Tasks at the embedded edge are getting more and more complex. Supporting up to 8 cores and 16 threads, the recently-launched AMD Ryzen Embedded V2000 processors are paving the way for x86-based embedded designs with unprecedented compute density and performance per watt. congatec is offering the new Zen 2 based processor generation, which complements V1000 processor technology with 6- and 8-core versions, on COM Express Compact modules.
As the trend towards more IIoT connectivity and digitization continues, embedded systems must handle a growing number of tasks. Additional tasks include the collection of application data that needs to be transcoded, and sometimes analyzed locally using artificial intelligence (AI). In parallel, data must be exchanged with customers’ enterprise clouds and other OEM apps, and this communication must be highly secure, because new pay-per-use models turn the devices, machines, and systems into an OEM revenue source that needs to be safeguarded. Ideally, embedded systems are therefore constantly monitored to preempt malicious attacks. All this requires many processes to run in parallel, which is why every additional core in an embedded system is very welcome, since it can then also be assigned dedicated tasks using real-time capable hypervisor technology.
As an add-on to V1000 processor technology, the new AMD Ryzen Embedded V2000 processors set completely new standards, making it possible to deliver COM Express Type 6 modules with 8-core processors and 16 threads for the very first time. What is more, the TDP scales up to 54 watts, and down to as little as 10 watts. This is crucial for many fanless embedded systems and subsequently creates both an immense performance leap and significantly increased multi-tasking capacities. With the new AMD Embedded Ryzen V2000 series, users benefit from double the processing power per watt, approximately 15% more instructions per clock (IPC), and twice as many cores compared to the previous generation.
These significant performance gains over the AMD Ryzen Embedded V1000 processors are made possible by the new Zen 2 processor architecture, which is now manufactured in 7 nm technology. AMD has retained many innovations of the first Zen microarchitecture: for instance, the CPU Core Complex (CCX) combining up to 4 cores. Within a CCX, each core can access the shared L2 and L3 cache with the same low latency. However, compared to the previous V1000 series, AMD has doubled the memory from 2 MB to 4 MB L2 cache and 8 MB per CCX for the L3 cache. While the L1 cache has not increased in size, bandwidth was doubled from 16 to 32 Kbytes throughput per clock cycle. Floating-point bandwidth, too, was doubled from 128 to 256 bits, which significantly speeds up the processing of AVX2 instructions, for instance.
One effect of this is a 15% increase in the IPC rate. Combined with the smaller manufacturing technology, which has shrunk from 14 to 7 nm, these modifications have resulted in twice as much multi-thread performance per watt and up to 30% higher single-thread performance.
AMD has used renowned benchmarks to prove the performance of its AMD Ryzen Embedded V2000 processors. The AMD Ryzen V2718 processor with a nominal TDP of 15 watts, 8 cores, and a base clock rate of 1.7 GHz was compared with the 6-core Intel Core i7 10710U processor with a base clock rate of 1.6 GHz and the quad-core Intel Core i7 10510U processor with a nominal clock rate of 1.8 GHz. Both are 10th generation Intel Core processors and have a TDP of 15 watts. While the 6- and 4-core Intel Core processors are on par with the AMD Ryzen V2000 in single-thread performance, the picture shifts significantly when it comes to multi-core performance. Here, the AMD Embedded Ryzen V2000 processor is more than twice as fast as the quad-core i7 processor, and it also has a clear lead of around 33% over the hexa-core i7. The pendulum swings in favor of the AMD Ryzen V2000 processor in terms of graphics performance, too. Compared to both Intel Core i7 processors, it scores more than twice as high in the 3DMark Time Spy benchmark. Even when pitched against a processor with a much higher TDP, like the 6-core Intel Core i7 9750H processor with a TDP of 45 watts and 2.6 GHz base clock rate, the Intel Ryzen V2000 impresses with outstanding performance. It offers an almost identical single-thread performance, but about 45% more multi-thread and almost 40% more graphics performance.
Next to the pure computing power, the graphics performance is truly impressive. The AMD Radeon RX Vega graphics integrated into the powerful AMD Ryzen Embedded V2000 SoC provides up to 7 GPU processing units. The GPU performance of the integrated Radeon graphics, which is based on the 5th generation GCN architecture and traditionally high at AMD, has increased by 40% compared to the predecessor. It, too, benefits from the more energy-efficient 7 nm technology and at 1.6 GHz, its operating clock rate is also 300 MHz faster.
The AMD Embedded Ryzen V2000 family can therefore control four independent 4k displays in parallel – with a seamless 60 Hz refresh rate and lifelike high dynamic range (HDR) reproduction featuring 10-bit color depth per channel. Next to immersive digital signage and gaming applications, this specifically benefits high-quality medical diagnostic systems, where 10-bit color depth is required for the display of digital X-ray images. DirectX 12 and OpenGL 4.4 provide support for 3D graphics, and the integrated video engine enables hardware-accelerated streaming of 4k HEVC (H.265 and VP9) video at 60 Hz in both directions without straining the CPU. To ensure that graphics signals arrive at the display in full bandwidth, the Vega graphics supports the latest interface technologies with up to 4x DisplayPort 1.4, HDMI 2.1 as well as eDP 1.4. Thanks to HSA and OpenCL 2.0 support, deep learning workloads can also be assigned to the GPU.
AMD has optimized its interface offering quite significantly and expanded the I/Os to meet the requirements of high-performance edge systems and IPCs, which are in particularly high demand today. The AMD Ryzen V2000 processors support 20 PCIe Gen 3 lanes. That’s 8 more than the V1000 series, which even with the 4 additional PCIe lanes only offers a total of 16 lanes. What is more, the V2000 processors enable full USB-C implementation out of the box. That’s because 2 of the 4 USB 3.1 Gen 2 capable USB ports with 10 Gbit/s also support the USB modes Power Delivery and Alt-DP with DisplayPort 1.4 signals. In addition, the new SoCs support 4x USB 2.0. As before, up to 2x SATA ports are available for data storage. However, storage media are best connected via the fast NVMe interface, which with 4x PCIe offers significantly more bandwidth. The AMD Ryzen V2000 SoCs further supports a range of interfaces that are specifically adapted for the embedded market, such as legacy UART, I²C, SMBus, SPI, and GPIOs.
Next to improved performance and efficiency, AMD Ryzen Embedded V2000 series processors also provide advanced security features for greater protection against unauthorized access to stored data or critical software. As in the AMD Ryzen Embedded V1000 and R1000 series, AMD Memory Guard supports several security features in the AMD Ryzen Embedded V2000 processors, among them:
- Secure Memory Encryption for encrypting the main memory. It prevents physical attackers from accessing confidential data and helps defend against cold start attacks.
- Secure Boot protects the boot process so that no unauthorized software or malware can take over important system functions.
- UEFI Secure Boot prevents malicious code or unauthorized software from being loaded at system startup.
To ensure that performance is not impacted by the encrypted data encoding and decoding, AMD Ryzen Embedded V2000 processors support AES-NI for hardware-accelerated encryption and decryption of AES-encoded data. Zen 2 also provides AMD processors with hardware-based protection against the well-known Spectre and Spectre v4 exploits. Consequently, OEMs and users no longer have to rely on firmware or OS-based security mechanisms.
AMD has launched the new AMD Ryzen Embedded V2000 series in four processor variants, which are all supported by congatec’s new conga-TCV2 COM Express Type 6 Compact modules. They feature 4 MB L2 cache, 8 MB L3 cache, and up to 64 GB for fast and energy-efficient 64-bit Dual Channel DDR4 memory with up to 3200 MT/s and ECC support for maximum data security. The integrated AMD Radeon graphics with up to 7 processing units supports applications and use cases that require high-performance GPU data processing. The conga-TCV2 Computer-on-Modules therefore support up to four independent displays with a resolution of up to 4k60 UHD via 3x DisplayPort 1.4/HDMI 2.1 and 1x LVDS/eDP. Further performance-oriented interfaces include 1x PEG 3.0 x8 and 8x PCIe Gen 3 lanes, 2x USB 3.1 Gen 2, 8x USB 2.0, up to 2x SATA Gen 3, 1x Gbit Ethernet, 8 GPIOs, SPI, LPC, and 2x legacy UART provided by the board controller.
The RTS hypervisor as well as the Microsoft Windows 10, Linux/Yocto, Android Q, and Wind River VxWorks operating systems are further supported. Applications with high-security demands benefit from the integrated AMD Secure processor, which helps with hardware-accelerated encryption and decryption of RSA, SHA, and AES. Onboard TPM support is also provided.
Together, these impressive features, led by the currently highest compute core count and performance density for COM Express Type 6 modules, predestine the new Computer-on-Modules for digitization and parallel edge analytics – including workload balancing and consolidation enabled by virtual machines based on congatec’s RTS real-time hypervisor implementations. With up to 16 threads, the AMD Ryzen V2000 series based conga-TCV2 modules enable high-performance embedded system designs that can execute twice as many tasks within a given TDP range. This really is great news for edge computing as it helps address the rising number of parallel tasks. The integrated graphics, which continues to provide excellent 3D graphics quality on up to four independent 4k60 displays, is also impressive – as is the fact that all this is scalable from 54 watts TDP to ultra low power configurations consuming as little as 10 watts. However, this doesn’t mean that the AMD Ryzen Embedded V1000 processor series has become obsolete with the launch of the new processors; for lower performance ranges with fewer cores, the V1000 series is still the best choice.
Further use cases for the new modules include all standard embedded applications – from industrial box PCs and thin clients to embedded systems with impressive computing and graphics performance. The new modules also find an important market in smart robotics, e‑mobility, and autonomous vehicle applications that use deep learning to optimize situational awareness.
Further information on the new high-performance conga-TCV2 COM Express Type 6 Compact module is available at https://www.congatec.com/en/products/com-express-type-6/conga-tcv2/