➀ e-con Systems has launched a new embedded camera designed for integration with the NVIDIA Holoscan platform, enabling real-time AI-based object detection and recognition.

➁ The camera is ideal for industrial automation, robotics, and medical imaging applications, featuring high-speed and low-latency imaging with e-con's proprietary TintE Image Signal Processor (ISP).

➂ It leverages the NVIDIA Holoscan Sensor Bridge powered FPGA board to simplify setup and offload image processing tasks, optimizing GPU resources for AI inference.

➀ Ed向他的日记透露，执政不到一年的工党政府正在削减福利预算；

➁ 总理认为Ed是一个多面手，经常找他解决问题，Ed成了他的跑腿；

➂ 总理希望Ed帮助解决长期病假人员的问题，以减少福利预算开支，并避免政府形象受损。

➀ Solidigm showcased its liquid-cooled NVMe SSDs at NVIDIA GTC 2025;

➁ The new Solidigm D7-PS1010 E1.S 9.5mm SSD features a cooling design that can cool both sides of the PCB on the drive;

➂ This design will be important for future AI servers due to increasing power density and the need for more efficient cooling solutions.

➀ NVIDIA recently unveiled its RTX Pro 6000 Blackwell graphics cards with impressive specifications for modern workstations and servers.

➁ The RTX Pro 6000 Blackwell Workstation Edition and RTX 6000 Blackwell Max-Q Workstation Edition models are priced over $8,000.

➂ The RTX Pro 6000 Blackwell features 24,064 CUDA cores, 96GB of GDDR7 memory, and a 600W TDP, while the Max-Q version has a 300W TDP.

➀ Nvidia's GTC 2025 unveiled a roadmap with powerful new AI GPUs and systems, including Rubin GPUs for 2026 and Rubin Ultra in 2027.

➁ Intel's restructuring began with new CEO Lip-Bu Tan and a reorganization of Intel Foundry operations.

➂ Storage solutions for AI use cases were showcased, including Kioxia's 122.88TB SSD and Seagate's NVMe HDDs.

➀ The data shows that the average selling price of RTX 5090 on eBay has dropped to $4,000.

➁ The prices have stabilized after the launch volatility, showing a slow but consistent downward trend.

➂ The average price of RTX 5090 in March has also dropped, which may indicate lower prices in the future.

➀ Gelsinger discusses the difference between throughput computing and scalar computing, highlighting NVIDIA's focus on GPU-based computing for AI.

➁ He argues that GPUs are overpriced for AI inference, suggesting a need for more cost-effective solutions.

➂ Gelsinger hints at the potential for 'NPUs' as a more efficient alternative for AI inference.

➀ The upcoming GeForce RTX 5060 Ti graphics card from NVIDIA is expected to be launched in the middle of next month. It is one of the two mainstream RTX 5060 series cards expected in the coming weeks, with options for 8GB and 16GB VRAM.

➁ The GeForce RTX 5060 Ti will feature a GB206-300 Blackwell GPU with a PG152 board, 4,608 CUDA cores, and a memory bandwidth of 448GB/s.

➂ The card is expected to have a 180W TGP rating with a single 8-pin power connector, which is 15W higher than the GeForce RTX 4060 Ti's 165W TGP rating.

➀ The launch of GeForce RTX 5060 Ti and 5060 has been allegedly delayed;

➁ The initial announcement was planned for late March, but has been postponed to early April;

➂ The new launch date is expected to be around mid-April for the 5060 Ti and mid-May for the 5060.

➀ The NVIDIA DGX Spark is a compact AI mini PC designed for scale-out clustering;

➁ It features 20 Arm cores, 128GB of LPDDR5X memory, and a Blackwell GPU;

➂ Priced at $3999, it offers 200GbE NVIDIA ConnectX-7 networking and is compatible with clustering configurations.

➀ ASUS ROG Astral RTX 5090，Blackwell旗舰显卡的高端版本，采用金色外壳，并由NVIDIA首席执行官黄仁勋亲自签名。

➁ 预计该GPU的售价约为20,000美元，并可能通过慈善拍卖进行出售。

➂ 尽管ASUS ROG Astral旗舰显卡本身价格不菲，但这款金色签名版的ROG Astral RTX 5090的售价将远超数千美元。

➀ The ASUS Ascent GX10 is a NVIDIA GB10 Mini PC featuring 128GB of memory and 200GbE connectivity.

➁ It is designed to be portable and compact, suitable for various applications.

➂ The system includes an NVIDIA ConnectX-7 NIC for 200GbE clustering and a 10GbE NIC for basic networking.

➀ ASUS introduced the ASUS Ascent GX10, a petaflop-scale AI supercomputer powered by the NVIDIA Grace Blackwell Superchip. It provides unprecedented computing power, high-speed memory integration, and easy scalability for enterprise-level AI capabilities.

➁ The device delivers 1,000 TOPS of AI processing power, supporting AI models with up to 200 billion parameters. It includes Blackwell GPU with fifth-generation Tensor Cores and Grace 20-core Arm CPU for efficient data orchestration and real-time inferencing.

➂ The supercomputer integrates NVIDIA NVLink-C2C connectivity, offering 5x the bandwidth of PCIe 5.0. It supports seamless transitions between local development and scalable deployment, aiding developers in AI model prototyping and fine-tuning.

➀ The server market grew 91% year-over-year in Q4 2024, with revenue reaching a record $77.3 billion. Revenue from x86 servers increased 59.9% to $54.8 billion, and non-x86 servers increased 262.1% to $22.5 billion. Servers with embedded GPUs saw a 192.6% year-over-year growth.

➁ Nvidia dominated the server GPU market with over 90% of shipments with embedded GPUs in 2024Q4. The adoption of servers with embedded GPUs by hyperscalers and cloud service providers has been a significant driver of market growth.

➂ The U.S. was the second fastest-growing region in the server market, with China also experiencing significant growth. Dell Technologies and Supermicro tied for the number one position in the market, followed by Hewlett Packard Enterprise, IEIT Systems, and Lenovo.

➀ AI has become an ever-accelerating treadmill, with costs increasing exponentially.

➁ Jensen Huang highlighted the need for significantly more computation power due to agentic AI, with a requirement 100 times greater than last year's expectations.

➂ Nvidia is increasing its GPU generation frequency to keep pace, as the AI treadmill is driven by new iterations with enhanced capabilities, such as Perception, Generative, Agentic AI, and the upcoming Physical AI.

NVIDIA has introduced its DGX Station workstation platform, featuring the GB300 Desktop Superchip, which combines a Grace CPU with a Blackwell GPU optimized for AI workloads.

The platform is designed for software developers, researchers, and data scientists, with a focus on high-speed networking and scalability for extensive AI models.

The specifications of the 'Desktop Superchip' are yet to be disclosed, but it is expected to be a combination of Grace CPU and Blackwell Ultra GPU components optimized for desktop PCs.

➀ Micron and SK hynix unveil new LPDDR5X SOCAMM memory modules;

➁ The modules are designed for AI and low-power servers, debuting on Nvidia GB300 systems;

➂ The memory offers high capacity, performance, and low power consumption.

➀ Huang emphasized the need for significantly more computation for agentic AI;

➁ Nvidia announced the Blackwell Ultra GPU with Dell's 20-PetaFLOPS workstation;

➂ The company also introduced the GROOT humanoid robot and discussed collaborations with GM and T-Mobile for autonomous driving and AI-native 6G networks.

➀ Texas Instruments (TI) has released new power-management chips aimed at meeting the rising power demands of modern data centers, particularly for high-performance computing and AI applications.

➁ These chips, including the TPS1685 and integrated GaN power stages, offer enhanced protection, higher density, and improved efficiency, with a focus on reducing size by 50% compared to traditional hot-swap controllers.

➂ TI's innovations are designed to support data center engineers, power system architects, and hardware developers in creating more sustainable and energy-efficient data center infrastructures.

➀ We are covering the NVIDIA GTC 2025 keynote live, expecting a number of announcements in GPUs, AI, networking, and robotics.

➁ The event is held at the SAP Center, San Jose, due to the large number of attendees.

➂ NVIDIA has seen massive growth with the AI boom, becoming a leader in major data center categories.