2025 Consumer Electronics Show (CES 2025) was held from January 7 to 10, and Nvidia CEO Jensen Huang took the stage this morning to deliver the opening keynote, officially unveiling the new RTX 50 series graphics cards, including the $1,999 RTX 5090, $999 RTX 5080, $749 RTX 5070 Ti, and $549 RTX 5070.
The RTX 50 series is scheduled to start shipping in January, and according to the Nvidia website, the RTX 50 series graphics card prices in Taiwan are as follows:
- GeForce RTX 5090: Starting at NT$ 71,990
- GeForce RTX 5080: Starting at NT$ 35,990
- GeForce RTX 5070 Ti: Starting at NT$ 26,990
- GeForce RTX 5070: Starting at NT$ 19,990
According to Huang, the RTX 50 series has 4000 TOPS (Tera Operations per Second) of AI computing power, 380 TFLOPS of floating-point performance, 125 Shader TFLOPS, 92 billion transistors, Micron's GDDR7 memory, 1.8TB/s of memory bandwidth, and an AI management processor.
Huang stated that the Blackwell architecture, which combines AI-driven neural rendering and ray tracing technology, is the biggest graphics innovation since the introduction of programmable shading 25 years ago. The new generation not only has powerful AI computing power, but can also improve performance through NVIDIA DLSS 4.
The flagship RTX 5090 has 92 billion transistors and a computing power of 3,352 TOPS, which is twice the performance of the previous flagship RTX 4090, providing strong support for gaming and design users. With ray tracing and DLSS 4 enabled, it can run Cyberpunk 2077 at a smooth 238 fps, compared to only 106 fps for the RTX 4090.
In terms of specific configurations, the RTX 5090 is based on the Blackwell GB202-300 core, with a base clock of 2.01GHz and a boost clock of 2.41GHz, 21,760 CUDA cores, 170 ray tracing cores, 32GB of 28Gbps GDDR7 video memory, PCIe 5.0×16 interface, 600W TBP, and 1x16Pin power supply.
It's worth mentioning that the latest mid-range RTX 5070, priced at $549, is about one-third the price of the previous flagship RTX 4090, and not only has improved computing power, but also relatively lower pricing. According to Huang's slides, even with the RTX 5070 using DLSS 4 technology, it can achieve gaming performance equivalent to the RTX 4090.
Building a Massive Chip to Surpass the World's Fastest Supercomputer
Additionally, the external world is generally concerned about the specific progress of the Blackwell chip. Huang stated that Blackwell is in full-scale production, with all major cloud service providers having established systems, offering about 200 different models and configurations, and the chip series comes from about 15 hardware manufacturers.
Huang also introduced Nvidia's various computing network systems, such as NVLink 36 and NVLink 72, which can meet the needs of almost all data centers worldwide. Currently, they are produced in about 45 factories, and Nvidia's goal is to build a massive chip, the Grace Blackwell NVLink72, which will use 72 Blackwell GPUs or 144 chips, and its performance will surpass the world's fastest supercomputer.
Nvidia Likely to Regain the Market Cap Leader
Huang's keynote is expected to have a significant impact on Nvidia's stock price. Nvidia's stock price rose 3.43% to $149.43 on the 6th, hitting a new all-time high, with a latest market capitalization of $3.66 trillion, which is likely to surpass the current $3.7 trillion market cap of Apple, becoming the world's largest company by market value.
