Semiconductor Storm Rages: Memory, CPU, GPU Unleash Major Upgrades

As the CES 2025 consumer electronics show kicks off in full swing, AI remains the “ultimate weapon” showcased by major tech giants—continuing last year’s trend—while related products have taken further leaps in both performance and innovation.

In the semiconductor sector, the latest products and technologies showcased by companies such as Qualcomm, MediaTek, NVIDIA, Intel, AMD, SK Hynix, and Black Sesame Technologies—spanning SoCs, CPUs, GPUs, memory, and more—are all deeply integrated with AI themes, delivering significantly enhanced performance and attracting widespread attention.

Qualcomm: Launches the New AI Chip Snapdragon X

Qualcomm comprehensively showcased its innovations across PC, automotive, smart home, and enterprise application domains, and announced the launch of its new AI chip—the Snapdragon X. Built on a 4nm process and featuring Qualcomm's Oryon CPU with eight cores and a peak clock speed of up to 3GHz, it delivers robust computing performance for the next generation of PCs.

According to available information, Project DIGITS is a personal AI supercomputer initiative launched by NVIDIA, designed to allow users to enjoy supercomputing performance at home. The Project DIGITS device is equipped with the GB10 superchip and primarily targets AI researchers, data specialists, scientists, and similar professionals.

Prior to this collaboration, MediaTek had already partnered with NVIDIA to integrate NVIDIA's next-generation GPU-accelerated AI computing and NVIDIA RTX graphics processing technologies into the Dimensity Automotive Cockpit Platform. Additionally, MediaTek has also integrated the NVIDIA TAO toolkit for AI model training and optimization with the MediaTek NeuroPilot SDK, accelerating the development of applications in the IoT domain with advanced edge AI capabilities.

Qualcomm believes that the era of AI presents significant opportunities for the edge side. Cristiano Amon, President and CEO of Qualcomm, stated: AI is driving a transformative shift in the technology landscape. In 2025, AI processing will continue to migrate to the edge, enabling and enhancing AI-first experiences. Qualcomm is collaborating with a broad ecosystem of partners to deliver AI-first experiences across diverse end-device categories for both consumers and enterprises.

MediaTek X NVIDIA: Collaboratively Launch the Dimensity GB10 Super Chip

During CES 2025, MediaTek announced its collaboration with NVIDIA to co-design the NVIDIA GB10 Grace Blackwell Superchip, which will be used in NVIDIA's personal AI supercomputer, the NVIDIA® Project DIGITS.

According to available information, Project DIGITS is a personal AI supercomputer initiative launched by NVIDIA, designed to enable users to experience supercomputer-level performance from their own homes. The Project DIGITS device is equipped with the GB10 Superchip and primarily targets AI researchers, data specialists, scientists, and similar professionals.

Prior to this collaboration, MediaTek had already partnered with NVIDIA to integrate NVIDIA's next-generation GPU-accelerated AI computing and NVIDIA RTX graphics processing technologies into the Dimensity Automotive Cockpit Platform. Additionally, MediaTek has integrated the NVIDIA TAO toolkit for AI model training and optimization with the MediaTek NeuroPilot SDK development suite to accelerate the application development of advanced edge AI capabilities in the IoT field.

NVIDIA: Launches GeForce RTX 50 Series

In addition to the GB10 Superchip co-developed with MediaTek, NVIDIA's major product announcements at CES 2025 also include the new GeForce RTX 50 Series featuring the Blackwell architecture. These GPUs are designed to drive technological advancements in fields such as gaming, autonomous driving, and robotics.

Information indicates that the GeForce RTX 50 series features 92 billion transistors, delivers computing power of up to 4000 TOPS, and achieves AI performance of 4 PFLOPS, representing a threefold performance increase compared to the previous Ada architecture.

The GeForce RTX 50 series delivers ray tracing performance of 380 TFLOPS and parallel shading capabilities of 125 TFLOPS, further enhancing the visual quality experience. Additionally, the Blackwell series introduces the new "Neural Texture Compression" and "Neural Material Shading" technologies, enabling shaders to directly process neural networks, resulting in more efficient and detailed graphics performance.

Additionally, according to NVIDIA CEO Jensen Huang, the performance of the RTX 5090 will be twice that of the RTX 4090, while the performance of the RTX 5070 will be comparable to that of the RTX 4090.

Intel: First Intel 18A Process Chip Unveiled.

Targeting enterprises, creators, and gaming enthusiasts, Intel unveiled the new Core Ultra 200V series processors, along with the 200HX and 200H series processors, as well as the 200U series mobile processors. The company also showcased GPU products such as the Arc B580 and B570.

At the CES 2025 keynote session, Intel’s Interim Co-CEO Michelle Johnston unveiled the first chip built on the Intel 18A process and announced that the Panther Lake processor, manufactured using Intel 18A technology, will be officially launched in the second half of 2025, with the chips currently in the testing phase.

According to available information, Panther Lake is designated as Core Ultra 300, featuring 16 cores and is expected to succeed Intel's Arrow Lake-U/H series.

AMD: Continuously Advancing AI PC Development

At CES 2025, AMD announced multiple processor products to further solidify its position in the AI PC market. Among them, the new Ryzen AI Max series processors meet the demand for high-performance computing in premium thin-and-light laptops. These processors feature up to 16 "Zen 5" architecture CPU cores, up to 40 AMD RDNA 3.5 graphics cores, and an AMD XDNA2 NPU delivering up to 50 TOPS of AI computing power. They support up to 128GB of unified memory, of which up to 96GB can be allocated to graphics processing. The Ryzen AI Max PRO series processors are specifically designed to redefine thin-and-light workstations.

The new Ryzen AI 300 series processors are equipped with up to 8 "Zen 5" architecture CPU cores and the latest RDNA 3.5 graphics architecture. Powered by an NPU driven by AMD XDNA 2 technology, the Ryzen AI 300 series processors deliver up to five times the performance improvement compared to the first-generation NPU, providing significantly enhanced AI computing capabilities.

For daily business operations, the new Ryzen AI 7 PRO 350 and Ryzen AI 5 PRO 340 processors are specifically designed to support the next-generation Microsoft Copilot+ application experience. Commercial systems equipped with the Ryzen AI 300 PRO series processors feature leading AI performance of up to 50+ NPU TOPS, providing enterprises with the powerful computing capabilities necessary for transitioning to an AI-driven work model.

SK hynix: Cutting-edge AI Memory Products Make Their Full Debut

SK hynix has established an artificial intelligence data center exhibition area at this year's CES, showcasing cutting-edge, AI-oriented memory technologies including HBM, server DRAM, eSSD, CXL, and PIM.

Among them, the 16-layer HBM3E sample has garnered significant attention. The 48GB 16-layer HBM3E is specifically optimized for AI training and inference. Currently, this product remains in the development stage. DDR5 RDIMM and MCR DIMM, these two high-speed server DRAM modules are tailored for data center environments, providing rapid data processing capabilities and high-capacity memory support.

SK hynix has specifically showcased its innovative eSSD solutions, including the PS1010, PEB110, and PE9010. These products are tailor-made for data centers, offering exceptional reliability and fast read/write speeds to handle the massive volumes of data generated by AI applications.

In addition, SK hynix also showcased CXL technologies including CMM-DDR5 and CMM-Ax, GDDR6-AiM and AiMX PIM solutions, as well as edge AI solutions.

Black Sesame Technologies: HuaShan A2000 Makes Its Debut

Influenced by changes in the international landscape, the number of mainland Chinese chip manufacturers participating in CES 2025 was limited, with AI-related chipmakers being particularly scarce. Black Sesame Technologies was one of them. At this year's exhibition, the sample chip of Black Sesame Technologies’ HuaShan series A2000 made its debut appearance.

The HuaShan A2000 chip is a custom-designed AI chip tailored for next-generation AI algorithms, offering higher performance and greater efficiency. Manufactured using a 7nm process, it incorporates Black Sesame Technologies' proprietary "Jiu Shao" NPU core and supports full-scenario general autonomous driving. Last month, Black Sesame Technologies launched its HuaShan A2000 family chip platform, which includes the HuaShan A2000, A2000 Lite, and A2000 Pro.

According to the official announcement, the HuaShan A2000 chip not only demonstrates powerful performance for high-level autonomous driving scenarios but also supports multiple fields such as embodied intelligence and general computing.

Concluding Remarks

Observing the CES activities of the aforementioned semiconductor manufacturers, it is evident that the growth momentum in servers and data centers remains strong in the AI era. At the same time, AI PCs, intelligent vehicles, and other fields continue to advance, with competition among major semiconductor players intensifying compared to last year. Looking ahead, AI will continue to be a crucial driving force behind the development of the semiconductor industry.