Accelerate high-speed secure connection on your KR260 via TLS10GC-IP

As edge computing becomes more critical across industries—from smart factories to autonomous systems—ensuring high-speed secure communication without sacrificing performance is no longer optional.

At Design Gateway, we’re excited to introduce our latest hardware-accelerated TLS solution: ✅ TLS10GC-IP, working alongside TOE10GLL-IP and LL10GEMAC-IP, delivers ultra-fast 10GbE secure communication directly on the AMD Kria KR260 platform.

Why It Matters

Edge devices typically rely on CPU-based encryption, such as OpenSSL, which can severely bottleneck data throughput, especially during TLS handshake and encrypted transfers. In our new demo, we show how you can offload both the TCP/IP stack and TLS tasks to dedicated hardware.

This demo showcases secure communication between the Lynx web browser on KR260 and a web server. The setup includes: A KR260 board, A test PC running an example Node.js server, A switch and gateway connected together via 1G and/or 10G Ethernet connections. — Overview system

Key Highlights from the Demo

🔹 Platform: AMD Kria KR260
🔹 Browser: Lynx (Text-based, lightweight)
🔹 Encryption Engine: OpenSSL modified to interface with hardware accelerator
🔹 Connection Options: Toggle between 1GbE and 10GbE, with or without hardware acceleration
🔹 Performance: Up to 2x speed improvement while reducing CPU usage

💡 This means more CPU resources are freed for other critical tasks, making your edge device truly efficient.

Performance speaks for itself:

1GbE + Software TLS: 0.7 Gbps (100% CPU usage)
10GbE + TLS10GC-IP: 1.4 Gbps (same CPU usage, double the speed!) This proves that hardware acceleration is the key to unlocking next-gen edge security without compromising performance.

To compare download speeds, Lynx can connect to an example Node.js server and download a data pattern.
With 1G Ethernet, the download speed averages 0.7 Gbps while utilizing 100% of one CPU core and 60% of another.
In contrast, with 10G Ethernet, the download speed averages 1.4 Gbps with 100% usage of a CPU core.
These results highlight that Lynx’s performance is CPU-bound.
By offloading the TCP/IP stack to TOE10GLL-IP and TLS tasks to TLS10GC-IP, the CPU can focus on other tasks, boosting transfer speeds without altering the browsing experience. — Connecting with an example server

3-step integration process:

1️⃣ Minimal modification of Lynx and OpenSSL in user space
2️⃣ Seamless communication via /dev/mem and /dev/udmabuf in kernel space
3️⃣ Load accelerator bitstream via provided tools

🎥 Watch the full video demo here