NVIDIA releases RTX 30 series graphics cards with double the performance and a lower price

In the early morning of September 2nd, Beijing time, NVIDIA released the GeForce RTX 30 series graphics cards based on the NVIDIA Ampere architecture - RTX 3090, RTX 3080 and RTX 3070. NVIDIA said that compared with the previous generation Turing architecture products, the performance of the RTX 30 series has been increased to 2 times and the energy efficiency has been increased to 1.9 times.

The RTX 30 series graphics cards using the Ampere architecture also have many new features: support for the second-generation RT Core (ray tracing core), support for the third-generation Tensor Cores (tensor core), the world's fastest GDDR6X video memory, customized Samsung 8nm process, and more.

New streaming multiprocessor: 30 Shader-TFLOP/s processing power and 2x FP32 throughput than the previous generation.

Second-generation RT Cores: The computing power is 58 RT-TFLOPS, which is twice that of the previous generation, and supports ray tracing, shading and calculation.

Third-generation Tensor Cores: With twice the throughput of the previous generation, it can run AI-driven technologies such as NVIDIA DLSS faster and more efficiently, with a computing power of up to 238 Tensor-TFLOPS.

NVIDIA RTX IO: Enables fast GPU-based loading and decompression of game resources, accelerating input/output performance by up to 100 times compared to hard disks and traditional storage APIs. Combined with Microsoft's new Windows version of DirectStorage API, RTX IO offloads the work of dozens of CPU cores to the RTX GPU, increasing frame rates and achieving near-instantaneous game loading.

The world's fastest video memory: GDDR6X video memory, developed by NVIDIA and Micron, supports PAM4 signaling and can provide data transmission speeds of nearly 1TB/s for RTX 30 series graphics card applications.

Next-generation process technology: The new 8N NVIDIA custom process from Samsung enables higher transistor density and higher efficiency, with a reported integrated transistor count of up to 28 billion.

Dual-axis flow-through cooling solution: RTX 3090 and RTX 3080 adopt an integrated front and rear dual-fan design, which increases the cooling performance by 2 times, the airflow by 55%, the efficiency by 30%, and the quietness by 3 times.

Mechanical and circuit design: A more robust mechanical structure, with a new low-profile leaf spring and a new 12-pin power interface, leaves more space for components and heat dissipation, and is compatible with the 8-pin connector of existing power supplies through the included adapter. It is reported that the PCB circuit board is half the size of the previous generation.

HDMI 2.1: Supports higher bandwidth, allowing for the first time a single cable connection to an 8K HDR TV for ultra-high-resolution gaming.

AV1 Decode: The first discrete GPU to support the new AV1 decoder, allowing you to watch 8K HDR web video with 50% less bandwidth than before.

The specific specifications of the three products are as follows:

GeForce RTX 3090: has up to 10496 NVIDIA CUDA cores, an acceleration frequency of 1.70GHz, a standard video memory configuration of 24GB GDDR6X, a video memory width of 384 bits, and a power of 350W. The RTX 3090 is equipped with a three-slot dual-axis streamlined design radiator, which is 10 times quieter than the TITAN RTX and can reduce the GPU temperature by up to 30°C. NVIDIA said that the RTX 3090 is 50% faster than the current ultimate PC graphics card TITAN RTX, allowing players to reach 60fps in many top games at 8K resolution.

GeForce RTX 3080: It has 8704 NVIDIA CUDA cores, an acceleration frequency of 1.71GHz, a standard video memory configuration of 10GB GDDR6X, a video memory bit width of 320 bits, and a power of 320W. Nvidia said that the RTX 3080 is 2 times faster than the RTX 2080, running at up to 19Gbps, and brings a stable gaming experience of 60 frames per second at 4K resolution.

GeForce RTX 3070: It has 5888 NVIDIA CUDA cores, a boost frequency of 1.73GHz, a standard video memory configuration of 8GB GDDR6, a video memory bus width of 256 bits, and a power of 220W. The RTX 3070 is faster than the RTX 2080Ti and 60% faster than the RTX 2070. It is a sweet spot for playing 4K games and 1440p games.

Product price and release date: GeForce RTX 3090 is priced from 11,999 yuan and will be available from September 24; GeForce RTX 3080 is priced from 5,499 yuan (1,000 yuan cheaper than the initial price of RTX 2080), and will be available from September 17; GeForce RTX 3070 is priced from 3,899 yuan (100 yuan cheaper than RTX 2070), and will be available from October.

NVIDIA also mentioned that graphics card suppliers such as ASUS, Colorful, Gainward, GALAXY, GIGABYTE, INNOVATION, MSI and Zotac will release standard and overclocked versions of GeForce RTX 3090, GeForce RTX 3080 and GeForce RTX 3070 graphics cards in China. The GeForce RTX 30 series GPUs will also be sold through Acer, Alienware, ASUS, Dell, HP, Lenovo and MSI products.

In addition to releasing a new generation of graphics cards, NVIDIA founder and CEO Jensen Huang also announced at an online press conference that "Fortnite" will support RTX real-time ray tracing, adding four new ray tracing functions: reflection, shadow, global illumination and ambient occlusion.

NVIDIA has also prepared three new technologies to enhance the gaming experience for gamers, streamers, and professional esports players:

NVIDIA Reflex: Gamers seek the lowest possible latency in competitive games so that their PCs can respond faster to inputs, improving players’ shooting accuracy in games. NVIDIA Reflex is a new set of technologies that optimize and measure system latency.

NVIDIA Broadcast: Livestreamers can transform their homes into livestreaming studios with the NVIDIA Broadcast universal plugin, which enhances microphones and webcams with RTX-accelerated AI effects like noise removal, virtual backgrounds and auto-framing.

NVIDIA Omniverse Machinima: Omniverse Machinima brings path tracing viewing tools and engines designed to achieve physical accuracy, simulate light, physics, materials, and AI, making it easier for games to become blockbusters through artistic processing. Users can take assets from games that support NVIDIA Omniverse Machinima and use webcams and AI to create characters, add high-fidelity physics and facial and voice animations, and use the outstanding rendering capabilities of GeForce RTX 30 series GPUs to create movie-level movies.