Connect with us

PC / Laptop

AMD Ryzen 7 3700X: Zen 2 in all its glory

Advertisement
Samsung J7 V just $5 mo. New device payment purchase req'd. Plus, free shipping.

The first processor microarchitecture Zen was released in March 2017. Since it’s only been two and a half years, but today, AMD is updating the lineup of its desktop processors that inherit this microarchitecture for the second time. Moreover, it is not about the formal updates. Let appeared in last year Ryzen second generation and can be calculated by a simple translation of the initial design on rails for a more advanced technological process, but now we are talking about much more significant change. New Ryzen third generation, we are talking about today, is not just advanced technological process, but also significant changes in the topology and microarchitecture.

AMD acts decisively and quickly: time after time she makes strides toward improving their proposals. And the results are not forced himself to wait. The company is steadily increasing its share of the processor market, and Ryzen second generation deservedly reputed to be the best proposals for the mass market on a combination of price and performance. And even the skeptics now believe that AMD has managed to seriously shake up the CPU market and make at least so that announcements of new processors from mediocre updates turned into the main events in the computer industry.

However, now the company wants even more. While Intel continues to torment the 14-nm process technology and microarchitecture Skylake come from 2015, AMD is going to finally seize the initiative. At Ryzen third generation has the mission to demonstrate the technological superiority of AMD and translate it from second place to first. But will the third attempt improved AMD to create the best in recent history processor for desktop systems successful?

We write these lines, when you already know the test results. And I can say for sure, Ryzen in 3000 there is a lot of positive changes that put them head and shoulders above its predecessors. However, challenges remain, which subject the story of the novelties turns out not too simple.

For this reason, the material on testing Ryzen 3000 we broke into two parts. In the first part we talk about the new osmeteria Ryzen 7 3700X, the example which is easiest to analyse design compared Ryzen previous generation Intel Core. The second part, which will be released following the first, will be devoted to testing the 12-core flagship Ryzen 9 3900X with which AMD is going to put in the mass market segment a series of absolute records.

#What you should know about the new microarchitecture Zen 2

If the release of processors Ryzen first and second generation AMD wanted to convey the idea that she finally returned to the top League of developers and manufacturers of x86 processors, today’s announcement Ryzen 3000 brings already a completely different message. Now the company puts before itself a more ambitious goal – to become the leader of the processor market, which offers the fastest, most energy efficient and most technologically advanced chips.

And this task seems impossible. In the past year, AMD managed to build a very solid Foundation from which she is able to confidently start up. Thanks to the cooperation with one of the leading contract manufacturers of semiconductors, Taiwan’s TSMC, the first company in the PC industry shifted the production of its processors for the 7-nm technology, which allowed it to increase the density of the crystals to raise their operating frequencies and, in parallel, to improve energy efficiency. In addition to this, AMD has introduced another innovation and moved to a new mnogoshipovyh (cabletow) the layout of processors, which suggests the Assembly of final products of several semiconductor crystals, which allows to circumvent many manufacturing complexity and significantly reduce the cost of complex multi-core processors.

But Ryzen third generation mark so highly not only because they are able to offer the users a lot of working at high frequency cores for a relatively small amount. Something similar already was in the range of AMD before. But based on previous arm processors, there were many claims with low single-threaded performance with serious delays in the internuclear interaction and the inefficient memory controller. Now, all these deficiencies in varying degrees, should be eliminated. Speaking about the improvement of the performance of new Ryzen in comparison with the predecessors, AMD operates two-digit percentages, and it really seems very serious progress against the background of development in recent years Intel.

However, you need to understand that this major increase in performance Ryzen third generation is largely driven by the low base effect. The new processor microarchitecture is not something fundamentally new: 2 Zen is different from Zen/Zen+ is, in fact, carries with it a set of fixes for the most critical problems of its predecessors. But because of the problems of a different sort were many, and many of them caused considerable damage to the overall efficiency of the microarchitecture, their removal ultimately leads to significant productivity gains.

And yet, to belittle the merits of AMD, we would not like. The ultimate in Zen 2, there have been many positive changes: increased throughput on-chip all major highways, increased download of available processor cores of the computing resources become more amounts of data that the processor can operate locally, as well as significantly increased key performance indicator specific performance microarchitecture – the number of executable per clock instructions (IPC).

Detailed analysis the analysis of the architectural innovations and improvements in Ryzen 3000 we have dedicated a separate article in it about the structure of micro Zen 2 describes in great detail.

Here we only recall the main causes that determine the growth of the notorious figure of the IPC. They need to know at least in order to better understand the results of the tests representatives of the family Ryzen 3000. So this:

  • Increase the width of the block floating point unit (FPU) from 128 to 256 bits. Thanks Zen 2 can execute a 256-bit AVX2 instructions in one go, that is twice as faster than previously.
  • A two-fold increase in cache decoded micro-operations, which should reduce the downtime of the Executive part of the conveyor due to the lack of performance of the decoder of x86 instructions.
  • Significantly improved the prediction of transitions in the mechanism which is now used new TAGE predictor (Tagged geometric) and increased by the amount of buffers for the purposes of branchings of the first and second levels. All of this reduces the error probability of the branch prediction and minimizes the number of situations when the processor is forced to reset the pipeline state from-for incorrectly made predictions of branching code.
  • The appearance of an extra (third) of the block address generation (AGU), which allows the execution units more timely access to necessary data even at high loads.
  • Doubled bus width of the cache memory that also allows you to eliminate bottlenecks when handling the actuators to the data.
  • Double volume cache of the third level, the total amount of which made up 32 MB for each eight-chiplet.
  • Advanced algorithms pre-fetch data, enabling transfer of data from memory into the cache before they are requested during the execution of the program code.
  • Increased the size of the queue schedulers that allowed us to improve the efficiency of the SMT technology.
  • Increased the size of the register file, which gives the processor the ability to process more commands in parallel without any delay.
  • Additional fixes in microarchitecture that help to neutralize the attacks, the Spectre V4 without impacting performance.

To illustrate the micro-to improve the practical examples is quite simple. For this we usually use a simple synthetic benchmark test of AIDA64 utility: they allow you to see how performance has changed in the performance of certain standard algorithms. In the figures below we compare the past generation Ryzen (Pinnacle Ridge) with the current (Mattisse) for example shestnadcatiletnih and eight-core chips running at the same clock frequency of 4.0 GHz. In addition, the chart is placed the results of eight Lake Coffee Refresh, also clocked at 4.0 GHz.

In fact, all these results are very curious. First, they show that for some algorithms, microarchitecture Zen 2 provides almost two-fold performance increase, while in other cases, the performance remained at the same level. Second, they allow to say that in terms of a relatively simple computational algorithms, which are well rasparallelivanija and do not need in active work with external data from memory, microarchitecture Zen 2 has not only grown to the efficiency of the microarchitecture of Intel Skylake, but even surpassed it.

The most impressive progress Matisse shows in those algorithms which use floating point. Specifically, where used instructions AVX2, FMA3, and FMA4. After all, it is their performance in Zen 2 has doubled.

As for integer calculations, with them, no problems there in previous processors Ryzen. Now there was a small change in performance, associated primarily with changes in the caching and decoding of instructions with a decrease of the L1I cache and increase the cache decoded micro-operations. Not to mention the relatively poor result of Matisse in the CPU Photoworxx test. The fact that it is the only benchmark in which among other things plays a role in the performance of the memory subsystem. And with it a new Ryzen really is not as good as a class. But let’s not get ahead of ourselves.

#Bus Infinity Fabric and the speed of internuclear interaction

If we talk about osmeterium and shestiyaderny, the processors Ryzen third generation has retained its traditional basic structure – they are composed of two Quad-core complexes CCX (Core Complex), which are placed in one eight-core processor chip-Deplete CCD (Core Complex Die) and connected inside tire Infinity Fabric. However, the difference from previous processors is that the OCTA core CPU is no longer a single monolithic crystal. The memory controller, PCI Express controller and the SoC elements removed from the CCD chiplet in a separate I/O-chipset manufactured for 12-nm technology at GlobalFoundries companies. This double-crystal arrangement does not affect the communication between cores and L3 cache – everything here stays the same.

In processors with 12 and 16 cores all adds another level of hierarchy – they use the same OCTA core CCD chipley, but a double number. This direct connection with each other CCD chipley do not have. They are connected by bus Infinity Fabric only with I/O-capleton, so all the interaction between nuclei in different chiplet, going through the smart host – I/O chipset.

In the end, it turns out that even in the case of eight cores cores of unequal relative to each other: there is a “close” of the kernel (in one CCX-complex), and is – “distant” (located in different CCX and having the ability to communicate with each other only via Infinity Fabric). In processors based on CCD pair are still “very far” cores physically located in different crystals. For this specific delay when communicating between cores is different depending on, they are United in one CCX or are different. And it is quite disturbing moment: in the processors Ryzen past generations delay encountered when communicating cores from different CCX became quite noticeable and in some cases hindered the performance.

Ryzen in 3000, this issue should have been partially fixed. First, AMD worked with Microsoft and was able to ensure that the scheduler of the operating system now account for the topology of the processor and first loaded the kernel from one CCX-complex, moving to the next CCX, available only when the kernel is already loaded in the previous job. Such a strategy is inherent to the scheduler in the new version of Windows 10 May 2019 Update in relation to the processors Ryzen this allows to reduce the number of calls inter-core bus-Infinity Fabric high delays and focus calculations, if they do not load all the processor cores inside the CPU the smallest unit.

Secondly, the tyre Infinity Fabric Ryzen new generation of markedly accelerated by itself: its width is doubled from 256 to 512 bits. Much it improves the situation? Positive effect it is easy to check that, we did our traditional test delays when transferring data between cores. For comparison, the following are the results of the measurements made, not only for the OCTA-core processor Ryzen third generation, but also for osmeteria last generation (Pinnacle Ridge), and eight Coffee Lake Refresh. All processors during the test was reduced to a single clock frequency of 4.0 GHz, the memory of all CPU operated in the mode DDR4-3466, and that means the tire is Infinity Fabric to compare Ryzen used the same frequency 1733 MHz.

The situation in Ryzen 3000 really has improved markedly. Core belonging to the CCX-complex, is now able to exchange data 25% faster, and the kernel belonging to different CCX are “closer” to each other by a third. Thus, Ryzen 3000, at least if we talk about CPUs with cores not more than eight, problems with high latency inter-core cooperation will be subject to significantly lesser extent. Moreover, the speed of connections between nuclei belonging to the CCX-complex, new members of the family Ryzen has surpassed even Coffee Lake Refresh, which uses a ring bus, which is considered the most successful variant of the connection parts of the processor into a single unit.

The positive impact of high speed Infinity Fabric should be manifested not only when transferring data between cores. It is worth Recalling that each CCX-complex Ryzen processor has its own cache of the third level, and all the large 32 MB L3 cache in eight Ryzen 3000 actually represents two cache to 16 MB. Therefore, treatment using Infinity Fabric occur in the case when the kernel of the same CCX-complex need of data in L3-cache of the second CCX-complex. Therefore, the observed speedup Infinity Fabric should have a positive impact on performance in a wide range of situations, including when working with data.

However, another problem associated with the speed Infinity Fabric remained unresolved: the frequency of this bus continues to be associated with the frequency of the memory controller. Although the new AMD processors and implemented the asynchronous mode of operation Infinity Fabric, the frequency of this bus still can not exceed the frequency at which the memory controller and the DDR4 SDRAM will continue to have a significant impact on the performance of Ryzen 3000.

#Speed cache memory

If we talk about working with the data, then the subsystem of the cache memory in processors Ryzen 3000 has not changed. The cache memory of the first (L1D) and the second level will retain the same size, organisation and latency, and the only novelty is the increased cache of the third level. Due to the transition to the 7-nm technology AMD allowed themselves to significantly increase the transistor budget CCX-complexes, and thus the third level cache has been doubled to 16 MB for every four cores. But even this area, which Ryzen 3000 is a CCX-a set of 7-nm semiconductor crystal is only 31,3 mm2, while in the processors of the previous generation, which are produced at 12-nm technology, CCX-complex is a 60 mm2.

But the increase in L3 cache there was not so much from the generosity of the developers. This is partly a forced measure. The new processors chilenas layout memory controller “drifted away” from the cores, and cache as much data – a method is needed in order to try to reduce the number of situations when processor cores are idle waiting for data from memory. AMD representatives say that in the first place, this should help resolve performance issues in games, but we’ll see.

Now I want to talk about another point: the increase in the volume of the cache memory is always accompanied by an increase in its delay. So what happened this time, but in fairness it should be noted that the increase in latency was quite small, with 38-39 to 41-42 cycles.

In the graphs below, we compared the latency of the cache memory eight cores Ryzen second and third generations, as well as relevant representatives of the family of Intel Core. All processors during the measurement was reduced to a single frequency of 4.0 GHz.

The cache memory of the first and second level Ryzen 3000 in comparison with processors of the previous generation has not changed its key performance parameters. The latency L1 and L2 cache remains at 4 and 12 bars. However, to say that the nearest to the computing cores, the cache memory is not changed, it would be wrong. The first-level cache Ryzen in 3000 is actually faster, because now he is able to serve two 256-bit read and one 256-bit write each cycle, which means increased throughput compared with previous processor family Ryzen in half.

As a result, the speed of L1 and L2 cache Ryzen 3000 are fully comparable with the speed of the cache memory levels lower in the actual mass processors competitor. And the cache memory of the third level in the new Ryzen, though, and increased its latency can still offer lower latency compared to the L3-cache in Intel processors Coffee Lake Refresh. However, we should not lose sight of the fundamentally different algorithms L3 cache in processors from different manufacturers. Zen/Zen+ and Zen 2 cache of the third level is very simple and victimization, and also independent for each CCX-complex. At the same time, consumer Intel platform LGA 1151 implemented more intelligent of inclusive cache, with writeback that is shared between all cores. In other words, the practical efficiency of the L3 cache in AMD and Intel are very different.

However, graphs are given of the latency give some reasons for concern. Namely, the concern is the terminal part of the curve of latency to Matisse, which shows the characteristics of the memory subsystem. As you can see, there are no reasons for optimism: Ryzen third generation were worse than their predecessors and, as a consequence, really lost on memory latency processors competitor. What’s the matter?

#Work with memory

Implemented in Ryzen 3000 Chilena layout was shared between the computing cores and the memory controller. While CCX-complexes with cores and L3 cache are in the 7-nm CCD-chiplet, the memory controller along with a PCI Express controller and elements of the SoC issued in another I/O-chiplet. The connection between chipsetati mounted the processor on a single printed substrate, using tires Infinity Fabric, which means data path from memory to CPU core has an additional stage. Although AMD says that the external bus Infinity Fabric similar high-speed characteristics of the bus that bind CCX-complexes inside the CCD chiplet, all of this somehow had an impact on the delays that occur when accessing the memory.

In other words, when it was discovered that the latency of memory Ryzen 3000 became worse than before, we are not surprised. More interesting: how deteriorated the speed of the memory in the new AMD processors. Clearly the answer to this question indicators Cachemem test from the AIDA64 utility (for correct measurements, all processors have the same frequency of 4.0 GHz in all cases, the systems have dual channel DDR4-3466 SDRAM timings 16-16-16-36-1T).

As shown, worsening memory latency Ryzen in 3000 compared with the previous generation processor is about 11 %. In addition, deteriorated, and the throughput speed of the recording, which shows the memory controller Ryzen 3000, became one and a half times lower than they were before. In other words, miracles do not happen: just as occurred in the mid-2000s, years of moving the memory controller from the chipset into the processor to accelerate work with memory, reverse the separation of the memory controller from cores naturally led to the opposite result.

Moreover, for AMD processors, the increase in memory latency is not just annoying little thing, it is really a very unpleasant moment. Speed of work with memory Ryzen last generation and so much lost to competitor processors. Now, with the release of Ryzen 3000, the situation will only get worse. Although throughput when reading and copying data from Ryzen 3000 and Intel Refresh Coffee Lake remain comparable, the speed of entry and from the point of view of the latency of memory, the new AMD processors are inferior competitors in 1,6-1,8 times.

But not everything is so dramatic. For potential buyers Ryzen third generation there is good news. The most important – in the newer processors used significantly redesigned memory controller, which is not so capricious as its predecessor. This is reflected in the passport characteristics: new Ryzen 3000 has received official support DDR4-3200 SDRAM, which had not formally been proposed. Moreover, the performance of the memory in the mode DDR4-3200 is guaranteed for all pairs of modules, regardless of their organization and the component base.

In addition, if we talk about implemented in a new memory controller capabilities, it is worth mentioning a couple of important things. First, Ryzen 3000 will now be supported by 32 GB modules, which means that in a system based on the new processors can be installed in a total of 128 GB memory. Second, the memory controller supports ECC. However, the ability to use this feature will depend on the motherboard, and, as experience shows, in regular consumer platforms, manufacturers usually do not activate.

However, the main advantages of the new controller becomes clear when its practical use. It is no exaggeration to call it unproblematic: it is omnivorous in respect to the memory modules and much more stable, not requiring tedious selection of timings to achieve stability at a relatively high frequency. At the time, as with processors Ryzen in previous generations the memory modules rarely could run in the modes faster DDR4-3466, with the new controller does not cause the problem and run the memory in higher speed modes. Coupled with the increased volume cache of the third level, this largely offsets the increase in latency of the memory subsystem as a whole.

However, AMD wouldn’t be himself if positive change didn’t come with the list of limitations and reservations. So, despite the possibility of considerable acceleration of memory, the maximum a rational mode of operation of memory Ryzen 3000 supports DDR4-3600. In this case, achieving maximum productivity, a more rapid frequency modes is meaningless from the point of view of performance.

Ryzen 7 3700X c DDR4-3600

Ryzen 7 3700X c DDR4-3600

The reason is that relations between the frequencies of the memory, memory controller and bus Infinity Fabric. They have complicated the lives of fans of AMD before, and will continue to do the same in processors Ryzen 3000, although certain changes have occurred. The most important thing: AMD has been able to decouple the bus frequency Infinity Fabric from the frequency of the memory: they are in the newer processors can vary independently. However, there is an important caveat: frequency Infinity Fabric must be either equal to or lower than the frequency of the memory. This means that the choice of memory modules will continue to have a significant impact on the processor performance in General.

The second caveat concerns the fact that the maximum number Infinity Fabric Ryzen 3000 is 1800 MHz, and at higher values the processor can not function. Also there is a third caveat. It concerns the fact that when using faster memory modules DDR4-3600, the clock generator of the memory controller automatically switches to 2:1, that is, starts to operate at half the frequency.

The memory frequency (mclk) The frequency of the memory controller (uclk) Bus frequency is Infinity Fabric (fclk)
Up to DDR4-3600 To 1800 MHz uclk = mclk fclk = mclk
DDR4-3600 1800 MHz uclk = 1800 MHz fclk = 1800 MHz
After DDR4-3600 Above 1800 MHz uclk = mclk/2 fclk = 1800 MHz

All of this leads to the fact that the use of memory in the faster modes DDR4-3600 no practical sense: when crossing the border into the subsystem memory comprising asynchrony adds additional and very significant delays.

Ryzen 7 3700X c DDR4-3866

Ryzen 7 3700X c DDR4-3866

As you can see by the above screenshot, the latency of the memory in the mode DDR4-3866 is approximately 9 NS is higher than when you select DDR4-3600 with the same settings, timings. To compensate for this increase in delay with a further increase in the frequency of DDR4 SDRAM, if we talk about the usual non-extreme overclocking, almost unreal.

Little hope remains that the frequency at Infinity Fabric in the serial processor under some conditions still to be raised above 1800 MHz, because in theory the motherboard have the appropriate setting with a rich choice of frequency for this bus. In this case, the processor systems Ryzen 3000 can make sense to use and faster than the DDR4-3600 modules.

However, we cross the line in the 1800 MHz to Infinity Fabric failed: select higher values will inevitably lead to complete failure of the test system.

#X570 chipset and compatibility with old boards

Ryzen testing the 3000 we were armed with a platform based on the chipset X570. AMD has prepared this chipset specifically for the release of their processor microarchitecture Zen 2, however, the fee for X570 – dispensable companion to the new Ryzen. Like their predecessors, Ryzen 3000 is compatible with the usual socket Socket AM4, capable of working in boards, as issued during the first and second generation Ryzen.

However, it is not so simple. For the new CPUs in older boards needed their support at the BIOS level, but its implementation is not so smooth for marketing reasons. In fact, Ryzen 3000 will be probably compatible with any Board on the X470 and B450, with all other platforms the situation is at the mercy of motherboard manufacturers. Therefore, support for new processors in any specific motherboards with chipsets X370, B350 A320 and may or may not appear.

The criterion of compatibility is simple: to ensure that she or other fee to work with Ryzen 3000, its BIOS needs to be rebuilt using the libraries AMD AGESA PI Combo_AM4 1.0.0.1 or later. If the motherboard manufacturer has released an update to the firmware Plata Ryzen 3000 will fit.

However, to use now for testing the new processors the old boards, it would not be a good idea. The fact that the motherboard manufacturers put all the efforts on optimizing the BIOS fresh generation Socket AM4 platform and support Ryzen 3000 to older platforms is implemented by a residual principle. It appears that almost all of the available BIOS updates for old boards are based on the code AGESA version 1.0.0.1 or 1.0.0.2, and these versions do not disclose performance Ryzen 3000 in full.

For proper operation of the new CPU and achieve maximum performance in the BIOS code needs to be applied library Combo_AM4 AMD AGESA PI 1.0.0.3, and this condition is currently performed only for a few boards, mostly with the chipset X570. For this reason, the tests were conducted us to the X570 which is due to better optimizations to offer better performance in a pair of Ryzen 3000. However, this situation is temporary: as code updates the BIOS in old motherboards with their performance Ryzen 3000 will have to catch up to the same level, which today provide a fresh platform.

The chipset X570 are no particularly popular in this moment of opportunity in the platform Socket AM4 adds. The main reason why users should pay attention to is the appearance in boards based on PCI Express 4.0. If you use 3000 Ryzen in such fees this interface is supported both graphics PCIe x16 slots and slots M. 2 NVMe drives, but also any other PCIe slots. In addition, charges of new generation usually are equipped with a large number of ports USB 3.1 Gen2: processor and chipset X570 unable to provide up to 12 ports.

Processors Ryzen 3000 have 24 lines PCI Express 4.0. Four lines of that number are utilized to connect to the chipset, and four more lines are given to work with the system NVMe SSD. The remaining 16 lines is an interface with the graphics card.

Chipset X570 has at its disposal 20 PCI Express 4.0, four of which need to communicate with the processor. The remaining 16 lines of the manufacturer of the motherboard may allocate the PCIe, M. 2, or to configure them as additional SATA ports.

At this stage it all seems not so popular, although devices that support PCI Express 4.0 gradually penetrate into the market. So looking interface with twice as high throughput will use the graphics card Radeon 5700 RX and RX 5700XT. In addition, in the near future will begin to appear and SSD based on Phison controller PS5016-E16 (for example, the Gigabyte AORUS Gen4 NVMe SSD or Corsair Force Series MP600) that will also be able to take advantage of the increased bandwidth of the interface.

However, if you are thinking about purchasing motherboards X570, you should keep in mind that this chipset is very hot chip, the dissipation of which lies in the range of 11 to 14 watts under peak loads. Technically it is a reconfigured I/O-chiplet from server processors EPYC Rome, meaning it is based on a 14 nm chip manufactured at the facilities of GlobalFoundries. And so it is not surprising that she, as the processors need active cooling: the vast majority X570-motherboard used chipsety cooler with a fan.

In addition, Socket AM4-boards that use the X570 will be among the flagship platforms. And this means that they will cost a lot to expect that the price of the cheapest boards of the new generation will start from $200-$250.

#Read more about Ryzen 7 3700X

The range of processors Ryzen 3000 (code name Matisse) consists of six versions: two shestiyaderny, two osmeteria, dvenadtsatietazhnogo and shestnadtsatiyaderny processors. For the first trial review we chose the average model Junior OCTA core CPU Ryzen 7 3700X.

This was done primarily because it is easy to pick up competitors – OCTA core processor as a competitor, and among Ryzen past generations. In addition, Ryzen 7 3700X seems to be one of the most popular new products. It is possible to conclude, if you look at the composition of the lineup completely.

Cores/ Threads Base frequency, MHz Turbo frequency MHz L3 cache MB TDP, watts Chipley Price
Ryzen 9 3950X 16/32 3,5 4,7 64 105 2×CCD + I/O $749
Ryzen 9 3900X 12/24 3,8 4,6 64 105 2×CCD + I/O $499
Ryzen 7 3800X 8/16 3,9 4,5 32 105 CCD + I/O $399
Ryzen 7 3700X 8/16 3,6 4,4 32 65 CCD + I/O $329
Ryzen 5 3600X 6/12 3,8 4,4 32 95 CCD + I/O $249
Ryzen 5 3600 6/12 3,6 4,2 32 65 CCD + I/O $199

The attractiveness of Ryzen 3000 is due to four factors: high performance, affordable price, reasonable power consumption and heat and hope for a non-zero overclocking potential.

And indeed, Ryzen 7 3700X – a full eight-core Socket AM4 processor microarchitecture Zen 2, collected from a single 7-nm CCD-chiplet with a complete set of active cores and 12-nm I/O chiplet. It has a slightly lower frequency in comparison with the older osmeterium Ryzen 7 3800X, however, the difference in maximum frequency is only 100 MHz. No other fundamental differences have Ryzen 7 3700X in stock and a full-sized L3 cache for a total volume of 32 MB, and the L2 cache capacity of 512 KB per core.

That as the base frequency for the CPU is specified at 3.6 GHz, it is possible not to pay attention – in reality due to the technology of Precision avtoradgona Boost 2 CPU almost always comes at a much higher speed. For example, when testing in Cinebench R20 with load for different number of cores our instance showed real frequencies in the range from 4.1 to 4.4 GHz, which not only looks good, but exceeds the typical operating frequencies of last year’s flagship, Ryzen 7 2700X.

However, AMD is clearly disingenuous, speaking of energy efficiency Ryzen 7 3700X and relating it to 65-watt thermal package. To understand this, it helps to see how the system behaves with this CPU in the nominal mode in the stress tests, for example, in Prime95.

Raises questions are literally everything. And we should start with high operating temperature, which for our instance Ryzen 7 3700X test in Prime95 29.8 reached 90 degrees at that the AMD considers to be the maximum possible heating of CPU only to 95 degrees. But this picture in our case, there is not even a boxed Wraith Prism, but with much more powerful Noctua NH-U14S.

Of course, the processor crystal, performed on a 7-nm standards, has a very small “contact patch” with the heat-spreading lid, and therefore, cooling Ryzen 3000 really harder than the 12 – and 14-nm processors. However, the 90-degree heat CPU makes me doubt that this CPU under load shows the energy consumption at 53 W warn about all of its internal sensors. The impression is that AMD deliberately and very much lowers rates of consumption so that the processor is automatically overclocked to a higher frequency in the technology of Precision Boost 2, which for 65-watt Ryzen 7 3700X sets the upper threshold of consumption of 88 watts.

About what the real consumption Ryzen 7 3700X, to let the sensor Converter power the motherboard. According to their testimony, on the processor, which supposedly creates electrical load 53 W, current is supplied with a power of 106 watts plus about 15 watts per SoC. The system as a whole at this time shows consumption of the order of 185-190 watts, so that no doubt remains: 65-watt Ryzen 7 3700X under load can consume about twice the declared thermal package. In other words, energy efficiency Ryzen 7 3700X is a lie, slander and provocation.

Of course, this consumption can be attributed to misconfiguration technology Precision Boost 2 in the BIOS specific motherboard or AMD initiated by the most wilful disregard of the part of the heating package, but you need to understand that if the manufacturer and decide to return Ryzen 7 3700X promised energy efficiency, it will inevitably hurt its performance. Otherwise here is not given.

But then Ryzen 7 3700X not to take, it’s a very attractive price. Ryzen 7 3700X – not just the cheapest cosmedent architecture Zen 2, is also the processor with the lowest in the new model range of the specific value of the kernel. In addition, its cost is lower prices Junior OCTA core proposals of the competitor. All of this could easily be done by Ryzen 7 3700X “choice of millions”, despite any of its shortcomings.

#Acceleration

Ryzen 7 3700X – Junior cosmedent in the lineup of new products, and such induction usually means the processor is effective to disperse at least reach the frequency of senior representatives of the model range. Moreover, AMD traditionally did not repair any obstacles overclockers. Multipliers Ryzen in 7 3700X are not blocked and heat-dissipating cover continues to pripiats to the surface of semiconductor crystals, despite the fact that under it their is now two.

However, dispersal is still not about Ryzen 7 3700X. AMD in each new generation has steadily squeezed from the processor frequency potential all the juice, and to date has achieved this perfection. We can say that Ryzen 7 3700X operated close to the limit of their capabilities even in the nominal mode due to the technology of Precision Boost 2, which is clearly implying the observed operating temperatures.

Anyway, the maximum frequency we managed to “squeeze” when manually overclocking Ryzen 7 3700X, was only 4.2 GHz. If you increase the voltage to 1.4 V CPU at this frequency was stable and passed stress tests in Prime95, but temperatures under load increased to 105 degrees, which is hardly a normal operational mode.

The result is of more theoretical value, and to resort to such a crackdown on the practice makes no sense. The performance boost in multi-threaded load will be a few percent despite the fact that at partial load of cores the CPU will work even slower than in the nominal mode.

While AMD offers enthusiasts another way to increase performance – adjustment settings Precision Boost 2 so that the processor yourself out on the higher frequencies as a built-in algorithm avtoradgona. This technology allows you to change the key supporting constant – consumption limits current (PPT) and electric power (TDC and EDC), together with the increase in the upper limit frequency, and can be used for acceleration. However, some noticeable effect by changing these limits in the case of Ryzen 7 3700X us to achieve and failed. Even in the nominal mode Precision Boost 2 controls the frequency Ryzen 7 3700X very aggressive, and the main problem standing in the way of dispersal is not limiting, consumption and currents, and high temperatures.

SOURCE

Advertisement
Samsung J7 V just $5 mo. New device payment purchase req'd. Plus, free shipping.
Code: VZWDEAL. Enter this coupon code at checkout to get $100 discount on Samsung Galaxy Note 8. Includes free shipping. Restrictions may apply. Device payment purchase required.
Continue Reading
Click to comment

Leave a Reply

Your email address will not be published. Required fields are marked *

PC / Laptop

AMD Ryzen 3900X 9: a split personality

Advertisement
Code: GKBADWJ7-5. You must use promocode. There is no minimum order value.

Last week AMD gave the press a computer a nightmare. The company’s desire to release their new 7-nm products certainly in “beautiful” date July 7 (7.07) led to the fact that the preparation for their announcements was carried out in a wild rush and very messy. For example, samples Ryzen 3000 arrived in our lab just five days before the “hour X”, but even during these five days we have been updating the BIOS of the test motherboard based on set logic X570, which not only corrected mistakes and added platform stability, but also seriously affect performance.

Because of this, we were faced with a difficult choice: either to engage in continuous retesting of the new processors at the next firmware versions and abandon any attempt to catch with the publication of the review of the first media micro Zen 2 at the end of information embargo, or for use in publications are not the most current performance data. But in the end was chosen as a compromise: the review was divided into two parts. The first part, which is dedicated to eight Ryzen 7 3700X, you most likely have already read – it appeared on our website in the very moment when the world started selling Ryzen 3000. And this article is the second part, and it will focus on the older 12-core representative of a new family, the processor Ryzen 9 3900X.

This separation allowed us again to recheck all the results on the last (at the current moment) version of firmware for the motherboard and ensure that the first buyers Ryzen 3000 will have at its disposal is the level of performance, referred to in this material.

However, no certainty that this performance level does not undergo regular adjustments in the very near future, we have. The fact that the five different versions of the BIOS of our test motherboard that we have tested for five days of experiments, change in the system is quite fundamental. Them AMD not only in optimizing memory controller, but “tighten up” the variable Boost 2 technology Precision, increasing the real frequency of the processor at the expense of efficiency. It is quite possible that later the company will want to change the operating parameters of their processors again. In other words, today’s tests Ryzen 3000 is only a preliminary acquaintance, which occurs at an early stage in the life cycle of new products.

Perhaps in the circumstances it would be right to wait a few weeks and thoroughly test the new chips, when the situation with them already ustakanilos: motherboard manufacturers will prepare a stable version of firmware, based on the AGESA libraries that have all the essential optimization and AMD will come to a final understanding, a balance between performance and energy efficiency it wants to achieve in the end.

However, to do so would be not too rational. The hype created around the new version of the microarchitecture of authorship and the first AMD desktop processors based on it, is so great that the computer community is ready to swallow any information, even if it is in the nature of rough estimates. Therefore, the review Ryzen 9 3900X still goes on our website after the review Ryzen 7 3700X, despite the clear “dampness” of these products. Later, to put the final point in the subject’s performance of new products, we’ll just test them again.

#Of eight — twelve: topology Ryzen 9 3900X

Together with new microarchitecture Zen 2 AMD introduces into use the new topology. Processors now are not a monolithic crystal, and are assembled from several components. This allows the company under the old ecosystems Socket AM4 to offer a new powerful CPU, with 12 and 16 cores. The budget for such multi-core processors consists of nearly 10 billion transistors, but they are distributed over several semiconductor crystals, similar to the way it was done in the HEDT CPUs Ryzen Threadripper.

However, unlike Threadripper desktop CPUs are not constructed of the same components, they use semiconductor crystals of two different types. First, it is OCTA-core “computing” CCD chipley produced for 7-nm process at TSMC companies. Each chipset combines two CCX-complex, containing four cores and 16 MB cache in the third level. Second, an additional 12-nm I/O-chipset controller memory controller, PCI Express 4.0 and elements of the SoC, which is responsible for input / output.

Given the fact that Ryzen 9 3900X – dwenadzatiperstnuu processor, it is constructed from three semiconductor crystals: two 7-nm CCD-chipsetov, with 3.9 billion transistors with an area of 74 mm2, and 12-nm I/O chiplet area of approximately 125 mm2, consisting of 2.09 billion transistors. Connect chipley tyre Infinity Fabric, the same as that used for communication between CCX-complexes within a single CCD chiplet. And, most importantly, chipley nuclei have no direct link connecting them, and all internuclear interaction, is constructed using the I/O-ciplet, who also plays the role of the switch.

It is easy to calculate that a 12-core CPU Ryzen 9 3900X needs to be active not all the cores available. And it’s true: one computational kernel in each CCX available-complex hardware locked, which makes any of the initial 16-core stock CPU with 12 cores. In this case, limiting the number of cores is partly a forced measure, as 16 cores hard enough to enter in a valid Socket AM4 limits energy consumption. But AMD does not abandon the idea of release and 16-core Socket AM4 product. This project is scheduled for the autumn, it will be released under the name Ryzen 7 3950X, but the manufacturer will have to resort to a special education careful selection of semiconductor crystals.

Chipley approach allows to obtain several advantages. First, it allows to simplify the design and production of processors, but also offers ways for easy scalability of the products. Second, because of the separation of functions chiplets different types of AMD had the opportunity to avoid duplicating nodes, we need in a single number, for example, memory controller or bus controller PCI Express 4.0. Due to the fact that all these blocks are put into a single I/O-chiplet to which the CCD chipley have equal access, a logical processor is a monolithic structure with centralized mechanisms of memory accesses and to external devices. Any hints on the usual users Threadripper NUMA-modes in the AM4 Socket system, all cores in 12 – and 16-core processors have exactly the same access to the array RAM.

Everything said is illustrated by the results of practical measurements. As for memory access, then it is all really OK. Ryzen 9 3900X shows approximately the same latency that OCTA-core processor Ryzen 7 3700X. This means that the connection to the I/O chiplet with the memory controller not one, but two CCD chipsetov does not entail any negative effects. Thanks cipitol the layout of memory for all cores do seem to be a single array with the same delay.

And by the way, please note, Ryzen 9 3900X is not subject to the same problem with lower half-bandwidth of the recording, which we found Ryzen 7 3700X. It turns out that the memory controller I/O coplete Ryzen 3000 just optimized to work with two CCD chipsetati and when you connect one chiplet full performance can not give.

But much more disturbing than the speed of the memory cause a latency of internuclear interaction. It seems that the location of the nuclei in different CCD-chiplet should impose a substantial penalty in data transfer between them. For example, the latency in the exchange of information between cores in different crystals in the processors Threadripper, one and a half times higher than when transfers between different cores CCX-complexes within a single crystal. However, surprisingly, such problems have Ryzen 9 3900X does not exist. Here the internuclear interaction between the nuclei belonging to different CCX-complexes, leads to the same delays regardless of whether we are talking about the CCX in the same or in different CCD-chiplet.

And this is a really serious achievement. Thanks to the centralized scheme of the mutual connection of cores, multi-processor Ryzen 9 3900X from the side looks like a completely solid solution. No additional penalties for mercapital communication do not. And therefore, there is no reason to compare Ryzen 9 3900X with Threadripper. What AMD offers at this time – a full 12-core processor, not the Assembly of the two shestiyaderny in a single housing.

#Read more about the characteristics of Ryzen 9 3900X

The review Ryzen 7 3700X we compare it with the older LGA1151-Intel. It was logical, based on the number of cores and threads: Ryzen 7 3700X – shestnadtsatiletnij and OCTA-core CPU, like Core i9-9900K. In fact, however, AMD continues to follow his principle “we will have more cores for the same money than the competitor” and contrasts the Core i9-9900K completely different your processor. The recommended price of $499 in the new lineup got a 12-core, 24-threaded Ryzen 9 3900X, and it is positioned as alternatives to the five hundred-dollar osmeteria Core i9-9900K.

Cores/ Threads Base frequency, MHz Turbocheetah, MHz L3 cache MB TDP, watts Chipley Price
Ryzen 9 3950X 16/32 3,5 4,7 64 105 2×CCD + I/O $749
Ryzen 9 3900X 12/24 3,8 4,6 64 105 2×CCD + I/O $499
Ryzen 7 3800X 8/16 3,9 4,5 32 105 CCD + I/O $399
Ryzen 7 3700X 8/16 3,6 4,4 32 65 CCD + I/O $329
Ryzen 5 3600X 6/12 3,8 4,4 32 95 CCD + I/O $249
Ryzen 5 3600 6/12 3,6 4,2 32 65 CCD + I/O $199

Already in the simple comparison of the characteristics Ryzen 9 3900X against the Core i9-9900K AMD processor looks very impressive. Time and a half because in addition to the superiority in number of cores and threads Ryzen 9 3900X can offer and huge L3 cache with a total capacity of 64 MB, while in a competing processor, the cache memory of the third level less than four times. In addition, Ryzen 9 3900X can boast official support for DDR4-3200 and accelerated version of PCI Express 4.0.

The only thing Ryzen 9 3900X has not yet been able to surpass Intel CPUs is the clock frequency. They are for 12-core are declared in the range of 3.8-4.6 GHz, although in reality it will strive to get closer to its upper bound due to aggressive technology Precision Boost 2. However, lower frequencies are quite Ryzen 3000 kompensiruet his outstanding IPC (the number of executable per clock instruction): as we have seen, in many cases at the same frequency core 2 Zen work even faster cores Coffee Lake Refresh.

In the end, to find in the camp of Intel suitable opponent for Ryzen 9 3900X with the same number of cores, we will inevitably have to turn to heavy HEDT platform LGA2066. Only in its composition there are 12-core Core i9-9920X, but it is estimated at $1 199, that is 2.4 times more expensive than AMD’s flagship news related to the mass platform Socket AM4. Here, of course, one could argue that the HEDT platform offering Quad-channel memory controller and more PCI Express, however, such possibilities can hardly be attributed to the vital for most users things. And that means Ryzen 9 3900X actually blurs the line between HEDT and Socket AM4. Thanks to this gift from AMD users conventional bulk systems can now get the level of multithreading, which recently was only available to owners of heavy and expensive platforms LGA2066 or Socket TR4.

Ryzen 9 3900X Core i9-9920X Core i9-9900K Threadripper 2920X
Platform Socket AM4 LGA2066 LGA1151v2 Socket TR4
Manufacturing process, nm 7/12 14 14 12
Cores/ threads 12/24 12/24 8/16 12/24
Frequency (nominal/turbo) GHz 3,8/4,6 3,5/4,4 3,6/5,0 3,5/4,3
L3 cache MB 64 19,25 16 32
TDP, watts 105 165 95 180
Memory 2 × DDR4-3200 4 ×DDR4-2666 2 ×DDR4-2666 4 ×DDR4-2933
Lines PCIe 24 ×Gen4 44 ×Gen3 16 ×Gen3 64 ×Gen3
Price $499 $1 199 $488 $625

Let’s look at the characteristics of Ryzen 9 3900X more. Among the processors in the consumer segment this processor looks like an alien from another world. So it turns out due to the increased number of cores and huge cache, and these features are acquired thanks to the introduction of kiplenge design. In other words, it is legitimate to say that Ryzen 9 3900X – a vivid demonstration of the correctness of the course taken by AMD. The company’s engineers “flick of the wrist” collected from the two CCD-chipsetov mass processor to catch up with that in the baseline characteristics of the Intel in the near future obviously will not be able to.

 

Another feature Ryzen 9 3900X – relatively good frequency. Typically processors with more cores, lower clock frequency, but in this case the opposite happened. And nominal, and maximum frequency Ryzen 9 3900X superior cosmedent Ryzen 7 3700X, and even ahead of eight Ryzen 7 3800X terms of the maximum frequency. This means that Ryzen 9 3900X will be as good in multithreaded applications or when a more simple load on a limited number of cores. That is, the processor must be universal, not a niche solution for demanding work loads, this is technology Precision Boost 2 in aggressiveness which Ryzen 3000 we have already seen.

For example, look how are real working frequency Ryzen 9 3900X in Cinebench R20 load on different number of cores. The minimum frequency observed at full load CPU rendering, start from 4,025 GHz.

And for example, when OCTA core load, the processor is capable of operating at a frequency of 4.1 GHz, that is at least not slower than Ryzen 7 3700X. In the case of downloading a small number of cores Ryzen 9 3900X comes on frequencies above a 4.3-4.4 GHz.

When meeting with Ryzen 7 3700X we drew attention to its high operating temperatures. Ryzen 9 3900X, in which nuclei and a half times more, was also affected by this issue, and even more.

The stress load is received by the utility Prime95 29.8, raises the temperature of the CPU in the nominal mode up to 95 degrees (with the cooler Noctua NH-U14S) and the Converter power the motherboard reports the CPU consumption of about 190 watts of electricity.

It looks somewhat normal, and we were able to get unofficial confirmation that in the latest versions of the BIOS that came for X570-motherboard before the announcement of Ryzen 3000, technology of Precision Boost 2 is configured with a strong overestimation of the allowable limits for the consumption and heat dissipation that automatically provides an additional increase of operating frequencies. However, apparently, the initiator of this “overclocking out of the box” is the AMD, and it is unclear whether this mode will now become standard, whether it is imposed only on the output time of the first wave of reviews to get more positive reviews in the press, and will be removed.

#Acceleration

About acceleration Ryzen 9 3900X not much to tell. Here everything is done to us. Technology parameters Precision Boost 2 12-core unscrewed to the limit, and this CPU works at the maximum possible under dynamic acceleration frequencies without any additional configuration.

If to speak about static acceleration, the power voltage setting To 1.2, it was possible to achieve stable operation at a frequency of 4.0 GHz. Unfortunately, the processor is strongly heated, and it almost completely ties the hands of overclocking.

It is doubtful that in this static acceleration there is some practical sense. Technology Precision Boost 2 even at 100% load on CPU provides a slightly lower frequency, but when loaded by the operation of a part of cores it is able to provide for Ryzen 9 3900X noticeably better performance.

SOURCE

Advertisement
YYD2-821660510629CA Lenovo ThinkPad T430 Notebook PC - Intel Core Dual Core i5-3320M 2.6GHz CPU, 8GB RAM, 320GB HDD, HD Graphics 4000, 14" HD 1366x768 LED AG, DVDRW, Win 10 Pro 64-bit, Refurbished - YYD2-821660510629CA
Code: FASTWIN35. You must use promocode. The promotion applies to Fast SSD Windows VPS & Dedicated Servers with 24/7 Heroic Support. The promotion is available for all customers of the store.
Continue Reading

PC / Laptop

Microarchitecture Zen 2: that’s why we are waiting for Ryzen 3000

Advertisement
Code: VZWDEAL. Enter this coupon code at checkout to get $100 discount on Samsung Galaxy Note 8. Includes free shipping. Restrictions may apply. Device payment purchase required.

In two weeks we, apparently, expecting a miracle. Such a conclusion can be made, if to summarize all the assumptions expressed by the user in anticipation of the upcoming announcement of Ryzen processors of the third generation. But even the most bold statements about that in the second half of the year in the market of processors for PC we will see a change of leader (in performance), has not been entirely groundless. At the beginning of the year, at CES 2019, AMD promised that its next-generation processors will increase specific performance (at constant frequency) at least 15 %. And now we have learned that this will make a noticeable increase in clock frequencies, dramatically increasing the number of cores and reduced heat dissipation.

Each of these promises separately, it seems at least very brave. But all at once?! However, all this is possible. Held in the framework of the exhibition E3 2019 special event Next Horizon, AMD explained in detail how it happened that the microarchitecture Zen 2, which originally was supposed to be a trivial translation of Zen on the rails 7-nm process technology, could be a real breakthrough, having the chance to turn the whole CPU market.

Since the release of the first processor microarchitecture Zen took a little over two years. During this time, AMD has managed to release the intermediate generation microarchitecture, Zen+. However, we saw almost no improvement. The essence of the past the updates were actually reduced to a transition from 14-nm to 12-nm production technology, and only. New microarchitecture Zen 2 meeting which awaits us in July, again implies a change in the manufacturing process — from 12 nm to 7 nm with a simultaneous change of production-contractor: now CPU, the company will produce not GlobalFoundries, and TSMC. But that’s not all: along with the process technology changes dramatically and a lot of other things.

To understand how Ryzen 3000 will be unlike their predecessors, just look at any photo of these processors with the removed heat-dissipating cover. One glance is enough to understand that AMD are moving away from the use of monolithic semiconductor crystal. The kernel they are distributed across multiple semiconductor crystals – capleton, some chiplet will be submitted and all the controllers I / o. It should be added that simultaneously with the introduction of radical changes in the construction of AMD reworked the internal structure of the cores and made sure to address the major bottlenecks of previous CPU microarchitecture Zen and Zen+.

In addition, with the advent of Ryzen 3000 changes will affect the whole ecosystem in which to operate such processors. Compatibility of new products with traditional Socket AM4 with it will remain, but completely all their benefits may be felt only in the newer motherboards that will support PCI Express 4.0.

All the numerous improvements and optimizations made in the new generation Zen 2, clearly deserve more than a mere enumeration. Therefore, following the event, AMD’s Next Horizon, which is able to be the representative of our website, we decided to prepare a separate detailed material and elaborate on why the Zen 2 is really cool.

#7 nm technology is the key to everything

The goals set by the company AMD while working on a new microarchitecture Zen 2, was quite evident. The main objective was to improve the performance of processors for both the desktop and server segment, with the compulsory preservation of continuity and compatibility with existing platforms. In other words, it was about the future scalability of existing CPU families Ryzen and EPYC comprehensive and improving their consumer qualities.

A solid Foundation for the design of a Zen 2 was to adapt the new process. In the transition from 14 – to 12-nm standards, which occurred in April last year, processors Ryzen only slightly won the clock frequency and was able fairly to slightly increase its specific capacity. But a new process technology with a resolution of 7 nm was supposed to catalyze a much more significant progress in improving the whole set of consumer characteristics. Due to the fact that long-standing manufacturing partner AMD, GlobalFoundries, abandoned the development of the 7-nm technology, the chip maker had to shift to cooperation with TSMC. And in the end, AMD has made the right decision. Speaking in favor of this number: the basic CPU building block — Quad-complex CCX (Core Complex) with L3-cache 8 GB MB — the production of 12-nm GlobalFoundries, had an area of 60 mm2. A similar complex Zen 2 enhanced with four cores and twice more capacious, 16-megabyte L3-cache, manufactured on TSMC for 7-nm process, occupies almost half the area of 31,3 mm2.

The total CPU crystal (chiplet) Zen 2, as before, is formed of two CCX. That is, it contains eight cores and a cache memory of the third level capacity of 32 MB. The total area of such crystal is only 74 mm2, which is significantly less than 213 mm2, which is the processor chip design, Zen/Zen+, for example, the same Ryzen 7 2700X. Such a noticeable gain in density of transistors has opened the doors for developers AMD opportunities for the improvement of the microarchitecture, which could be carried out without any substantial damage for the cost of the new processors.

At the beginning of this year, AMD announced that the microarchitecture Zen 2 will provide a 15 percent performance advantage compared to the Zen+ due to one only micro-improvements, that is, at the same clock frequency. However, a lot of advantages given and new advanced semiconductor process. For example, at identical energy consumption for Zen 2 promised at least 1.25 times higher performance than predecessors, and with the same performance, the new processors should be almost twice cheaper. Furthermore, AMD does not hesitate even to say that in some situations the advantage of the new processors Zen 2 will be more than 75% compared with the previous Zen+ of the same class and more than 45 % compared to the equivalent solutions of a competitor.

Of course, all these calculations have yet to be be tested independent tests and reviews, which will be released July 7. In the framework of the event AMD is actively operated indicators Cinebench R20, which suggests that if you compare the Zen 2 and Intel with the same number of cores, the AMD wins as single-threaded and multi-threaded performance and power consumption and price.

Just one example: according to AMD, senior Ryzen 3800X OCTA core 7 with a price of $400 is very close to a 500 dollar eight-core Core i9-9900K in single-threaded and multithreaded rendering, but its power consumption is comparable with the consumption of Core i7-9700K.

#Kernel Zen 2: “tik” and “tak” at the same time

According to the original plan, microarchitecture Zen 2 was supposed to be a simple transference of the old Zen design to new process technology. But later, analyzing the weaknesses of its first generation of processors Zen and Zen+, AMD engineers decided to bend and the underlying microarchitecture. And I must say, this plan apparently worked perfectly. Despite the fact that in Zen 2, there are no drastic alterations, the increase in IPC (average number of executed per cycle instructions) by 15 % — a perfect illustration of the fact that everything was done correctly.

At the same time, you need to understand that Zen 2 — microarchitecture very similar to the original Zen/Zen+. All the basic elements of a processor core remained unchanged, and the alterations relate only to improve the efficiency of existing function blocks. Accordingly, the internal configuration of the kernel has not changed: it is able to decode up to four instructions and execute up to six instructions per clock. In addition, remained unchanged support SMT technology: each core Zen 2 can execute two threads simultaneously.

What has changed? As usually happens when working on the refinement of existing microarchitectures, the first place of application of forces engineers unit fetching instructions and predicting transitions. However, the changes are not very obvious, because the basis of this unit continues to lie “neural” algorithm based on the perceptron. Although in General this scheme does not give very impressive results when working with a buffer the objectives of the branch of the first level it provides good energy efficiency, therefore, hence, to abandon it and just added more multi-stage statistical mechanism TAGE (Tagged geometric) working with buffer purposes of branching the second level.

At the same time was increased and the size of the buffer goals of the branch. Table of first level in Zen 2 includes 512 entries instead of 256, and the second level – 7K instead of 4K records. With regard to the zero level, the corresponding buffer, as before, includes 16 entries, but an array of addresses of indirect transitions has expanded to 1K entries. In other words, the new microarchitecture transitions are predicted clearly better than the original Zen/Zen+. This means that situation when the processor needs to completely reset the Executive pipeline because of the incorrectly predicted transition will happen much less frequently.

Another improvement is the Zen 2 was the fact that AMD decided to significantly reallocate resources to caching instructions. The micro-operation cache that stores already-decoded x86 instructions has been doubled to 4096 entries. This classic statement cache of the first level, which keeps the team prior to their decoding, on the contrary, decreased. While the earlier volume was 64 KB, with 4-channel associative, Zen 2, he was cut to 32 Kbytes, while increasing the degree of associativity to 8.

Simulations of AMD, demonstrated that such changes have a positive impact on performance. And judging by what happened to the growth of the IPC, it really is. Interestingly, as a result of changes in the size of the cache memory, the Zen 2 become the processor with the largest cache micro-operations. For example, in the Skylake microarchitecture this cache is 1.5 To operations, while at the Sunny Cove Intel engineers just extended it to 2.25 To operations.

Changes in the input part of the Executive pipeline does not entail any significant changes in the organization of the work of planners. As before, the Zen decoder 2 is able to supply four instructions per clock cycle together with the cache micro-operation, which can do up to eight related instructions, they fill a queue of micro-operations from which the user choose two schedulers: one for integer operations and another for operations with floating point numbers. In this case the integer scheduler can submit to the execution of six micro-operations per clock, and veselinovsky – four.

But a noticeable change in the microarchitecture occurred at the stage of execution of instructions. If to speak about the execution of integer instructions, here — bargain to increase the size of the buffers (like the scheduler, so the register file and reorder buffer) by about 10-15 % — there are additional generating block addresses (AGU). In sum, it means that the number of Executive ports in the Zen 2 has increased from six to seven: four ports for arithmetical and logical operations (ALU) and three port – operations for generating addresses (AGU). As a result, the microarchitecture Zen 2 can initiate two 256-bit reads and one 256-bit write operation every clock cycle. Last version of micro was, for obvious reasons, is limited to only two such operations per clock, and only a width of 128 bits.

But more importantly, in Zen 2, AMD has doubled the throughput of the block floating point operations. He was now fully 256-bit, which means the possibility of direct enforcement AVX2 instructions. In the original architecture of Zen/Zen+ such teams, working with 256-bit registers, before executing crashed onto a pair of 128-bit instructions and were processed in two steps, therefore, from Zen 2 you can expect a doubling in the pace of work with AVX2 code. The execution units in the FPU remained old. There are two devices for the operations of addition and two of multiplication that gives Zen 2 the ability to simultaneously execute two 256-bit FMA-team. There is a very useful ability, a new microarchitecture to initiate a 256-bit forwarding operations data: as a result, the performance of the AVX2 code can occur without any delay. Moreover, in Zen 2 AMD was able to ensure that the processing of AVX2 instructions can be carried out without any reduction in clock frequency, as is the case in Intel.

Along the way, AMD reported that it was able to increase the speed of the multiplications of floating point numbers from four to three cycles. Ultimately, this also contributes to increase the performance of processors with a new microarchitecture.

As follows from the above, microarchitecture Zen 2 was a bit “wider” Zen in the sense of ability of parallel execution of instructions. But at the same time she became “wider” in the sense of working with data. Although the subsystem of the cache memory, a data not structurally changed, she got a bus with more capacity, which allow to obtain the necessary data, without delaying the implementation of AVX2 commands. More specifically, it means that L1 data cache retained size 32 KB per core 8 way set associativity, and the L2 cache, as before, has a volume of 512 KB per core 8 way set associativity, but now the cache can handle two 256-bit reads and one 256-bit write operation per clock cycle at level L1, and one 256-bit read and write per clock cycle at the L2 level. The latency of the cache memory has not changed and is 4 cycles for L1 and 12 clock cycles for L2.

Despite the immutability of the structure of the cache memory, the Zen 2 was improved L2 TLB (buffer address translation). In the first generation of processors Zen the size of this table was 1.5 K, now it increased to 2K, and the latency at the same below. But most importantly, now the L2 TLB supports page 1 Gbyte, which in previous versions of the microarchitecture had not been implemented.

Another notable change in the Zen 2 was the doubling of the size of the cache memory of the third level. In new processors of its size is not 8, but 16 MB on each Quad CCX. So AMD tried to compensate for the dismemberment of the processor into several independent crystals. Developers Zen 2 believe that the growth in the volume of L3 cache will reduce the amount of data transfer between the chipset core and capleton with the memory controller. Maybe so, but do not forget that the increase of the cache memory is almost always accompanied by an increased latency. And she had L3 cache in the Zen 2 is really increased to 40 cycles, while in Zen processors L3 cache had a latency of approximately 5 cycles below.

#From the core to the CCX and CCD, and further to the CPU

We have already mentioned that the design of the processors Ryzen 3000 is markedly different from what was arranged all the past Ryzen. However, CCX-complexes collected from the cores 2 Zen exactly the same as before. In one unit CCX merged 4 cores and 16 MB of shared cache in the third level.

CCX pair within the 7-nm semiconductor crystal and generates CPU chipset, received the acronym CCD (Core Complex Die). In addition to cores and caches, in CCD-chiplet also includes a bus controller Infinity Fabric, by which it shall be ensured connection CCD required for any Ryzen 3000 capleton IO.

In coplete input / output (I/O) processors generation Zen 2 are the so-called extra-nuclear components as well as elements of the North bridge and SoC. In it, among other things, the memory controller and bus controller PCI Express 4.0. Also in the I/O coplete implemented and two of the tires Infinity Fabric required for connection with the CCD chipsetati.

Depending on what processor family Ryzen 3000 it is, it can consist of either two or three chipsetov. In processors with eight cores and less used one CCD-ciple and one I/O-chiplet. In processors with cores more than eight CCD-chipsetov becomes two. However, you need to understand that the processor is still a single entity. Due to the fact that in any Ryzen 3000 memory controller is in I/O-Deplete and he is only one, either kernel can smoothly access any of its areas: no-NUMA configurations, which spoiled the lives of the owners of processors Threadripper, in the case of Zen 2 will not.

It is worth mentioning that the Zen 2 is not the first attempt to go to layout multi-chip processors. Previously, manufacturers had resorted to such an approach. For example, was based on two semiconductor crystal Quad-core Core 2 Quad, and before the same technique was used in the creation of a dual core Pentium D. But then the producers still went on to have a monolithic structure processors, as it was more effective as the number of cores and transferring the processor components of the North bridge. However, the new Ryzen 3000, which includes two or three ciplet, – not a step back. On the contrary, it is the transition to the next level because of AMD new generation of processors is not a simple extensive, increasing the number of cores by adding additional crystals and uses a much more intelligent approach, introducing the use of chipley with different features and combining them into a single entity the specialized high-speed bus Infinity Fabric.

The advantages of using a multi-chip layout is quite obvious. First and foremost, it allows you to reduce costs. Production of chipsetov having a relatively small die area, much simpler than the production of large monolithic processor. Smaller crystals not only allow you to obtain a higher yield of chips, but also more effective placed on a circular semiconductor substrate, which further reduces the amount of waste. In the end, it Chipita configuration allowed AMD to create a very complex processors Ryzen 3000 is relatively inexpensive, despite the fact that their production is organized at the facilities of TSMC’s most advanced and new to industry process technology with the norms of 7 nm.

The distribution functions of the processor in various capleton allowed AMD to save money and even in one aspect. New process technology was not necessarily to use in the manufacture of all parts of the processors. “Thin” advanced standards are important to processor cores, because they directly affect the frequency potential and energy consumption, but there is no need to use them to craft more simple chiplet responsible for the functions of input / output. The I/O chiplet Ryzen in 3000 is made the old-fashioned way – in factories GlobalFoundries for 12-nm process used in the manufacture of Ryzen processors of the second generation.

However, you need to keep in mind that Chipita design gives rise to certain difficulties. For example, in modern processors, very high demands are placed on how connect and interact with each other in different parts of the CPU. To implement such a bus when mnogoshipovyh the layout is a bit trickier. However, this problem was solved by engineers of AMD. Processors Ryzen first and second generations, although they were based on monolithic kernel, used for connection the CCX and the memory controller, North bridge and elements of the SoC specialized tyre Infinity Fabric. New Ryzen 3000 processors used the second version of this bus: it is responsible for data transfer between all chipsetati.

Frankly, to how to operate the Infinity Fabric, had previously been well-founded claims: it is not always able to provide the required level of performance when interacting processor cores with L3 cache and memory controller. In processors Ryzen 3000 AMD has tried to correct the main shortcomings of the Infinity Fabric.

First, this bus has been expanded twice: now its width is 512 bits, which means a doubling of the capacity and the transfer of 32 bytes per clock in each direction. The developers claim that this step they went in the first place because of the emergence of Ryzen in 3000 support PCI Express 4.0, but it is obvious that a more productive bus, which connects all key components are CPU, will play a positive role in many other cases.

Secondly, Infinity Fabric is now “unleashed” with the memory controller frequency. Earlier the frequency of this bus was synchronized with the memory clock, which on the one hand, leads to a strong dependence of the performance of processors Ryzen speed assigned to the DDR4 SDRAM, and the other – let overclocking memory above 3466-3600 MHz. Now the tyre Infinity Fabric will be able to work with the memory controller not only synchronous, but less than half relative to the frequency with the use of divider 2:1. It is — at least theoretically — means a much greater freedom in the choice of speed of memory, though AMD continues to insist that the synchronous mode to Infinity Fabric will still provide better performance, and more optimal with Ryzen 3000 to use modules DDR4-3600 with low timings.

However, it is already known that the memory in Socket AM4-systems, processors Ryzen 3000, will do much to disperse.

For example, AMD showed working memory module in the mode DDR4-5100 in the system built on Socket AM4-motherboard MSI MEG X570 Godlike.

SOURCE

Advertisement
Samsung J7 V just $5 mo. New device payment purchase req'd. Plus, free shipping.
Click this link to get rebate on your Care Pack. Restrictions may apply, see website for details.
Continue Reading

PC / Laptop

Which processor is needed for GeForce RTX 2060 1660 and GeForce GTX Ti: Ryzen vs AMD Intel Core

Advertisement
Code: VZWDEAL. Enter this coupon code at checkout to get $100 discount on Samsung Galaxy Note 8. Includes free shipping. Restrictions may apply. Device payment purchase required.

In this article we will talk about what is often called the potential of the GPU, or, to speak more correctly, on processorepiemonte in games. On our website, and in other sources, you can find a lot of reviews of processors and graphics cards that explore the relative performance of different types of CPU and GPU. However, the question of what types of processors and graphics better suited to each other, often remains unanswered. Meanwhile, the buyers are forced to look for the answer on it: having a fixed amount, they should optimally allocate it between the processor and the graphics card, and here on this subject there is a clear shortage of some ready and well-reasoned prescriptions.

It is worth Recalling that during the reign on the market with Quad core processor was a popular theory with the conditional name “the potential revealed”, in which it was assumed that for each card there is a ceiling in processor performance, above which rise does not make sense – all the same frame rates in games then have will not grow. The explanation of this effect was based on the assumption that at some point the card stops to display the time frames with the speed with which they are preparing for her CPU, and at this point you can stop chasing faster CPU – better performance to achieve still does not work.

But later in the manifest fallacy of this theory has convinced many of her followers. The fact that the CPU and GPU in modern games are responsible for different stages of the training image. The CPU handles the reaction of the player, recalculation of internal game logic and behavior of the secondary characters, as well as modeling of the whole game environment and physics of the environment. In addition, one of the most important tasks of the CPU in most games is also flow control and transmission of necessary information to the GPU. The GPU is thus engaged exclusively on visualization based on the data received from the CPU the data it builds a picture, which then appears on the screen. Obviously, this is largely the successive stages of one process, so ultimately the frame rate in any case should affect the power of the graphics card and CPU.

However, this does not mean that by increasing the CPU power, you can compensate for a weak GPU, and Vice versa. The optimal ratio between performance Central and graphics processor exists precisely because these components meet in the games for different things. Therefore, in order to obtain adequate performance of the system and deal with sudden FPS dips, if you have any specific game situations, with its design should seek to maximize the characteristics of both GPU and CPU. The limit here in most cases is allocated for the purchase of the equipment budget, which makes selection of suitable processors and graphics cards in the optimization problem with multiple criteria.

Proponents of the theory of “potential” can bring here the argument that in practice at some point, gaming performance with increasing CPU power to grow almost stops. But, if you do not take into account any exorbitant cases with the installation of ultra-high resolution in which the performance of the SRI securely masked behind a lack of GPU performance, it will be connected not with the fact that above a turn of the performance of the processor becomes quite important. In reality, the fact is that “CPU power” is too abstract and collective term, which many people associate with a model number or price of the chip. In fact, it consists of many factors such as number of cores, clock speed, amount and speed of the cache memory, the speed of the memory controller and so on. More powerful processors better less powerful only for some part of these parameters, while from the point of view of other characteristics of any improvement when driving on the lineup up may not be observed. The result is some of the features in the various families of processors can become a bottleneck, which does not allow them to show higher performance in games, despite the fact that the other options are much better.

A good illustration for the given thesis can be the situation with the gaming performance of the members of the family Ryzen. Even the most senior and the most expensive models of AMD processors almost always give significantly lower frame rate in comparison with what provide offers from Intel, and to fix it until AMD could not increase the frequency or increase number of cores or roomy L3-cache. Performance Ryzen, obviously, depends on some other characteristics such as the speed of the memory subsystem, latency, or inter-core interaction in single-threaded performance. In other words, if the power increase of processors does not increase the number of FPS in games, it’s just talking about what we felt the bottleneck of the architecture and not ran into a limit that is predefined by the system-installed graphics card.

Unfortunately, all these arguments systematically lead us to the conclusion that the selection of a balanced combination of CPU and video card is a very trivial task. After all, no universal table that describes the performance of all possible combinations of configurations of CPU and GPU does not exist in nature. While there was such a fanatical researchers, not frightened would be a huge number of options to be included in such a comparison. And we also do not belong to them. Therefore, under the present experimental material, we will reply only to private question about which processor is better to choose NVIDIA’s latest generation mid – RTX 2060 and GeForce GeForce GTX Ti 1660. And then, if the results of this study will be useful to our audience, we will conduct similar tests and some other common graphics cards.

#Why GeForce RTX 2060 1660 and GeForce GTX Ti

Video card average level is one of the most popular options for use in modern gaming systems. Performance is enough for to run modern games with maximum quality settings in the standard Full HD resolution, which, according to statistics collected by the digital distribution system Steam, choose two-thirds of the total number of players.

It is therefore not surprising that, according to the same statistics, the most popular discrete graphics card is still the GeForce GTX 1060 – a mid-level solution from the previous generation of NVIDIA accelerators. Now replaced Pascal Turing come, and if you start from the positioning, it is 2060 and RTX GeForce GeForce GTX Ti is 1660 should gradually take the place of the most common graphics cards.

It is quite natural that now GeForce RTX 2060 1660 and GeForce GTX Ti are the three accelerators with the fastest growing shares. Only last month the user base of these cards increased by almost 40 %. In absolute terms, the share of owners GeForce RTX 2060 1660 and the GeForce GTX Ti is still not too noticeable, but the background of how a growing number of systems with other graphics cards of the family of Turing, no question remains that these GPUs will soon become the most popular choice of the gamers.

It should be noted that the positions of the GeForce RTX 2060 in this case look a bit better, although this card and a quarter as much. However, buyers apparently willing to pay $70 more for the possession of a special RT-kernel, allows the graphics hardware to accelerate algorithms for ray tracing.

#Optimal processor for GeForce RTX 2060 1660 and GeForce GTX Ti: I advise usually

The lack of detailed and thorough tests of processorepiemonte popular graphics cards has led to the fact that buyers have developed a number of empirical approaches to how to relate spending on basic components of the game system. The most popular rule of this kind suggests that the price of the video card and the price of the processor should be as two to one, that is, in a balanced system the graphics card should be roughly twice as expensive as the CPU.

If you apply this rule to video cards, which we are going to talk about today, that is GeForce RTX 2060 1660 and GeForce GTX Ti with manufacturer’s recommended price of $349 and $279, respectively, it turns out that they need to look for processors that stand about $175 and $140. In the price list Intel for these amounts are available Core i5-and Core i3 9400-9300 and AMD, with a similar budget, you can choose considerably cheapened recently Ryzen Ryzen and 5 2600 5 2500X.

Almost the same recommendations formulated in his “Computer of the month” and our resident expert, Sergey Plotnikov. For several months in optimal Assembly 3DNews uses the GeForce RTX 2060 the system with which it is proposed to install processors Ryzen 5 2600X or Core i5-9400F.

In other words, almost all estimations indicate that for medium level graphics cards need a six-core processor, but not necessarily senior – will go and the easier option. However, convincing illustrations, how fair are the assumptions still did not exist. That’s why we undertook detailed tests GeForce RTX 2060 1660 and GeForce GTX Ti in combination with different processors AMD and Intel. Try to use a practical approach and understand what the CPU is the best choice for medium gaming configurations.

#Description of test systems and test methodologies

To study processorepiemonte GeForce RTX 2060 1660 and GeForce GTX Ti was chosen graphic cards in the performance of the NVIDIA (Founders Edition) and Gigabyte (1660 GTX Ti OC). The tests were conducted in two resolutions relevant to these GPU – 1080p and 2560p. Paired with these video cards we tried to test the widest possible range of processors for platforms LGA 1151v2 and Socket AM4. Therefore, tests were selected for major and minor modifications within each class of CPU that made it possible to gather very heterogeneous and representative company of a variety of Core ninth-generation and second-generation Ryzen with the number of cores from two to eight nominal frequencies from 2.9 GHz to 4.0 GHz and the amount of L3 cache from 4 to 16 MB.

In the end, the list involved in the testing of components were as follows:

  • Processors:
    • AMD Ryzen 7 2700X (Pinnacle Ridge, 8 cores + SMT, 3,7-4,3 GHz, 16 MB L3);
    • AMD Ryzen 5 2600X (Pinnacle Ridge, 6 cores + SMT, the 3.6-4.2 GHz, 16 MB L3);
    • AMD Ryzen 5 2500X (Pinnacle Ridge, 4 cores + SMT, 3,6-4,0 GHz, 8 MB L3);
    • AMD Ryzen 3 2300X (Pinnacle Ridge, 4 cores, 3.5-4.0 GHz, 8 MB L3);
    • Intel Core i9-9900K (Refresh Coffee Lake, 8 cores + HT, 3,6-5.0 GHz, 16 MB L3);
    • Intel Core i7-9700K (Refresh Coffee Lake, 8 cores, the 3.6-4.9 GHz, 12 MB L3);
    • Intel Core i5-9600K (Lake Coffee Refresh, 6 cores, 3.7 V-4.6 GHz, 9 MB L3);
    • Intel Core i5-9400F (Lake Coffee Refresh, 6 cores, a 2.9-to 4.1 GHz, 9 MB L3);
    • Intel Core i3-9350KF (Lake Coffee Refresh, 4 cores, 4.0 to 4.6 GHz, 8 MB L3);
    • Intel Core i3-8100 (Coffee Lake, 4 cores, 3.6 GHz, 6 MB L3);
    • Pentium Gold G5600 (Coffee Lake, 2 cores + HT, 3.9 GHz, 4 MB L3).
  • CPU cooler: Noctua NH-U14S.
  • Motherboard:
    • ASRock X470 Taichi (Socket AM4, AMD X470);
    • ASRock Z390 Taichi (LGA1151v2, Intel Z390).
    • Memory: 2 x 8 GB DDR4-SDRAM 3466, 16-16-16-36 (G. Skill Trident Z RGB F4-3466C16D-16GTZR).
  • Graphics card:
    • NVIDIA GeForce RTX 2060 (TU106, 1365/14000 MHz, 6 GB GDDR6 192-bit);
    • Gigabyte GeForce GTX Ti OC 1660 (TU116, 1500/12000 MHz, 6 GB GDDR6 192-bit).
  • Disk subsystem: Samsung 960 PRO 1TB (MZ-V6P1T0BW).
  • Power supply: Thermaltake Toughpower DPS G RGB Titanium 1000W (80 Plus Titanium, 1000 watts).

All Intel CPUs were tested with enabled Multi-Core Enhancements, that is, without any restrictions on energy consumption. We are aware that this mode is somewhat at odds with the specifications of Intel, but most users Willy-nilly uses the processors that way. The fact is that without exception, all motherboard manufacturers will activate the Multi-Core Enhancements by default, and some even hide the settings to disable it, and any shift in this trend is not expected.

Testing was performed in the operating system Microsoft Windows 10 Enterprise (v1809) Build 17763.503 installed patches against the vulnerabilities of Meltdown and the Spectre, and using the following set of drivers:

  • AMD Chipset Driver 19.10.0429;
  • Intel Chipset Driver 10.1.1.45;
  • Intel Management Engine Interface Driver 11.7.0.1017;
  • NVIDIA GeForce 430.64 Driver.

To check gaming performance of the platforms used the following games and settings:

  • Assassin’s Creed Odyssey. The 1920 × 1080 Resolution: Graphics Quality = Ultra High. The Resolution Of 2560 × 1440: Graphics Quality = Ultra High.
  • Ashes of the Singularity. Resolution 1920 × 1080: DirectX 12, Quality Profile = Extreme. The Resolution Of 2560 × 1440: DirectX 12, Quality Profile = Extreme.
  • Civilization VI: the Gathering Storm. Resolution 1920 × 1080: DirectX 12, MSAA = 4x, Performance Impact = Ultra, Memory Impact = Ultra. The resolution of 2560 × 1440: DirectX 12, MSAA = 4x, Performance Impact = Ultra, Memory Impact = Ultra.
  • Far Cry New Dawn. The 1920 × 1080 resolution: Graphics Quality = Ultra, HD Textures = On, Anti-Aliasing = TAA, Motion Blur = On. The resolution of 2560 × 1440: Graphics Quality = Ultra, HD Textures = On, Anti-Aliasing = TAA, Motion Blur = On.
  • Grand Theft Auto V. the Resolution of 1920 × 1080: DirectX Version = DirectX 11, FXAA = Off MSAA = x4, NVIDIA TXAA = Off, Population Density = Maximum Population Variety = Maximum Distance Scaling = Maximum Texture Quality = Very High, Shader Quality = Very High, Shadow Quality = Very High Reflection Quality = Ultra, Reflection MSAA = x4, Water Quality = Very High Particles Quality = Very High, Grass Quality = Ultra, Soft Shadow = Softest, Post FX = Ultra In-Game Depth Of Field Effects = On Anisotropic Filtering = x16 Ambient Occlusion = High, Tessellation = Very High, Long Shadows = On, High Resolution Shadows = On, High Detail Streaming While Flying = On, Extended Distance Scaling = Maximum Extended Shadows Distance = Maximum. The resolution of 2560 × 1440: DirectX Version = DirectX 11, FXAA = Off MSAA = x4, NVIDIA TXAA = Off, Population Density = Maximum Population Variety = Maximum Distance Scaling = Maximum Texture Quality = Very High, Shader Quality = Very High, Shadow Quality = Very High Reflection Quality = Ultra, Reflection MSAA = x4, Water Quality = Very High Particles Quality = Very High, Grass Quality = Ultra, Soft Shadow = Softest, Post FX = Ultra In-Game Depth Of Field Effects = On Anisotropic Filtering = x16 Ambient Occlusion = High, Tessellation = Very High, Long Shadows = On, High Resolution Shadows = On, High Detail Streaming While Flying = On, Extended Distance Scaling = Maximum Extended Shadows Distance = Maximum.
  • Hitman 2. Resolution 1920 × 1080: DirectX 12, Super Sampling = 1.0, Level of Detail = Ultra, Anti-Aliasing = FXAA, Texture Quality = High Texture Filter = Anisotropic 16x, SSAO = On, Shadow Maps = Ultra Shadow Resolution = High. The resolution of 2560 × 1440: DirectX 12, Super Sampling = 1.0, Level of Detail = Ultra, Anti-Aliasing = FXAA, Texture Quality = High Texture Filter = Anisotropic 16x, SSAO = On, Shadow Maps = Ultra Shadow Resolution = High.
  • Kingdom Come: Deliverance. Resolution 1920 × 1080: Overall Image Quality = Ultra High. The Resolution Of 2560 × 1440: Overall Image Quality = Ultra High.
  • Metro Exodus. Resolution 1920 × 1080: DirectX 12, Quality = Ultra, Texture Filtering = AF 16X, Motion Blur = Normal, Tesselation = Full, Advanced PhysX = Off, Hairworks = Off, Ray Trace = Off, DLSS = Off. The resolution of 2560 × 1440: DirectX 12, Quality = Ultra, Texture Filtering = AF 16X, Motion Blur = Normal, Tesselation = Full, Advanced PhysX = Off, Hairworks = Off, Ray Trace = Off, DLSS = Off.
  • Shadow of the Tomb Raider. Resolution 1920 × 1080: DirectX12, Preset = Highest, Anti-Aliasing = TAA. The resolution of 2560 × 1440: DirectX12, Preset = Highest, Anti-Aliasing = TAA.
  • The Witcher 3: Wild Hunt. The 1920 × 1080 Resolution: Graphics Preset = Ultra Postprocessing Preset = High. The Resolution Of 2560 × 1440: Graphics Preset = Ultra Postprocessing Preset = High.
  • Total War: Warhammer II. Resolution 1920 × 1080: DirectX 12, Quality = Ultra. The Resolution Of 2560 × 1440: DirectX 12, Quality = Ultra.
  • Watch Dogs 2. The 1920 × 1080 resolution: Field of View = 70°, Pixel Density = 1.00, Graphics Quality = Ultra, Extra Details = 100%. The resolution of 2560 × 1440: Field of View = 70°, Pixel Density = 1.00, Graphics Quality = Ultra, Extra Details = 100%.
  • World War Z 1920 × 1080: DirectX 11, Visual Quality Preset = Ultra. The Resolution Of 2560 × 1440: DirectX11, Visual Quality Preset = Ultra.

In all gaming tests as the outcome is the average number of frames per second and the 0.01-quantile (first percentile) for the values of FPS. The use of 0.01-quantile instead of the minimum FPS due to the desire to clear the results from the random bursts of productivity that was triggered not directly associated with the operation of the main components of the platform causes.

SOURCE

Advertisement
Code: VZWDEAL. Enter this coupon code at checkout to get $100 discount on Samsung Galaxy Note 8. Includes free shipping. Restrictions may apply. Device payment purchase required.
Promo code not required. The deal applies to a specific group of goods. The deal is available for all customers of the e-store. The deal is available without amount restrictions
Continue Reading

Deals

Advertisement
Code: FASTMWP50. You must use promocode. The promotion applies to Managed Word Press. The promotion is available for all customers of the store.
SoundPath Accessories
You love your iPad so much. Is there anything that could make you love it more? We bet there is: this 8-Piece iPad Accessory Kit & Organizer Case.æGet the most out of your iPad with this super-handy 8-piece kit that will bring out the best in your beloved tablet.æ

Trending