Table of contents Table of content memory Innovations on the way to KI telephones the interplay of memory and AI, which goes beyond the RAM capacity.
One of the most obvious – and frankly the most boring – in recent years in the smartphone industry has been the incessant speech about AI experiences. Silicon Warriors in particular have often advertised how their youngest mobile processor would enable AI processes such as videoogenization.
We are already there, if not complete. In the middle of the hype show with hit-and-Miss-Ki-tricks for smartphone users, the debate hardly went beyond the glittering presentations beyond the new processors and constantly developing chatbots.
Only when the absence of the Gemini -Nano on Google Pixel 8 raised the eyebrows that the masses experienced the crucial importance of the RAM capacity for AI on mobile devices. Apple soon made it clear that Apple Intelligence held at least 8 GB RAM of devices.
The image of “AI phone” is not just about the storage capacity. How well your phone with AI-powered tasks can also be done depends on the invisible RAM optimizations and the memory modules. And no, I'm not just talking about the ability.
Memory innovations that are directed to AI telephones
Micron / digital trends
Digital trends were with Micron, a worldwide leading provider of memory and storage solutions to break up the role of RAM and memory for AI processes on smartphones. The progress made by Micron should be on your radar if you buy a first -class phone.
The latest of the company based in Idaho includes the G9 Nand Mobile UFS 4.1 memory and 1 & GGR; (1-gamma) LPDDR5X RAM modules for flagship smartphones. How exactly do these memory solutions push the cause of AI on smartphones, apart from the increase in capacity?
Let's start with the G9 NAND UFS 4.1 memory solution. The Overarching promise is economical power consumption, lower latency and high bandwidth. The UFS 4.1 standard can achieve a sequential reading and writing speed of 4100 Mbit / s, which corresponds to a profit of 15% compared to the UFS 4.0 generation, while the latency numbers also cut off.
Another decisive advantage is that the mobile memory modules from Micron carry up to 2 TB capacity. In addition, Micron has managed to reduce its size, which makes it an ideal solution for foldable phones and slim telephones of the next generation such as the Samsung Galaxy S25 Edge.
Micron / digital trends
Micron has merged with RAM progress and has developed what it is as 1 & ggr; LPDDDR5X -RAM modules. They deliver a top speed of 9200 million MT/S, 30% can pack more transistors due to the reduction in size and consume 20% lower power during time. Micron has already served the somewhat slower 1β-RAM solution (1-beta) in the smartphones of the Samsung Galaxy S25 Series.
The interaction of memory and AI
Ben Rivera, director of product marketing in Micron's mobile business unit, tells me that Micron has made four important improvements to her latest storage solutions to ensure faster AI operation on mobile devices. This includes zones -calls, data surveys, pined writebooster and intelligent latency tracker.
“With this function, the processor or host can identify and insulate and isolate or” pin “the most frequently used data of a smartphone in a area of the storage device called Writebooster buffer (similar to a cache) to enable quick, quick access,” explains Rivera via the feeder writebooster function.
Micron / digital trends
Every AI model thinking of Google Gemini or Chatgpt-, which is to be carried out on tasks of on-device tasks, requires its own instruction files that are stored locally on a mobile device. For example, Apple Intelligence needs 7 GB memory for all Shenanigans.
In order to carry out a task, you cannot deactivate the entire KI package on the RAM, as it needs space to deal with other critical tasks, e.g. B. calling or interaction with other important apps. In order to deal with the restriction of the micron memory module, a memory card is created that only loads the required AI weights from the memory and on the RAM.
When the resources become scarce, you need faster data exchange and read. This ensures that your AI tasks are carried out without influencing the speed of other important tasks. Thanks to Pinned Writebooster, this data exchange is accelerated by 30%to ensure that the AI tasks are treated without delays.
So let's assume that you need Gemini to draw a PDF for analysis. The quick memory exchange ensures that the required AI weights are quickly moved from the memory to the RAM module.
Next we have data questions. Imagine it as a desk or Almirah organizer, which ensures that objects in different categories are properly grouped and placed in their unique cupboards so that it is easy to find them.
Micron / digital trends
In connection with smartphones, since more data is stored over a longer usage period, this is usually stored in a rather random matter. The net effect is that it becomes more difficult to find them all if the onboard system is needed to access a certain type of files, which leads to a slower operation.
According to Rivera, the data call not only helps with the proper storage of data, but also changes the interaction path between storage and device controller. Included Improves the reading speed of data by impressive 60%Which of course accelerates all kinds of user machine interactions, including AI workflows.
“This function can help accelerate AI functions, e.g.
Intelligence Latency Tracker is another feature that essentially keeps an eye on delay events and factors that may slow down the usual pace of your phone. Then it helps to debug and optimize the performance of the phone to ensure that regular and AI tasks do not get to speed.
Micron / digital trends
The final memory improvement is the UFS in zones. This system ensures that data with a similar E/A nature is saved in an orderly manner. This is of crucial importance, since it makes it easier for the system to locate the necessary files instead of wasting time that browsing through all folders and directories.
“Micron's trial function helps organize data so that the system, if the system has to find specific data for a task, is a faster and more smooth process,” said Rivera.
Go beyond the RAM capacity
When it comes to AI workflows, you need a certain amount of RAM. The more, the better. While Apple set the baseline for its Apple Intelligence Stack at 8 GB, players in the Android ecosystem have moved to 12 GB as a safe standard. Why so?
“AI experiences are also extremely data intensive and therefore powerful. In order to promise the promise of AI, memory and storage must offer low latency and high performance with the greatest performance efficiency,” explains Rivera.
With its LPDDR5X RAM solution for smartphones (1 & GGR; 1 & Gamma), Micron has reduced the operating voltage of the memory modules. Then there is the too important question of local performance. According to Rivera, the new memory modules can sum up to 9.6 gigabits per second to ensure first-class AI performance.
Micron / digital trends
According to Micron, improvements in the extreme ultraviolet (EUV) lithography process not only open the doors for higher speeds, but also a healthy jump of 20% of energy efficiency.
The way to private AI experiences?
The RAM and memory solutions of microns for the next generation smartphones are not only geared towards improving AI performance, but also for the general pace of their daily smartphone tasks. I was curious whether the G9 Mobile UFS 4.1 memory and 1 & GGR; (1-gamma) LPDDR5X RAM improvements would also accelerate the offline AI processors.
Smartphone manufacturers and AI laboratories are increasingly changing towards local AI processing. This means that instead of sending your queries to a cloud server on which the process is carried out and the result is sent with an internet connection to your phone, the entire workflow is carried out locally on your phone.
Nadeem Sarwar / Digital trends
From transcripting calls and language notes to the processing of your complex research material into PDF files, everything happens on your phone, and no personal data ever leave your device. It is a safer approach that is also faster, but at the same time strong system resources are required. A faster and more efficient memory module is one of these requirements.
Can microns of the next generation help with local AI processing? It can. In fact, processes are accelerated that require a cloud connection, e.g. B. Videos using Google's VEO model, for which powerful computer servers are still required.
“A native KI app that is carried out directly on the device has the largest data traffic, since you not only read user data from the storage device, but also the AI degree on the device. In this case, our functions would help optimize the data flow for both,” says Rivera.
How quickly can you expect telephones to land on the shelves with the latest micron solutions? According to Rivera, all major smartphone manufacturers will take Micron and memory modules from Micron. With regard to the arrival of the market, “flagship models that will be launched at the end of 2025 or early 2026” should be on their shopping radar.
