MSquare Technology
MSquare Technology
COMPANY PROFILE
MSquare VISION
MEMORABILIA
HONORS
PRODUCTS OVERVIEW
HBM3 IP
D2D (UCIe) IP
LPDDR4X IP
LPDDR5X IP
ONFI 5.0 IP
ONFI 5.1 IP
eDP IP
PCIe 4.0 IP
USB 3.2 IP
ML100 IO Die
MSquare Technology, incorporated in 2021, is a leading provider of integrated circuit IPs and Chiplets, dedicated to
addressing the challenges of chip interconnectivity and vertical integration in the smart economy era. MSquare operates
offices in Shanghai, Taipei, Tokyo, Sydney, and San Jose, boasting a team of a few hundred employees, with 80% dedicated
to research and development. MSquare strives to foster an open ecosystem service platform for AI and Chiplets, providing
comprehensive support for innovation and development within the IC and Chiplet industry.
MSquare is named after the 2MASX occulting pair, a twin galaxy system comprising overlapping spiral galaxies whose
brilliant illumination extends six times their radius, guiding astronomers in observing surrounding galaxies. Inspired by this
phenomenon, MSquare aims to lead the semiconductor industry with cutting-edge products and services, striving to light
up the industry and cultivate a thriving, open ecosystem in collaboration with industry stakeholders.
COMPANY PROFILE
01
As a key technological leader upstream in semiconductor industry, MSquare scales the frontiers of chip design
technology through integrating system design, Chiplet Die, and custom services within the 2.5D OSAT alliance.
MSquare aspires to be pioneer in the AI era, focusing on the innovative research and development of
interconnect interface IPs and Chiplets across various fields, including AI, Data Centers, Automotive Electronics,
the IoT, and Consumer Electronics. By providing cutting-edge R&D and customized services for IPs and Chiplets,
MSquare is dedicated to leading industry standards and driving deep innovation in the hard core technology.
MSquare aims to forge a global leading Chiplet brand and pioneer the interconnected IP frontier. Leveraging a
comprehensive portfolio of IP products, MSquare is actively constructing a leading IP infrastructure platform to
drive AI and Chiplet technology innovation. MSquare's IP products have been successfully validated by notable
foundries process nodes and brought to mass production, spanning from 5nm to 180nm and covering over 400
different process nodes across 5 leading foundries. The R&D team has launched interconnect interface IPs
including HBM, LPDDR, ONFI, UCIe, eDP, PCIe, USB, as well as Chiplet solutions represented by M2LINK. MSquare
aims to build an open ecosystem service platform, continuously empowering digital transformation.
PRODUCT VISION
INDUSTRY VISION
02
MEMORABILIA
03
2021
MSquare Technology
Incorporated
Aug.
Office in San Jose, U.S. Established
Sept.
Pre-A Financing Completed
Nov.
PSRAM PHY Delivered
Dec.
MSquare Joined MIPI Alliance
Jan.
PCIe 4.0 12nm Tapeout
Feb.
PSRAM PHY First Customer Tapeout
Mar.
MSquare Joined UCIe Alliance
Apr.
MSquare Joined USB-IF Association
Jun.
MSquare Obtained ISO 9001
Certification
Aug.
ONFI 5.0 12nm Tapeout
Oct.
USB 3.2 12nm Tapeout
Jun.
Series A financing Completed
Dec.
2022
04
MSquare Joined CCITA Chiplet
Working Group
Jan.
LPDDR4X 12nm Tapeout
Feb.
Chiplet IO Die Ready
Jun.
ONFI5.1 5nm Delivered
Aug.
LPDDR5X 6nm Tapeout
May.
UCIe 12nm Tapeout
Sept.
MSquare Open Day
Jun.
Office in Brisbane, Australia Established
Aug.
ONFI5.1 12nm Delivered
LPDDR5X 6nm Delivered
HBM3 12nm Delivered
Oct.
MSquare Awarded
“Top 10 IP Company on
the Fabless 100 List”
Apr.
2023
2024
MSquare Awarded
“2023 Venture 50 List”
Jan.
MSquare Obtained ISO 26262:
2018 Certification
Feb.
CEO Awarded “2023 Most Noteworthy
Female Entrepreneur”
Mar.
MSquare Awarded “China's Top 100
Future Unicorns” and “Hard & Core
Technology Innovators”
Dec.
HONORS
05
Innovative IP Company of
the Year 2022 Top 20 Tech Cool Vendors
with the Most Commercial
Potential in the AI Field
2022 The Most Investment
Value Innovative Enterprise The Best Product Award
of 2023
2023 Most Noteworthy
Female Entrepreneur The Most Innovative
Products of 2023
The Best Employers of
2023
Hard & Core Technology
Innovators
2023 Venture 50
2023 Science & Technology
Good Company
2023 The Most Investment
Value Company
China's Top 100 Future
Unicorns
49+
Invention patent
10+
Layout patent
17+
Copyright & Trademark
30+
Awards
M2
Chiplet
D2D/UCIe M2LINK M2Chip IO Die Integration Service
PRODUCT OVERVIEW
High Speed
Interface IP
HBM LPDDR ONFI eDP PCIe USB
Processor Core Hardening+OPPA Interface IP Integration
IP Integration
and Harden
Service
Foundation
IP
Memory Compiler GPlO, Specialty IO Standard Cell
06
HBM3 IP PRODUCT INFO
07 For more process nodes, please contact us
OVERVIEW
MSquare's HBM3 IP (the 3 generation of High-Bandwidth Memory) is specifically tailored for applications
that require high memory throughput and low latency, complying with the JESD238 memory standard. It
includes both PHY and Memory Controller components, supporting HBM3 SDRAM speeds ranging from
4.8Gbps/pin to 6.4Gbps/pin. Flexible configurations are available, including PHY Only, PHY + Controller,
and Controller Only, to accommodate diverse customer design specifications. Additionally, the chip's
footprint and power consumption are highly competitive within the industry.
HIGHLIGHTS
Designed for high memory throughput and low latency applications
Figure 1: Block Diagram
Consists PHY and Memory Controller
Optimized for 12nm process
Supports speed up to 6.4Gbps/pin
Features integrated PLL and IO
Flexible configurations available: (PHY only) or (PHY + Controller) or (Controller only)
Supports both firmware-based training and hardware-based training
Optional Add-on IPs to achieve best performance
Minimized footprint and power consumption
NOC
NOC
NOC
Optional Add ons
X2C
CMDPATH
DATAPATH
CCS Host Interface (AXI4) PHY Interface (DFI/HFI)
HBM3MCTL HBM3PHY
PLL
ZCAL
DLL
PHY Micro
Controller
Aword Path
Dword Path
HBM3 I/O DFI/HFI Decoder
HBM3
DRAM
Stack
Host
Host
ASIC/FPGA
M2 HBM3 Host IP
rd
OVERVIEW
The Die-to-Die interface is a functional block that provides a data interface between two chip dies within
the same package. MSquare's D2D solution, compatible with the UCIe v1.1 specification, includes both a
Die-to-Die adapter layer and a physical layer. Each layer features a sideband interface that provides a
back-channel for link training and access to the registers of the link partner, as shown in Figure 1. This
unique hybrid analog/digital architecture offers low power consumption, compact footprint, and robust
performance, making it well-suited for target applications such as high-performance computing, AI, or
multimedia SoCs, and Die-to-Die interconnections.
Compatible with UCIe v1.1 specification
Features single-ended, source-synchronous, and DDR I/O signaling
Supports 32-bit (16-bits TX + 16-bit RX) data bus per module for standard packages
Offers a high clock frequency up to 16GHz
Provides data transfer rate up to 32Gbps per lane
Delivers 1Tbps (512Gbps TX + 512Gbps RX) bandwidth per module for standard packages
High energy efficiency with ~0.8pJ/bit for standard packages
Built-in test and diagnostics [Functional (PRBS), scan, at-speed external loopback]
The advanced 12nm/6nm process
Figure 1: Architecture Diagram Figure 2: Block Diagram
DIE 0 DIE 1 (R180 of DIE 0)
UCle PHY
TX 16lane
RX 16lane
UCle PHY
TX 16lane
RX 16lane
UCle PHY
TX 16lane
RX 16lane
UCle PHY
TX 16lane Module3
RX 16lane
Module2 Module1 Module0
UCle
Adapter
Customer
Core
Logic
UCle PHY
UCle PHY
TX 16lane
RX 16lane
TX 16lane
RX 16lane
UCle PHY
TX 16lane
RX 16lane
UCle PHY
TX 16lane
RX 16lane
Module3 Module2 Module1 Module0
UCle
Adapter
Customer
Core
Logic
Substrate
32Gbps/lane
512Gbps TX + 512Gbps RX
Protocol Adapter DPHY APHY
Protocol Layer Adapter Layer Physical Layer
Sideband_
Prot
Sideband_
Adap
Sideband_
Phy
Flit aware
D2D interface
(FDI)
Raw D2D
interface
(RDI)
x16 MB Data
Forward CLK
SB Data
Forward CLK
08
D2D (UCIe) IP PRODUCT INFO
For more process nodes, please contact us
HIGHLIGHTS
LPDDR4X IP PRODUCT INFO
MSquare's LPDDR4X PHY is a transceiver physical layer IP interface solution designed for ASICs and
SOCs. Compatible with the universal IP protocol DFI 4.0, it operates at a data rate of up to 4267Mbps with
a 16-bit data width per channel. As depicted in Figure 1, the LPDDR4X PHY block diagram for system
application includes a data and address transmit path, a receiver data path, a PLL block, and more.
Featuring a hybrid analog/digital architecture, MSquare's IP delivers low power consumption, a compact
footprint, and robust performance, making it well-suited for LPDDR4X applications.
Compatible with LPDDR4X, with a maximum rate up to 4267Mbps
Compliant with DFI 4.0 for PHY and control interfaces
Flexible channel architecture
Supports PHY-independent training mode using an embedded processor
Supports dual rank
DFS supports 4 trained frequencies
Supports BIST and loopback modes
Supports background tracking for PVT compensation
Integrated low-jitter PLL and DLL
Compliant with JEDEC standard JESD209-4C
Figure 1: System-Level Block Diagram
DRAM
Device
Interface
MC
DFI 4.0
LPDDR4X
Soft_PHY DFI Interface 4.0
ACP2S
DP2S
DS2P
AC_TX
DATA_TX
DATA_RX
PIPE_AC Address Macro
Data Macro
PLL,DLL,other logic
PIPE_TX
PIPE_RX
BIST etc.
09 For more process nodes, please contact us
OVERVIEW
HIGHLIGHTS
LPDDR5X IP PRODUCT INFO
OVERVIEW
MSquare's LPDDR5X/4X PHY is a transceiver physical layer IP interface solution designed for ASICs and
SOCs. Compatible with the universal IP protocol DFI 5.0, it operates at data rates up to 8533Mbps with a
16-bit data width per channel. With flexible configuration options, the LPDDR5X/4X PHY supports a
variety of mobile applications using LPDDR5X, LPDDR5, and LPDDR4X SDRAMs. Featuring a hybrid
analog/digital architecture, MSquare's IP delivers low power consumption, compact footprint, and robust
performance, making it well-suited for LPDDR5X/4X applications.
HIGHLIGHTS
Compatible with JEDEC standards LPDDR4X, LPDDR5 and LPDDR5X SDRAMs
Supports for data transfer rate up to 8533Mbps
DFI 5.0 for PHY and controller interfaces
Supports both firmware-based training and hardware-based training
Supports multiple gears of DFS
High performance, low jitter DDR IO
Single Tap DFE for Channel Robustness
Supports 4 trained frequencies
User-customizable arbiter (scheduler)
Compliant with JEDEC standards JESD209-4D and JESD209-5B
Figure 1: System-Level Block Diagram
For more process nodes, please contact us 10
LPDDR LPDDR PHY
Controller
CK
Domain
WCK
Domain
Memory
Array
LPDDR5X
Device
BIST
HW
Training
Engine DLL
FSP
DCM
Power
Mqr
MCU
CMD/ADDR
Block
Data Block
DDR IO DFI LogicDFI Interface DRAM CA and Data
PLL
AXI TX
AXI RX
Arbitration
System
Control
Logic
ONFI PHY
ONFI 5.0 IP PRODUCT INFO
OVERVIEW
MSquare offers a silicon-proven ONFI 5.0 PHY IP that supports all modes of the Open NAND Flash
Interface (ONFI) 5.0 specification. ONFI is an interface that connects external flash particles (NAND
Flash) and internal flash controller (NAND Controller), commonly used in Solid State Drives (SSD) and
embedded memory (eMMC) products. To facilitate customer implementation, the MSquare's ONFI 5.0
PHY supports expandable channels and customizable CE/RB Pad Numbers.
HIGHLIGHTS
Compliant with ONFI 5.0 specification
Supports NV-DDR2 mode
Supports NV-DDR3, NV-LPDDR4, with a maximum rate of 2400MT/s
Supports PHY Independent TX/RX Training mode
Supports DFI with 1:1 / 1:2 clock ratio mode
Supports self-test mode with built-in BIST
Supports loopback test with built-in PRBS
Supports ODT programmable and ZQ calibration
Integrated low-jitter PLL & DLL
Figure 1: Architecture Diagram Figure 2: Block Diagram
PHYD PHYA IO
Processor or Main chip
eMMC I/O,UFS,SATA and etc
SSD/eMMC
Flash ONFI interface
controller Flash memory
array
For more process nodes, please contact us
PHY_reg
Build-in
Bist Training
Ctrl path
Data path
Ctrl slice
Ctrl IO
Data slice
DLL PLL Data
IO
11
ONFI 5.1 IP PRODUCT INFO
OVERVIEW
HIGHLIGHTS
Compliant with ONFI 5.1 specification
Supports NV-DDR3/NV-LPDDR4, with a maximum rate up to 3600MT/s
Supports matched or unmatched DQS
Supports WDCA/Per-Pin VREFQ Training for NAND devices
Supports WT Monitor
Supports Separate Command Address (SCA) to improve NAND IO efficiency
Supports 2 channels with a single PHY PLL & flexible floorplan with double or single row IO
Supports DFI with 1:1 / 1:2 clock ratio mode
Supports TX/RX Training mode
Delay-Chain supports background tracking for PVT compensation
Test Mode: Self-test mode with built-in BIST & Loopback test with built-in PRBS & BIST-Delay-Chain
Figure 1: Architecture Diagram Figure 2: Block Diagram
Processor or Main chip
eMMC I/O,UFS,SATA and etc
SSD/eMMC
Flash ONFI interface
controller Flash memory
array
ONFI PHY
PHYD PHYA IO
Ctrl IO
PHY_reg
Build-in
Bist Training
Ctrl path
Data path
Ctrl slice
Data slice
DLL PLL Data
IO
For more process nodes, please contact us
MSquare's ONFI 5.1 PHY is a high-performance and low-power PHY IP, which supports all modes of the
Open NAND Flash Interface (ONFI) 5.1 specification. ONFI is an interface connecting external flash
particles (NAND Flash) and internal flash controller (NAND Controller), widely used in Solid State Drive
(SSD) products. To simplify the implementation, MSquare ONFI 5.1 PHY supports expandable channels
and customizable CE/RB pad numbers.
12
eDP IP PRODUCT INFO
13 For more process nodes, please contact us
OVERVIEW
HIGHLIGHTS
Compatible with embedded display port v1.5/v1.4b specification
Figure 1: System-Level Block Diagram
Four-lane main link with support for 8.1/5.4/4.32/3.24/2.7/2.43/2.16/1.62Gbps
Supports Enhanced Framing Mode
Automatic Link-Training with an option for firmware-controlled Link-Training procedure
Supports Fast Link Training/No Link Training
Supports AUX Channel for access to DPCD and EDID
Supports a maximum 144Hz refresh rate
Supports variable refresh rate and Adaptive-Sync for VESA Adaptive Sync
Supports Panel Self Refresh (PSR) function and PSR with selective update (PSR2) function
Supports VESA DSC v1.01, v1.1, v1.2a with FEC
Supports Panel Replay (PR) and Adaptive Refresh Panel (APR)
Image buffer
Logic
AUX CH
Com
Logic
Decode
DPCD
EDID
PSR
M-Touch
BackLight
RX Bit Recovery Lock
Image
Framebuffer
Display
Main Link
Hot Plug
Detect
AUX
TX
Driver
MSquare's eDP PHY is a highly reliable solution for display interface requirements, supporting resolutions
of 4K and higher. By operating at substantially higher bit rates, this eDP solution reduces both the number
of wires and pins required compared to interfaces like LVDS, DVI, and HDMI. It is fully compliant with eDP
v1.5 and v1.4b standards, and capable of driving up to 8.1Gb/s per lane in configurations up to 4 lanes. It
consists of electrical sub-blocks of main-link and AUX channel of PHY for eDP. MSquare provides eDP with
superior PPA and offers customization based on customer specifications.
OVERVIEW
MSquare's PCI Express 4.0 PHY IP includes a high-speed, highly-efficient, and cost-effective transceiver
to turbocharge today's high-performance and high-bandwidth applications such as data centers,
automotive, enterprise computing, and high-performance embedded systems.
The PCIe IP boasts a low power consumption and compact silicon footprint. Its robust PHY architecture
tolerates process, voltage, and temperature (PVT) variations. This IP integrates high-speed mixed-signal
circuits to support PCIe 4.0 traffic at 16Gbps. It is backward compatible with PCIe 3.1 data rates at
8.0Gbps, PCIe 2.1 at 5.0Gbps, and PCIe 1.1 at 2.5Gbps. The multi-tab transceiver design, accompanied
by a robust BIST, an embedded bit error rate (BER) tester, and an internal eye monitor, enables designers
to control, test, and monitor signal integrity without the need for expensive test equipment.
HIGHLIGHTS
Compiles with PCIe 4.0, 3.1, 2.1, 1.1 and PIPE 4.4.1 specifications
Supports all power-saving modes (P0, P0s, P1, P2) as defined in PIPE 4.4.1 specifications
Supports L1 PM/CPM substates with CLKREQ#
Supports the separate REFCLK Independent SSC (SRIS) architecture
Accessible register control TX PLL bandwidth, TX de-emphasis level, CDR bandwidth, and EQ strength
Supports both FOM and DIR modes for Link-EQ Training
Supports robust BIST/DFT functions for mass production tests
Supports bifurcation with 4-lane x 1 or 2-lane x 2 configurations
Figure 1: Block Diagram Figure 2: Chip Layout
pattem
checker
RX Logic 8b/10b128b
/130b
Decoder
Elastic
Buffer Deserilizer EQ/CDR
TX
Driver Serializer
Clock Multiplying Units
8b/10b
128b/130b
Encoder
TX Logic
Pattem
Generator
RXDP
RXDM
REFCLKM
TXDP
TXDM
REFCLKP +
-
MUX
32 bit @500MHz
32bit @500MHz
Control
/status
Rate
-
+
PCIe 4.0 IP PRODUCT INFO
For more process nodes, please contact us 14 PCIe 4.0 Controller
OVERVIEW
MSquare's USB 3.2 PHY IP is designed based on the USB 3.2 Gen2x1 and USB 2.0 specifications from
the USB Implementers Forum and delivers bandwidth up to 10Gbps. This high bandwidth capability is
particularly beneficial for consumer and industrial video applications, as well as display, docking, cloud
computing, and automotive applications.
As a leading supplier of USB IP, MSquare offers customers low-power, compact, and high-performance
IPs featuring a unique hybrid analog/digital architecture. This IP supports all USB functions, including
transceivers for 10G, 5G, 480M, 12M, and 1.5Mbps. It provides complete deliverables that simplify SoC
integration and silicon validation. Available in a wide variety of process nodes from 40nm to 6/7nm, this IP
features a compact die size and low pad count, enhancing product competitiveness in the USB market.
HIGHLIGHTS
Fully compliant with USB 3.2 Gen2x1 featuring PIPE 4.3 and USB 2.0 with a UTMI+ interface
Figure 1: Architecture Diagram Figure 2: Chip Layout
Patternche
Checker
RX Logic 8b/10b
128b/132b
Decoder
Deserilizer
8b/10b
128b/132b
Encoder
TX Logic
Pattem
Generator
MUX MUX DEMUX MUX
CC1
SSRXP0
SSRXM0
SSRXP1
SSRXM1
SSTXP0
SSTXM0
SSTXP1
SSTXM1
CC2
USB 2.0 PHY DP DP
DP
PLL And
Clock Generator
Crystal/
Coreclkin
DP
PIPE
Elastic
Buffer
UTMI+
Crystal /
Coreclkin
Serializer TX Driver
Serializer TX Driver
EQ/CDR
CC SSCG-PLL Module
Supports host, peripheral, and dual-role applications
Supports 10/12/25/30/19.2/24/27/40MHz crystal oscillators or clock inputs
Supports TX 3-Tap FFE and RX CTLE+1-Tap DFE for SS+
Integrates an active switch to support orientation-less connections with the USB Type-C connector
Provides an auxiliary CC module IP for USB Type-C related functions
Offers robust BIST functions for mass production tests
Supports a crystal pad for clock source, compatible with MSquare's XTAL module IP
Supports OTG applications, compatible with MSquare's IDPAD module IP
Available in 7nm, 12nm, 16nm, 28nm, and 40nm processes
Supports both wire-bond and flip-chip package types
USB 3.2 IP PRODUCT INFO
15 For more process nodes, please contact us
For more process nodes, please contact us 16
ML100 IO Die PRODUCT INFO
OVERVIEW
HIGHLIGHTS
Figure 1: Block Diagram
SOC Die
(High Speed
UCIe Interface) HBM3 6.4G
DRAM
HBM3
PHY
HBM3
Ctrl
AXI 4.0
BUS
Moniter
VT
Sensor
I2C QSPI Jtag
GPIO UART QSPI I2C
Nor
Flash
RV-CPU
TX
RX
TX
RX
TX
RX
TX
RX
TX
RX
TX
RX
TX
RX
TX
RX
TX
RX
TX
RX
TX
RX
TX
RX
TX
RX
TX
RX
TX
RX
TX
RX
Adapter Protocol Adapter Protocol Adapter Protocol Adapter Protocol
APB
Mainband
Sideband
APB AHB-BUS
MSquare's ML100 IO Die is a high-bandwidth memory solution that integrates efficient UCIe (Die-to-Die)
interconnect IP and HBM3 IP. The UCIe IP follows the UCIe 1.1 Specification and supports both standard
and advanced packaging technologies, offering up to 1 TB/s of transfer bandwidth in a single module
configuration. The UCIe supports the AXI4.0 interface standard, enabling ultra-low latency, high-speed
interconnectivity between two dies. And the integrated HBM3 IP adheres to the HBM3 JESD238 standard
and supports IO transfer rates of up to 6400 Mbps.
Integrates 16 complete UCIe modules, following the standard UCIe 1.1 standard
Integrates one complete HBM3 IP, including the controller and PHY, following the HBM3 JESD238 standard
Supports interconnectivity with the SoC via the AMBA AXI4.0 standard
The AXI4.0 bus supports bandwidth traffic and latency statistics
Supports various peripherals, i.e., NOR Flash, QSPI, I2C and UART
Maximum bandwidth is 819.2 GB/s, matching the bandwidth of 6400 Mbps HBM3 DRAM
Utilizes a 2.5D packaging solution, interconnected with the SoC through substrate
Decouples the HBM from the SOC, reducing the thermal effect on HBM DRAM
www.m2ipm2.com
E-mail: marcom@m2ipm2.com
MSquare E-Book
Shanghai|Taipei|Tokyo|Sydney|San Jose
MSquare LinkedIn
Chiplets Interconnection Blazing,
Massive Computing Wellspring.