Smartphones with Snapdragon 8 Gen 3 and Dimensity 9300 chipsets can now run on-device LLMs (Large Language Models), giving them the same capabilities equivalent to significantly bigger machines. Unfortunately, running AI-related programs on handsets is a resource-intensive task, one that not only requires computing horsepower but also ample amounts of memory and storage, with an earlier report stating that 20GB RAM on future smartphones might become a standard to support this functionality. As for storage, Samsung has a solution for that, and it is rumored to be a new version of the UFS 4.0 standard that is optimized for AI.
It is difficult to determine how much storage needs to be allocated for AI operations to be carried out on smartphones. Still, Revegnus may have a clue, thanks to the claims of Omdia Senior Researcher Yeon Seung-hoon, who states the following while talking about the requirements:
“Next year, with the integration of AI features in smartphones, there will be growth opportunities for storage. To drive AI operations, including learning and inference, more than 15% of the total smartphone storage must be utilized.”
We have seen iPhones ship with 1TB of internal storage, but Android flagships typically feature 256GB of onboard memory as standard, with the less expensive models featuring half that amount. However, with AI operations running, releasing 512GB storage variants might become compulsory, as 15 percent of that will be a whopping 76.8GB. Keep in mind that the operating system storage requirements are separate, not to mention that continuous app updates will gobble that onboard memory instantly.
To help remedy this, Samsung is said to be working on a new version of UFS 4.0 optimized for AI operations,
Read more on wccftech.com