In recent years, AMD has continued to consistently lack innovation in the CPU segment, trying to make up for lost market share with sales of hybrid CPUs that have not found much popularity among enthusiasts and gamers. In the meantime, Intel has been systematically increasing its already huge lead over its main competitor thanks to the well-established release of new generations of chips every one to two years. The tick-tock strategy, in which the chipmaker took turns introducing a new technological process and then releasing the next processor core architecture, has proven its effectiveness and ensured the presence of Intel “stones” in almost every gaming PC or workstation.
Understanding their unenviable position, the management of Advanced Micro Devices in 2012 returned the talented engineer Jim Keller to the ranks of the company, under whose leadership the legendary AMD K7 (Athlon, Athlon XP) and K8 (Athlon 64, Opteron) processors were designed. Wasting no time, Mr. Keller immediately set about developing a new microprocessor architecture, called Zen, from scratch. When it was created, AMD set a goal to achieve a 40% increase in the number of instructions executed per clock compared to Excavator processors. Although Jim left AMD in 2015 to join Tesla Motors, where he still works, the Sunnyvale chipmaker still managed to complete the work he started and the new microarchitecture was used in Ryzen solutions, which will be discussed today.
AMD Ryzen Desktop Processors (Summit Ridge Family) are manufactured to 14nm process standards using FinFET transistors at GlobalFoundries facilities. The number of active computing cores on a CPU chip can vary from four to eight, and the maximum amount of total L3 cache reaches 16 MB. The new processors feature a dual-channel DDR4 memory controller as well as 24-lane PCI Express 3.0 interface controllers, two SATA 6 Gb/s ports, and four-port USB 3.1 first generation on the die. Thus, the new CPUs are rather full-fledged single-chip systems, rather than processors in the usual sense of the word.
Unlike solutions based on the Excavator architecture, which used cluster multithreading technology (CMT), Zen chips switched to using simultaneous multithreading (Simultaneous Multi-threading), one of the branches of which is the long-familiar Intel Hyper-Treading technology. Also, one of the important evolutionary steps of the Zen architecture is the presence of a floating point unit for each core. Recall that the Excavator processors used one FPU for two cores, combined into a so-called module. In addition, thanks to the 14nm process technology, the top octa-core Ryzen processors have a nominal TDP of 95W, instead of the 220W of their FX predecessors.
In Ryzen processors, AMD engineers have implemented a whole set of new technologies, called SenseMI, designed to provide the most optimal and efficient operation of all chips based on the Zen architecture. It includes:
- “Smart” caching system Smart Prefetch, which pre-loads the data into the cache, necessary for the program to work at the next stage.
- Neural Net Prediction is an advanced branch prediction block that functions like a self-learning neural network and predicts which instructions the application will use in the course of its further work.
- Pure Power monitors the operating mode, frequency and power consumption of the processor, providing dynamic adjustment of the CPU voltage depending on the loads.
- Precision Boost dynamic overclocking technology that controls the operating frequency of the processor in 25 MHz steps and a speed of one millisecond.
- Extended Frequency Range can raise the operating frequency beyond the limit set by Precision Boost when using powerful cooling systems.
The last three technologies are closely related to each other and rely on hundreds of sensors built into the processor chip that monitor the voltage, current and temperature of various functional blocks of the CPU. This system, based on the information received, can automatically adjust the clock frequency of each computing core individually, ensuring maximum performance of the busiest cores.
The Infinity Fabric bus is used to exchange information between many sensors and system-on-a-chip functional blocks, which are Ryzen processors and the upcoming Raven Ridge APUs. The new interface is almost universal and, due to its modular structure, can be easily scaled to provide the necessary throughput. In addition to Zen-based chips, the Infinity Fabric bus will also be used in Vega GPUs, where its throughput can reach 512 GB / s, while Raven Ridge mobile APUs will be content with its variant with a speed of about 50 GB / s.
With proper cooling, XFR technology can increase the operating frequency of all processor cores by an additional 50 MHz (100 MHz for models with the X index) above the nominal. At the same time, the frequency of the boost-mode, at which only one computing core is able to work simultaneously, will also be increased. For example, when using a highly efficient cooling system, all cores of the Ryzen 7 1800X processor (3.6 / 4 GHz) will operate at a frequency of 3.7 GHz, and during single-threaded workloads, the clock frequency of the used core will increase to 4.1 GHz.
Before moving on to the detailed specifications of Ryzen processors, you should pay attention to the new notation. The names of the new AMD CPUs and APUs will be divided into three parts. The first is responsible for the brand, in this case it is Ryzen. The second, consisting of only one digit, determines belonging to the market segment. The company did not reinvent the wheel, and used, by analogy with Intel Core processors, the numbers 7, 5 and 3 already familiar to most users, where a larger number corresponds to a higher product segment.
The third part of the marking consists of four digits and one optional character. The first digit is responsible for the generation of the processor, the second determines the performance level of a particular chip model. The two following are responsible for the revision and small increases in the operating frequency of new versions of the CPU. The letter at the end indicates the level of power consumption of a particular sample, as well as the presence / absence of an integrated graphics core. For example, the X symbol at the end indicates a nominal TDP of 95 W and the presence of full support for XFR technology.
In the near future, AMD plans to launch Ryzen desktop processors for all market segments. At the moment, three eight-core models of the Ryzen 7 line are available for free sale, designed primarily for computer enthusiasts and demanding gamers. In total, the new family will include more than a dozen different models. The release of six- and four-core processors is scheduled for the second and third quarters of this year.
Separately, it is worth noting AMD’s concern for fans of overclocking. Absolutely all AMD Ryzen processors have an unlocked multiplier, and under the metal heat-distributing cover that covers the semiconductor crystal, there is nothing more than solder. In addition, the chipmaker has also prepared a special Ryzen Master overclocking utility, which provides all the tools needed for overclocking directly from the operating system.
Specifications for the first trio of AMD Ryzen processors are shown in the table below, including some competitors.
|Processor||AMD Ryzen 7 1800X||AMD Ryzen 7 1700X||AMD Ryzen 7 1700||Intel Core i7-6950X||Intel Core i7-7700K||AMD FX-9590|
|Core||Summit Ridge||Summit Ridge||Summit Ridge||Broadwell-E||Kaby Lake||Vishera|
|Process technology, nm||14||14||14||14||14||32|
|Number of cores (threads)||8 (16)||8 (16)||8 (16)||10 (20)||4 (8)||8|
|Rated frequency, GHz||3,6||3,4||3||3||4,2||4,7|
|Boost-mode frequency, GGc||4||3,8||3,7||3,5||4,5||5|
|Multiplier unlocked up||+||+||+||+||+||+|
|L1 cache, KB||8 x (32 + 64)||8 x (32 + 64)||8 x (32 + 64)||10 x (32+32)||4 x (32+32)||8 x 16 + 4 x 64|
|L2 cache, KB||8 x 512||8 x 512||8 x 512||10 x 256||4 x 256||4 x 2048|
|L3 cache, MB||16||16||16||25||8||8|
AM4 platform and new chipsets
Together with Ryzen processors, the AM4 platform will debut in the consumer market, which previously had access exclusively to AMD OEM partners. All new desktop processors of the chipmaker in the next couple of years will have a Socket AM4 design with 1331 pins.
Due to the increase in the number of pins by more than three hundred compared to the Socket AM3 + connector, the complexity of motherboard wiring has increased, as a result, the new platform uses the original cooler mounting system. This fact has already managed to cause a wave of indignation among the computer community, but not everything is so bad: eminent CPU cooler manufacturers have already begun distributing new mounting kits for free, and some boards (like the ASUS Crosshair VI Hero) have inherited compatibility with old cooling system mounts.
|Model||AMD X370||AMD B350||AMD A320||AMD X300||AMD A300|
|PCI-Express 3.0 configuration||x16
|Number of additional PCI-Express 3.0 lanes||4||4||4||8||8|
|Number of PCI-Express 2.0 lanes||8||6||4||0||0|
|Port||2x USB 3.1 G2 + 10x USB 3.1 G1||2x USB 3.1 G2 + 6x USB 3.1 G1||1x USB 3.1 G2 + 6x USB 3.1 G1||4x USB 3.1 G1||4x USB 3.1 G1|
|Total USB 3.1 + USB 2.0 ports||18||14||13||4||4|
|Serial ATA||6x SATA 6Gb/s||4x SATA 6Gb/s||4x SATA 6Gb/s||2x SATA 6Gb/s||2x SATA 6Gb/s|
In total, AMD has prepared five chipsets for the new platform: X370, B350, A320, X300 and A300. Moreover, the last two are designed specifically for Mini-ITX motherboards and various kinds of ready-made mini-PCs. The most interesting from the point of view of an overclocker and an advanced user are the X370, B350 and X300 chipsets. Motherboards based on them will provide their owners with all the necessary tools for overclocking and tweaking Ryzen processors.
The Socket AM4 CPU chipset includes a single chip that performs the tasks of the “south bridge”. Thus, the X370, B350 and A320 chipsets provide additional USB 3.1 ports of the first and second generations, USB 2.0, SATA 6 Gb / s ports, and eight PCI Express 2.0 interface lines for connecting expansion cards and additional controllers. Four PCI Express 3.0 bus lines are used to communicate with the processor and the peripherals built into the three above-mentioned chipsets.
It should be borne in mind that when using the SATA 6 Gb / s controllers built into the CPU, only two of the four PCI Express 3.0 interface lines intended for connecting an NVMe drive will be available. For this reason, it is worth considering the configuration of the storage system in advance. Another unpleasant feature of Ryzen processors and new chipsets is the low operating frequencies of DDR4 memory when four modules are installed at once. So, when using only two DIMM strips, they will work in DDR4-2400/2667 mode, and when all four slots are used, the RAM mode will deteriorate to DDR4-1866/2133. Of course, this problem can be solved by resorting to overclocking, but how well the memory controller will behave during overclocking remains to be seen.
As already mentioned, the company has prepared two special chipsets specifically for Mini-ITX boards: X300 and A300. These chipsets do not provide any additional ports or interfaces for connecting peripherals. Nevertheless, since Ryzen processors are single-chip systems, motherboards based on these chipsets will be able to offer quite a good set of interfaces by modern standards. It includes four USB 3.1 ports of the first generation, PCI-E 3.0 x4 for connecting NVMe drives, two SATA 6 Gb / s ports, SATA Express, as well as an additional four PCI Express 3.0 lanes for various peripherals, which in all other cases were used to communicate with the built-in controllers of “normal” sets of logic.
More details about the new platform and AMD chipsets can be found in further materials on our website, because due to logistics, the new items did not arrive on time. Stay with us!