➀ Zotac announces new mini-PCs powered by AMD's Strix Halo APUs and Nvidia RTX 5060 Ti/5070 GPUs, set to debut at Computex;

➁ Models include the Magnus-E lineup with Intel Core Ultra 7 and desktop GPUs, plus an AMD-focused Magnus EA variant with Ryzen AI Max+ 395 and up to 128GB unified memory;

➂ AMD's Strix Halo APUs offer 16-core Zen 5 CPUs and 40 RDNA 3.5 CUs, rivaling RTX 4060/4070 laptop GPUs.

➀ AWS plans to invest over $5 billion to establish an 'AI Zone' in Saudi Arabia, featuring data centers for AI training and services like SageMaker and Bedrock;

➁ AMD and Humain will invest $10 billion to deploy 500MW of AI compute capacity using AMD's Instinct GPUs, EPYC CPUs, and ROCm software ecosystem, targeting multi-exaflop capacity by 2026;

➂ NVIDIA and Humain will build data centers with up to 500MW capacity powered by advanced GPUs, including a Grace Blackwell AI supercomputer, and collaborate on AI talent development through the Omniverse platform.

➀ Intel's dual-GPU Arc B580 with 48GB VRAM is rumored to debut at Computex 2025 as a cost-effective AI/ML solution;

➁ The card combines two 24GB B580 GPUs via PCIe bridge, lacking shared memory pooling but offering 48GB VRAM at ~$800;

➂ Targets developers needing affordable high VRAM, though performance trails NVIDIA's Blackwell GPUs and requires manual optimization.

➀ NVIDIA has reportedly raised GPU prices by 5-15% across most product lines, driven by rising manufacturing costs and tariffs;

➁ AI GPUs like H200/B200 face the steepest hikes (up to 15%), while gaming GPUs see 5-10% increases, with RTX 50 series prices doubling in some channels;

➂ Despite export restrictions impacting $5.5B in Chinese AI chip sales, strong global AI demand and cloud service investments are expected to keep NVIDIA's earnings stable.

➀ Global semiconductor revenue grew 25% YoY to $683 billion in 2024, driven by strong demand for AI-related chips, particularly HBM, which fueled a 74% surge in memory segment revenue;

➁ Memory market recovery, led by improved supply-demand balance and higher ASPs, lifted Samsung, SK Hynix, and Micron into the top seven semiconductor companies by revenue;

➂ Despite overall growth, automotive and industrial semiconductor revenues declined, highlighting market imbalances, while NVIDIA surpassed Samsung as the top semiconductor company due to GPU dominance.

➀ NVIDIA introduced the MGX 44RU Manifold for liquid cooling in its GB200/GB300 NVL72 AI racks, enhancing OCP ORv3 racks with standardized power and cooling;

➁ Eight vendors (Pinda, LiteOn, Delta, Foxconn, etc.) showcased customizable manifolds at GTC 2025, highlighting industry collaboration;

➂ The design supports blind-mating connections for compute trays, NVLink switches, liquid tubing, and busbars, streamlining high-density AI deployments.

① A prototype of NVIDIA's unreleased RTX Titan Ada GPU, featuring 18,432 CUDA cores and 48GB GDDR6X VRAM, was tested by overclocker Der8auer;

② The GPU outperformed the RTX 4090 by 15% in benchmarks but consumed 14% more power, while the upcoming RTX 5090 showed an 11% speed advantage at higher power draw;

③ Manufacturing costs and potential overlap with high-end workstation GPUs likely led to its cancellation, as the Titan Ada’s fully enabled AD102 die would have been prohibitively expensive.

➀ Asus showcased its latest NUC mini-PCs and ROG gaming laptops at its Taipei HQ, featuring advancements in cooling, noise reduction, and AI-driven design;

➁ The ROG NUC 2025 integrates Intel Arrow Lake HX CPUs and NVIDIA RTX 50-series GPUs, with a focus on quiet operation under 38dB;

➂ Asus highlighted its 2025 laptop lineup, including ultra-slim Zephyrus models and AI-centric Strix Halo APUs, targeting both gamers and professionals.

① NVIDIA confirms withholding RTX 5060 pre-release drivers from reviewers, delaying reviews until its May 19 launch date;

② The timing overlaps with Computex 2025, complicating testing for media teams and leaving budget gamers without performance benchmarks;

③ Speculation arises over NVIDIA's motives, including potential driver tweaks or avoiding unfavorable comparisons with Intel's Arc B580 GPU.

➀ Despite U.S. warnings, Malaysia has been breaking records in chip imports, with nearly $2 billion worth of chips imported in March.

➁ In April, Malaysia imported nearly $3 billion worth of GPUs, setting a new record and showing a trend of triple-digit monthly increases since January 2025.

➂ Malaysia's GPU imports have reached $6.45 billion year-to-date, accounting for around 13% of Nvidia's total revenue in Q1 2024.

➀ The NVIDIA RTX 5090 demonstrates a 33% to 2x performance boost over the RTX 4090 in password hash cracking, breaking 8-digit numeric passwords in 3 hours with a single GPU.

➀ Hive Systems' benchmarks reveal that 12 RTX 5090 GPUs working in tandem can crack simple 8-character passwords in 15 minutes, but complex passwords with symbols may take millennia.

➂ The study highlights the critical importance of password complexity, as adding symbols and mixed-case letters exponentially increases cracking difficulty.

➀ Intel's XeSS 2 expands support to 10 new games, surpassing 200 supported titles;

➁ Leveraging Battlemage architecture, XeSS 2 combines upscaling and latency reduction, competing with Nvidia DLSS and AMD FSR;

➂ Performance data shows up to 4x FPS boosts in AAA titles, with Xe Low Latency (XeLL) reducing input lag by 45% on average.

➀ Lenovo's Legion 9i gaming laptop features an 18-inch dual-mode 3D display, up to an RTX 5090 GPU, 192GB RAM, and 8TB storage;

➁ The 3D display uses eye-tracking and lenticular lenses, targeting both gamers and 3D designers;

➂ Launching first in China with a fall release in North America, pricing remains undisclosed but expected to be premium.

➀ The GeForce RTX 5060 family launch brought the price of Blackwell below $500 on paper;

➁ The GPU situation is currently frustrating, but there is optimism with more models listed at MSRP;

➂ The first GeForce RTX 5060 listing on Best Buy shows a price below $300, with one model from NVIDIA's partners adhering to the baseline MSRP.

➀ U.S. lawmakers propose mandating AI chips (e.g., NVIDIA) to embed location tracking to curb smuggling to China.

➁ Export bans since 2022 have faced criticism for inefficacy, with Chinese firms reportedly using smuggled chips for AI development.

➂ The bill also explores geo-blocking to disable chips without proper licenses, raising technical and privacy concerns.

➀ NVIDIA opposes the AI Diffusion Rule, warning it undermines U.S. AI leadership by pushing China to build rival ecosystems;

➁ Huawei’s CANN platform emerges as a potential CUDA competitor, despite current challenges like documentation and reliability;

➂ U.S. export restrictions risk accelerating China’s AI hardware independence, fragmenting global standards and weakening Western influence.

➀ NVIDIA announced the RTX Pro 6000 Blackwell family at GTC 2025, featuring three variants (Server Edition, Workstation Edition, and Max-Q Workstation Edition) tailored for AI and graphics workloads;

➁ The GPUs offer 96GB GDDR7 ECC memory, with power configurations ranging from 400W to 600W, and are positioned as cost-effective alternatives to H100/H200/B200 data center GPUs;

➂ The Server Edition supports 8x GPU configurations with up to 768GB memory, targeting high-density AI and rendering tasks in 4U/5U systems.

➀ Chip smugglers have found creative ways to sneak semiconductors into China, such as hiding them with live lobsters, inside fake baby bumps, and other means designed to evade customs and import/export restrictions.

➁ A U.S. lawmaker is pushing for a bill to require NVIDIA and other chip companies to equip their semiconductors with tracking technology.

➂ The bill aims to track chips more generally, allowing NVIDIA and the U.S. government to determine if advanced AI semiconductors are finding their way to China, and also introduces a kill switch to prevent chips from booting if they evade export controls.

➀ NVIDIA's RTX Pro 6000 Blackwell workstation GPU, priced over $11,000, targets professionals in AI, HPC, and 3D rendering with 24,064 CUDA cores and 96GB GDDR7 memory;

➁ Despite higher specs than the consumer-grade RTX 5090, early benchmarks show it slightly trailing in performance, likely due to unfinished drivers and power limitations;

➂ The GPU's niche appeal lies in enterprise applications, with its cost justified by specialized workloads rather than consumer gaming.

➀ IBASE introduces EC3100, a rugged edge AI system powered by NVIDIA Jetson Orin NX/Nano modules, featuring 8-core Arm CPU and 1024-core Ampere GPU for industrial automation and robotics.

➁ Designed for harsh environments with aluminum-steel chassis, wide temperature support (-20°C to 70°C), and 9V-36V DC input, ensuring reliability in demanding applications.

➂ Offers extensive I/O connectivity including Gigabit Ethernet, CANBus, PoE support, and expansion slots for 4G/5G/Wi-Fi modules, targeting smart cities and mobile robotics solutions.