implementing edge-specific process correction methods, Intel can extract more high-quality and sellable dies from a single wafer. These improvements are node-independent and have been attributed to disciplined execution improvements under new manufacturing leadership. Intel's efforts have led to better output and demand for CPUs, with even lower-quality chips now being sold as viable products.

Tom's Hardware•
Chinese GPU maker Lisuan Tech becomes only the fourth GPU maker ever to earn Microsoft WHQL certification — LX 7G100 GPU joins Nvidia, AMD, and Intel as it crosses the WHQL driver finish line, first Chinese firm to earn certification

Chinese GPU maker Lisuan Tech becomes only the fourth GPU maker ever to earn Microsoft WHQL certification — LX 7G100 GPU joins Nvidia, AMD, and Intel as it crosses the WHQL driver finish line, first Chinese firm to earn certification

Lisuan Tech has become the fourth GPU maker to earn Microsoft's WHQL certification, alongside Nvidia, AMD, and Intel. This certification for its 6nm graphics cards marks a significant milestone for the Chinese semiconductor industry. While Chinese graphics card designers have shown prowess in hardware, they have historically faced challenges in software, making this achievement by Lisuan Tech noteworthy. The WHQL certification ensures software reliability and compatibility, potentially opening up global markets for the company. This certification also signifies a step towards expanding beyond China's borders and offering a seamless plug-and-play experience for users.

Tom's Hardware•
Nvidia quietly launches 12GB RTX 5070 laptop GPU — midrange mobile gaming gets more VRAM amid the RAMpocalypse

Nvidia quietly launches 12GB RTX 5070 laptop GPU — midrange mobile gaming gets more VRAM amid the RAMpocalypse

Nvidia has quietly released a new 12GB RTX 5070 laptop GPU, offering 50% more memory capacity than the existing 8GB version. This new option aims to address the ongoing memory shortage crisis, providing gamers and creators with more VRAM for improved performance. The 12GB RTX 5070 will sit between the 8GB and Ti 12GB versions in Nvidia's mobile GPU lineup, offering enhanced capabilities for AAA gaming and higher graphical settings. This move marks a rare instance of Nvidia offering higher VRAM options amidst industry trends of reducing memory capacities on consumer hardware.

Tom's Hardware•
China announces CPU-only exascale supercomputer with 47,000 homemade processors, record 2 Exaflops of performance without GPUs — Lingshen super said to use Huawei Kunpeng servers and no foreign-made components

China announces CPU-only exascale supercomputer with 47,000 homemade processors, record 2 Exaflops of performance without GPUs — Lingshen super said to use Huawei Kunpeng servers and no foreign-made components

China's National Supercomputing Center in Shenzhen unveiled the Lingshen supercomputer project, aiming for over 2 ExaFLOPS performance using 47,000 homemade processors without GPUs or foreign components. The system, designed to surpass the current fastest supercomputer, El Capitan, would utilize Huawei Kunpeng servers and Arm-based Taishan cores. The project includes a pilot phase with 100 servers and a full production system with 1,580 blade servers. While China's claims of achieving 2+ ExaFLOPS are ambitious, questions remain about the feasibility of surpassing existing supercomputing benchmarks without GPUs or foreign-made CPUs.

Tom's Hardware•
Legendary ZSNES Nintendo emulator rewritten from scratch with GPU-acceleration, no vibe coding — new Super ZSNES has ‘far more accurate CPU and audio cores than the original’

Legendary ZSNES Nintendo emulator rewritten from scratch with GPU-acceleration, no vibe coding — new Super ZSNES has ‘far more accurate CPU and audio cores than the original’

The legendary ZSNES Nintendo emulator has been completely rewritten with GPU acceleration, resulting in the new Super ZSNES. This reboot, created by zsKnight and Demo, offers more accurate CPU and audio cores than the original version. The emulator now utilizes a GPU-powered PPU core and features a Super Enhancement Engine for improved performance. Despite being in its early stages with bugs to be fixed, Super ZSNES is available for Windows, Mac, Linux, and Android, with an iOS release on the way.

Tom's Hardware•
The GeForce RTX 30-series upgrade matrix — does your Ampere GPU need an upgrade in 2026?

The GeForce RTX 30-series upgrade matrix — does your Ampere GPU need an upgrade in 2026?

Nvidia's RTX 30-series graphics cards are approaching their sixth birthdays, and while they continue to receive support, newer software features like DLSS Frame Generation are not available for Ampere GPUs. The limitations of these cards, particularly in terms of VRAM, may be felt when playing the latest games at high resolutions. The article provides upgrade recommendations for different Ampere cards based on performance improvements and features like DLSS 4.5 and MFG. It suggests considering factors like monitor resolution and refresh rate before deciding on a GPU upgrade. For cards like the RTX 3080, RTX 3070, and RTX 3060, the article recommends upgrading to newer models for better performance and VRAM capacity.

Tom's Hardware•
Intel reportedly says it boosted yields by selling what would normally be 'scrap' or 'low-expectation' CPUs — customers more willing to accept lesser chips due to overwhelming CPU demand

Intel reportedly says it boosted yields by selling what would normally be 'scrap' or 'low-expectation' CPUs — customers more willing to accept lesser chips due to overwhelming CPU demand

Intel has seen an unexpected increase in profit margins by selling CPUs that would typically be considered scrap or low-quality due to overwhelming demand for CPUs. The company's first-quarter earnings surpassed expectations, with revenue hitting $13.6 billion and gross margins reaching 41%. Intel has been able to sell chips that would normally be discarded as lower-tier SKUs at usable prices, contributing to the revenue boost. The high demand for CPUs, particularly for AI-driven infrastructure, has led customers to purchase even lower-quality chips, providing Intel with additional revenue.

Tom's Hardware•
Intel posts very strong Q1 2026 earnings as the AI boom starts to focus on CPUs

Intel posts very strong Q1 2026 earnings as the AI boom starts to focus on CPUs

Intel's Q1 2026 earnings exceeded expectations, with $13.58 billion in revenue, up 7.2% year over year. The data center and AI divisions contributed significantly to this success, generating $5.1 billion in revenue. Intel's foundry business is gaining momentum, with $5.4 billion in revenue, positioning itself as a strong competitor to TSMC. Recent partnerships with Tesla and Google, along with new hardware releases, have further boosted Intel's performance. The increasing demand for AI-focused CPUs is driving growth for companies like Intel, prompting them to raise chip prices and forecast Q2 revenue of $13.8 to $14.8 billion.

TweakTown•
Intel's upcoming Xeon 7 "Diamond Rapids" server CPUs reportedly delayed to 2027 — Next-gen Coral Rapids lineup lands 2028 but can be accelerated, according to new leak

Intel's upcoming Xeon 7 "Diamond Rapids" server CPUs reportedly delayed to 2027 — Next-gen Coral Rapids lineup lands 2028 but can be accelerated, according to new leak

Intel's Xeon 7 "Diamond Rapids" server CPUs have been delayed to 2027, with the 8-channel variant canceled in favor of focusing on 16-channel SKUs. The lineup will feature up to 256 cores initially, with a 512-core version to follow shortly after. Diamond Rapids will lack hyperthreading support, unlike the next-gen Coral Rapids lineup, which is expected to launch in mid-2028 with 8-channel variants. Intel may accelerate the rollout of Coral Rapids based on market demand. AMD's EPYC Venice CPUs are still on track for a 2026 release.

Tom's Hardware•
Intel has reportedly cancelled discrete gaming GPUs for the upcoming Xe3P Arc "Celestial" family — gaming GPU remains uncertain even for the next-gen Xe4 "Druid" lineup that lands in 2027

Intel has reportedly cancelled discrete gaming GPUs for the upcoming Xe3P Arc "Celestial" family — gaming GPU remains uncertain even for the next-gen Xe4 "Druid" lineup that lands in 2027

Intel has reportedly scrapped plans for discrete gaming GPUs in the upcoming Xe3P Arc "Celestial" family, leaving the fate of gaming GPUs uncertain even for the Xe4 "Druid" lineup expected in 2027. The Celestial GPU was originally intended for a 2025 launch but was replaced by Battlemage, with Xe3P now serving other purposes. Intel's focus seems to be shifting towards AI applications, with leaks suggesting a potential late-2027 release for the Druid architecture. The future of dedicated gaming GPUs from Intel remains speculative, with the possibility of a revival with the Druid lineup.

Tom's Hardware•
Intel Core Ultra 7 270K Plus vs AMD Ryzen 7 7800X3D — Can Intel finally beat X3D?

Intel Core Ultra 7 270K Plus vs AMD Ryzen 7 7800X3D — Can Intel finally beat X3D?

Intel's Arrow Lake Refresh CPUs, the Core Ultra 7 270K Plus and Ryzen 7 7800X3D, are compared for gaming and productivity. The 270K Plus, with 24 cores and 24 threads, offers better gaming performance, but the 7800X3D leads in gaming benchmarks by 10.4%. In productivity tests, the 270K Plus excels with a 29% lead in single-threaded performance and a significant advantage in multi-threaded tasks. Intel's CPU also wins in overclocking capabilities, offering more performance-enhancing features. However, the 7800X3D is more power-efficient and costs less, making it a strong competitor.

Tom's Hardware•
Intel VP claims up to 30% of CPU performance is untapped by modern games — software optimization is critical to unlocking full potential of hybrid CPUs

Intel VP claims up to 30% of CPU performance is untapped by modern games — software optimization is critical to unlocking full potential of hybrid CPUs

Intel's VP Robert Hallock emphasizes that up to 30% of CPU performance in modern games remains untapped due to software optimization issues with hybrid CPUs. Hallock discusses the impact of disabling E-cores on Intel CPUs for better performance, noting that both E-cores and P-cores contribute equally to overall operation. He highlights the importance of software optimization in maximizing CPU performance, mentioning Intel's binary optimization feature in Arrow Lake chips as an example. Hallock suggests that software optimization is crucial for extracting more performance from existing silicon, with potential for significant gains in efficiency.

Tom's Hardware•
AMD's memory-boosting EXPO 1.2 is here, adds support for three Chinese memory vendors — performance gains could be muted until Zen 6

AMD's memory-boosting EXPO 1.2 is here, adds support for three Chinese memory vendors — performance gains could be muted until Zen 6

AMD has released EXPO 1.2, a memory-boosting technology for AM5 motherboards, enhancing memory overclocking capabilities and supporting three Chinese memory vendors. Notable improvements include support for module geometry, MRDIMMs for higher bandwidth, and Ultra Low Latency mode for reduced memory latency. However, the full performance gains may only be realized with AMD's upcoming Zen 6 processors. Asus has begun rolling out beta firmware with EXPO 1.2 support, signaling a commitment to enhancing the AM5 platform, though the full impact may not be felt until Zen 6 processors arrive in late 2026 or early 2027.

Tom's Hardware•
Intel Xeon 7 Diamond Rapids reportedly delayed to mid-2027, followed by Coral Rapids in 2028

Intel Xeon 7 Diamond Rapids reportedly delayed to mid-2027, followed by Coral Rapids in 2028

Intel's Xeon 7 Diamond Rapids has reportedly been delayed to mid-2027, with Coral Rapids expected in 2028 according to hardware enthusiast Jaykihn. The Xeon 6+ "Clearwater Forest" is set to launch in the first half of 2026, featuring energy-efficient E-cores based on the Darkmont architecture and Foveros 3D packaging. Intel's Granite Rapids will continue production as the 16-channel DMR faces delays, with the company focusing on high-end enthusiast and enterprise markets. The delay could impact Intel's DCAI earnings and profit margins, especially with AMD's upcoming "Venice" launch. Intel's return to SMT with Coral Rapids aims to succeed Diamond Rapids, with details on these CPUs yet to be revealed.

TweakTown•
NEO Semiconductor's revolutionary 3D X-DRAM for AI processors has passed proof-of-concept validation — company secures funding to develop next-gen memory HBM alternative

NEO Semiconductor's revolutionary 3D X-DRAM for AI processors has passed proof-of-concept validation — company secures funding to develop next-gen memory HBM alternative

NEO Semiconductor's 3D X-DRAM technology has successfully completed proof-of-concept validation, showcasing a new high-density DRAM class achievable through existing 3D NAND infrastructure. The company has secured funding for further development, with a strategic investment led by Stan Shih. The 3D X-DRAM technology offers enhanced density, lower power consumption, and AI workload suitability by leveraging 3D NAND manufacturing techniques. Industry experts view this advancement as a significant milestone in overcoming traditional DRAM scaling limits, particularly in the context of increasing demands from AI workloads. While the technology shows promise, it remains in the proof-of-concept stage, with the journey to commercial viability ahead.

Tom's Hardware•
SemiWiki Acquires IPnest!

SemiWiki Acquires IPnest!

SemiWiki has acquired IPnest, known for its IP reports, with Eric Esteve staying on until 2026 for a smooth transition. This acquisition will expand SemiWiki's coverage of Interface IP and Design IP reports. Esteve's extensive experience in the semiconductor industry and IP market has been instrumental in the success of IPnest. The growth of the Interface IP market, driven by AI advancements, underscores the importance of robust interconnect protocols for building powerful AI systems. Esteve expresses confidence in SemiWiki's ability to elevate IPnest reports to a higher level.

SemiWiki•
Elon Musk Needs to Put His Fab Money Where his Mouth is!

Elon Musk Needs to Put His Fab Money Where his Mouth is!

Elon Musk plans to use Intel's 14A manufacturing process for chips at his Terafab project in Austin. Musk expressed concerns about global chip production meeting future demands during a presentation, emphasizing the need for Terafab. Intel's move into foundry competition is seen as a win for TSMC and the semiconductor industry, with potential impacts on innovation and supply chain strength. Musk's collaboration with Intel and Lip-Bu Tan is crucial for success in semiconductor design and manufacturing, signaling a positive disruption in the industry.

SemiWiki•
Intel stock jumps 28%, setting a record, after it posts strong Q1 with rising forecasts — Intel says yields are improving faster than expected with new nodes

Intel stock jumps 28%, setting a record, after it posts strong Q1 with rising forecasts — Intel says yields are improving faster than expected with new nodes

Intel's stock surged by 28% to a record high after a strong Q1 performance exceeding expectations, driven by increased demand for data center CPUs and improved productivity at Intel's fabs. Despite a $3.7 billion loss due to Mobileye goodwill write-downs and restructuring charges, the company reported $1.5 billion in net income on a non-GAAP basis. Intel's revenue reached $13.6 billion in Q1, with notable growth in its Client Computing Group and Data Center and AI segment. The company's Foundry division, while showing revenue growth, posted an operating loss, mainly attributed to ramping up new nodes like 18A. Intel's outlook for Q2 includes modest growth but warns of potential supply constraints impacting PC sales due to memory and SSD shortages.

Tom's Hardware•
Intel CEO says he can 'think of no better partner than Elon Musk' to explore 'unconventional' ways to improve chip manufacturing — TeraFab partnership aims to rethink how chips are made to reduce costs

Intel CEO says he can 'think of no better partner than Elon Musk' to explore 'unconventional' ways to improve chip manufacturing — TeraFab partnership aims to rethink how chips are made to reduce costs

Intel CEO Lip-Bu Tan expressed strong support for a partnership with Elon Musk to explore innovative ways to enhance chip manufacturing, specifically through the TeraFab project. The collaboration aims to revolutionize chip production processes to reduce costs. While the exact nature of the partnership remains unclear, it is speculated that TeraFab may license Intel's 14A node technology. Tan hinted at potential unconventional approaches to improve manufacturing efficiency, emphasizing the need to address the growing demand for semiconductors. The partnership with TeraFab marks a significant development for Intel's Foundry business, which has faced financial challenges in recent years. Intel's stock saw a notable increase following the announcement of the partnership.

Tom's Hardware•
First Intel Wildcat Lake laptop spotted in the wild, geared to compete with MacBook Neo — features an aluminum chassis with 11W fanless mode

First Intel Wildcat Lake laptop spotted in the wild, geared to compete with MacBook Neo — features an aluminum chassis with 11W fanless mode

Intel's Wildcat Lake laptop, a potential competitor to the MacBook Neo, has been spotted in a reference design featuring a mix of Cougar Cove P-cores, Darkmont E-cores, and Xe3 integrated graphics cores. The device offers various power modes, including an 11W fanless mode, and boasts an aluminum chassis. With 16GB of RAM and superior performance compared to the Neo, this laptop showcases the capabilities of Intel's latest CPU lineup. Despite limitations in testing, it's evident that Wildcat Lake devices may benefit from enhanced cooling solutions for optimal performance.

Tom's Hardware•
Intel shutters open-source evangelism program and archives key community projects — closures point to significant shift in open-source leadership

Intel shutters open-source evangelism program and archives key community projects — closures point to significant shift in open-source leadership

Intel has closed its Open Ecosystem Community and Evangelism initiative, signaling a shift away from its open-source advocacy role. The move is part of a broader trend of discontinuing internal open-source efforts amid restructuring. The company has archived key projects on GitHub, including AI, infrastructure, and developer tools, due to limited activity and maintenance challenges. This shift reflects Intel's evolving open-source strategy towards a more selective, product-focused approach, impacting developer engagement and platform visibility. The company's move away from broad ecosystem cultivation may have long-term implications for its open-source posture.

Tom's Hardware•
Carbon in the Age of AI Chips: What the Semiconductor Industry Needs to Know This Earth Day

Carbon in the Age of AI Chips: What the Semiconductor Industry Needs to Know This Earth Day

The semiconductor industry is facing increasing pressure to address carbon emissions, with a new report from TechInsights highlighting the impact of AI chips on the environment. The report projects a significant rise in fabrication emissions, particularly from advanced technologies and AI accelerators. It emphasizes the importance of considering manufacturing carbon as a strategic variable in decision-making processes. The study also examines the carbon footprint of consumer and enterprise devices, revealing that memory and storage components contribute significantly to embodied carbon. To mitigate emissions, the report suggests actions such as pursuing cleaner electricity sources and optimizing platform configurations.

SemiWiki•
Bolt Graphics tapes out its first Zeus GPU test chip on TSMC 12nm — firm touts 17x lower cost of compute

Bolt Graphics tapes out its first Zeus GPU test chip on TSMC 12nm — firm touts 17x lower cost of compute

Bolt Graphics has successfully taped out its first Zeus GPU test chip on TSMC's 12nm process, transitioning from FPGA emulation to manufactured silicon. The company claims a significant 17 times lower cost of compute with this move. The test chip will undergo customer benchmarking before a planned production increase in Q4 2027, targeting markets like HPC and graphics rendering. Bolt's Zeus GPU, initially introduced in 2025, aims to offer high performance and efficiency, with plans to utilize more advanced nodes for final production. The company's product pipeline exceeds $500 million, with significant interest from various stakeholders, though specific funding details remain undisclosed.

Tom's Hardware•
TSMC unveils process technology roadmap through 2029 — A12, A13, N2U announced, A16 slips to 2027

TSMC unveils process technology roadmap through 2029 — A12, A13, N2U announced, A16 slips to 2027

TSMC unveiled its process technology roadmap through 2029, introducing A12, A13, and N2U while pushing back A16 to 2027. The company emphasized a segmented strategy for nodes based on end-market requirements rather than a one-size-fits-all approach. A13 and N2U are geared towards client applications, offering incremental improvements and IP reuse, while A16 and A12 target AI and HPC applications with significant performance enhancements. Notably, A13 and A12 will not require High-NA EUV lithography tools, showcasing TSMC's innovative scaling approach.

Tom's Hardware•
Where to buy AMD's Ryzen 9 9950X3D2 — more L3 cache and dual 3D V-cache-stacked CCDs

Where to buy AMD's Ryzen 9 9950X3D2 — more L3 cache and dual 3D V-cache-stacked CCDs

The new AMD Ryzen 9 9950X3D2 processor is now available for purchase at various retailers for $899. It is designed for high-end tasks like Blender, 3D rendering, and content creation, with gaming performance similar to the 9800X3D. The 9950X3D2 features 3D V-cache technology, a larger L3 cache of 192MB, dual CCDs with stacked V-cache, and a 16-core, 32-thread setup with a 200W TDP. Despite its higher price compared to the 9950X3D, stock availability is not expected to be an issue. The processor excels in multi-threaded performance and is a solid choice for demanding tasks.

Tom's Hardware•
Congress moves to strip the DoC of chip-export discretion with the MATCH Act — DUV lithography machines among those targeted in chipmaking tool crackdown

Congress moves to strip the DoC of chip-export discretion with the MATCH Act — DUV lithography machines among those targeted in chipmaking tool crackdown

A bipartisan group of U.S. lawmakers introduced the MATCH Act to restrict the sale and servicing of advanced chipmaking equipment to China, targeting DUV lithography systems and specific Chinese semiconductor firms. The bill aims to eliminate loopholes exploited by Chinese companies through front companies and third-country routing. The legislation would transfer authority from the executive branch to Congress, making any future relaxation of restrictions require congressional approval. The bill's path through Congress is uncertain, with equipment makers lobbying against it, but the bipartisan support and current trade tensions between the U.S. and China may lead to its passage.

Tom's Hardware•
AMD SDK suggests 4x and 6x frame generation multipliers are in the works — driver-level tweak will let gamers upgrade existing titles to FSR4 ML-based frame generation

AMD SDK suggests 4x and 6x frame generation multipliers are in the works — driver-level tweak will let gamers upgrade existing titles to FSR4 ML-based frame generation

The latest AMD SDK update hints at upcoming 4x and 6x frame generation multipliers for FSR4 ML-based frame generation. A driver-level tweak will allow gamers to enhance existing titles with FSR4's AI-based frame generation. This feature will provide smoother gameplay and improved visual quality, especially beneficial for users with lower-end AMD graphics cards. The addition of higher frame generation ratios aims to bring AMD closer to its competitors like Nvidia and Intel in terms of performance and feature parity. The SDK update suggests that these enhancements may be available soon, offering gamers more flexibility and optimization options.

Tom's Hardware•
Enthusiast builds his own RAM in garden shed cleanroom — fledgling array of memory cells groundwork for much larger future project

Enthusiast builds his own RAM in garden shed cleanroom — fledgling array of memory cells groundwork for much larger future project

An enthusiast has successfully built his own RAM in a garden shed cleanroom, marking the first time RAM has been made at home. The project aims to address the ongoing DRAM crisis and disruptions in the industry caused by high demand and AI-driven price increases. The process involves intricate semiconductor steps like silicon chip preparation, oxide layering, and transistor formation. The DIY RAM cells were successfully created and tested, with plans to scale up for future PC integration. This innovative project showcases the potential for DIY solutions in the tech industry.

Tom's Hardware•
SemiEngineering

Blog Review: Apr. 22

Siemens EDA, Synopsys, Cadence, Keysight, Arm, and SEMI experts discuss various tech topics in the latest blog review. Topics include coverage closure challenges in verification, electromagnetic simulation for accurate resistance models, verification complexities of CXL 4.0, ATM security risks, Arm Performance Libraries updates, sensor evolution, and more. Additionally, there are insights on advanced nodes, GaN transistors, DRAM performance, and MLF packaging methods in the Manufacturing, Packaging & Materials newsletter.

SemiEngineering•
Intel teases wider range of overclockable CPUs for future platforms — VP Robert Hallock says budget builders 'deserve the same level of features' as more well-heeled enthusiasts

Intel teases wider range of overclockable CPUs for future platforms — VP Robert Hallock says budget builders 'deserve the same level of features' as more well-heeled enthusiasts

Intel is planning to expand its range of overclockable CPUs to include more affordable options, breaking away from its previous strategy of limiting unlocked CPUs to mid-range and high-end segments. VP Robert Hallock emphasized the importance of offering overclocking capabilities to budget builders, stating that all PC enthusiasts deserve access to these features. Intel aims to introduce more unlocked SKUs over time to compete with AMD in the budget CPU market and improve its relationship with enthusiasts. While this shift won't happen immediately, it signals a positive change for budget-minded DIY desktop users who have faced barriers to CPU overclocking in the past. Intel's move may involve providing multiplier overclocking support on B-series chipsets to enhance competitiveness with AMD.

Tom's Hardware•
Is Intel About to Take Flight?

Is Intel About to Take Flight?

Elon Musk's order for Terafab chips from Intel Foundry highlights the supply constraints in the industry, with TSMC and Samsung already committed. Intel's partnership with Musk could be a strategic move to meet the demand for AI, robotics, and space-related chips. The article draws parallels between this collaboration and Pan Am's order of the Boeing 747, emphasizing the importance of audacity and foresight in shaping the tech industry. While Intel has the capacity and expertise to support Musk's vision, questions remain about execution timelines and yield ramp expectations. The article underscores the critical role of process engineering expertise in semiconductor manufacturing and the potential impact of AI talent migration on the industry's supply chain.

SemiWiki•
AMD Ryzen 9 9950X3D2 review: More cache, more cash

AMD Ryzen 9 9950X3D2 review: More cache, more cash

The AMD Ryzen 9 9950X3D2 Dual Edition CPU is the first to stack cache on both CCDs, offering 192 MB of L3 cache. Priced at $899, it's AMD's most expensive consumer CPU, bridging the gap between consumer and HEDT ranges. While it excels in specialized workloads like rendering and encoding, its gaming performance is not significantly better than lower-priced models. The additional cache reduces latency and enhances performance in data science tasks, with a 25% performance increase in specialized workloads. The CPU's higher thermal and power demands are balanced by its ability to maintain a high-end thermal design and support overclocking.

Tom's Hardware•
Cerebras files for IPO — company remains unprofitable despite 20x revenue growth

Cerebras files for IPO — company remains unprofitable despite 20x revenue growth

Cerebras, a wafer-scale AI processor supplier, has filed for an IPO despite remaining unprofitable, with most revenue coming from just two customers. The company positions itself as an AI infrastructure provider, offering full-stack solutions. While revenue has grown substantially, Cerebras faces risks due to its heavy reliance on a few large clients. Recent agreements with Amazon Web Services and OpenAI aim to diversify revenue streams. The company's IPO is expected to raise around $3 billion, reflecting its growth and AI infrastructure ambitions.

Tom's Hardware•
Original Task Manager creator explains why it lies to you about CPU usage — former Microsoft engineer shows unique solution to a seemingly simple, but actually complicated, task

Original Task Manager creator explains why it lies to you about CPU usage — former Microsoft engineer shows unique solution to a seemingly simple, but actually complicated, task

Former Microsoft engineer Dave Plummer, creator of the original Task Manager, explains the complexity behind measuring CPU usage accurately. Plummer reveals that the Task Manager provides an interpretation of CPU usage between refreshes, not a real-time view. He details a unique solution where the tool calculates CPU consumption based on total time and processes, offering a more precise measurement. However, with modern processors' dynamic behaviors, the traditional method may feel less accurate due to factors like frequency scaling and thermal throttling. Plummer suggests a new approach where CPU usage reflects work accomplished compared to potential work, but acknowledges his retirement from Microsoft limits his influence on future developments.

Tom's Hardware•
Why Proof Convergence Matters

Why Proof Convergence Matters

The article discusses the importance of proof convergence in semiconductor verification as chip complexity increases. With more cores, interactions, and reliance on AI for AI chip development, achieving a definitive answer in verification becomes challenging. Ashish Darbari, CEO of Axiomise, emphasizes the impact of functional interactions on safety and security in designs. By identifying common patterns to prevent bugs and improve coverage, time to sign-off can be reduced.

SemiEngineering•
The upcoming AMD Ryzen 9 9950X3D2 Dual 3D V-Cache CPU may feature a 250W PPT

The upcoming AMD Ryzen 9 9950X3D2 Dual 3D V-Cache CPU may feature a 250W PPT

The AMD Ryzen 9 9950X3D2 Dual Edition CPU is set to be a powerhouse with a unique dual 3D V-cache design. While AMD's official product page lists a 200W TDP, a recent leak suggests the CPU may have a 250W PPT, the highest among current AMD desktop CPUs. This higher power envelope could pose challenges for overclocking and thermal management, potentially requiring robust cooling solutions. While these details are based on leaks and not officially confirmed by AMD, they hint at the CPU's potential performance and power requirements.

TweakTown•
Intel's Nova Lake will unify L2 cache and feature new 'D' and 'DX' lines for enthusiasts, claims leaker

Intel's Nova Lake will unify L2 cache and feature new 'D' and 'DX' lines for enthusiasts, claims leaker

Intel's upcoming Nova Lake CPUs will introduce major changes to the cache hierarchy, featuring a shared L2 cache and new 'D' and 'DX' lines for enthusiasts. The flagship Nova Lake CPU will boast a dual-compute tile layout with up to 52 cores and specialized 'bLLC' models offering up to 288MB of L3 cache to compete with AMD's X3D offerings. Intel is shifting away from its private L2 cache design for the first time in 17 years to improve core-to-core communication and increase core counts without bus latency issues. The leaked specifications also reveal a diversification in product segmentation, with single-tile unlocked variants falling under the 'D' series and dual-tile models under the 'DX' series. Expect more details on Nova Lake CPUs at Computex in June.

TweakTown•
New AMD Zen 7 leak surfaces, up to 25% IPC upgrade over the unreleased Zen 6

New AMD Zen 7 leak surfaces, up to 25% IPC upgrade over the unreleased Zen 6

A leak has surfaced about AMD's upcoming Zen 7 architecture, codenamed "Prometheus," which is expected to offer up to a 25% improvement in IPC over the unreleased Zen 6. The leak details the internal layout, expected performance, and products that will feature Zen 7 cores manufactured on TSMC's A14 process node. EPYC "Florence" server CPUs could scale up to 288 cores, while desktop CPUs in the "Grimlock Ridge" lineup and laptop CPUs in the "Grimlock Point" and "Grimlock Halo" series are also mentioned. Production for EPYC Zen 7 is anticipated to start in mid-2028.

TweakTown•
Nvidia RTX 3060 comeback in 2026 could alleviate soaring GPU prices and memory shortages — rumored RTX 5050 9GB abruptly shelved amid speculation

Nvidia RTX 3060 comeback in 2026 could alleviate soaring GPU prices and memory shortages — rumored RTX 5050 9GB abruptly shelved amid speculation

Nvidia is rumored to be reintroducing the RTX 3060 graphics card in June 2026 to address GPU price hikes and memory shortages. The company has reportedly paused the launch of the RTX 5050 9GB variant, potentially due to these market challenges. The RTX 3060, with its 12GB of GDDR6 memory, could offer a cost-effective solution for gamers, especially with the current scarcity of GPUs. The decision to bring back the RTX 3060 may be influenced by the lower cost of GDDR6 VRAM and the ease of manufacturing using Samsung's 8nm process.

Tom's Hardware•
Minisforum's beastly N5 Max NAS comes powered by AMD Strix Halo — $2,899 for 'AI NAS' with pre-installed OpenClaw, supports up to 200TB capacity

Minisforum's beastly N5 Max NAS comes powered by AMD Strix Halo — $2,899 for 'AI NAS' with pre-installed OpenClaw, supports up to 200TB capacity

Minisforum is set to release the N5 Max NAS on April 23, 2026, featuring an AMD Ryzen AI Max+ 395 processor with 126 TOPS of AI compute power, 64GB LPDDR5X RAM, and support for up to 200TB storage. Priced at $2,899, the N5 Max includes pre-installed OpenClaw and aims to empower users with AI capabilities for tasks like semantic photo search and smart editing. The device offers extensive connectivity options, including 10GbE ports and multiple USB4 ports. While initial storage is limited to a 128GB system disk, users can expand storage up to 200TB with HDDs and M.2 SSDs.

Tom's Hardware•
Intel Core Ultra 250K Plus vs AMD Ryzen 5 9600X faceoff — battle for the best $200 CPU

Intel Core Ultra 250K Plus vs AMD Ryzen 5 9600X faceoff — battle for the best $200 CPU

The article compares Intel's Core Ultra 5 250K Plus with AMD's Ryzen 5 9600X in the $200 CPU segment. The Intel chip, part of the Arrow Lake refresh lineup, offers 18 cores, higher clock speeds, and faster DDR5 memory support, making it a strong contender for gaming and productivity tasks. On the other hand, the AMD Ryzen 5 9600X, based on Zen 5 architecture, features fewer cores but competitive clock speeds and lower power consumption. In gaming benchmarks, both CPUs perform well, with the Intel chip slightly edging out in FPS in most games. In productivity benchmarks, the Intel Core Ultra 5 250K Plus outperforms the AMD Ryzen 5 9600X significantly in multithreaded tasks. When it comes to overclocking, the Intel chip offers more flexibility and better performance gains, while the AMD chip leans towards smart tuning features like Precision Boost Overdrive. Overall, the Intel Core Ultra 5 250K Plus emerges as the winner due to its superior performance in both gaming and productivity workloads.

Tom's Hardware•
Nvidia CEO Jensen Huang ‘nearly lost his composure’ when pressed on selling chips to China — ‘You’re not talking to someone who woke up a loser’

Nvidia CEO Jensen Huang ‘nearly lost his composure’ when pressed on selling chips to China — ‘You’re not talking to someone who woke up a loser’

Nvidia CEO Jensen Huang engaged in a heated debate about selling chips to China, emphasizing that China already has significant compute power and could develop advanced AI models regardless. He argued that restricting Nvidia's chips would not prevent China's AI development and could lead to a fragmented tech ecosystem. Huang stressed the importance of maintaining innovation and nurturing the tech ecosystem to compete effectively. He highlighted the complexity of AI technology, emphasizing the need for all layers of the industry to succeed, not just focusing on one aspect.

Tom's Hardware•
Elon Musk pushing forward with Terafab at 'light speed' — staff reaching out to various suppliers and are reportedly willing to pay a premium to gain priority

Elon Musk pushing forward with Terafab at 'light speed' — staff reaching out to various suppliers and are reportedly willing to pay a premium to gain priority

Elon Musk's Terafab project is moving quickly, with staff actively seeking out suppliers and willing to pay a premium for priority. They have engaged with companies like Applied Materials Inc., Tokyo Electron Ltd., Lam Research Corp., and Samsung Electronics Co. to inquire about pricing and delivery times for various components. Musk's ambition to build his own chips is evident, despite warnings from Nvidia's Jensen Huang about the complexity of semiconductor manufacturing. The project, launched in March 2026 with a $20 billion investment from Musk, aims to produce 1 TW/year of compute, attracting partnerships from companies like Intel. While the timeline for chip production remains uncertain, the industry is closely watching Terafab's progress.

Tom's Hardware•
NVIDIA is reportedly bringing back the RTX 3060 12GB in June, as RTX 5050 9GB gets pushed back

NVIDIA is reportedly bringing back the RTX 3060 12GB in June, as RTX 5050 9GB gets pushed back

NVIDIA is rumored to reintroduce the RTX 3060 12GB in June, delaying the release of the RTX 5050 9GB model. The decision to prioritize the older GPU architecture is linked to production capacity and sourcing considerations. While the RTX 3060 may lack newer features and require more effort compared to recent models, its 12GB VRAM capacity remains attractive for budget-conscious users. The pricing for the RTX 3060 is yet to be disclosed but is expected to fall between the RTX 5050 and RTX 5060. NVIDIA's strategy of reviving older GPUs in times of need is not new, as seen with the RTX 2060's comeback in 2021.

TweakTown•
Nvidia RTX 5070 spotted at rare MSRP price in Woot flash sale — move fast to secure a bargain, 38% off saves you $340

Nvidia RTX 5070 spotted at rare MSRP price in Woot flash sale — move fast to secure a bargain, 38% off saves you $340

A rare Nvidia RTX 5070 was spotted at MSRP in a Woot flash sale, offering a 38% discount of $340, bringing the price down to $549 from $889. This mid-range graphics card provides excellent performance for 1080p and 1440p gaming, with 12GB of GDDR7 VRAM and access to Nvidia's software features like DLSS 4 and 4.5. The card outperforms previous-gen models and competes well against other GPUs in its class. With the current scarcity of GPUs at reasonable prices, this deal is expected to sell out quickly, so act fast if interested.

Tom's Hardware•

We use cookies

We use cookies to ensure you get the best experience on our website. For more information on how we use cookies, please see our cookie policy.