AMD announces today during a live streaming event more is known about the long-awaited Zen processors. In addition to new benchmarks, and some new details about the structure of the chip, AMD also announces the processors actually called go. The high-end desktops targeted Zen versions, known under codename Summit Ridge, will see the light under the official name AMD Ryzen. This continues the now very popular codename Zen retained as part of the new name.
In recent months, AMD has increasingly revealed details of the highly anticipated Zen processors. Recently, we have all the news summarized thus far in an enlarged AMD Zen Preview. Today some missing pieces of the puzzle is completed. We got the evening information published last week told during an exclusive event in Sonoma, California.
To deal with the bad news from today’s announcement but immediately: who hoped for a formal introduction of Zen in the first week of January, bad luck during the CES. The AMD Ryzen CPUs come to market slightly later. AMD talks about himself “production in the first quarter of 2017”. Or that his carefully chosen words, which one gets the opportunity for an even later actual introduction in the second quarter ajar, remains conjecture.
Final AMD Ryzen chips do not we see so unfortunately at CES, but AMD does have announced that several motherboard manufacturers at CES for AMD Ryzen required AM4 AMD motherboards will show. That’s at least something to look forward to. These motherboards are AM4 according to AMD right up to date with support for DDR4, PCI Express 3.0, USB 3.1 and Gen 2 NVMe. According to AMD use different motherboard manufacturers hard on products aimed specifically at gamers, overclockers and other hardware enthusiasts.
Besides the name will be announced today on the chips. Where the expected clock frequencies for the Zen CPUs until now been a mystery, AMD will today officially announced that the flagship AMD Ryzen processor receives a basic clock frequency of at least 3.4 GHz. That is still the exclusive turbo / boost mode that get the chips. What we already knew, but once again confirmed that the flagship AMD Ryzen processors have eight cores and can handle simultaneously tasks 16 program threads with simultaneous multi-threading. The total amount of cache memory is further made explicitly clear: 20 MB, 4 MB L2 cache (512KB per core) and 16 MB of shared L3 cache.
In August, during the Intel Developer Forum, AMD demonstrated its first comparison test between Zen and Intel’s 8-core Broadwell-E processor. one had both CPUs put on an identical, fixed clock frequency at that moment. During the demonstration last week, AMD went a lot further. Again Zen was compared with the 8-core Intel Core i7 6900K, but this time it worked 6900K at its default clock frequency, including armed Turbo – mode. The shown configuration Zen worked at 3.4GHz, but still without boost. First of all, AMD showed a test with the popular HandBrake video compression software, with an Ultra HD file was converted to Full HD H.264. AMD Ryzen CPU completed this task a fraction faster than the Core i7 6900K, which again showed that AMD Zen in any case withstand Intel Broadwell-E.
However, there has now been shown more than just performance. As in August, we got to see a test based on the Blender 3D rendering benchmark, but now with the power there. AMD Ryzen idle test system consumed about 93W, while the “equivalent configured” (exact specs we did not) Core i7 6900K system consumed about 106W idle. During the Blender benchmark, which fully taxed all CPU cores, the AMD system consumed a maximum of about 187 watts and the Intel system about 191 watts. Again the AMD machine was faster finished. AMD’s message is clear: we are faster and more efficient than Intel. AMD Knowing you can assume that that list at launch is made up to faster, more efficient and cheaper.
During the presentation in August AMD already made known about the architecture of the AMD processors. This week came another other aspects discussed, including a set of technologies called AMD SenseMI, which should improve performance, efficiency and overclocking potential.
In addition, AMD made known to the Zen-processors make use of a new type of internal connection, called AMD Infinity Fabric. This is a new and very fast way of connecting various parts within a SoC. Infinity fabric is not only used in the Zen processors, but also in the future Vega AMD GPUs and (almost) all other AMD chips can be expected in the near future. Inifinity Fabric allows for faster and better income secure connections within a chip. It has to be standardized for all successor at present a wide range of existing intra-chip connections. This makes Infinity Fabric must also ensure that AMD can be more flexible to develop custom chips.
Back to the announced SenseMI technology where the Zen processors use. It consists of five parts. Pure Power is a technology that allows AMD Ryzen chips and other Zen variants as economical and thus operate as efficiently as possible. According to AMD, the Zen CPUs equipped with hundreds of sensors that constantly monitor temperature, voltage and power consumption of all chip components. Through the Infinity fabricated sensors are all connected to a central control unit, the Infinity System Management Unit, which is based on all inputs constant adjustments can do the tuning of the processor.
This can be accomplished with significantly lower power consumption according to an equal level of performance AMD. It also works the other way: Precision Boost is the name for the Turbo mode that get the Zen processors. The same network of sensors constantly monitors how much space is there for extra performance and the chip parts can be increased. According to AMD’s adjusting clock frequencies without delay and with great accuracy in steps of 25 MHz. Precisely by lowering clock frequencies simultaneously with no or only limited use chip components and to increase heavily loaded components, precision can boost bidding AMD significantly better performance with similar power consumption.
Good news was there for overclockers. They can at AMD Ryzen processors in the future to enable the so-called Extended Frequency Range. This ensures that the turbo-ingrained mode completely independently doorschaalt at clock frequencies higher than the maximum standard values. The degree to which clock frequencies are increased, the cooling used with scales, according to AMD into liquid nitrogen to it. It makes better cooling AMD Ryzen chips also can provide improved performance without the user having to delve into overclocking.
Another part of SmartMI’s Neural Net Prediction. During the presentation in August AMD already announced that the Zen existing branch predictor (the code that predicts the path there with branches in code probably will be) has been greatly improved. Without giving much deeper detail, AMD indicated that due to the improved operation is AI-like algorithms. The branch predictor has a self-learning system on board, making it based on good or bad verspelde branches can improve its own performance. Branch predition is essential for the efficient operation of the processor; if that can work forward in a good way and can bring enough variety in instruction types, all excutions units present in the cores held as often and long as possible involved.
A similar self-learning system is also in pre-fetcher, the technology that predicts what data from the memory needs in the short term processor retrieves it and already placed in the caches. It speaks with Smart Prefetch. The better the work prefetchers, the less often the CPU has to wait before data is retrieved from relatively slow memory, thus the less delay may occur in the processing of instructions. This is the current CPU generation undoubtedly a weakness of AMD and why faster memory properly can ensure existing AMD processors for significant performance gains, while at Intel, though has been very good prefetch algorithms in its processors , is only limited.
Step by step makes AMD more is known about the new processor. Now we know that AMD Ryzen chips not only in performance but also in terms of power consumption are evenly Intel Broadwell-E chips will only lovers but even more looking forward to the final launch. Sadly we will there still be patient to have. AMD continues to talk about “first quarter 2017” for AMD Ryzen, thus it is not yet clear whether it refers to pure production or even launch. The first server version of Zen, called Naples processor must follow in the second quarter. For the second half of 2017 is Raven Ridge on the roll, a quad-core Zen version with integrated GPU, targeted at laptops and consumer PCs.