Get the newsletter
Samsung · Chip · Filed Apr 3, 2026 · USPTO published Aug 13, 2026 · Aug 13 release

Samsung Patents Chip That Automatically Translates Memory Requests Into the Right Commands

Samsung has filed a patent for a memory architecture where the controller, not the main processor, figures out exactly how to talk to memory, potentially saving the CPU from doing that translation work itself.

Architecture of a processing system routing memory requests through a controller to physical memory. Drawing from patent filing US 2026/0236396 A1.
Architecture of a processing system routing memory requests through a controller to physical memory.All 8 drawings ↓
Publication number
US 2026/0236396 A1(US20260236396A1)
Applicant
SAMSUNG ELECTRONICS CO., LTD.
Filed · Published
Apr 3, 2026 · Aug 13, 2026
Inventors
Hyunsoo KIM, Yuhwan RO
Status when we published
Waiting for an examiner
Document
20 claims
US classification
711/118
Parent application
Continuation of 18/350,074 (filed Jul 11, 2023)(18350074)

How Samsung's memory address trick offloads the CPU

Your phone or laptop's main chip currently has to handle a lot of bookkeeping when it reads or writes data to memory. It needs to know not just where data lives, but how to ask for it depending on what kind of memory is being used. That overhead adds up.

Samsung's patent describes a system where a separate memory controller takes on that job. The processor hands over a memory address, and the controller figures out which memory mode applies and translates that into the right command for the actual memory chip. The processor doesn't need to know the details of each memory type, it just makes a request and moves on.

For everyday users, this kind of design could mean a device that handles multiple memory types without the main chip getting bogged down in translation work. It's the sort of behind-the-scenes efficiency that shows up as faster, more responsive hardware, especially in devices that mix different memory technologies on the same board.

From the filing · CLAIM 1
… a memory controller, separate from the host processor, configured to generate the physical memory address and a memory device command according to the target memory mode mapped to the memory address based on the memory request and the memory address received from the host processor …

Translation: A dedicated controller sits between the main processor and memory to translate requests into specific hardware commands.

How the controller maps addresses to modes and commands

The patent describes a three-part system: a host processor, a memory controller, and a memory device.

The host processor generates a memory address, essentially a label saying 'I need to read or write data here.' But that address is special: it's part of an extended memory address space, meaning different address ranges are mapped to different memory modes (think of modes as different dialects that different types of memory chips speak).

The memory controller sits in the middle and does the heavy translation work. When it receives an address and a request from the processor, it looks up which memory mode that address belongs to, then generates both the correct physical memory address (the actual hardware location) and the specific command the memory chip needs to receive. Different memory technologies, such as standard DRAM versus newer or specialized memory types, may require different command formats entirely.

The memory device at the far end simply receives the translated command and physical address and executes them. The key architectural claim is that the host processor's operating system kernel doesn't need to understand the specifics of each memory type, that complexity is abstracted away into the controller layer.

From the filing · THE ABSTRACT
Disclosed are a device and method with memory request processing using an extension of a memory address space. An electronic device includes a host processor configured to generate a memory request and a memory address that is mapped to a target memory mode that is any one of available memory modes …

Translation: The system uses a special memory address map to tell the hardware exactly which mode to use for each data request.

What this means for chips handling mixed memory types

Chips are being asked to work with a growing variety of memory technologies at once, from traditional DRAM to high-bandwidth memory to newer non-volatile options. Managing all of those different protocols at the processor level adds software complexity and can slow things down. Offloading that translation to a dedicated memory controller keeps the main processor focused on actual computation.

For Samsung, which manufactures both the processors and the memory chips that go into phones, tablets, and data center servers, having a clean architectural patent in this area matters. The latest Big Tech patents in semiconductor memory architecture show a clear push toward separating memory management from core compute, and this filing fits squarely into that trend.

Source: US 2026/0236396 A1 at the USPTO (PDF). Full patent text and figures are in the official publication.

Get our take in your Top Stories

Liked this breakdown? Add Patentlyze as a preferred source on Google, and our plain-English take shows up more often in your Top Stories the next time Samsung patent news breaks.

There are more where this came from

We read the new patents from Big Tech, carmakers and more and send you the ones worth knowing. Plain English, free, every week.

The drawings · 8 sheets

8 drawing sheets from US 2026/0236396 A1 · click any drawing to enlarge

Processor memory routing architecture, drawing sheet 1 from US 2026/0236396 A1
Sheet 1 Processor memory routing architecture
Physical and virtual memory spaces, drawing sheet 2 from US 2026/0236396 A1
Sheet 2 Physical and virtual memory spaces
Memory allocation workflow, drawing sheet 3 from US 2026/0236396 A1
Sheet 3 Memory allocation workflow
Physical memory address bit layout, drawing sheet 4 from US 2026/0236396 A1
Sheet 4 Physical memory address bit layout
Application memory request path, drawing sheet 5 from US 2026/0236396 A1
Sheet 5 Application memory request path
New instruction set memory request path, drawing sheet 6 from US 2026/0236396 A1
Sheet 6 New instruction set memory request path
Neural network layer processing pipeline, drawing sheet 7 from US 2026/0236396 A1
Sheet 7 Neural network layer processing pipeline
Memory command execution workflow, drawing sheet 8 from US 2026/0236396 A1
Sheet 8 Memory command execution workflow
Patent filing page

Watchlists this filing belongs to

All watchlists →