For a long time, the video card remained an independent component of a personal computer, as well as the system logic. At the same time, advances in semiconductor technology were advancing so rapidly that it made it possible to accommodate more and more transistors per unit area, so that it was only a matter of time before the core components of a PC were combined. As a result, in 2009, the Intel Lynnfield chips saw the light, in which the north bridge was fully integrated into the CPU silicon die, but the system still required a discrete graphics accelerator to work. However, in the same 2009, energy-efficient Intel Atom processors based on the Pine Trail core were announced without too much fanfare, which, in fact, became the first CPUs that included a video accelerator. However, the silicone giant from Santa Clara at that time did not have a productive video core to ensure an acceptable level of performance even in the most primitive games. On the other hand, AMD had such accelerators, which, after the takeover of ATI Technologies in 2006, received full access to the most advanced developments in the field of gaming accelerators. Oddly enough, Advanced Micro Devices started small, namely with E-Series (Zacate) chips for low cost systems and netbooks popular at the time. Despite the lack of outstanding performance, the E-Series became an important milestone that revealed to the world the concept of APU — Accelerated Processing Unit, which consists in combining the main components of a computer inside a single semiconductor crystal.
The next stage in the evolution of AMD’s new ideology was Llano hybrid processors for desktop systems, which provided unprecedented performance for integrated graphics in games and outlined two unspoken trends in the development of APUs: a platform separate from high-performance solutions and an increase in the lag in performance of the computing part from the main competitor. In the meantime, Intel highly appreciated the idea of combining computing and graphics cores on a single silicon substrate, and starting from the Sandy Bridge generation and to this day, all mass CPUs from this vendor have a built-in video accelerator. Over time, somewhere after the release of quite competitive Trinity, there was a decline in the development of APUs, which can be explained by the lack of a powerful processor part at AMD’s disposal, as a result of which the same Kaveri, and even more so Bristol Ridge, could no longer compete in performance tests even with Pentium , especially after the latest support for Hyper-Threading.
At the same time, the graphics component of hybrid processors was still ahead of the competitor’s solutions, but its performance was no longer enough for modern gaming projects even at the minimum detail levels. It seemed that only a miracle could save the AMD processor business from a complete collapse … and this miracle happened in March 2017, along with the release of the revolutionary Ryzen, which marked the emergence of real competition in the CPU market and overnight returned Advanced Micro Devices to the segment of high-performance solutions. However, there was no place for a graphics accelerator in the new products, so the release of APUs combining the advantages of the latest architecture and a modern video core was only a matter of time. Finally, in February 2018, AMD showed the world their next creation — the Ryzen 3 2200G and Ryzen 5 2400G processors with integrated Vega graphics.
AMD Raven Ridge
The hybrid processors, codenamed Raven Ridge, will organically complement the entry-level Ryzen product line for the Socket AM4 platform, which, after the announcement of new products, has taken the following form:
|Processor||AMD Ryzen 3 2200G||AMD Ryzen 3 1200||AMD Ryzen 3 1300X||AMD Ryzen 5 2400G||AMD Ryzen 5 1400||AMD Ryzen 5 1500X|
|Core||Raven Ridge||Summit Ridge||Summit Ridge||Raven Ridge||Summit Ridge||Summit Ridge|
|Process technology, nm||14||14||14||14||14||14|
|Number of cores (threads)||4 (4)||4 (4)||4 (4)||4 (8)||4 (8)||4 (8)|
|Rated frequency, GHz||3,5||3,1||3,5||3,6||3,2||3,5|
|Boost-mode frequency, GGc||3,7||3,4||3,7||3,9||3,6||3,7|
|XFR frequency, GHz||–||3,45||3,9||–||3,45||3,9|
|Multiplier unlocked up||+||+||+||+||+||+|
|L1 cache, KB||4 x (32 + 64)||4 x (32 + 64)||4 x (32 + 64)||4 x (32 + 64)||4 x (32 + 64)||4 x (32 + 64)|
|L2 cache, KB||4 x 512||4 x 512||4 x 512||4 x 512||4 x 512||4 x 512|
|L3 cache, MB||4||8||8||4||8||16|
|Graphics core||Vega 8||–||–||Vega 11||–||–|
|Graphics core frequency, MHz||1100||–||–||1250||–||–|
|Number of unified shader processors||512||–||–||704||–||–|
First of all, the recommended prices of new products pay attention: now the Ryzen 3 2200G is the most affordable carrier of the Zen architecture with a price of only $99, which is even $10 cheaper than the Ryzen 3 1200. At the same time, the Ryzen 5 2400G model is offered for the same $169, which are asking for the Ryzen 5 1400, previously considered the best choice for mid-range systems. Like its closest relatives, the latest Raven Ridges have a TDP of 65W, but their clock speeds are even higher than similar Summit Ridges. In addition, the clock speed control mechanism has been updated and Precision Boost 2 allows you to change the speed of each core in 25 MHz increments, providing maximum performance depending on the nature of the computing load. For beginners, any of the existing Socket AM4 boards based on the AMD 300 series system logic, which may require an update to the UEFI control microcode, will be suitable. At the same time, the new processors have many differences from their predecessors, the main of which is the presence of an integrated Vega video accelerator, the carrier of the most advanced graphics architecture available to AMD. The introduction of such a massive element as a video accelerator forced engineers to reconsider the layout of the semiconductor crystal: Raven Ridge lost one quad-core CCX (CPU Complex), and the size of the L3 cache was reduced to 4 MB. At the same time, the transistor budget has increased from 4800 million for classic Ryzen to 4950 million with a slight decrease in area — 212.97 mm2 for «thoroughbred» CPUs versus 209.78 mm2 for APUs with the same 14-nm FinFET + process technology.
From a performance point of view, the micro-design embedded in the Zen processors is a noticeable step forward compared to its predecessors: the instruction scheduler window is increased by 1.75 times, and the decoder can decode up to 4 instructions per clock. In addition, it is now possible to load frequently used instructions directly, bypassing the L2 and L3 caches, and an improved branch predictor based on a neural network allows you to prepare optimal instructions with high accuracy to avoid idle pipelines. All these optimizations are aimed at improving single-threaded execution, and SMT — Simultaneous Multi-threading technology is used to speed up applications that support parallel computing. Cardinal changes aimed at increasing throughput also affected the cache memory subsystem: at the first level, 64 KB and 32 KB are allocated for storing instructions and data, respectively, while the L2 cache is a solid 512 KB. Unlike the rest of Ryzen, which have 8MB or 16MB L3 cache, the latest Raven Ridge processors have only 4MB L3 cache, but thanks to a complex, self-learning prefetch algorithm that speculatively places application data into memory, they are always ready for immediate execution.
But, of course, the key feature and pride of the new products is the integrated Vega graphics core with support for the DirectX 12 API and the Vulkan API, consisting of 8 CU (Compute Unit) or 11 CU in the case of Ryzen 3 2200G or Ryzen 5 2400G respectively. Each CU includes 64 ALUs and 4 ROPs, as a result, the older APU has 704 computing modules and 44 TMUs, while the younger one is content with 512 ALUs and 32 TMUs. The Vega 8 is clocked at 1100MHz and the Vega 11 can be clocked up to 1250MHz, giving the FP32 a theoretical performance of 1.126TFLOPS and 1.76TFLOPS respectively. In addition, the video core includes 16 rasterization units, two HWS (Hardware Schedule) scheduler units and four ACE (Asynchronous Compute Engine) modules responsible for hardware acceleration of encoding and decoding video streams, moreover, these modules are completely similar to those in AMD flagship products Vega 64. Thus, the graphics core built into Raven Ridge processors is capable of decoding video streams in 2160p resolution in VP9 10bpc format at 30 fps, and for HEVC 10bpc and H.264 — all 60 fps, which makes the new products an excellent choice for playing 4K content .
All Raven Ridge components are interconnected by a high-speed Infinity Fabric bus that connects the CCX, graphics core and imaging modules, a memory controller that has received official support for DDR4-2933 MHz modules, and an I / O Hub. In terms of peripheral connectivity, new items can offer four USB 3.1 Gen2 channels, one USB 3.1 Gen1 and USB 2.0 each, two SATA 6 Gb / s interfaces and eight PCI-E 3.0 lines for connecting system logic and various peripherals, as well as a total of eight PCI-E 3.0 lines to work with a discrete graphics accelerator.
Thus, the new products perfectly complement the line of AMD Ryzen processors of the first generation, offering users a unique combination of a relatively powerful integrated graphics subsystem and high-performance computing cores at a very favorable price.
However, enough theory, it’s time to move on to a direct study of the features of the Ryzen 3 2200G and Ryzen 5 2400G. In our first article, we will study the overclocking potential of new products, and extensive gaming testing will be in the next part.