A new era – so AMD describes access to RYZEN processor market. The architecture of the Zen, which are based on the new processors, AMD has to strengthen the position on all important markets. First RYZEN processors will come to the desktop. Then Zen architecture must show itself in the server market and in the mobile segment, the corresponding model can be expected in late 2017 or early 2018.
AMD quite successfully competes with NVIDIA graphics cards on the market, but the business of AMD processors plummeted in recent years. Of course, AMD has had some interesting models, but not for the desktop market, nor in the mobile market, not to mention the server, AMD was not able to get a significant share. Including, for the reason that AMD inferior Intel process technology. For a long time the company had to settle for planar technical processes 32 and 28 nm, which can not be considered competitive. A Bulldozer architecture does not live up to the high expectations ..
Of course, AMD has always tried to improve their designs. Integrated graphics has a strong advantage, and instruction sets were added to the processors after Intel. The only thing missing – a real improvement in CPU cores are also affected and the lag in process technology. With the new technical processes FinFET we received an improvement in efficiency, as AMD promises IPC performance improved by 40%. Hopefully, all this will help AMD to address the CPU market with a strong position.
Intel has always tried to consistently optimize the performance of your CPU generations, regardless of progress on process technology, but AMD plans to Zen architecture make a significant leap forward. That, in principle, it is logical, given the low base. The company converted the many areas of nuclear architecture. Added the support of SMT, the cache for each CPU core micro-operations, improved caches L1 and L2, and the FPU. All this amount should give a significant performance boost.
On AMD introduced the first three new processors last week. All are equipped with eight cores and can perform up to 16 threads simultaneously thanks to the support Simultaneous Multithreading. The difference there is at the TDP, clock frequencies, XFR (Advanced mode option “Turbo” from AMD) and, of course, the price.
Processor RYZEN 7 1800X – the fastest model with eight cores and 16 threads, the frequency is 3.6 GHz, but in Boost mode, it can reach 4.0 GHz. In XFR possible even mode frequency of 4.1 GHz. For RYZEN 7 1800X declared TDP (TDP) of 95 watts. AMD specifies the retail price of 559 euros, which is much cheaper than a competitor in the form of Intel Core i7-6900K, available for as low as 1.100 euros or 67 thousand. Rubles in Russia.
The second representative – RYZEN 7 1700X, it is in many ways identical to the older model, but it works with a base frequency of 3.4 GHz and 3.8 GHz, the Boost, which is less than 200 MHz. But XFR available via an additional 100 MHz. And the price of 439 euros in RYZEN 7 1700X much less. Processor RYZEN July 1700 – and the third is the “youngest” in the line RYZEN. The most important difference – the lower level of 65 watts TDP and clock speeds of 3.0 and 3.7 GHz, respectively. Do not XFR models-X technology is also present, allowing you to squeeze more clock frequencies. But RYZEN July 1700 growth is only 50 MHz. Price 359 euros is also very pleasing to the 8-core processor.
All these models are available since last week. But the service will only begin March 2. Processors RYZEN 7 1800X and 1700X are available without a cooler, in the case of boxed RYZEN July 1700 you get one of the new models of coolers Wraith. Below we shall return to this subject.
All processors RYZEN support DDR4. But the memory frequency depends on the number of DIMM slots, as well as the type or Dual Rank Single Rank. In the worst case, DDR4 modules will operate at 1.866 MHz rate, at best – on 2.667 MHz. Of course, the motherboard manufacturers will add a certain rate of acceleration that allows DDR4 modules to work with processors RYZEN at frequencies much higher than the 3,000 MHz.
So far in our lab did only one processor – RYZEN 7 1800X. AMD promises soon to send samples RYZEN 7 1700X and RYZEN July 1700, but exactly when is unknown. But we would not have been able to conduct tests of all three processors in time as the last days were quite hot – NVIDIA Editors Day event, which was presented to the GeForce GTX 1080 Ti, as well as the exhibition Mobile World Congress. Therefore, the time was little, we focused on RYZEN 7 1800X. Below are the details that we have already published from the event RYZEN Tech Day:
Full line of processors is already known by the time of RYZEN Tech Day. AMD officially only indicates RYZEN 5 1600X six cores and 12 threads, which is clocked at 3.6 GHz and 4.0 GHz. RYZEN 5 1500X works with six cores and 12 threads, but only at frequencies of 3.5 and 3.7 GHz. AMD has not indicated the exact information about the frequencies of the XFR.
AMD Tech Day at the event showed a small preliminary review of the performance RYZEN 5 1600X, and presenting the results of Cinebench:
Below we look at innovation RYZEN processors, including those related to architecture, process technology and other fields.
The biggest change in the architecture of Zen is that each core has regained more independence. The Bulldozer architecture used by many common components: FPU, SIMD, and the L2 cache used two CPU cores. All of these resources are now available exclusively for a single core, but if necessary it can be to handle two threads – in the case of Intel Hyper-Threading.
AMD An important innovation was the micro-cache (Micro-Op). Intel began to use it even with the architecture Sandy Bridge. Cache accelerates micro-decoding x86 instructions that come at him with a block sampling instructions and decode pipe. sample blocks of instructions and branch prediction 64-byte, in the case of a sample of instructions we get separation of 2x 32 bytes. branch prediction unit operates more rapidly, it can reassign the branch and change the priority in case of error.
In Zen kernel uses four integer units, which have 168 and registers 192 can process instructions at a time. Two blocks Load / Store is responsible for writing data back to the cache after the calculations. There are two block floating-point, 128 each FMAC (Floating Point Multiply Accumulators). instruction cache has a size of 64 KB 4-way, data cache – 32KB eight simultaneous calls (8-way). The L2 cache has a size of 512 Kbytes, it is also possible to simultaneously handling a read or write eight channels (8-way). Add to this the total L3 cache.
The new core Zen is based on three key “pillars”: optimized and fast kernel itself, improved caching system and lower power consumption. Support multi-threading should give an increase in their respective applications, it can be called a significant step forward. AMD also indicates improvements in the areas of branch prediction and allocation of computing resources that will lead to a smaller prediction error – and this, in turn, will reduce the number of passes and the waste of computing resources. A larger cache operations (Op Cache) here will also be useful.
micro-scheduling unit (Micro-Op Dispatch) has become more widely – from four to six operations. instruction scheduler (Instruction Scheduler) also increased considerably – up to 84 to 48 integer and 96 to 60 floating-point operations. Potentially erroneous instructions can now be removed in pieces of eight instead of four previously before they leave the conveyor. Queues also increased: 192 lifting operations (retire), 72 to load (load) and 44 to write (store).
L1 cache received the write-back function (write back), which provides a more efficient operation of individual nuclei. Cash K2 was also significantly accelerated, and the L3 cache is now working on the same frequency as the fastest CPU Complex core. AMD indicates doubling the bandwidth of caches L1 and L2, L3 cache was up to five times faster.
If you look at a snapshot of CPU Complex Zen chip architecture, then pay attention to the scanty size computing units ALU and FPU in comparison with the other components. Caches, planners and prediction blocks cover about 80% of the area.
AMD made significant progress on improving the L2 cache. Let’s start with the production of crystal, 512 Kbytes are placed on an area of 1.5 mm², although Intel managed to pack a 0.9 mm² only 256 KB. AMD also picked up the L2 cache bandwidth capacity more balanced than Intel, which has lost a lot of space on the interface – this approach is important for the scope of the server, but is not critical for the desktop. So AMD is talking about achieving a better balance.
This advantage AMD notes and L3 cache. The volume is 8 MB, associativity 16-way, the cache can share up to four cores CPU Complex. The frequency of L3 cache corresponds to the fastest in the core CPU Complex. To save energy L3 cache is divided into four areas, the power which is supplied separately to work only the desired area.
Support for SMT (simultaneous multithreading) gives hope to increase the multi-threaded performance. Here, of course, it is meant the ability to perform two threads per core. In the case of one stream to the core, it gets all of the resources. If two streams are performed, it is not required because some resources may be used only together without prioritization. The diagram shows the dedicated and shared resources.
AMD plans to introduce Zen architecture of Naples and processors on the server market. Whether AMD will be able to show themselves well here compared to Intel – an open question. In the case of Naples AMD focuses on one- and two-socket system. , AMD plans to promote Naples on certain markets not to spray in all areas of the server segment. Servers with a single socket will soon replace two-socket systems – although the latter will play a role for some time. However, with Naples AMD plans to capture both markets.
Forrest Norrod (Forrest Norrod) of AMD: “single-socket CPU platform win Due to the transition to more advanced process technologies, and increase the number of transistors on a chip, single-socket servers (1P) can already meet the needs of many of today’s server platforms 2P Great news for the IT market.”
Of course, with 32 cores and 64 threads per SoC AMD is focused on the tasks that will be able to use a large number of cores at the same time. “The number of nuclei is important. In the world of cloud computing opportunity to perform more work due to a greater number of cores and resources allows more effectively provide services to more customers and reduce the TCO. It’s simple,” Forrest said.
It seems that AMD has developed a broad strategy, and of SoC processors are only part of it. In almost all areas of GPU Computing will play an increasingly important role, with video cards Radeon Instinct AMD just planning to present a decision. At Tech Summit in 2016. AMD first showed Naples server with GPU-accelerators. By pure computational performance processors still can not compete with Intel Haswell architectures and Skylake for the server market. For example, the instructions AVX FMA Intel ensures that twice the number of FLOPS instructions per clock cycle, and provides twice the throughput capacity of the cache for the FPU and SIMD units. Using Intel AVX512 instructions give even more benefits. In the field of HPC AMD will not be easy to compete with Intel, so AMD decision to concentrate on certain areas seems quite logical.
“Heterogeneous systems are becoming widespread. GPU and other accelerators that complement the CPU, will form the building blocks of computing. GPU combination, CPU and FPGA accelerators in the data center will support a variety of new areas, including deep and machine learning networks, artificial intelligence, virtual and augmented reality, “Forrest said.
But we should not forget that in the server segment except pure CPU performance are important factors such as memory subsystem, I / O capabilities and security features. For example, certain areas of the workspace must be encrypted – there may be mentioned technologies Secure Memory Encryption (SME) and Secure Encrypted Virtualization (SEV). Both technologies are focused on in the memory data encryption. To support AMD technology uses AES128 encryption with the private key. So an attacker is unlikely to be able to access the data. But SME and SEV are working at different levels of encryption.
To save space and ensure a crystal high performance AMD had to make some sacrifices, concerning the use of servers. At the conference, International Solid State Circuits Conference 2017 (ISSCC) AMD has compared with Intel.
AMD at ISSCC comparing 4-core processor design RYZEN with a 4-core processor Kaby Lake from Intel. At that time, as the Intel chip takes 49 mm², AMD CPU area is 44 mm². At first glance, these 11 percent are unlikely to something can influence, but in fact it is a fundamental difference in the cost of production. Of course, these two CPU not directly comparable due to the differences in architecture and the cache size. Still, AMD managed to place the chip components smaller than it got from Intel.
It is interesting that the 4-core processor based on Intel architecture, modern Kaby Lake. Intel in its processor uses a 256 KB L2 cache and 8MB of L3 cache. On the other hand, AMD uses 512KB cache in core CPU Complex (CCX) and common 8MB third level cache on all four core. L3 cache, AMD is 16 mm², and it is smaller than Intel, with its 19.1 mm². In addition, twice the volume of the second level cache to 1.5 mm² “packed” denser than 0.9 mm² Intel.
But except for the size of the chip, the role played by other factors. Thus, AMD, apparently utilizes 12 metal layers, and Intel – 13 layers. This also reduces the production cost and makes more efficient architecture Zen economically. In the production of Intel processors it uses a proprietary process technology 14FF +, while AMD manufactures its chips at the facilities 14LLP Samsung company GlobalFoundries under license.
Zen architecture can scale performance at frequencies from 1.5 to 4.0 GHz, which allows AMD to use it not only for the production of an effective high-end desktop processors, but also APU or CPU for laptops. That may even be cooled passively. According to Lisa Sue, CEO AMD, the first mobile processors are already being tested and show good results.
Overall, AMD indicates an increase in efficiency of 270% over prior architectures, and 129% are achieved architecture 70% – transition to the process technology FinFET, 40% – Pure Power 31%, and physical design. On the next page we’ll look in more detail Pure Power.
On the report of the glider AMD RYZEN Tech Day the company has disclosed the number of transistors RYZEN 7 processor:. 4.8 billion The data presented in the table above, and published pictures of cores allow us to calculate the size of the crystal. It must be less than 200 mm², according to calculations we have received 195 mm². It is interesting to compare the result with Broadwell-EP processor chip which has an area of 246 mm² and contains 3.1 billion. Transistors.
Thus, the processor RYZEN 7 significantly smaller 8-core Broadwell-EP, but contains significantly more transistors, which is associated with a high packing density of transistors in the production of crystal on GlobalFoundries capacity.
In developing the architecture Zen “from scratch” AMD has been able to give up many blocks associated with previous architectures CPU. On the one hand, this led to an increase in productivity per cycle. With architecture Zen AMD mentions a significant increase in the number of instructions per clock (IPC). The company has received the order of 40% compared with the Excavator, but aiming for 52%. However, on RYZEN Tech Day had already given the other numbers.
The fine print was not mentioned increase + 52% compared to the Piledriver architecture. A comparison with Excavator gain should be + 64%. AMD has defined performance level based on the results of their own tests and Cinebench R15 1t SPECint06, but higher percentages relate solely to SPECint06. The test Cinebench R15 1t AMD + indicates increase of 76% compared with Piledriver and + 58% compared with Excavator. However, these figures are AMD’s, so they should be treated with skepticism. On the other hand, they emphasize the AMD effort that went to the development of Zen architecture.
SenseMI technology consists of Pure Power, Precision Boost, Extended Frequency Range, Neural Net Prediction and Smart Prefetch.
Pure Power is designed to guarantee the preservation of the former level of performance while reducing power consumption. Technology can not be called new, it just has been optimized for Zen architecture. AMD CPU installed on hundreds of crystal temperature sensors, voltage and frequency. In particular, the use of 48 sensors that monitor the voltage supply in different areas. There is also the temperature sensor 20 and the nine so-called droop sensor. They monitor the voltage drop under load. Hundreds of sensors mean that AMD has established not only the fastest sensors, but less speedy models in non-critical areas, which made it possible to keep track of thousands of critical paths.
Sensors are accompanied by another measuring electronics. On the basis of the collected data drawn curve of voltage / frequency – for each processor and multiple temperature ranges. This curve is the voltage / frequency is then sewn into the processor, which will use it in their work. Accordingly, for the value of the frequency will be selected optimum voltage.
Precision Boost allows finer control frequency, the technology works with Pure Power. It is based on the same sensors, and almost identical to the meter electronics, the goal is to obtain a higher frequency with the same power consumption. New infrastructure Infinity Control Fabric is used here. Precision Boost lets you increase the CPU clock speed to the level of up to 25 MHz. Classical state Boost States allow only eight gradations in steps of only 100 MHz. The new architecture is different Zen, available up to 100 gradations, that can be adjusted in increments of 25 MHz. Also, the data processor may change a state of less than 1 ms. It is also possible to change at runtime – previously it was not there.
On RYZEN Tech Day AMD talked a lot about efficiency and how to achieve it. Of course, there are important and voltage, and power distribution. AMD uses the term LDO (linear low drop out), although Intel prefers FIVR (Fully Integrated Voltage Regulator). Recall that FIVR has become one of the major innovations of the first processors “Haswell”. With the integration of Intel voltage stabilization system simplifies power system motherboard, which previously had to apply for a package of five different CPU stress: Vcore, Vgpu, VCCSA, VCCIO, and PLL. In addition, this move will allow Intel to more effectively manage processor power, leading to energy savings.
According to AMD, the LDO came to realize much simpler technology provides increased efficiency up to 95% on standard scenarios, and the voltage of each processor core is exposed independently. The same is possible with FIVR, but power / control circuit there are much larger, and therefore less effective. It will be interesting to look at the test results of energy consumption and compare it with the capacity to calculate the efficiency.
Additional layers in the crystal and other components ensure that the food goes through the LDO to the desired location. Of course, a stable supply voltage via the VRM is still important, but LDO is fundamental system that is located close to the powered components and know their needs best. Therefore, it can best manage power. If RYZEN 7 1800X LDO manages all physical cores individually, giving them needed voltage. This individual power system differs from previous approaches, when the “weak link”, that is the core of the highest voltage to operate at the desired frequency, set all the CPU power. AMD is now able to manage the nuclear power much more subtle, which increases efficiency.
To increase the overclocking potential of AMD added Extended Frequency Range function (XFR). Depending on the cooling system, the processor using this technology can operate at much higher frequencies. Frequencies XFR can be both above and below the values Precision Boost. Technology should provide substantial growth Zen CPU performance when using cooling water or liquid nitrogen.
We will try to explain the work of the different states and XFR P- and Boost States as an example RYZEN 7 1800X. Of the eight P-state is only interested in the top three. P2 / Pmin is a state of inactivity for Windows, the frequency is 2.2 GHz. Of course, the processor can operate at lower frequencies, but P2 / Pmin is the starting strip with which the processor can switch quickly to the faster condition. The processor P1 goes into the state at 3.2 GHz under light load, but it is not complete. Condition P0 corresponds to the base frequency of 3.6 GHz. From this point the operating system refuses to control P-states and passes control to the processor completely. Which begins on
More on AMD Tech Summit in December, AMD introduced the first information about Infinity Fabric, new high-speed interconnect that can deliver information on various processor blocks as quickly as possible. For example, if processors RYZEN Infinity Fabric interconnect is used for communication with memory controllers. Future products will also be used interconnect Infinity Fabric, which will replace the old Fusion Compute Link, used in the APU in 2011 for communication between the CPU and GPU. Infinity Fabric used in the processor RYZEN, but will also be used in future processors Vega, server processors Naples, mobile APU RYZEN and future SoC from AMD.
AMD working on Infinity Fabric four years. Fabric name means a web that indirectly indicates the interconnect structure. According to AMD, Infinity Fabric has a modular design and can be implemented with any degree of complexity. This flexibility allows you to use Infinity Fabric of all new processors and GPU.
Infinity Fabric divided into Control Fabric and Data Fabric. Control Fabric is responsible for managing different parts of the chip (Engine HUB). In Control Fabric-based working technologies such as power management, protection, and security features, reset, initialization and testing. Data Fabric, on the other hand, provides high-speed transfer of data within the architecture. Data Fabric is used for secure connection to the memory. In the case of GPU Vega interconnect chips HBM2 works with speeds of up to 512 GB / s. With mobile chips, which connects DDR4 memory, just enough speed 40-50 GB / s. All this demonstrates the flexibility of implementation of Infinity Fabric.
Infinity Fabric is a part of not only graphics architecture Vega, but also processors and Summit Ridge RYZEN, as well as mobile processors Raven Ridge, which will be released in the second half of 2017 also under RYZEN mark. In the case of architecture Vega Infinity Fabric interconnect is implemented in a full mesh topology. The reason is that the GPU contains thousands of stream processors, which is necessary to provide data and efficient data distribution just is best achieved through a fully meshed topology. As for the processor, they use less complex topology Infinity Fabric, because through this interconnect bind only part of the functional blocks. AMD did not disclose further details, but it is quite reasonable ring topology, which, for example, uses Intel.
With respect to processors RYZEN Infinity Fabric technology allows AMD to increase linearly multi-threaded performance with the number of cores. As shown by AMD internal tests, the 8-core processor shows almost twice the performance compared to the 4-core. Therefore, the bandwidth required for communication and data transfer different components seem quite high.
And fast interconnect is important not so much for desktop processors, as for server CPU based on Zen architecture. In the case of processors Naples we will get 32 cores and 64 threads, AMD plans to introduce the first models this summer, soon we plan to publish the details. But Infinity Fabric – not merely interconnect inside the CPU or the GPU chip. As pointed out by AMD, Infinity Fabric can be used for the connection sockets. Also interconnect is the physical basis for the AMD HyperTransport.
Infinity Fabric is also important for the future of AMD processors, interconnect is specified in the current plans to release processors and GPU. AMD is already working on Zen 2, but planned and 3. With Zen Zen 2, we are unlikely to get a revolutionary change, but AMD will likely number of optimizations of the current platform. As for the GPU, AMD says Navi, but so far no details.
AMD has revealed at this CES basic functions RYZEN chipsets. In addition, a little later it appeared additional information, to which we will return below. Now AMD has opened a full range of chipsets, even though some motherboard manufacturers have already demonstrated their models at CES. Although they did not lead the technical details – of which we are just now and we’ll talk.
Known models B350, A320, A300 and X370 added X300, oriented mini-ITX systems. X370, B350 and A320 support USB 3.1 Gen 2 speeds of up to 10 Gbit / s, which is not at the “junior” models. All processors RYZEN, regardless of performance, offer unlocked multiplier. overclocking potential in this case is limited to the motherboard. Overclocking is possible only on motherboards with chipsets X370, B350 or X300. Profiles CrossFire and SLI are available only for models based on the X370.
Only RYZEN processors offer 16 PCI Express 3.0 lines. On X370 motherboards with two graphics cards will be available at each of the eight lines. The processor supports protocols like NVMe two or four lines and 2x SATA. If the motherboard connects M.2 only two lines, the other two lines can be used in a different way. Furthermore, directly connected to the CPU 4x USB 3.0.
Thus, we can do without the chipset, but it offers additional connectivity slots and interfaces. These include 4x SATA or four lines of PCI Express 3.0. Eight PCI Express 2.0 lines can be used for additional slots, Ethernet, Wi-Fi, Bluetooth or controllers USB 3.1 Gen 2. Chipset X370 is equipped with 2x USB 3.1 Gen2, 6x USB 3.0 and 6x USB 2.0.
Meanwhile, many manufacturers have already presented their motherboards. Below is a gallery with them.
The new 8-core processors manufactured on 14-nm process technology, but the standard clock speeds are very high. Of course, the question arises about the potential of overclocking RYZEN processors. But how we get to own the results of the acceleration, it is necessary to remind once again that we are working with eight cores. Therefore, acceleration should be compared only with the appropriate models of Intel. And not with processors Kaby Lake.
If you look at the characteristics of the XFR and high frequency Boost, AMD RYZEN processor is unlikely to get called overclocker’s dream – at least, it concerns RYZEN 7 1800X. All three CPU models are technically very similar presented today, so it was logical that AMD selected the best crystals for RYZEN 7 1800X, less good – for RYZEN 7 1700X, even worse – for RYZEN July 1700 with a relatively low TDP. Therefore, the overclocking potential is different.
Already on the first demonstrations of the following picture: for air-cooled frequency reasonably expect eight cores from 4.1 to 4.2 GHz with a voltage of 1.45 to 1.55 – it will be a good result. Recall that the standard voltage of 1.1325 V. As for the single core, it can be overclocked to 4.35 GHz.
possible better results with water cooling. On RYZEN Tech Day the first experiments with liquid nitrogen were carried out. frequency of 5,202 MHz with the CPU voltage 1.875 V. was reached Which allowed to beat the previous world record in the Cinebench R15, which now amounts to 2.363 points. However, so far we have received only the individual results, and it was not clear how well the processors will be dispersed in the mass. We should wait a few days, when people begin to share their overclocking results.
Along with the new processors AMD also introduced a new cooler Wraith. Processors RYZEN 7 1700X and RYZEN 7 1800X are only available in the OEM-equipment without coolers, and with RYZEN 7 1700 attached Wraith Stealth. Wraith Max cooler will be available only in finished computers from collectors. In principle, Wraith Spire can also be delivered with a 65-watt processors. But Wraith Stealth is able to dissipate heat much quieter. Wraith Max is declared with a TDP of 95 watts, but it can cope with a lot of heat.
All coolers are using the new fixing mechanism. It is called the Spring Screw Clamping, and, in theory, the cooler is installed easier than with a simple mechanism Spring Screw. All three coolers are equipped with RGB-backlit, which can be controlled via the motherboard. AMD increased the distance between the holes, to give more space for the socket – he has 1331 contacts. Therefore, the old coolers, which are attached directly to the holes, no longer fit. Or, to look for an adapter for them. But cooler AM4 or FM2 +, which used AMD proprietary mounting system will be compatible.
However, many eminent manufacturers of coolers has announced the compatibility of its models with the new processors RYZEN via adapter kits. We published this separate news.
A few words about the TDP thermal envelope, which is often referred to in the article. TDP represents the processor heat and calculated as the TDP = (tCase – tAmbient) / HSF Θca
TCase – is on the heat spreader temperature, which is also used to control the XFR / Boost mechanism. TCase temperature is expressed in degrees Celsius. TAmbient – this temperature which enters from the outside through the cooler, it is also expressed in ° C. HSF Θca value expressed in ° C at the W and corresponds to the performance Clair. For AMD RYZEN 7 1800X formula is as follows:
(60 – 42) / 0.189 = 95.23 W
60 ° C for tCase we took on AMD specifications for RYZEN 7 1800X, as well as the temperature of 42 ° C tAmbient. 0,198 ° C / W indicates AMD for air coolers, with which you can get the maximum performance of the processor. If RYZEN 7 1800X maximum power from the socket is 128 watts. These 128 W limit the automatic overclocking function, such as the XFR and Precision Boost. Of course, as tCase temperature of 60 ° C.
As usual, the CPU overclocking is possible through the BIOS. All motherboards based on chipsets X370, B350 and X300 support overclocking. Interestingly, the CPU multiplier can be changed in increments of 0.25. You can also change the CPU voltage and memory timings. But if you want to change settings through the BIOS does not, we recommend that you take advantage of AMD RYZEN Master Tool utility.
After installing the utility request appears as to whether the user is willing to accept the loss of the guarantee. It should be remembered that in theory the use of tools may damage the CPU. After the first change the CPU goes to OC mode, it will disable features such as Precision Boost and XFR. Altering clock frequency and voltage are activated without rebooting. But to add / remove the number of active cores and memory changes restart delays longer required. In OC mode remains active C-states with low power consumption, but you can disable them.
RYZEN Master Tool Utility is interesting not only the ability to configure directly, bypassing your BIOS, but the rich and monitoring capabilities, with many convenient options are displayed in a graphical representation. It can be seen, for example, a CPU voltage varies depending on load and speed. The voltage is generally from 1.3625 to 1.2 V. For overclocking usually sufficient voltage of 1.35 V, but AMD also indicates that with good air cooling it is possible to raise up to 1.45 V.
Test new CPU was interesting, because we have not just a 10% increase, as it did in the case of test Core i7-7700K. AMD successfully launched on the market a new architecture and new processors to compete with Intel. Increase productivity RYZEN compared to previous AMD processors very decent, and we again got a competitor in the desktop CPU market.
Let’s start with the disadvantages. Platform RYZEN is that “raw”. If you are not afraid to deal with BIOS errors, avoid them and regularly set new version of BIOS, the new AMD processor will be interesting. The problem here is that not only the new architecture CPU, but also a platform. Therefore, if some settings will not be triggered, should not be particularly upset. Yes, and to the output of the new version of the BIOS, you can get used to it every week. If you are not sure, you should wait a month or two. Then the community of enthusiasts accumulate experience, publish manuals, and motherboard manufacturers release a stable BIOS.
In many processor AMD RYZEN 7 X1800 tests showed himself perfectly on Intel Core i7-5960X level or even in the vicinity of the 6950X, but in games it’s not so good, novelty noticeably inferior to the Intel processors. Unfortunately, the game is not as well optimized as multithreaded tests or applications that can share the load on multiple threads. So here the AMD position is not so strong – and we measured the performance of games without a clean background check Antivirus, teamspeak or other utilities. Therefore, we are not sure about the fact that the increase in the number of cores, and will continue to show themselves well in the future. Probably, Intel, and it does not go this path for desktop platforms. It should also be borne in mind that we conducted tests in a relatively low resolutions, to transfer some of the load on the processor with the graphics card. Many gamers probably will anyway.
But RYZEN 7 1800X has been very bad in other areas, and here we begin to talk about the benefits. Performance in multi-core tests is excellent, and the energy efficiency is also at altitude. The platform idle power works even more efficiently than an eight high-end CPU Intel, almost Kaby Lake level. In the coming weeks, motherboard manufacturers are sure to add any optimization, so AMD platform can be called one of the most effective on the market. But 1800X under load consumes more power than the 6950X, with a lower level of productivity. Energy efficiency is worse, but it can still be called normal.
X370 chipset, which is equipped with our motherboard ASUS Crosshair VI Hero, offers completely modern features, you can expect an abundance AM4 motherboards on the market, for all tastes and budgets in the near future. Of course, Intel Z270 platform is on the market longer (including previous versions), it is more polished, but after a few BIOS updates the differences are unlikely to be large.
In the coming weeks we will further explore the topic of overclocking, since the impression is difficult to make in a few days. We also plan to offer guidance on setting up the platform. Two these factors (acceleration and memory) we associate the possibility of further increasing productivity. In addition, you will need to evaluate the performance of the two younger 1700X and 1,700 processors.
A few words about the prices. In Europe, for AMD RYZEN 7 1800X asking from 559 euros, in Russia, he is just beginning to emerge as low as 37.3 thousand. Rubles. If the top model Intel Core i7-7700K Kaby Lake cheaper -. 365 euros, 10-core Core i7-6950X will have to pay 1750 euros.. What is interesting, the motherboard in our test platform ASUS Crosshair VI Hero is worth 270 euros in Europe, at the level of current models of X99, or Z270 with the best equipment. The rest of the system has been configured identically for both AMD and Intel, allowing you to focus on the CPU prices. Similar in price to RYZEN 7 1800X is the Core i7-6850K processor is the Intel, but he has just six cores, even if the frequency just above the Core i7-6950X. In this regard, AMD chose a good price / performance ratio: RYZEN 7 1800X can be called top-end processor that provides good performance for the money. From low-end models, especially RYZEN 7 1700, we expect even better terms.