In the last few years Intel has given us almost insignificant upgrades while steadily rising its prices. All in part because of AMD’s inability to offer a competitive processor. That’s over with Zen architecture. It’s all we expected and more and it will bring with it a renewed competition that will once again boost processors development, at least for the moment, in desktop systems.
The first models to hit the market are the Ryzen series 7. They will not be much different than inferior models, but they are the perfect example of what this architecture can offer to the domestic and professional users. These are eight-core processors where, for the first time in the history of AMD, the ability to process in two parallel threads by each of the cores is introduced.
But the great processing capacity of these new processors is not only due to their large number of cores or their capacity of double threads per core. AMD has developed a whole pile of features that make these processors shine in their single process capability.
Prediction has been improved so it’s able to locate the necessary instructions better and faster, a new Micro-op Cache has been introduced that reduces accesses to the second and third level cache. The instruction programmer window has been expanded to 1.75 times more capacity and the parallel execution resources have been expanded 1.5 times. This emphasis in parallel processing capability is what lays the foundation for the performance of this architecture and what gets it to scale so well in its multi-core configurations.
This architecture is powered by a revamped three-level cache hierarchy that is structured in a 64KB first level cache, a second level of 512KB per core and a third level cache of 8MB for every four cores, 16MB in eight-core models. This design gets more storage near the core by increasing the bandwidth by up to five times compared to previous generations.
Diagram of the second level cache of 512KB per core
The third-level cache behaves, using the prediction and access systems, as a second-level cache but with a capacity of 8MB per set, 16MB on 8-core processors such as the Ryzen 7 1800X
The other big step forward of these processors can be found in its manufacturing process and in its energy efficiency. The higher model, the 1800X, has a TDP of 90w, little more than the medium range of Intel and with a much higher power. In fact, the 1700, which competes directly with the Core i7-7700k in price, consumes only 65w while achieving more than 40% more performance than the Intel alternative.
They are made using a FinFET process of 14nm. This architecture allows to scale from the most efficient mobile models, to processors like the ones we are showing you, that compete with the best of the domestic market for desktop computers.
Zen is formed around groups of 4 cores with capacity for 8 processing threads. AMD can stack this structure to achieve 8-core, 16-threads or intermediate models with 6-core and 12-core by disabling some of the internal units of the processor. This also allows AMD to better take advantage of the wafers, using processors that have cores that do not function properly by deactivating it and selling it as a lower range. It is a way to cheapen the manufacturing process.
In this photo of the die we can see perfectly the four cores per set and the large amount of space that the cache system of this processor occupies (clearly seen among the four cores).
Each core can access all cache layers of the processor structure with the same latency as the cache physically closest to that core. In fact, the 8MB third level cache works the way a second level cache would in a more traditional design.
AMD has developed a new data bus for this architecture called “infinite fabric“. It is a scalable bus that allows AMD to develop models ranging from low-end designs to very dense processors for server applications. This includes system memory control, standard interfaces such as PCI Express, input and output, etc.
It is also complemented by AMD’s SenseMI technology, which is an array of ultra-sensitive sensors that, combined with new predictive algorithms, allows you to adjust the operation of key areas of the processor and even predict processor states dynamically. This helps AMD make its Turbo frequency system, for example, be more efficient by adjusting the frequency of each of the cores, voltages and other factors based on data such as the measured temperature, etc. The same can also be applied to frequency adjustments to reduce consumption or keep the processor in designed parameters for very specific solutions.
Precision Boost, the turbo of these processors, is the frequency-increasing technology of this architecture. It allows changes in very small steps, of only 25MHz. AMD has designed this technology to behave as the frequency adjustment of its GPUs. This allows high frequencies to be maintained in accordance with the thermal and consumption specifications of the processor.
These improvements mean that the ‘X’ models like the 1800X or the 1700X have at their disposal the XFR (Extended Frequency Range) technology that allows, with adequate cooling, the processor to adapt to its cooling conditions and increase the working frequency above the limits of the processor. The better the cooling of the processor, the more this technology will do for us automatically.
Neural Net Prediction is another jewel on the design of this architecture. Zen is able to predict the movements of the applications we run on it. It achieves this by learning in real time the application execution patterns in anticipation of their movements. This translates into a better prefetch of data limiting the accesses and taking better advantage of the massive amount of cache that the processor has .
With this processor comes a new socket. AMD’s Ryzen processors maintain their pin-socket design on the processor. The number has increased, largely due to the support for DDR4 memory and the revised consumptions of this architecture so it is necessary to change the motherboard.
They introduce a whole new series of chipsets that will act as connective bridge adding elements where the control unit of the CPU does not arrive. Basically we will find three models that differ in the number of USB connectors and storage that they support. They also add some extra PCI Express 3.0 lines and, of course, a Serial ATA controller in addition to more advanced storage features.
In the graph you can see the differences, but also very interesting data. The X370 is the only one that is ready for SLI and surely also for Crossfire. It also has more USB connectors and more SATA ports. Interestingly this element decreases to four ports to offer more PCI Express connectivity for more advanced storage solutions.
They natively support up to two USB 3.1 Gen2 Ports of 10Gbps, technology that Intel has not yet made native on their chipsets, also adds 8 PCI Express 2.0 lines and provides support for PCI Express storage solutions with support for NVMe protocol and PCI Express 3.0 4x links of 32Gbps bandwidth.
ASUS has renewed its Crosshair range for AMD processors with a new Hero. An excellent example of chipset 370X, high range, with a price that will be around 255 dollars.
The intention of AMD is to keep this socket architecture until 2020 but that will be a problem if they want to introduce new technologies such as PCI Express 4.0 or DDR5. Technologies that will surely not be implemented massively until future dates.
These benefits you see above are complemented with the ones that offers the input and output unit of the processor itself, which we will see now. 370X Chipset will be the reference for powerful systems and we can see its cost in motherboards of barely more than 200 dollars. Price that will clearly adjust while the competition puts more products on the shelves. B350 is a very capable alternative to 370X, except for the support of multiple graphs, and we can find motherboards with this chipset for a little bit more than 100 dollars.
These chipsets and these processors offers an excellent connective capacity.
The 370X and B350 chipsets will have support to take advantage of the unlocked settings of this processors. Essential for overclocking. They will also have added support to touch voltages.
Key information, models and prices
These processors are not only portents of calculation with many threads, approaching experiences that had been expensive to pay until now and that certainly rarely could we see in a domestic computer but rather in workstations for professional consumers.
These processors not only offer this, but also incorporate some complementary technologies of what chipsets that we have shown you before offers. These processors offer natively four USB 3.1 Gen1 of 5Gbps ports, 16 PCI Express 3.0 lines, 4 PCI Express 3.0 lines dedicated to storage with NVMe support and four PCI Express 3.0 extra lines so that the chipset disposes of them as it wants. They are a total of 24 PCI Express lines, even though some, as you have seen before, are specialized.
These processors have a dual channel memory bus and 128-bit with DDR4 memory support up to 2666MHz, although this “official” support frequency reduces to a little bit more than 2133 when we use the typical four memory slots. In this setting we could install up to 64GB of RAM.
All processors from the 7th series, even though this is transferable to all Ryzen range when it is available, comes unlocked. Its internal multiplied bus, which generates the final frequency, is variable and therefore it offer us an easy and accessible mechanism for overclocking. The only big limitations will be the voltages and, most of all, cooling.
The retail heatsink from the 7th series will have RGB illumination.
Socket counts with 1331 pins and have a thermal design of 128w TDP. This offer to AMD some room of maneuver to increase the settings of its domestic processors with more cores or higher frequencies.
Behavior and overclocking
With this processor I have the same fluidity and availability feeling that you find with the most powerful processors from Intel. It is no longer an issue of benchmarks, we will now see how it behaves in this area, but Windows’ proper behavior. I know is not easy to describe a feeling like that, but with the prices these new 8-cores processors from AMD will handle it is probable that more people has the chance to experiment themselves in their PCs.
The overclocking capacity of this AMD architecture depends a lot on the cooling capacity we can put on it, but it’s certainly not elevated in conventional cooling settings. With a good water cooling kit it gets better, but it’s not spectacular either. We can access frequencies up to 4.2GHz with the right cooling but also with somewhat high voltages, around 1.45v. AMD recommends to be more modest with voltage and certainly I think that one good target for these processors is been able to keep turbo frequencies constantly, 4.0-4.1GHz. On this sustained frequency, and a modest voltage, the performance is spectacular.
It maintains temperatures moderately well and always achieves some frequency above base frequency even in situations of high CPU stress.
Working temperature of this processor seeks to stay at a maximum of 70 degrees, turbo frequencies always adjust around that temperature. If we stress all processor threads out, with a conventional heatsink, we will have stability at a frequency of 3.73GHz in all cores. Quite interesting since it always keeps us somewhat more frequency, that base frequency, which make the performance improve persistently.
We can use traditional modes of overclocking, through bios or the programs from our motherboard manufacturer, or using the AMD Ryzen Master software which allows us to perform real-time overclocking. Multiplier settings, various voltages, frequencies, etc. We can set it on profiles and it’s really easy to use. We can even deactivate processor’s cores directly from this program.
The AMD Ryzen Master will be a good software to take advantage these processors. It offers monitoring, overclocking profiles and a very accessible control system.
We have compared on the habitual test battery this processor with some of the best Intel processors. We have done overclocking to this processor and the AMD processor to measure out all their potential. The results are interesting, now we share it.
Handbrake 1.0. In seconds, less is better.
Cinebench R15 1-Core
Cinebench R15 SMP
Sisoft Sandra 2016 CPU Arithmetic
Sisoft Sandra 2016 CPU Multimedia
Sisoft Sandra 2016 MEM Bandwidth
CPU consumption on load
Performance in games.
We´ve wanted to add on the results from the last generation from AMD because it is really amazing the performance improvement and consumption in compares to its previous models. With no doubt it surprises how it stands against or defeat much more expensive processors but in my own perspective the change can be notice in this data. It consumes half the energy and and runs as twice faster than the previous generation. That translates in a data that tell us that the new processors Ryzen are about four times more powerful per watts than the previous AMD models.
On our final tests, Intel suffers more than we expected and that will force the Company to bring out more powerful alternatives at the same Price or just cut the current prices. AMD has a great opportunity to recover market share on the domestic market and for all kinds of users.
AMD hopes to develop two more generations with this architecture over the next years. The meaning is to keep the socket so the actual motherboards could install the upgrades from the new Ryzen processors generations.
The closest novelty we´ll found in the 5 and 3 series with less numerous cores settings. These models will arrive later on, surely less than three months. They will compete with the most economic mid-range from Intel and they will do it with up to six core models with 2 process threads per core
Analysis and conclusion
AMD has the most interesting processor that has been introduced into the market over the last 5 years. This high-range model equals and surpasses on many tests to a processor as the core i7-6900k that is over 1000 dollars more expensive, almost half price. But, this means nothing because for less than 400 dollars we can have an eight-core with the potential to sweep out on any test a processor so powerful and demanded as the Intel core i7-7700k. an also recently presented processor
This processor innovations, the processors democratization, great number of cores and threads and its singular process power make that AMD has a winner product. Something that just didn’t seem on several years. It throws back ADM into the fight on the desktop market and it seems that will also be a good platform to bring back ADM in the laptops market or more compact formats
The future seems promising for the Brand, with good processors and excellent graphic chipsets and it will force their competitors to “move a gear up” and offer better products at more competitive prices. This eight core units that will be the ones we could buy for now, are perfect for both, players with the capacity to take off the best AMD graphics chipsets from Nvidia and AMD, as for design professionals who requires a process power and also for programmers which will see on these processors an economic way to have a great number of virtual machines.
They enjoy low very controlled temperatures, adjusted consumptions, and spectacular performance. An authentic architecture wonder that i have no doubt about it will be welcome with open arms by any who is searching for a pc exceptionally powerful with access to the latest connective and storage technologies for a real balanced price.
Best Gaming Monitor