Recent #performance news in the semiconductor industry

12 months ago
➀ TSMC will start producing 2nm GAA-based wafers in Q4 2025 at Fab 20 in Hsinchu; ➁ The production will be followed by Fab 22 in Kaohsiung in Q1 2026; ➂ N2P production will start in late 2026 without backside power delivery; ➃ N2 and N2P will use TSMC's NanoFlex technology; ➄ N2 is 15% denser than N3E and improves performance by 10% to 15%; ➅ N2E will be followed by N2P and N2X in 2026; ➆ TSMC's planned capacity for 2nm is more than for 3nm; ➇ 2nm wafers are expected to cost $30k+.
2nmGAAPower ConsumptionProcess technologyProductionTSMCperformancesemiconductor
12 months ago
➀ The Ampere AmpereOne A192-32X is reviewed, highlighting its performance, power consumption, and use in platforms like the Supermicro MegaDC ARS-211M-NR; ➁ The chip's importance in the context of the 2024 server CPU market is discussed, comparing it with Intel and AMD offerings; ➂ The cloud-native design and features of the A192-32X are detailed, including its core count, cache architecture, and support for PCIe Gen5 and DDR5.
ArmBroadcomCloud NativeKioxiaNVIDIAPower ConsumptionServer CPUsSupermicroperformanceserver CPU
about 1 year ago

➀ Samsung introduces the new 990 EVO Plus SSD, offering a 50% performance boost;

➁ The SSD is designed for gaming, business, and creative work, enhancing PC energy efficiency;

➂ Available in 1TB, 2TB, and 4TB capacities with suggested prices starting at NT$3,879;

➃ Features Samsung's latest 8th Generation V-NAND technology and 5-nanometer controller;

➄ Offers high read and write speeds, with the 4TB model having industry-leading random read/write speeds.

SSDSamsungperformancestorage
about 1 year ago
➀ The review introduces the CyberPowerPC Project Zero Infinity, a high-performance gaming PC with an Intel Core i9-14900KF CPU and an MSI RTX 4080 Super GPU. It highlights the system's impressive specifications, including the MSI Z790 Project Zero motherboard, 32GB of DDR5 memory, and a Corsair 3500X ARGB TG case. The article covers synthetic benchmarks, gaming performance, thermals, noise, and power consumption. ➁ The review includes detailed benchmarks with Cinebench R23, 3DMark Time Spy, and various games like Modern Warfare 3, Cyberpunk 2077, and F1 2024. ➃ The author provides their thoughts on the gaming experience and the overall performance of the system.
Gaming PCperformancereview
about 1 year ago

➀ GDDR memory is a graphics-specific memory optimized for GPUs with higher bandwidth than DDR memory.

➁ HBM memory is a high-bandwidth memory designed for GPUs with larger memory bus width than GDDR memory.

➂ The choice between GDDR and HBM depends on the specific application scenario and cost considerations.

GDDRHBMperformance
about 1 year ago

➀ The research on hollow-core fibers is active but not yet widely commercialized due to various reasons, such as incomplete theoretical exploration and immature mass production processes.

➁ Y10T283 employs three segments of hollow-core fibers spliced together for academic research.

➂ This year, Changfly manufactured 15km long continuous hollow-core fibers and discussed the impact on performance in large-scale production.

➃ The wall thickness of the tube for low-loss bands exhibits certain characteristic thicknesses to achieve low-loss design goals.

➄ Thinner wall thickness results in longer resonance wavelength spacing, wider low-loss bands, but also more sensitive pressure differences.

➅ Changfly evaluated the tolerance of the inner and outer diameter ratio and pressure difference based on a balance between manufacturing difficulty and performance.

➆ Changfly selected a low-loss window with a wall thickness of 1.1μm, achieving a minimum loss of 0.25dB/km for 10km hollow-core fibers and about 0.58dB/km for 15km fibers.

➇ The article also discusses the challenges of manufacturing thin glass tubes and their impact on reliability and cost.

Manufacturingperformance
about 1 year ago
➀ AMD's new EPYC 9005 Turin processors deliver exceptional performance with up to 768 threads per server, leveraging new Zen 5/ Zen 5c architecture and 25%+ TDP increase.➁ The processors are set to compete with Intel's server leadership, offering aggressive pricing and a range of SKUs with up to 192 cores.➂ Performance testing highlights the importance of using high-quality SSDs and NICs for optimal performance, with AMD's new platform providing a robust foundation for server environments.
AMDBroadcomEPYCZen 5performanceprocessorserver
about 1 year ago
➀ MediaTek unveiled the Dimensity 9400 processor at an event on October 9, 2024, marking a significant breakthrough in 5G and AI processing. The Dimensity 9400 is the company's second-generation all-large core design, featuring an Arm v9.2 CPU architecture, advanced GPU and NPU, and is manufactured using TSMC's second-generation 3nm process. It boasts 29.1 billion transistors, a 28% increase from the previous Dimensity 9300, making it the first 3nm chip for the Android system. The Dimensity 9400 features a second-generation all-large core CPU with 8 cores, including a high-frequency Arm Cortex-X925 ultra-large core and several other large and efficiency cores. It is the first to support global high-speed 10.7 Gbps LPDDR5X memory, significantly improving performance and efficiency. ➁ In tests, the Dimensity 9400 showed strong performance, with single-core Geekbench scores of 3040 and multi-core scores of 9600, a 35% and 28% increase respectively. Antutu scores reached nearly 2.85 million, and even exceeded 3 million in laboratory conditions. Compared to the previous Dimensity 9300, the Dimensity 9400 reduces power consumption by 40% and can save up to 32% in daily scenarios, offering better performance and battery life for smartphones. ➂ The first smartphones to use the Dimensity 9400 chip are expected to be released in the fourth quarter of 2024. Vivo and OPPO have announced that they will launch smartphones with the Dimensity 9400 chip. The first to be released is the Vivo X200 series on October 14th.
MediaTekperformancesmartphones
about 1 year ago
➀ The RTX5090 boasts 21760 CUDA Cores, a 33% increase over the RTX4090; ➁ It features 32 GB GDDR7 memory with a 512-bit bus, offering up to 2.00 TB/s bandwidth; ➂ The GPU uses the new Blackwell architecture and N4P process technology; ➃ It has a PCIe 5.0 x16 interface with double the theoretical bandwidth of the RTX4090; ➄ The TDP is a significant 600W, a 33% increase from the RTX4090.
GPUNVIDIAperformancespecs
about 1 year ago
➀ SiTime launched the industry's first fully integrated MEMS Clock System-on-Chip (ClkSoC) for the automotive industry, the Chorus Automotive MEMS Clock Generator, which includes a built-in fault detection mechanism suitable for the entire clock generation signal path in automotive design. ➁ The Chorus clock generator integrates MEMS resonators, oscillators, and advanced safety mechanisms into one package, offering performance improvements up to 10 times greater than quartz oscillators with half the size, while providing critical diagnostic coverage for easier achievement of automotive functional safety standards. ➂ The Chorus generator is designed to meet the timing needs of autonomous driving or advanced driver-assistance systems (ADAS) with its robust MEMS timing technology and high-precision timing networks used in data centers, ensuring reliable operation across a wide temperature range. It also features early fault reporting capabilities to minimize fault tolerance time intervals (FTTI), which are common requirements for automotive safety integrity levels (ASIL).
ADASSafetyperformance
about 1 year ago
➀ The vanilla version of Intel's Arrow Lake Core Ultra 9 part shows strong single-core performance in Geekbench but weak multi-core results. ➁ The chip matches the Core i9-14900K in single-core but falls behind in multi-core, even compared to the 13th Gen Core i7-13700. ➂ Suggests Intel might be lowering power limits for non-K variants of Arrow Lake processors.
Arrow LakeGeekbenchIntelcpuperformance