When Intel CEO Pat Gelsinger was interviewed by CRN recently, one of the topics talked about the competition with NVIDIA in the next few years. Looking ahead, one way Intel plans to differentiate itself from the GPU giants is to make the software that supports its components more friendly to the ecosystem.
He said: NVIDIA has become too proprietary, which is well-known in the industry, so we have to use one API to fill this stack, and in a more favorable and open way for the industry and their innovation.
one API is an open programming model that covers different types of architectures, including Intel’s chip products such as CPU, GPU and FPGA. Although acknowledging that there is a gap with NVIDIA in the graphics business, the company hopes to surpass in terms of functionality, performance and power consumption, and provide it to consumers at a convincing price point.
Gelsinger attributed NVIDIA’s success to the company’s operations over the past decade, but also pointed out that they are really lucky in the field of artificial intelligence. Gelsinger explained.
NVIDIA basically has 10 times or better performance leadership in ten years. If you have the ability to take a 10-fold lead in 10 years, people will take advantage of this. Then they were really lucky: AI happened. After 30 years of overnight fame, they harvested it at that time and it was really good. Therefore, they worked hard and won the market, and they were lucky in this regard.
Gelsinger further detailed how Intel intends to provide attractive products in the GPU market. He talked about how they will achieve seamless integration from integration to discrete on Intel platforms. So what are we going to do? To provide excellent products in these areas (…), the market is eager for us to provide them with an alternative. We need to provide it with unique and differentiated value-added.
Join RealMi Central on Telegram, Facebook & Twitter
In the GPU business, we went to the customer and said, Well, guess what, we happen to be the undisputed leader in the integrated graphics field. You have already identified all our products, and we will be on the Intel platform To achieve seamless connection from integration to independence.
Better than this, we will make integration and separation work together. Therefore, if you have the value of three [execution units, EU] in the integrated [GPU], and then you have the value of 10 EUs in the independent, we will give you the value of 13 EUs, and you only need to buy 10 With an EU independent GPU, you will be eligible for a product that works seamlessly between these two products.
Behind his words, Intel has dug up Vineet Goel, AMD’s former top GPU architect, to oversee its Xe GPU architecture, and future GPU products will be based on this. He will lead a team of architects and design engineers to structure, design and verify Intel’s Xe IP roadmap.