Samsung's research and development division is investing heavily to renew the design of processors, and also wants to drastically improve the definition of OLED displays.
Young Sohn, in charge of Samsung's technology strategy, explains wanting a new generation of processors on the market land, "x86 architecture is limited in memory channels, why limit the memory that speed bandwidth is highly crucial? "He explained to EE Times newspaper.
A model: the brain
The same Sohn Young-called start-up to get into the new trend of the moment: CPU whose architecture mimics the brain, improving memory bandwidth and massive parallel computing. It calls for the development of processors consuming less than one Watt, 300 mW would be ideal, he said, while merging maximum calculation functions and memory game.
NVIDIA compete
Samsung claims to have invested $ 200 million in 60 startups last year, especially for this purpose. One of them is GraphCore, which is developing a co-processor for servers on the model of the human brain for Artificial Intelligence: an IPU for Intelligent Processing Unit, with memory bandwidth advertised 100 times the competition. Samsung and clearly seeks to compete with NVIDIA, very advanced in the area.
10K OLED screens
Young Sohn explains also want to improve the definition of Samsung OLED screens for virtual reality, a project that has yet to put in place: "We certainly need to 10K screens, it is an investment of 5 to 10 billion dollars because each new plant costs this price, especially for OLED. "
Young Sohn, in charge of Samsung's technology strategy, explains wanting a new generation of processors on the market land, "x86 architecture is limited in memory channels, why limit the memory that speed bandwidth is highly crucial? "He explained to EE Times newspaper.
A model: the brain
The same Sohn Young-called start-up to get into the new trend of the moment: CPU whose architecture mimics the brain, improving memory bandwidth and massive parallel computing. It calls for the development of processors consuming less than one Watt, 300 mW would be ideal, he said, while merging maximum calculation functions and memory game.
NVIDIA compete
Samsung claims to have invested $ 200 million in 60 startups last year, especially for this purpose. One of them is GraphCore, which is developing a co-processor for servers on the model of the human brain for Artificial Intelligence: an IPU for Intelligent Processing Unit, with memory bandwidth advertised 100 times the competition. Samsung and clearly seeks to compete with NVIDIA, very advanced in the area.
10K OLED screens
Young Sohn explains also want to improve the definition of Samsung OLED screens for virtual reality, a project that has yet to put in place: "We certainly need to 10K screens, it is an investment of 5 to 10 billion dollars because each new plant costs this price, especially for OLED. "
0 commentaires:
Enregistrer un commentaire