Support portal Careers Contact Investors
EN
EN · English JA · 日本語 KO · 한국어 ZH · 简体中文
Request a Demo
Request a Demo
Request a Demo
Support portal Careers Contact Investors
Language EN JA KO ZH

Logic and memory semiconductor solutions: TCAD and design for advanced CMOS and memory

A memory semiconductor program can't absorb a failure first found in silicon, so Silvaco lets you model the 3D structure and verify yield across the whole memory before you commit a wafer.

Talk to our Technical Experts Browse the documentation

A 3D problem with a yield budget counted in sigma

On an advanced CMOS or memory program, you can't settle the process first and design the device afterward. A FinFET, a GAA nanosheet or a 3D NAND stack is a 3D object, and each etch and deposition shapes how it behaves electrically. A 2D shortcut throws away the very effects that make or break a new node.

Variation is the second problem. Every bit cell in an array is a copy of the same cell, so even a one-in-a-billion failure rate turns into lost yield. Simulate one cell or a slice, extrapolate, and you're guessing. You need the full memory in view before silicon tells you the same thing at a far higher price.

Where Silvaco fits in your flow

Begin with the 3D structure and end with yield for the full memory. When the question grows past one device, FTCO takes it to the fab.

  1. Your process and bit cell The recipe, the 3D structure and the cell design
  2. Victory Process Turns your process recipe into the 3D structure: FinFET, GAA or memory stack
  3. Victory Device Simulates each structure's electrical behavior, RRAM switching included
  4. VarMan Finds the bit cell's rare failures
  5. VarMan XMA Estimates yield over the entire memory
  6. SmartSpice Pro Handles SRAM arrays full SPICE can't
  7. SmartDRC/LVS Checks memory layouts against the foundry's rule deck
  8. FTCO Picks up the fab-level questions one device can't answer

Memory semiconductor and logic technologies covered

What Silvaco models, cell by cell and across the array.

  • Memory

    DRAM, 3D NAND, flash and HBM, plus emerging memory such as resistive RAM (RRAM).

    Victory Process etch models cut deep holes and reproduce the bowing and twisting real etches leave behind. Victory Device's ReRAM model captures the switching physics inside the cell.

  • Logic

    Planar and advanced CMOS, FinFETs and gate-all-around nanosheet FETs (GAAFETs).

    Fins get their shape from geometric etch models, and Victory Device simulates the finished transistor in 3D. On the design side, Silvaco's standard cells and memory compilers are built for you to tune to your chip, not take as they ship.

  • Yield and variation

    Some bit cells have to be proven at 7 sigma or higher. VarMan gets there at up to 30X the speed of brute-force Monte Carlo.

    VarMan XMA runs the entire memory through your golden SPICE simulator and estimates yield from all of it. No sampled slice, no single critical path.

  • Verification at array scale

    When an SRAM array outgrows full SPICE, SmartSpice Pro still simulates it.

    SmartDRC/LVS runs physical verification on memory designs, and Jivaro Pro reduces post-layout parasitics so sign-off simulation of a dense array stays within reach.

Applications

The work that brings memory and logic teams to Silvaco.

Evidence from memory makers and research groups

Micron, Nanya and PSMC have signed multiyear agreements for Silvaco's sign-off accepted verification, on some of the hardest process nodes in production. SK Hynix signed off HBM4E with help from capacity and performance gains in Jivaro-Pro-DX.

  • CustomerApril 2024

    Micron Technology

    Widened its Silvaco partnership to co-develop FTCO digital twin models for memory and storage. It also put $5.0 million into FTCO's development.

    “Silvaco’s AI and digital twin solution is enabling us to accelerate our groundbreaking advancements in memory and storage.”

    Dr. Gurtej Sandhu, Principal Fellow and Vice President, Micron Technology

  • Research partnerJune 2019

    CEA-Leti

    Ran VarMan XMA on an ultra-low-power 256 Kb SRAM to find its fails and estimate yield. Measured STMicroelectronics 28 nm UT-FDSOI dies tested the predictions.

  • Customer case studyNovember 2019

    Mobile Semiconductor

    Took its memory IP characterization to SmartSpice and SmartView, where a single 22 nm node called for 56,000 SPICE runs. Characterization times improved 50% after the switch.

  • Talk · SURGE 2026

    Advanced Logic FinFET/Nanosheet Simulation with Silvaco Victory Tools

    Professor Zhenhua Wu, Zhejiang University, on how Victory tools model FinFETs and the GAA nanosheets that carry scaling past 3 nm.

  • Webinar · Emerging memoryJune 2020

    Resistive RAM Switching Analysis with Victory TCAD Solution

    How a ReRAM model built on Victory Device's Chemistry Module explains OXRAM and CBRAM switching, for memory past the scaling limit of flash.

  • Webinar · FinFET and memoryAugust 2022

    Learn How to Efficiently Achieve Accurate Experimental Etch Profiles in FinFET and Memory Applications with Victory TCAD Solution

    Geometric etch models matched to measured fin and 3D NAND profiles, bowing and twisting included.

More on logic and memory

Read the Simulation Standard

Working on a new node or memory stack?

Tell our engineers what it is and which step is holding it up. They'll show you what changes, using a structure or memory close to yours.

Talk to our Technical Experts Documentation Support and Downloads

Questions from memory and logic teams

TCAD for memory development simulates how a memory device will be built and how it will behave, before a wafer runs. Victory Process grows the structure step by step in 3D, deep-hole etch and its non-ideal profiles included. Victory Device then simulates it electrically, including resistive switching in RRAM.

A memory repeats one bit cell across the whole array. A failure too rare to show up in one simulated cell still costs yield in silicon. A slice or critical path misses most of the memory, and the yield extrapolated from it can be off. VarMan XMA analyzes all of it instead.

Silvaco's tools cover DRAM, 3D NAND, flash, high-bandwidth memory and emerging RRAM. Silvaco has published deep-hole etch work for 3D NAND and ReRAM switching models. Its full-memory yield analysis is verified on FinFET, FDSOI and planar CMOS foundry technologies, and CEA-Leti tested it on an SRAM project.

High-sigma analysis tells you whether a design holds its spec at failure rates far out in the tail, usually 4 to 6 sigma or more. You need it wherever one cell repeats many times. A memory bit cell may have to hold at 7 sigma, which brute-force Monte Carlo reaches only after billions of runs. VarMan needs far fewer.

For FinFETs, gate-all-around nanosheets and planar or advanced CMOS, Victory Process and Victory Device build the transistor in 3D and simulate it. VarMan covers variation on the same nodes. From there, SmartSpice Pro simulates circuits at array scale and SmartDRC/LVS runs physical verification.

Yes. VarMan and VarMan XMA drive whichever golden SPICE simulator you already trust, SmartSpice or third-party, using the models from your foundry kit. Nothing in your flow has to change. SmartDRC/LVS reads foundry rule decks, so ask our technical experts whether a deck exists for your node.

Related

All solutions