Samsung’s 3nm process technology represents a major milestone in advanced semiconductor manufacturing, as it introduced Gate-All-Around (GAA) transistor architecture to high-volume production ahead of competitors in some respects. Samsung Foundry pioneered this shift from the traditional FinFET design used in earlier nodes (and still used by TSMC at their initial 3nm node).

Introduction and Key Features

Samsung’s 3nm process, often referred to under nodes like SF3 (or variants such as SF3E for the first generation and SF3 for the second), relies on MBCFET (Multi-Bridge Channel FET), Samsung’s specific implementation of GAA technology. In GAA transistors, the gate fully surrounds the channel (using nanosheets instead of fins), allowing better control over the current flow. This design helps reduce leakage, improve drive current, and enhance overall power-performance-area (PPA) metrics compared to FinFETs.

Samsung officially began initial production of chips using its 3nm GAA process in mid-2022 (specifically June 2022), marking it as one of the first companies to ship GAA-based 3nm chips. The company has described the node as delivering significant improvements over its prior 5nm and 4nm processes, including:

• Up to ~45-50% reduction in power consumption.
• Around 30% improvement in performance.
• Approximately 30% reduction in area (or higher density).

These gains come from the wider channel area provided by nanosheets, which can be stacked and adjusted for better electrostatic control and efficiency. Samsung’s official foundry page highlights the SF3 node (mass production started in 2022) as using special nanosheets to achieve these PPA benefits.

Generations of Samsung’s 3nm Process

Samsung’s 3nm efforts are divided into generations:

• First-generation (3GAE or SF3E) — Initial version, entered risk production around 2022 and moved to mass production/shipping shortly after. It targeted low-power, high-performance applications.
• Second-generation (SF3 or similar) — An enhanced iteration with optimizations like variable nanosheet channel widths within the same cell for greater design flexibility, improved performance (e.g., ~22% higher at iso-power vs. prior 4nm), and better power efficiency (~34% reduction).

The second-generation aimed for further refinements but faced significant challenges in scaling.

Production Status and Challenges as of March 2026

By early 2026, Samsung’s 3nm process has been in production for several years, but it has not achieved the widespread commercial success or yield stability seen at rival TSMC’s 3nm nodes (which use FinFET initially before transitioning to GAA at 2nm).

Key issues have centered on yield rates (the percentage of functional chips per wafer):

• First-generation yields reached around 50-60% in some reports (an improvement over early lows but still below industry targets for high-volume, cost-effective production).
• Second-generation yields have been reported as significantly lower, often in the 20% range or below 30% in various industry analyses from 2024-2025. This is well short of internal goals (e.g., 70%) and has limited external customer adoption.

These yield struggles contributed to delays, such as pushing mass production at Samsung’s Taylor, Texas fab to 2026 and causing some customers to prefer TSMC (whose 3nm yields were much higher, often 80-90%+). Samsung’s 3nm production has remained largely internal, used for certain Exynos processors (e.g., in wearables or select Galaxy models), rather than attracting major external clients like those for AI, mobile, or HPC chips.

Despite these hurdles, Samsung has continued refining the technology. The company has shifted focus toward its 2nm GAA process (SF2), where yields have improved to 50-60% ranges by late 2025, with mass production announcements and claims of 5-8% better performance/efficiency over the 3nm second-generation. Some reports indicate Samsung is applying lessons from 3nm challenges to accelerate 2nm progress, including potential use in next-gen Exynos chips and even HBM memory base dies.

Comparison to Competitors

TSMC’s 3nm (N3 family) entered high-volume production in 2022 using FinFET, achieving strong yields and capturing major clients (e.g., Apple, NVIDIA). TSMC’s GAA transition comes at 2nm (expected broader ramp in 2025-2026). Samsung’s earlier GAA adoption gave it a theoretical edge in transistor control, but practical yield and ecosystem issues have limited its foundry market share gains.

In summary, Samsung’s 3nm process was groundbreaking for introducing GAA at scale in 2022, offering strong theoretical PPA advantages through nanosheet design. However, persistent yield challenges (particularly in the second-generation) have constrained its broader adoption and profitability in the foundry business as of 2026. The company has pivoted emphasis to maturing its 2nm node, where early signs show better progress and potential to compete more effectively in the future.

---

Gate-All-Around (GAA) transistor design

Gate-All-Around (GAA) transistor design, also known as GAAFET (Gate-All-Around Field-Effect Transistor), represents the most advanced evolution in transistor architecture for leading-edge semiconductor nodes (primarily 3nm and below, with widespread adoption at 2nm-class processes as of 2026). It builds directly on the principles of earlier designs like planar MOSFETs and FinFETs but addresses their fundamental scaling limitations through superior electrostatic control.

Evolution Leading to GAA

To understand GAA, consider the progression of transistor structures:

• Planar MOSFET (pre-2010s dominant): The gate sits flat on top of the channel (a thin silicon layer between source and drain). As transistors shrank below ~20-30 nm, short-channel effects became severe — the gate lost effective control over the channel, leading to high leakage current (especially subthreshold leakage) even when the transistor was supposed to be off.
• FinFET (introduced ~2011–2012, dominant from 16/14nm through initial 3nm nodes): The channel is raised into a vertical “fin” structure. The gate wraps around three sides of the fin (top and both sidewalls), dramatically improving gate control compared to planar. Multiple fins can be placed side-by-side to increase drive current. FinFET enabled scaling down to ~5–3 nm while keeping leakage manageable and performance high. However, as dimensions continued shrinking (especially fin width/height), challenges grew: fin quantization (drive current only in discrete multiples of fins), increasing parasitic capacitance, process variability, and residual short-channel effects at the fin base (where the gate does not fully wrap).

GAA solves these by taking gate wrapping to its logical extreme.

Core Structure and How GAA Works

In a GAA transistor, the gate completely surrounds the channel on all four sides — top, bottom, and both sidewalls — providing 360° electrostatic control.

The channel itself is no longer a single vertical fin. Instead, modern commercial GAA implementations use one of two main forms:

1. Nanosheet (or nanoribbon/sheet-based) GAA — The dominant approach in high-volume production.
   • The channel consists of multiple thin, flat, horizontal silicon sheets (nanosheets) stacked vertically.
   • These sheets are typically 5–10 nm thick and spaced apart.
   • The gate material (high-k dielectric + metal gate) wraps completely around each individual nanosheet.
   • Source and drain regions connect to the ends of all nanosheets in the stack.
2. Nanowire GAA — An earlier/research-focused variant.
   • The channel uses cylindrical or rod-like nanowires (horizontal, suspended between source/drain).
   • Gate wraps around each nanowire.
   • Nanowires provide excellent control but lower drive current per structure compared to wider nanosheets; thus, nanosheets became preferred for balancing performance and manufacturability.

Key physical implementation steps (simplified):

• Alternating layers of silicon (for channels) and sacrificial material (e.g., SiGe) are epitaxially grown.
• The sacrificial layers are selectively etched away, leaving suspended silicon nanosheets or nanowires.
• High-k dielectric and metal gate are deposited to fully encircle each suspended channel.
• Inner spacers, source/drain epi, and contacts complete the device.

This results in a vertically stacked, horizontally oriented channel structure within the same footprint as older designs.

Major Advantages Over FinFET

GAA delivers several critical improvements that enable continued scaling:

• Superior electrostatic control — Full 360° gate wrap eliminates weak spots (especially at the fin base in FinFETs), strongly suppresses short-channel effects, reduces subthreshold swing degradation, and dramatically lowers leakage (both subthreshold and gate-induced drain leakage).
• Significantly reduced leakage current — Leads to much lower static power consumption, crucial for mobile, AI, and always-on devices.
• Higher drive current (Ion) at the same leakage level — Nanosheets provide a larger effective channel width without relying on multiple discrete fins. Stacking 3–5 nanosheets vertically increases current density in the same area.
• Flexible drive current tuning — Unlike FinFETs (where current scales in coarse steps of 1–2–3 fins), GAA allows fine-grained adjustment by varying nanosheet width (horizontal dimension) within the same cell library. This enables precise optimization for power, performance, or area (PPA) in different parts of a chip (e.g., high-performance logic vs. low-power memory periphery).
• Better area scaling and density potential — The vertical stacking of channels allows more drive strength in a smaller footprint, supporting higher transistor density. Future extensions (e.g., complementary FET or CFET, stacking NMOS over PMOS) build on this.
• Improved subthreshold slope and DIBL (drain-induced barrier lowering) — Translates to sharper on/off transitions and better voltage scaling.

Reported typical gains (vs. equivalent FinFET at similar node) include ~20–40% better power efficiency, ~10–30% performance uplift at iso-power, or significant area reduction, depending on the specific foundry optimization and comparison point.

Commercial Implementations (as of March 2026)

Different foundries use proprietary names but share the same core GAA principles:

• Samsung → MBCFET (Multi-Bridge-Channel FET) — Nanosheet-based; first introduced in production at 3nm (2022), refined in later generations.
• TSMC → Nanosheet GAA (often just called “nanosheet”) — Adopted starting at N2 (2nm-class) node, following FinFET at N3.
• Intel → RibbonFET — Nanosheet/ribbon-based; introduced at Intel 20A/18A nodes.

All major players converged on horizontal stacked nanosheets (rather than nanowires) for production due to better manufacturability and performance balance.

Challenges and Trade-offs

While revolutionary for scaling, GAA introduces complexities:

• Much more intricate fabrication (selective etching of sacrificial layers, uniform gate deposition around suspended structures, inner spacer formation).
• Higher process variability risks in early generations.
• Increased parasitic capacitances if not carefully engineered.
• Yield and cost challenges during initial ramp-up.

Despite these, GAA has become the standard beyond ~3nm FinFET limits, enabling the industry to continue Moore’s Law-like density and efficiency scaling into the angstrom era (sub-2 nm).

In summary, GAA transistor design achieves unprecedented gate control by fully encircling nanosheet or nanowire channels, overcoming FinFET’s remaining limitations in leakage, variability, and scaling headroom. This architecture underpins the most advanced logic chips today and will remain central for the foreseeable future, with ongoing refinements pushing toward even denser stacked variants like CFET.

---

Samsung’s 3nm: Gate-All-Around (GAA) Transistors

Samsung’s 3nm process nodes (primarily referred to as SF3E for the first generation and SF3 for the second/enhanced generation) employ a Gate-All-Around (GAA) transistor architecture under Samsung’s proprietary branding: MBCFET™ (Multi-Bridge-Channel FET). This marks a pivotal shift from the FinFET transistors used in prior nodes (such as Samsung’s 5nm and 4nm) and in competitors’ initial 3nm offerings (like TSMC’s N3 family). MBCFET is Samsung’s specific implementation of GAA, optimized for high-volume production starting in 2022, making Samsung the first foundry to commercialize GAA at the 3nm class.

Fundamental Principle of MBCFET in Samsung’s 3nm

In MBCFET, the transistor channel consists of multiple thin, flat, horizontally oriented nanosheets (also called nanoribbons or nano-sheets in some contexts) stacked vertically. The gate electrode completely wraps around all four sides of each individual nanosheet, providing full 360° electrostatic control over the channel. This contrasts with FinFET, where the gate covers only three sides (top and two sidewalls) of a vertical fin, leaving the bottom connected to the substrate and susceptible to some residual leakage and control issues at ultra-scaled dimensions.

The core advantage stems from this full encirclement: it suppresses short-channel effects far more effectively than FinFET, allowing lower supply voltages, reduced subthreshold leakage, sharper on/off transitions (better subthreshold slope), and minimized drain-induced barrier lowering (DIBL). The nanosheets are typically 5–10 nm thick, with spacing between them, and the gate dielectric (high-k material) and metal gate fill the gaps to encircle each sheet uniformly.

Key Structural Features Specific to Samsung’s MBCFET

Samsung deliberately chose a nanosheet-based (sheet/ribbon) design over a nanowire (cylindrical rod) approach for several reasons:

• Wider effective channel area — Nanosheets provide a larger perimeter and surface area for current flow per stack compared to narrower nanowires. This boosts drive current (Ion) without needing excessive stacking or area penalties.
• Adjustable channel width — A standout feature of MBCFET is the ability to vary the horizontal width of the nanosheets within the same standard cell library or even the same design block. Designers can tune this width independently for different transistors (e.g., wider for high-drive logic paths, narrower for low-power sections). This offers granular control over drive strength, power, and performance — far more flexible than FinFET, where current scales only in discrete multiples of fins (e.g., 1-fin, 2-fin, 3-fin devices).
• Stacked configuration — Typically 3–5 nanosheets are stacked vertically (exact number can vary by optimization and generation). Stacking increases total effective channel width in a compact footprint, supporting higher transistor density.
• Orientation — The nanosheets lie horizontal (parallel to the substrate), with source/drain regions epi-grown at the ends to connect all sheets in parallel electrically. Inner spacers (dielectric barriers) separate the gate from source/drain to reduce parasitics.

This structure evolved from Samsung’s early GAA research (starting in the early 2000s, with serious development from ~2017). Early demos (e.g., 256Mb SRAM test chips around 2021) validated functional operation at low voltages.

Performance, Power, and Area (PPA) Benefits on Samsung’s 3nm

Samsung positions MBCFET as delivering substantial gains over its prior 5nm FinFET node (comparisons are often vs. 5nm or 4nm):

• First-generation (SF3E / 3GAE) — Up to 45% lower power, 23% higher performance, and 16% smaller area vs. 5nm.
• Second-generation (SF3) — Further refined with additional process optimizations and variable nanosheet widths, achieving ~22% speed improvement, ~34% power reduction, and ~0.79× logic area vs. the prior 4nm FinFET platform.

These gains arise from:

• Better electrostatic integrity → lower leakage and ability to scale voltage down.
• Higher effective channel width per footprint → more current drive.
• Design flexibility from tunable widths → optimized PPA across logic, SRAM, analog, and mixed-signal blocks (especially beneficial for high-density SRAM cells, where balancing pull-up/pull-down/pass-gate strengths improves stability margins without area penalties).

Visual Illustrations of the Architecture

Here are selected diagrams illustrating the evolution and specifics of MBCFET compared to prior generations:

semiconductor.samsung.com

3nm GAA MBCFET™: Unrivaled SRAM Design Flexibility | Samsung Semiconductor Global

This shows the progression: Planar → FinFET → MBCFET (stacked nanosheets with full gate wrap).

semiconductor.samsung.com

GAA MBCFET™ PPA optimization through DTCO | Samsung Semiconductor Global

This compares nanowire GAA (narrower channels) vs. nanosheet (wider, stacked sheets) — Samsung’s preferred MBCFET approach.

sammobile.com

Samsung Foundry starts production of 3nm semiconductor chips with GAA architecture – SamMobile

A cross-sectional-style view highlighting the nanosheet stacking and gate encirclement in MBCFET.

cdrinfo.com

Samsung Electronics Announced MBCFET Transistor Structure For Advanced Foundry Nodes | CdrInfo.com

This illustrates the evolutionary path and performance/power benefits curve, peaking at nanosheet-based GAA (MBCFET).

semiconductor.samsung.com

3nm GAA MBCFET™: Unrivaled SRAM Design Flexibility | Samsung Semiconductor Global

This depicts SRAM bitcell flexibility enabled by variable nanosheet widths in MBCFET designs.

Manufacturing and Implementation Notes

The fabrication sequence involves epitaxial growth of alternating silicon (channel) and sacrificial (e.g., SiGe) layers, selective removal of the sacrificial material to suspend nanosheets, deposition of inner spacers, high-k/metal gate wrapping each sheet, and source/drain formation. This is more complex than FinFET but enables the superior control needed at 3nm-class dimensions.

As of March 2026, Samsung’s 3nm GAA (MBCFET) remains the core architecture for its SF3 family, though yield maturation has been challenging compared to FinFET nodes. The technology’s design flexibility — especially variable nanosheet widths — continues to be highlighted as a key differentiator for custom and high-performance applications.

In essence, Samsung’s 3nm core architecture via MBCFET realizes the full potential of GAA by combining 360° gate control, stacked wide nanosheets, and tunable dimensions to push beyond FinFET scaling limits in power efficiency, performance, and density. This positions it as a foundational step toward even denser nodes like 2nm.

---

Samsung’s 3nm process family

Samsung’s 3nm process family, built around its proprietary MBCFET (Multi-Bridge-Channel FET) Gate-All-Around (GAA) transistor architecture, consists of multiple variants developed iteratively to refine performance, power efficiency, yield, and manufacturability. As of March, 2026, the primary variants in Samsung Foundry’s 3nm lineup are the first-generation and second-generation nodes, with no widely confirmed third-generation 3nm variant in production (third-generation GAA advancements have shifted toward the 2nm family).

Samsung Foundry uses the “SF” prefix for its logic nodes (SF = Samsung Foundry), with specific designations for 3nm-class processes. The family has evolved amid challenges like yield maturation, limited external customer adoption, and some roadmap adjustments (including past reports of rebranding discussions around the second-generation node, though official Samsung materials and recent analyses retain the 3nm-class labeling for these).

1. First-Generation: SF3E (also known as 3GAE or 3nm Early)

• Introduction and Status — This is Samsung’s inaugural commercial 3nm process, marking the world’s first high-volume production of GAA-based transistors. Risk production began around early 2022, with mass production/shipping starting in June 2022. As of 2026, it remains in use, though primarily for internal Samsung applications and select simpler designs due to yield and ecosystem factors.
• Key Characteristics — Uses MBCFET with stacked horizontal nanosheets (typically 3–4 in early implementations). The gate fully encircles each nanosheet for superior electrostatic control compared to FinFET. Nanosheet widths are adjustable, but design rules and flexibility are more conservative than later generations.
• PPA Metrics (as claimed by Samsung vs. its own 5nm FinFET baseline):
  • Power consumption reduction: Up to 45%.
  • Performance improvement: Around 23%.
  • Area reduction: Approximately 16%.
• Target Applications — Initially positioned for high-performance, low-power computing (e.g., mobile SoCs, wearables). In practice, it has seen limited external foundry tape-outs, with production mostly internal (e.g., certain Exynos variants or test chips).
• Challenges — Early yields were lower than industry targets (reports of 50–60% in mature phases), contributing to customer preference for TSMC’s FinFET-based 3nm nodes. This variant represents the baseline GAA introduction but with more conservative optimizations.

2. Second-Generation: SF3 (also referred to as 3GAP or enhanced 3nm)

• Introduction and Status — This is the refined, enhanced iteration of Samsung’s 3nm GAA technology. Development focused on addressing first-gen limitations through process tweaks, better yield targeting, and architectural refinements. Trial production began around 2023–2024, with mass production ramp targeted for 2024–2025 (some reports indicate full-scale production in late 2024 or 2025). By March 2026, it is in active use, including in products like the Exynos W1000 wearable processor (confirmed via die analysis).
• Key Improvements Over SF3E — Builds on the same MBCFET nanosheet foundation but introduces significant enhancements:
  • Greater flexibility in nanosheet channel widths — Variable widths can be used within the same standard cell or design block, allowing precise tuning of drive current for different circuit types (e.g., high-performance paths vs. low-leakage sections). This is a major differentiator for custom logic and mixed-signal designs.
  • Process optimizations (e.g., improved gate stacking, reduced parasitics, better uniformity in nanosheet suspension/etching).
  • Refined high-k/metal gate deposition and inner spacer engineering for lower variability.
• PPA Metrics (as claimed by Samsung, often vs. its 4nm FinFET baseline like SF4):
  • Performance uplift: Up to 22% higher at iso-power.
  • Power reduction: Around 34%.
  • Area scaling: Approximately 0.79× logic area (or ~21% reduction).
  • Overall vs. first-gen 3nm: Incremental gains in efficiency and density, with better suitability for complex SoCs.
• Target Applications — Aimed at broader adoption, including mobile application processors (e.g., potential Exynos for Galaxy devices), wearables, and some HPC/AI edge cases. It has seen more internal Samsung System LSI usage than the first-gen, though external foundry traction remains modest compared to competitors.
• Challenges — Yield maturation has been slower than hoped (reports from 2024–2025 indicated figures in the 20–50% range during ramp-up, with goals of >60%). By 2026, improvements have been made, but the node has not captured major third-party clients like Apple’s or NVIDIA’s leading-edge designs.

Other Notes on the 3nm Family

• No Confirmed Third-Generation 3nm Variant — Samsung’s official roadmap and recent reports do not detail a distinct “SF3P” or third-gen 3nm node in widespread production as of March 2026. Some earlier mentions of “SF3P” (or 3GAP+) referred to potential further tweaks or specialized versions of SF3, but these appear folded into the second-gen refinements or redirected toward 2nm-class nodes (e.g., SF2 family, where GAA generations advance further with backside power delivery in variants like SF2Z).
• Roadmap Context — Samsung’s overall advanced-node strategy has emphasized iterative GAA improvements. Lessons from the 3nm family (especially yield and process control challenges) have directly informed the 2nm nodes (SF2, SF2P, etc.), which are positioned as offering even greater PPA gains (e.g., 5–8% better efficiency/performance over mature 3nm in some claims). The 3nm family thus serves as the foundational GAA platform, with two clear generations: SF3E (introductory) and SF3 (enhanced/optimized).
• Overall Positioning — Both variants deliver the core advantages of MBCFET GAA — full gate encirclement of nanosheets for reduced leakage, higher drive current, and design flexibility — but SF3 represents a more mature, production-ready evolution. Adoption has remained largely internal due to competitive dynamics (e.g., TSMC’s higher-yielding FinFET 3nm and upcoming nanosheet 2nm), but the technology demonstrates Samsung’s early leadership in commercializing GAA at scale.

In summary, Samsung’s 3nm family comprises two main variants: the pioneering SF3E (first-gen, 2022 mass production, conservative GAA intro) and the improved SF3 (second-gen, 2024–2025 ramp, with enhanced nanosheet tunability and PPA). These form the backbone of Samsung’s transition to GAA transistors, paving the way for subsequent angstrom-era nodes.

---

GAA Advantages Over FinFET

Gate-All-Around (GAA) transistors, including Samsung’s MBCFET implementation at 3nm and beyond, represent a significant architectural advancement over FinFET transistors, which dominated from roughly the 16/14nm node through initial 3nm offerings (e.g., TSMC’s N3 family). The core difference lies in gate-channel interaction: FinFET gates wrap around three sides of a vertical fin-shaped channel (top and two sidewalls), while GAA gates fully encircle the channel on all four sides — top, bottom, and both sidewalls. This 360° encirclement, typically achieved with stacked horizontal nanosheets (or nanoribbons), provides fundamentally superior electrostatic control.

As of March 2026, GAA has become the standard for leading-edge nodes (Samsung’s 3nm and 2nm families, TSMC’s 2nm-class N2, Intel’s 18A/20A RibbonFET), enabling continued scaling where FinFET reaches physical and electrical limits around 3–5nm. Below is a detailed breakdown of the main advantages of GAA over FinFET, grounded in transistor physics, foundry claims, and industry analyses.

1. Superior Electrostatic Control and Suppression of Short-Channel Effects

• Mechanism — In ultra-scaled transistors (sub-5nm gate lengths), short-channel effects (SCE) degrade performance: the drain voltage increasingly influences the source-channel barrier, causing threshold voltage (Vt) roll-off, increased subthreshold swing (SS), and higher off-state leakage. FinFET mitigates this by wrapping the gate on three sides, but the channel bottom remains connected to the substrate, creating a leakage path and weaker control at the fin base.
• GAA Advantage — Full gate encirclement maximizes gate influence over the entire channel volume, dramatically reducing SCE metrics like drain-induced barrier lowering (DIBL, often ~60–80 mV/V in FinFET vs. significantly lower in GAA) and subthreshold swing (closer to the theoretical 60 mV/decade limit, with 20–25% better SS in GAA nanosheet devices).
• Outcome — Sharper on/off transitions, better low-voltage operation, and reduced variability in threshold voltage. This is critical for maintaining performance at lower supply voltages (Vdd), especially in power-constrained mobile and AI edge devices.

2. Significantly Reduced Leakage Current (Static Power)

• Mechanism — Leakage includes subthreshold leakage (when off), gate-induced drain leakage (GIDL), and junction leakage. FinFET leakage worsens at smaller nodes due to incomplete gate control and fin quantization effects.
• GAA Advantage — Enhanced gate control suppresses leakage paths, with reports of up to 40–75% lower leakage at equivalent voltages (e.g., TSMC N2 GAA shows ~75% reduction at 0.5V). Samsung’s MBCFET at 3nm achieves substantial static power savings through better off-state control.
• Outcome — Lower static power consumption (critical for always-on devices, battery life in mobiles, and thermal management in high-density AI chips). Combined with dynamic power reductions from lower Vdd, overall power efficiency improves markedly (e.g., 25–50% lower total power at iso-performance in various foundry nodes).

3. Higher Drive Current (Ion) and Performance at Iso-Power or Iso-Area

• Mechanism — Drive current determines switching speed and performance. FinFET drive scales coarsely by adding side-by-side fins (1-fin, 2-fin, etc.), incurring area penalties and parasitics.
• GAA Advantage — Nanosheets stack vertically (typically 3–5 layers) with gate wrapping each sheet, providing larger effective channel width (Weff) per footprint. Stacking boosts Ion without lateral area increase. Additionally, nanosheet width (horizontal dimension) is continuously tunable, unlike FinFET’s discrete fin multiples.
• Outcome — Higher performance at the same power (e.g., Samsung claims 22–30% uplift in SF3 vs. prior 4nm/5nm FinFET; TSMC N2 ~10–20% faster at iso-power). Or equivalent performance with 20–40% lower power. This flexibility enables fine-grained optimization across logic paths.

4. Greater Design Flexibility and PPA Optimization

• Mechanism — FinFET limits designers to discrete drive strengths (multiples of fin count), complicating balancing pull-up/pull-down ratios in SRAM or analog circuits.
• GAA Advantage — Variable nanosheet widths allow independent tuning of NMOS/PMOS drive currents within the same cell library or block. This is especially powerful for SRAM (e.g., better stability margins at lower voltages, hundreds of mV write-voltage reduction in Samsung demos) and mixed-signal/analog blocks.
• Outcome — Superior power-performance-area (PPA) trade-offs. Samsung emphasizes MBCFET’s “unrivaled SRAM design flexibility” and ability to optimize entire SoCs (e.g., wider channels for high-drive sections, narrower for low-leakage). Overall PPA gains include ~16–35% area reduction, 23–50% power savings, and 23–30% performance uplift vs. prior FinFET nodes (per Samsung/others).

5. Better Scalability and Path to Future Nodes

• Mechanism — FinFET scaling stalls due to increasing variability, parasitics, and leakage as fins narrow/tall.
• GAA Advantage — Nanosheet stacking and full-wrap gate enable density scaling beyond 3nm without excessive variability. It paves the way for extensions like backside power delivery (BSPDN in Intel 18A, TSMC variants) and complementary FET (CFET, stacking NMOS/PMOS vertically).
• Outcome — Sustains Moore’s Law-like density/efficiency gains into the angstrom era (sub-2nm). Industry consensus: GAA is “mandatory” at 2nm-class nodes, retiring FinFET for leading-edge logic.

Summary Comparison Table

Aspect	FinFET (3-sided gate)	GAA (4-sided gate, nanosheet)	Typical Improvement (GAA vs FinFET)
Gate Control	Good (3 sides)	Excellent (360° encirclement)	Superior SCE suppression
Leakage Current	Moderate at sub-5nm	Significantly lower	40–75% reduction
Drive Current Tuning	Discrete (fin multiples)	Continuous (variable nanosheet width)	Greater flexibility
Effective Channel Width	Limited by fin count/area penalty	Higher via vertical stacking	Better Ion per footprint
Power Efficiency	High vs planar, but limits at scale	Even higher (lower Vdd, less leakage)	25–50% better at iso-performance
Area Scaling Potential	Good to ~3nm	Extends beyond 3nm	16–35% reduction
SRAM/Analog Flexibility	Constrained by quantization	Superior (tunable ratios)	Better margins, lower Vwrite

In practice, realized gains depend on process maturity (e.g., early Samsung 3nm GAA had yield challenges, while TSMC’s FinFET 3nm excelled in volume). However, the fundamental physics of GAA — full gate control, reduced leakage, tunable drive, and stacking — make it decisively superior for sub-3nm scaling, power-constrained applications (mobile, AI accelerators), and future density needs. This is why all major foundries adopted GAA by 2025–2026, marking the end of the FinFET era for cutting-edge logic.

---