TSMC vs Samsung Foundry: Chip Manufacturing Compared

TSMC and Samsung Foundry are the two leading contract chip manufacturers, and their manufacturing options differ mainly in process node naming, transistor architecture, yield maturity, and customer support. This article compares their offerings—from 5nm and 3nm to upcoming 2nm—so you can understand which foundry might be a better fit for different chip designs.

Background: Who They Are

TSMC (Taiwan Semiconductor Manufacturing Company) is the world’s largest dedicated semiconductor foundry, producing chips for Apple, AMD, NVIDIA, Qualcomm, and many others. Samsung Foundry is a division of Samsung Electronics, making chips for its own devices as well as for external customers like Qualcomm and IBM. Both companies have invested heavily in advanced nodes, but they have different approaches to process technology and business strategy.

Process Node Naming and Reality

Neither company’s node names (e.g., 5nm, 3nm) reflect actual physical dimensions anymore; they are marketing labels. TSMC’s 5nm (N5) and Samsung’s 5nm (5LPE) are different in transistor density and performance. TSMC has historically offered higher density and better power efficiency at the same node name. Samsung’s 5nm is less dense than TSMC’s 5nm, but Samsung’s 4nm (4LPE) is closer to TSMC’s 5nm in density. This naming gap can confuse buyers, so it’s essential to look at real specs like transistor density (MTr/mm²) and performance per watt.

Key Differences in Manufacturing Options

The main differences between TSMC and Samsung foundry options are:

• Transistor architecture: TSMC uses FinFET for its 5nm and 4nm nodes, while Samsung uses Gate-All-Around (GAA) starting at 3nm (3GAE). GAA offers better electrostatic control and lower leakage, but it’s newer and less proven.
• Yield and maturity: TSMC’s N5 has been in high-volume production since 2020, with excellent yields. Samsung’s 5nm and 4nm have faced yield challenges, especially for high-performance chips like the Snapdragon 8 Gen 1.
• Design ecosystem: TSMC has a vast ecosystem of IP and design tools, making it easier for customers to design chips. Samsung offers a similar ecosystem but with fewer third-party IP options.
• Customer support: TSMC is known for its “shuttleâ€� services and close collaboration with customers. Samsung has improved its foundry services but still lags in customer trust.

Advanced Nodes: 5nm, 4nm, 3nm, 2nm

TSMC’s N5 (5nm) is used in Apple’s A14, M1, and many other chips. Its N4 (4nm) is an enhanced version of N5 with minor improvements. TSMC’s N3 (3nm) started production in late 2022, using FinFET, and is used in Apple’s A17 Pro and M3. Samsung’s 5nm (5LPE) is used in some Exynos chips, and its 4nm (4LPE) is used in Snapdragon 8 Gen 1 and Exynos 2200. Samsung’s 3nm (3GAE) is the first GAA node, but it’s only used in a few low-volume products like Samsung’s own wearable chip. TSMC’s N2 (2nm) is scheduled for 2025, using GAA, while Samsung’s 2nm is expected around 2025-2026.

How to Choose Between Them

Choosing between TSMC and Samsung Foundry depends on your priorities:

• Performance and power: If you need the best power efficiency and performance, TSMC’s mature nodes are generally safer. For example, the Snapdragon 8+ Gen 1 (TSMC) was more efficient than the Snapdragon 8 Gen 1 (Samsung).
• Cost: Samsung often offers lower prices for wafers, but lower yields can increase effective cost per good die.
• Supply chain security: Some companies prefer Samsung to diversify away from Taiwan’s geopolitical risks.
• Design complexity: If your design uses advanced packaging like CoWoS, TSMC has more experience. Samsung offers similar 3D packaging but with fewer proven cases.

Common Problems and Misconceptions

A common misconception is that a smaller node number always means a better chip. In reality, the same node name from different foundries can have very different performance. For example, Samsung’s 4nm is comparable to TSMC’s 5nm in density, but TSMC’s 4nm is denser than Samsung’s 4nm. Another issue is yield: Samsung’s 3nm GAA has faced low yields, making it expensive for large chips. TSMC’s 3nm FinFET has better yields but still costs more than 5nm. Also, some people think that Samsung’s GAA is inherently better than FinFET, but GAA only shows its advantages at very small nodes; at 3nm, FinFET is still competitive.

Cost and Pricing Considerations

Exact wafer prices are confidential, but industry reports suggest that TSMC’s 5nm wafers cost around $10,000-$15,000, while Samsung’s 5nm is slightly cheaper. At 3nm, TSMC’s wafers are estimated to cost over $20,000. Samsung’s 3nm is likely cheaper per wafer, but low yields can make the cost per good die higher. For low-volume designs, the mask costs (non-recurring engineering) are significant—millions of dollars—so the choice of foundry also affects upfront costs.

Future Outlook

TSMC is set to mass-produce N2 (2nm) with GAA in 2025, and it has already secured major customers like Apple. Samsung is also developing 2nm with GAA, but its timeline is less clear. Both companies are investing in advanced packaging, which is becoming as important as the transistors themselves. TSMC’s CoWoS and InFO are widely used, while Samsung’s I-Cube and X-Cube are emerging. The competition between the two is intense, and each has strengths: TSMC in reliability and ecosystem, Samsung in innovation and vertical integration.

Frequently Asked Questions

Is Samsung’s 3nm better than TSMC’s 3nm?

Not necessarily. Samsung’s 3nm uses GAA transistors, which are theoretically superior, but TSMC’s 3nm FinFET has better yields and is already in mass production for Apple. Samsung’s 3nm is only used in a few products, so it’s hard to compare directly. In practice, TSMC’s 3nm is more mature and reliable.

Why did Qualcomm switch from Samsung to TSMC for the Snapdragon 8+ Gen 1?

Qualcomm switched because Samsung’s 4nm process had lower yields and higher power consumption, leading to thermal issues in the Snapdragon 8 Gen 1. TSMC’s 4nm offered better efficiency and yield, so the 8+ Gen 1 was made on TSMC. This is a well-known example of foundry choice affecting real-world performance.

Can I design a chip for both TSMC and Samsung?

Yes, but it requires a dual-track design that meets both processes’ design rules. This is costly and time-consuming, so most companies choose one foundry per product. Some companies like Qualcomm use both for different products, but not for the same die.

Which foundry is cheaper for small batches?

Samsung is often cheaper for small batches because of lower mask costs and more flexible pricing. However, TSMC offers better support for small customers through its “shuttle� programs. The total cost depends on yield and volume, so it’s best to get quotes from both.

What is the difference between FinFET and GAA?

FinFET uses a fin-shaped channel that is wrapped by the gate on three sides. GAA (Gate-All-Around) uses nanosheets that the gate wraps around completely, providing better control over the channel and reducing leakage. GAA is more difficult to manufacture but allows for further scaling.

Related

• Balcony Solar vs Portable Power Stations for Apartments
• ARK Survival Ascended Leveling Guide: Fast XP
• Replace Range Names in Formulas: How To