Why does Nvidia want to kill itself?

Nvidia has set off on another self-revolution in the past four years.

Nvidia’s new-generation graphics card stabbed its previous-generation products, but its rival AMD is what it really wants to hit.

Nvidia released the GeForce RTX 30 series of graphics cards based on the ampere architecture on September 2, including three models: RTX 3070, RTX 3080 and RTX 3090. The GeForce RTX 30 series not only duplicated the previous generation’s performance and energy efficiency. Key parameters such as process, video memory, ray tracing, and AI game experience have also made great changes.

More importantly, the price remained unchanged for this generation of graphics cards. Compared to the “price-performance ratio” when the previous generation product was released, the pricing of the GeForce RTX 30 series runs counter to the previous high-end attitude, hitting the “high-end price” so that it does not seem to leave any way for its previous generation products. NVIDIA can consider this kind of “cost-effective” approach as another self-revolution in the last 4 years.

Huang Renxun said, “The GeForce RTX 30 series graphics cards have made the greatest leap in history. If the development of the past 20 years is amazing, then the next 20 years will be simply science fiction.” GeForce RTX 30 series graphics cards assessment and expectations are extremely high.

Moving out of the “feat” of Nvidia 21 years ago may be the first press conference of Huang Jiao.

Before the press conference, Nvidia posted a short video preview and promotional poster on Twitter with the slogan “21 days, 21 years”, which drew everyone’s appetite, and showed that the new GeForce RTX 30 series graphics cards are unusual and have epoch-making significance. Only by revisiting history can we perceive the present. For Nvidia 21 years ago, it was a year worth remembering, and the significance of that year was precisely Nvidia’s key breakthrough in business competition.

Tenology rolling is always the last word

Nvidia launched GeForce 256 in August1999. Although GeForce 256 is not Nvidia’s first-generation product, it has created many historic nodes. GeForce 256 created a number of world premieres, and the GPU concept was first proposed. NVIDIA hailed GeForce 256 for being the world’s first GPU product. GPU has since kept pace with CPU as an independent computing unit; the card family’s first series of products named after GeForce has become N The Star Product.

SimplifiedIn particular, GeForce 256 introduced the Hardware 1 T&L ( light and shadow conversion ) technology to the user market for the first time, which was nothing like it at the time. The cpu has been responsible for the calculation and operation of processing 3D making, driving and making come into existence polygonal graphics. however, even the most advanced cpus were not up to the work at the time, so they had no selection but to reduce the complex part of graphics. It was not until the emergence of GeForce 256 that this question for which decision is hard was changed

GeForce 256 has powerful graphics processing capabilities, more powerful than the fastest CPU at the time, and accelerates the 3D rendering of computing from CPU to GPU in the industry.

At that time, Nvidia faced threats from many competitors including 3Dfx, S3, and ATI in the market. GeForce 256 established Nvidia’s dominant position in the GPU field. And AMD’s acquisition of ATI, playing a double arena with Nvidia is still a story.
So, what is the difference between the GeForce RTX 30 series and the GeForce 256?

Has the visual quality under the Architecture of Ampere made an incredible leap? Enhanced Version of Nvidia Marbles

The GeForce RTX 30 series uses a custom 8N process from Samsung, integrates 28 billion transistors, and is equipped with the new GDDR6X from Micron ‘s fastest storage ever.

The process is upgraded, and the speed of reading and writing is doubled. “The era of scaling only on transistor performance is over,” Huang Renxun said.

Because the GeForce RTX 30 series adopts the latest ampere architecture, shader calculation, performance, and triangle intersection throughput, ray tracing is two to three times that of the previous generation Turing architecture, and the Tensor Core (tensor computing core) processing sparse network speed is also high Twice one generation, that is, “three doubles.” The more ray tracing, the faster the ampere architecture accelerates, which greatly enhances the immersion and realism of the gaming experience.

“Nvidia uses every engineering lever to extract system performance, architecture, custom process design, circuit design, logic design, video memory, software and algorithms… Full-stack engineering design and exquisite craftsmanship are the hallmarks of Nvidia GPUs. Performance, performance Power consumption and low power consumption reach world-class levels.” Huang Renxun concluded, “The leap in performance between generations is the biggest highlight of the new GPU.”

An industry analyst believes that “the biggest highlight of the GeForce RTX 30 series is the architecture upgrade, and the introduction of a new generation of Tensor Core (tensor computing core) and RT Core (ray tracing core). The graphics card released this time will bring NVIDIA soon The GPU technology of the former data center A100 was introduced into game graphics cards.”

In addition, in terms of AI technology, NVIDIA has also made a series of updates, such as removing image noise, reducing the noise of the anchor background sound, and changing the background theme. Among them, the host background sound noise reduction can be described as the gospel of the host, no matter how noisy the background sound, press the noise reduction function, the world can instantly quiet down, leaving only the host’s own voice.

Although Huang Renxun focused on the improvement of the performance and energy efficiency parameters of the GeForce RTX 30 series at the press conference, many people in the industry believe that the GeForce RTX 30 series has been significantly improved because of its technology, while also making its price more attractive. . Embarked on the route of high performance instead of low price and high cost performance.

??The aforementioned industry analysts told Geek Park, “AMD has always taken the cost-effective route, attacking the market with lower process technology and lower prices. Now NVIDIA has upgraded to 8nm and doubled the performance, but the price has not risen. This makes AMD’s GPU The graphics card has a bad outlook.”

??Blocking the path of AMD’s rise

??A month ago, AMD’s 2020 Q2 financial report showed that AMD’s quarterly revenue increased year-on-year and quarter-on-quarter, setting a new high in single-quarter revenue and profit. Ryzen (PC processor) and EPYC (server processor) sales more than doubled, directly driving the substantial growth of various indicators in Q2.

??After Intel’s 7nm process process bounced again, AMD’s stock price rose rapidly, surpassing Intel, and doubled its share price from the beginning of the year. The capital market is optimistic about it. AMD returned to the highlight moment after going through twists and turns.

??AMD has both a CPU business and a GPU business, and it competes directly with Intel and Nvidia respectively. Although AMD’s 2020 Q2 financial report set a record, it mainly relied on the growth of the CPU business sector to make up for the decline in the GPU business. But in fact, AMD’s GPU business still exerts some pressure on Nvidia.

??According to data from Jon Peddie Research, in 2020 Q2 GPU discrete graphics market, NVIDIA will account for 80% and AMD will account for 20%; and (discrete graphics + integrated graphics) of the overall market share, AMD will account for 18% and NVIDIA will account for 19% , The gap between the two is not obvious.

AMD’s product strategy is inseparable from AMD’s CPU business in recent years. After AMD launched the new Zen chip architecture in 2017, it was bound to TSMC and closely followed the advanced process. While chip performance is improved, the price is lower than that of competitors. For example, the Ryzen 3000 series CPU released by AMD in 2019 has a 7nm process that is more advanced than Intel’s 14nm, but the price is half that of Intel.

The same product and pricing strategy is also fully reflected in the GPU business.

In 2019, AMD’s Radeon RX 5700 series graphics cards, based on a new architecture, are ahead of Nvidia in adopting the 7nm process, and the price is lower than the previous generation product, which is half of Nvidia. According to the product roadmap announced by AMD, in 2020, AMD will further update its architecture and chip manufacturing process to enter the high-end GPU market where Nvidia is located. Undoubtedly, AMD is targeting user groups that Nvidia cannot cover, that is, users who are more sensitive to the cost-effectiveness of chips.

This is understandable, why this time Nvidia GeForce RTX 30 series pricing is so “radical”, which completely blocked AMD’s attack from both performance and price. Of course, this is not the first time NVIDIA has changed its pricing strategy due to AMD. In 2017, in the face of AMD’s offensive, NVIDIA released the performance monster GTX 1080 Ti graphics card at a moderate price.

However, compared with Nvidia, AMD still has a big gap in the key performance of graphics cards, such as ray tracing, software ecology, and the scale of the discrete graphics market.

In fact, as early as the 1990s, Nvidia overwhelmed 3Dfx and ATI by virtue of cost performance. This time, NVIDIA regained its cost-effective strategy. AMD will probably be slammed in all directions. After the only price advantage no longer exists, AMD’s counterattack will be doomed to be full of bumps.

And t may be the goal of the “scheming” Lao Huang-it is good for Nvidia to have a competitor around the market that will open tisthe market, but never let the opponent have the opportunity to change this pattern. Seeing the other party’s routine and position taking shape, he launched a wave of surprise attacks, even if it hurts his previous generation of products, but the impact of “beating cattle” on AMD is obviously more strategic.