In addition to Weibo, there is also WeChat
Please pay attention
WeChat public account
Shulou
2025-03-29 Update From: SLTechnology News&Howtos shulou NAV: SLTechnology News&Howtos > IT Information >
Share
Shulou(Shulou.com)11/24 Report--
Thank CTOnews.com netizens soft media Xinyou 1933769 for the clue delivery! CTOnews.com, Nov. 20, according to Joongang.co.kr, SK Hynix has begun recruiting designers for logic semiconductors such as CPU and GPU, hoping to integrate HBM4 directly on the chip through 3D stacking.
It is reported that SK Hynix is discussing its HBM4 integrated design method with several semiconductor companies, including Nvidia.
Foreign media believe that Nvidia and SK Hynix will probably jointly design this kind of integrated chip and carry out contract manufacturing with the help of TSMC, and then stack SK Hynix's HBM4 chip onto the logic chip through TSMC wafer bonding technology. In order to realize the integration of memory chip and logic chip, joint design is inevitable.
If SK Hynix succeeds, it could significantly change the way the industry works, because it will change not only the way logic and new chips are interconnected, but also the way they are made.
At this stage, the HBM stack is mainly placed on the intermediary layer next to CPU or GPU and connected to the logic chip using the 1024bit interface. The goal of SK Hynix is to stack the HBM4 directly on the logic chip, completely eliminating the intermediary layer.
To some extent, this approach is similar to AMD's 3D V-Cache stack, which encapsulates the L3 SRAM cache directly on the CPU chip, while HBM can achieve higher capacity and cheaper (CTOnews.com Note: HBM is obviously slower than cache).
At present, one of the main factors troubling the industry is that HBM4 needs to use 2048bit interface, so the intermediary layer of HBM4 is very complex and expensive. Therefore, if memory and logic chips can be stacked together, it is economically feasible, but it also raises another problem: heat dissipation.
Modern logic chips, such as Nvidia H100, are equipped with huge HBM3 memory, which not only brings huge VRAM bandwidth, but also generates hundreds of watts of thermal energy. Therefore, cooling logical and memory packaged assemblies may require very complex methods, even considering liquid cooling and / or immersion cooling.
"if the heat dissipation problem is solved in two or three generations, then HBM and GPU will be able to operate as one without an intermediary layer," said Kim Jung-ho, a professor in the department of electrical and electronic engineering at the Korean Institute of Science and Technology.
"in the next 10 years, the 'rules of the game' of semiconductors may change, and the difference between memory and logical semiconductors may become negligible," an industry insider told Joongang.
Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.
Views: 0
*The comments in the above article only represent the author's personal views and do not represent the views and positions of this website. If you have more insights, please feel free to contribute and share.
Continue with the installation of the previous hadoop.First, install zookooper1. Decompress zookoope
"Every 5-10 years, there's a rare product, a really special, very unusual product that's the most un
© 2024 shulou.com SLNews company. All rights reserved.