A New Era Begins in Phones with Samsung Exynos 2800
Samsung is developing a new formula to bring server-grade HBM memory technology to smartphones. The technology is expected to take flight on-device artificial intelligence with Exynos 2800.
According to information leaked from South Korea-based sources, Samsung is rolling up its sleeves to integrate high-bandwidth memory (HBM) technology, which is currently used only in massive artificial intelligence servers and information centers, into portable devices. This attack promises to increase the local artificial intelligence capabilities and performance of smartphones to a level never seen before.
-
Server Grade Memory Performance:Samsung is developing a new packaging system to fit the HBM memory architecture that powers the great computers in information centers into the tight physical spaces of smartphones.
-
30 Percent Bandwidth Increase:Thanks to the advanced packaging technology developed, data transmission speed increases by 15 to 30 percent compared to standard portable memories, and memory stacking capacity increases 1.5 times.
-
Fully Customized Architecture:Exynos 2800, which is expected to be used in the Galaxy S28 series, makes a difference not only with this memory problem, but also with the special graphics unit (GPU) architecture developed entirely by Samsung.
Hardware Bottleneck in Mobile Devices is Overcome with HBM
Today, artificial intelligence features on smartphones often work in the background by connecting to cloud servers, as they require high processing power. Artificial intelligence models running within the device are limited by the bandwidth limits of existing portable RAM (LPDDR) technologies. In order to solve this problem radically, Samsung is taking the technology called vertical copper pole stack (VCS) even further.
The company increases the aspect ratio of the copper poles that provide the relationship between vertically stacked DRAM layers from the industry standard of 3:1 to a record level of 15:1 to 20:1. In this way, while data paths become narrower, the amount of data passing through the unit area increases tremendously. Advanced molding techniques are used to prevent thin copper structures from bending or breaking.
How to Overcome Thermal Ends and Dimensional Maniacs?
The biggest reasons why HBM memories cannot be used in smartphones are high power consumption, thickness and excessive heat generated. To overcome these physical obstacles, Samsung engineers use the Fan-Out Wafer Level Packaging (FOWLP) packaging technology, which is also used in Exynos processors.
This advanced structural reinforcement both protects the copper contact points and prevents the phone from getting too hot by optimizing heat dissipation. Reducing server-class performance to phone sizes paves the way for portable devices to run massive artificial intelligence models in milliseconds, even without an internet connection.
Galaxy S28 Series Transforms into an Artificial Intelligence Monster
Industry analysts predict that this technology, which is in the development phase, will be commercially available for the first time in Exynos 2800 or Exynos 2900 chipsets. Samsung’s plans for the future include switching to a proprietary GPU architecture that it has designed, completely independent of its AMD subsidiary, with Exynos 2800. Since graphics processing units play a critical role in artificial intelligence calculations, an Exynos 2800 equipped with its own GPU and HBM memory architecture has the potential to completely change the stability in the market.
This major hardware breakthrough not only improves smartphone performance, but also reinforces Samsung’s leadership in the semiconductor and memory market. Being one of the leading suppliers in the portable HBM market gives the company a great financial advantage on a global scale. In the future, smart glasses, autonomous driving systems and advanced robotic technologies will also be fed by this new architecture and become smarter without the need for the internet.
