Get the newsletter
Samsung · Chip · Filed Jul 29, 2025 · USPTO published Aug 13, 2026 · Aug 13 release

Samsung Patents a Processor That Lets Different Memory Types Work Together Seamlessly

Modern chips often have to talk to two or more completely different kinds of memory at once, and right now that conversation is clunky and slow. Samsung's new patent describes a processing core designed to sit in the middle and translate fluently between them.

High-level block diagram showing the connection between a host device and a memory device. Drawing from patent filing US 2026/0236190 A1.
High-level block diagram showing the connection between a host device and a memory device.All 7 drawings ↓
Publication number
US 2026/0236190 A1(US20260236190A1)
Applicant
Samsung Electronics Co., Ltd.
Filed · Published
Jul 29, 2025 · Aug 13, 2026
Inventors
Zongwang Li, Yang Seok Ki, Rekha Pitchumani
Status when we published
Waiting for an examiner
Document
20 claims
US classification
711/154
Examiner
PARIKH, KALPIT (Art Unit 2137)
Parent application
Claims priority from a provisional application 63757201 (filed Feb 11, 2025)

What Samsung's memory-translation core actually does

Every time your phone or laptop runs an app, several different kinds of memory are working at the same time, passing instructions back and forth. When those memory types speak different "languages," the chip has to stop and work out the translation, which burns time and energy.

Samsung's patent describes a dedicated processing core built specifically to handle that translation job. When it receives an instruction written for one type of memory, it figures out the right way to rewrite that instruction for a different type of memory and sends it along. While the slower memory is doing its job, the core uses its own small, fast internal buffer to keep things moving without waiting around.

The result is that a chip can work with multiple memory types without the usual speed penalty. Think of it like a real-time interpreter at a meeting: instead of the whole room grinding to a halt while one person explains what another said, the interpreter keeps the conversation flowing.

From the filing · CLAIM 1
… generating, by the processing circuit, a second command for a memory controller associated with a second memory type different from the first memory type; …

Translation: The chip translates commands so one type of memory can talk to a completely different type.

How the processing circuit converts and buffers commands

The patent describes a memory-communications core, a dedicated processing circuit that acts as a translator between different memory types inside a chip system.

Here is how the core handles a single operation:

  • It receives a command written for one type of memory (say, a fast, on-chip cache format).
  • It determines the correct processing protocol for that command, essentially the rulebook for how to convert it.
  • It generates a new, equivalent command formatted for a second, different memory type (such as main DRAM), and sends that translated command to the appropriate memory controller.
  • Simultaneously, it performs whatever local bookkeeping is needed on its own internal buffer, a small, very fast scratchpad that is quicker to access than the external memory being targeted.

The key idea is that the core does not sit idle while the slower memory processes the translated command. By offloading local tasks to the on-chip buffer and dispatching translated commands in parallel, the circuit avoids the back-and-forth stalling that normally happens when two memory systems have to coordinate.

The patent covers both the method (the sequence of steps) and the system (the hardware circuit that carries them out), which means Samsung is staking out territory over both the process and the physical implementation.

From the filing · THE ABSTRACT
… wherein the buffer has a lower latency for performing the task than the second memory type.

Translation: It uses a fast internal cache to handle tasks much quicker than standard memory can.

What this means for chips juggling multiple memory types

As chips pack in more types of memory, including different cache layers, high-bandwidth memory, and traditional DRAM, the overhead of coordinating between them eats into performance. A dedicated translation core could make that overhead much smaller, which matters for everything from AI accelerators to server processors where memory bandwidth is often the main bottleneck.

For Samsung, which makes both memory chips and the processors that use them, this kind of patent sits at a natural intersection of its product lines. The company publishes a steady stream of filings in the chip architecture space, and plain-English patent summaries of that work show a consistent push toward tighter integration between processing and memory, a direction the whole semiconductor industry is moving in.

Source: US 2026/0236190 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 · 7 sheets

7 drawing sheets from US 2026/0236190 A1 · click any drawing to enlarge

System architecture overview diagram, drawing sheet 1 from US 2026/0236190 A1
Sheet 1 System architecture overview diagram
Detailed internal controller connections, drawing sheet 2 from US 2026/0236190 A1
Sheet 2 Detailed internal controller connections
Core memory processing components, drawing sheet 3 from US 2026/0236190 A1
Sheet 3 Core memory processing components
Command processing decision flowchart, drawing sheet 4 from US 2026/0236190 A1
Sheet 4 Command processing decision flowchart
Command translation process flow, drawing sheet 5 from US 2026/0236190 A1
Sheet 5 Command translation process flow
Read operation execution sequence, drawing sheet 6 from US 2026/0236190 A1
Sheet 6 Read operation execution sequence
Write operation data flow, drawing sheet 7 from US 2026/0236190 A1
Sheet 7 Write operation data flow
Patent filing page

Watchlists this filing belongs to

All watchlists →