AMD Puts an Anti-Tamper Shield on the Back of Its Chips

AMD has announced Versal tamper-resistant (TR) devices. They protect sensitive information against sophisticated physical attacks. They also keep the processing power of the AMD Versal AI Edge Series Gen 2.

AMD says these are the first adaptive SoC or FPGA-based devices with built-in protection against physical attacks on the silicon backside. The claim rests on AMD’s own analysis from November 2026 of competing devices.

How the protection works

A sensor-based protection die is bonded to the main compute die using 3D active-on-active technology. It covers the whole backside of the device. It protects sensitive information and detects physical attacks. It also stops the device from booting if someone tampers with it or removes the protection die.

AMD says customers can strengthen the protection of security-critical information without giving up performance, size, weight, power or time to market.

The problem it targets

Physical attacks on sensitive data are a priority across many markets. They matter in data centre networking, fintech, defence systems and critical industrial infrastructure. AMD says customers who want stronger device security today usually build custom solutions in-house. These use mechanical enclosures, ASICs and other parts. That adds development cost, specialist staff, verification work and the risk of a chip re-spin.

Versal TR devices combine processing and security features in one device. AMD says this reduces the need for separate compute, security and enclosure-based protection parts. It can save board space and simplify board design compared with custom builds. Customers also get AMD tools, technical support and a broad partner ecosystem.

Kirk Saban is corporate vice president of Product, Software and Solutions at AMD Embedded. AMD attributed a statement on the launch to him.

More security features

Beyond backside protection, the devices add expanded hardware acceleration for post-quantum cryptography. This includes new support for the Module-Lattice-Based Digital Signature Algorithm (ML-DSA) and the Stateless Hash-Based Digital Signature Algorithm (SLH-DSA).

Improvements to the application security unit key vault, plus a new memory element aging sensor, add further safeguards for cryptographic keys. The devices also widen glitch detection. They combine an improved glitch detector with a new orthogonal glitch detector. This gives better coverage, including glitches that hit two voltage rails at once.

Performance and the device range

Versal TR devices build on the Versal AI Edge Series Gen 2 and use Arm CPUs, AI Engines, programmable logic, hard IP and connectivity. They offer up to 229,000 DMIPS of scalar compute, up to 369 INT8 TOPS and up to 543,000 LUTs.

The portfolio has three devices: 2VE3358TR, 2VE3558TR and 2VE3858TR. They give a common platform across different performance and cost needs. Each has a footprint-compatible standard device available today. Customers can design on the standard Versal platform and later move to the TR device with little effort. They do not need to rerun synthesis or place and route.

Availability

The first Versal TR devices should sample in the second half of 2027. They should reach production in the first half of 2028. Customers can evaluate and build designs now on select footprint-equivalent standard Versal AI Edge Series Gen 2 devices.

AMD notes that tamper-resistant does not mean tamper-proof. Protection applies to specified physical attack methods. The product documentation sets out the threat model and limits.

Latest articles

Related articles