We tested and compared the Ryzen 7 3700X vs Ryzen 7 2700X in terms of Performance, Heating and Price. Below you will find the results of this Battle and the in-depth tests of each Processor.
Also we compared the the Ryzen 9 3900X versus Ryzen 5 2600X, which tests you will find below, too.
Performance Winner: AMDs Ryzen 7 3700X and Ryzen 9 3900X
- Best for 4K, VR Gaming
- Similar Performance to Intel, but cheaper
- Great Single and Multi Thread Performance
- Too expensive if you just need a Desktop for simple Games
It’s finally time: AMD blows to the big attack. The Zen 2 architecture apparently puts AMD in the situation to finally catch up with its competitor Intel. With Zen, Zen+ and the AM4 platform, AMD has created the basis for success over two years. Now they want to reap the laurels. How the Ryzen 7 3700X with eight cores and the Ryzen 9 3900X with twelve cores will beat the competition from Intel, we’ll clarify in our detailed review.
AMD has long-term plans – this should be known by now. The Zen microarchitecture was not a one-shot, but rather several generations were created from the beginning, which had different goals in combination with the expected manufacturing technologies. When Zen 2 was projected in 2015, the goal was already to achieve 7 nm production. At that time, however, it was not quite clear whether this would work. Zen 2 was planned as a server architecture with a low clock rate and was also designed accordingly. And this is exactly what we will see with the second generation EPYC processors and up to 64 cores from late summer.
At the end of 2015 AMD has seen that the processors and chips can be clocked much higher. So again it was decided that Zen 2 can also be used for the Ryzen desktop processors. From this the new Ryzen processors have been developed. AMD will not reveal whether there was an alternative plan and what it looked like.
For the Zen 2 architecture, AMD set itself some targets to help achieve an IPC plus of 8 to 10%. However, this would probably not have enabled AMD to catch up with Intel, even though this was not foreseeable in 2015. In the end, we ended up with an average of 15%, for the following reasons:
- the front end is now more balanced.
- the Micro Op Cache has been enlarged – many instructions that have already been processed can be kept in the Op Cache and reused
- an IPC plus of 15% means first of all that power consumption is increasing – the industry prediction has been improved to reduce overhead
Since the Next Horizon Tech Day in early June is known: There will be a 16-core Ryzen processor. It will be called Ryzen 9 3950X, has a basic clock rate of 3.5 GHz and can be booted to 4.7 GHz. It is also the fastest Ryzen processor when it comes to the highest boost clock. The thermal design power of this model is also 105 W. The Ryzen 9 3950X will be available in September. All other models, as well as the corresponding mainboards, will be available from next week.
Today we are testing the Ryzen 9 3900X with 12 cores. This clocks a bit higher at 3.8 GHz in the basic clock rate, but comes to “only” 4.6 GHz when booted. The TDP here is also 105 W. The Ryzen 7 3800X is more likely to be seen in the mainstream. It offers eight cores that operate at 3.9 or 4.5 GHz. Since only one Valhalla die is used here, the L3 cache is halved to 32 MB. The TDP is still at 105 W.
The Ryzen 7 3700X is a bit more economical – we’re testing it today. The clock rates are 3.6 and 4.4 GHz respectively, and the cache size is identical to the Ryzen 7 3800X. AMD offers this model with a reduced TDP of 65 W. The Ryzen 5 3600X is the first model with six cores operating at 3.8 or 4.4 GHz. The L2 cache is reduced to 3 MB due to only six active cores. The L3 cache remains identical at 32 MB. AMD gives the Ryzen 5 3600X a little more leeway in power consumption and sets it at 95 W.
The Ryzen 5 3600 is the entry-level model for the time being, but it also offers six cores, which clock slightly lower at 3.67 and 4.2 GHz. The TDP is reduced to 65 W. Later there will probably be Ryzen 3 variants. At the moment, only the six models just mentioned are official, five of which are to start immediately.
All processors use the AM4 platform, or the socket of the same name. AMD has revealed all the details on the Computex. Among other things, the third-generation Ryzen processors offer 24 PCI Express 4.0 lanes. Four of these are reserved for connecting the chipset. This leaves 16+4 for connecting the graphics card and other applications like SSDs. DDR4-3200 memory is now supported, but this generation should also benefit significantly from the higher memory clock rate.
The chipset itself is manufactured in 14 nm and has the same design as the I/O die of the processors. So AMD can also manufacture a chip that is used multiple times. However, the I/O die of the processor is manufactured in 12 nm and represents the master in communication, so to speak. The chipset is an I/O die from the 14 nm production. Components like the memory controllers are simply not used here.
A CCD (Core Complex Die) based on Zen 2 comes to 74 mm² with 3.9 billion transistors alone. In addition, there is still the IOD (I/O die) with 125 mm² and 2.09 billion transistors. For a Ryzen 5 or Ryzen 7 processor with six or eight cores, we arrive at a total size of 199 mm² with 5.99 billion transistors. The dies based on Zen 2 are thus a little smaller (CCD + IOD) combined, but have more than a billion additional transistors to offer. In the case of the Ryzen 9 models with 12 or 16 cores, we are even talking about 273 mm² and 9.89 billion transistors.
This can of course be taken a little further. An EPYC processor with eight CCDs and one IOD (assuming the IOD is identical for the server processors) can achieve 717 mm² and 33.29 billion transistors. For comparison: A Cascade Lake SP processor with 28 cores comes to 694 mm² and should contain about 8 billion transistors.
Ryzen and RAM
AMD’s Zen architecture, or Ryzen (Threadripper) processors, are in some ways extremely dependent on the memory used. Memory optimizations with regard to clock and timing can be extremely beneficial to the processors. On the other hand, the MCM designs of the Ryzen Threadripper processors have problems with memory accesses across the memory controllers of remote dies.
For the Ryzen processors with six and eight cores, AMD provides a CCD (Core Complex Die) and an I/O Die. An interconnect between the two chips is created via the Infinity Fabric. The data exchange takes place with 32 byte/cycle in both directions. The Ryzen processors with 12 or 16 cores use two CCDs. However, these are also connected to the I/O die via an interconnect and each of them has a 32 byte/cycle capacity.
The third generation Ryzen processors use the second generation of Infinity Fabric. This was already the backbone of the interconnect infrastructure in the Multi-Chip Module Processors (MCM). With the second generation, the bus has been doubled from 256 to 512 bits. Queries via the Infinity Fabric were accelerated in many areas.
The memory clock (mclk), the clock of the memory controller (uclk) and the clock of the Infinity Fabric (fclk) are coupled together via a fixed ratio up to a memory clock of DDR4-3600. If DDR4-3200 is selected, the clock of these three units is 1,600 MHz. If a memory clock is set via DDR4-3600, the three components operate in 2:1 mode. The Infinity Fabric clock is then always 1,800 MHz, but can be adjusted manually. The latencies increase by about 9 ns in 2:1 mode. For DDR4-4400 this means a memory clock of 2,200 MHz, the Infinity Fabric runs at 1,800 MHz and the memory controller operates at 1,100 MHz. The clock of the Infinity Fabric can be manually adjusted in 33 MHz steps, as already mentioned.
The third generation Ryzen processors support DDR4-3200 with a total capacity of 128 GB. ECC is supported, but AMD does not qualify an AM4 platform for it.
The measurements were made with the DDR4 memory at 2.667, 2.933 and 3.200 MHz at timings of CL17. The latencies and performance, however, benefit above all from the interaction of clock and timings. Accordingly, we took a closer look at this for the new processors.
Ryzen Master Software
New to the Ryzen Master are the settings at the top of the screenshot:
- Package Power Tracking (PPT): The PPT is expressed as a percentage and therefore the user can increase the power supplied to the socket in percentage terms. If Limit is displayed here, this describes the limit with activated PBO. Higher values can be set manually.
- Thermal Design Current (TDC): Each motherboard allows a maximum current flow to the socket via its VRMs. The TDC is given in percent. If limit is displayed here, this describes the limit with activated PBO. Higher values can be set manually.
- Electrical Design Current (EDC): EDC describes the peak values for the maximum current. If Limit is displayed here, this describes the limit with activated PBO. Higher values can be set manually.
Below you can find an overview of the cores . In the case of Ryzen 7 3700X, eight nuclei are shown here. A gray star indicates the fastest core of each CCX cluster. A gray circle indicates the second fastest. A golden star indicates the fastest nucleus of the entire chipset.
Besides overclocking individual cores, it is now also possible to overclock each CCX cluster individually. The CCX clusters and cores are now graphically assigned to the respective CCD.
In overclocking, the respective options such as Precision Boost Overclocking and Auto-Overclocking can then be selected. The clock of the individual cores is displayed as a number, but also graphically.
The Ryzen Master software offers not only numerous settings for the Ryzen processors, but also all the settings for the memory. Any user can let off steam and find the fastest settings.
In overclocking, the respective options such as Precision Boost Overclocking and Auto-Overclocking can then be selected. The clock of the individual cores is displayed as a number, but also graphically.
The Ryzen Master software offers not only numerous settings for the Ryzen processors, but also all the settings for the memory. Any user can let off steam and find the fastest settings.
The Ryzen threadripper processors have the advantage of a four-channel memory interface and clearly deliver the best memory bandwidth. But the Ryzen 9 3900X is also fast with read and write rates of almost 50 GB/s for a dual-channel memory interface. For the Ryzen 7 3700X, we measured surprisingly low values for reading data from memory. We don’t know the reason for this. Actually, the Ryzen 9 3900X and Ryzen 7 3700X should be about the same speed.
The latencies for memory accesses are about 75 ns. The current Intel processors can reach about 50 to 60 ns. AMD has reduced the internal latencies, but the memory interface and memory still have higher delays than the competition.
The Ryzen 7 3700X beats Intel’s Core i9-9900K in calculating pi and the Ryzen 9 3900X can play out its 12 cores here.
DigiCortex doesn’t necessarily benefit from many cores, so the Ryzen 3700X is a bit faster than the Ryzen 9 3900X – Intel’s Core i9-9900K doesn’t stand a chance against either of the new AMD models.
The three render benchmarks all behave quite similarly. The Ryzen 9 3900X benefits from its 12 fast cores. But Intel also has a few advantages with the Core i9-9900K, respectively the existing AVX512 support, and takes the lead over the Ryzen 7 3700X with its top model.
With its Ryzen 9 3900X, AMD is packing a proud twelve cores onto its AM4 socket, which is more than two years old. Starting in September, the Ryzen 9 3950X will even have 16 cores. Our first overclocking attempts show that you can quickly reach one or the other limit. Our first two Ryzen CPUs can hardly be overclocked at all, in fact they can only be tweaked: In our test, the new Ryzen CPUs seem to be running at their limits.
In the case of the AMD Ryzen 7 3700X, we were able to boot at 4.5 GHz on all eight cores, but the system didn’t run stable for long. Only when we reduced the clock to 4.3 GHz did the system run cleanly. Compared to the specified boost, this is 100 MHz less than AMD stated, but the 4,300 MHz on all cores are constantly present, which ultimately accelerates the performance significantly, especially in multi-core benchmarks. In Cinebench R20 the performance increases by about 6%. The Ryzen 7 3700X, which is accelerated in this way, increases considerably, especially in the power consumption. The hunger for power almost approaches that of its big brother and increases by almost 50%. This also shows that the new Ryzen CPUs are already running at their limit.
With the Ryzen 9 3900X, the result is even clearer: Here we could not achieve stable operation beyond the basic clock rate on all twelve cores. Even with 3.8 GHz on all cores, the system was not stable over time. The results are probably also the reason why the automatic overclocking via the Ryzen master tools only works to a limited extent, or rather stops after only a few seconds with the clock increase. In view of the 7 nm technology and the already very high basic voltage of about 1.475 V, we have refrained from increasing the voltage. A new BIOS update or even a second X570 mainboard could not improve our first attempts at overclocking with Ryzen 3000 either.
All in all, it remains to be said: The new Ryzen CPUs are already working at their limit and can only be slightly accelerated.
AMD currently succeeds with the Ryzen processors what seems to be difficult to achieve with graphics cards for several generations. They have caught up with their competitors in almost every respect and are beating them in many areas. However, the multi-core dominance of the first and second generation will be extended to other areas with the third generation of Ryzen processors. AMD is also on par with its competitors in terms of single-threaded or IPC performance thanks to the Zen 2 architecture.
According to AMD’s own statements, this was the plan for Zen 2, and the goals were exceeded. For the Zen 2 architecture, the problem areas of the Zen architecture were touched. The Micro Op Cache was increased, the same applies to the L3 Cache. Both should ensure a higher and more efficient utilization of the CPU pipeline. We see the result of this in the good results in the single-threaded benchmarks. By raising the memory clock to the current JEDEC standard DDR4-3200 plus a simple overclocking to DDR4-3600 this bottleneck is also loosened.
The chip design and manufacturing in 7 nm puts AMD in the position to be technologically one step ahead. This can be seen among other things in the power consumption. But also the processor clock reaches with 4.7 GHz in boost a much higher value than AMD itself would have expected. The production of many small CCDs has great economic advantages for AMD. A further advantage is the flexibility of the design. The IOD does the rest, reduces the dependencies on memory latencies and brings support for PCI-Express 4.0. In addition to the higher transfer rates for large data volumes, PCI-Express 4.0 still has to prove its advantages in the end customer area.
The Ryzen 9 3900X is currently unrivaled in the desktop class. 12 cores and 24 threads are unbeatable in applications where the cores put their power to the road. Even the eight fast cores of the Core i9-9900K don’t help there. It even gets headwind from the Ryzen 7 3700X, which also offers eight cores and can process 16 threads simultaneously. Due to the performance increase in architecture and manufacturing, the AMD Ryzen 7 3700X and Intel Core i9-9900K are now equal in terms of performance. Depending on the benchmark, the pendulum swings sometimes in the direction of the Ryzen 7 3700X and sometimes in the direction of the Core i9-9900K, but we haven’t seen such an equal race for a long time. The Ryzen 7 3800X should be even a bit better thanks to the higher TDP and beat the Core i9-9900K on the whole length.
However, the Ryzen 7 3700X and Ryzen 9 3900X can’t quite keep up in games yet. Intel is also clearly on Intel’s heels here, but it still doesn’t look quite as good as in the synthetic benchmarks. We tested in 1080p and 1440p. The higher the resolution, the sooner the GPU limit comes into play and the closer the processors move together.
Probably due to the 7nm manufacturing, the Ryzen 7 3700X has a clear advantage in terms of power consumption. Where the Core i9-9900K consumes almost 190 W under full load, the Ryzen 7 3700X manages with about 95 W. However, we take results here during a run in Cinebench – more load can hardly be demanded from a processor. The Core i9-9900K consumes a bit less in games, but AMD has the consumption advantage on its side. The Ryzen 9 3900X allows itself considerably more, but also offers considerably more multi-threaded performance.
The memory overclocking offers a lot of fun and potential. It doesn’t even have to be particularly fast memory to get the Ryzen processors out of the reserve. DDR4-3600 is already enough to achieve a significant performance boost. But if you want, you can let off steam with the memory. In addition to the clock, the timing also plays an important role.
The processor’s overclocking, however, is currently still a bit cumbersome. It remains to be seen whether this situation will change. At 4.3 GHz on all cores (more was not possible with our sample), the Ryzen 7 3700X offers a decent performance boost. The Ryzen 9 3900X was a bit bitchy, though. We’ll have to wait and see if the processor itself or the motherboard is the problem. The Ryzen 9 3950X was already presented with extremely high clock rates, so the platform and the design of 2x CCD can’t really be the problem.
Not only in terms of performance AMD puts its competitors under pressure, also in terms of price. But we already know this situation from previous generations of Ryzen. However, AMD was able to score points back then mainly because of the price, now they have a few more points on their side. The Ryzen 3700X, as a typical gaming processor, is probably the most frequently bought and costs $294. The Intel Core i9-9900K currently goes over the counter for just under $500.
The Ryzen 9 3900X is only over 100 Dollars more expensive at $432 and offers an additional four cores. If you have appropriate applications for this large number of cores, you get extremely multi-threaded performance for little money. An Intel Core i9-9920X with 12 cores costs over $1000 and also requires an LGA2066 board. AMD has remained faithful to the AM4 platform for over three generations now, even if there are minor compatibility limitations. Basically, however, what AMD has built up here on one platform is currently unparalleled.
Better Price: AMD Ryzen 7 2700X and Ryzen 5 2600X
- High Multi-Thread Performance
- Hight Tact levels
- Better Price than Ryzen 9 3900X
- High Energy use
About a year after the launch of the Ryzen processors, AMD brought the second generation of processors to market. This is intended to work faster and more efficiently and further increase the pressure on Intel. However, the motto is evolution instead of revolution, and instead of major changes, it is mainly fine-tuning. Whether the first two models, Ryzen 7 2700X and Ryzen 5 2600X, are really the “ultimate processors for gamers, creative people and enthusiasts”, the test shows.
Outwardly, AMD recently expressed satisfaction with the development of its own processor business. Probably not without reason, as various indications show. Various benchmarks showed an increasing number of runs with Ryzen and Co. in the past months, Amazon lists nine AMD processors among the 20 best-selling models. Among them are the two Raven Ridge APUs Ryzen 3 2200G and Ryzen 5 2400G, which were convincing in the test in large parts.
In retrospect, these turn out to be not only a combination of Zen CPU and Vega GPU, but also an intermediate step towards the second Ryzen generation. Not without reason AMD spoke at their presentation of Zen 1.5, as they anticipated small corrections to the architecture and the support of faster memory.
The second Ryzen generation, now also known as Ryzen 2000, ties in almost seamlessly with the APUs, but like its direct predecessors is designed as a classic CPU. There is therefore no integrated graphics unit.
4 Different Ryzen Processors in Second Generation
AMD does not change the previous division according to the current status, as the four first processors of the second Ryzen generation show. There are two six-core and two eight-core processors – the former belonging to the Ryzen 5 family and the latter to the Ryzen 7 family.
The top model, at least for the time being, will be the Ryzen 7 2700X with eight cores and 16 threads and a basic and boost clock of 3.7 and 4.3 GHz. The direct competitor is likely to be Intel’s Core i7-8700K, while the in-house predecessor is the Ryzen 7 1800X. The Ryzen 7 2700 with the same core and thread count, but clock rates reduced to 3.2 and 4.1 GHz, ranks directly below it. Accordingly, it should keep the Intel Core i7-8700 in check and replace the Ryzen 7 1700.
Those who want to spend less or get by with a lower performance will probably belong to the target group of the Ryzen 5 2600X and Ryzen 5 2600. Both offer six cores and 12 threads, the clock rates are between 3.6 and 4.2 and 3.4 to 3.9 GHz.
Initially, the Ryzen 7 2700X and Ryzen 5 2600X models were available for the test. The two non-X processors, Ryzen 7 2700 and Ryzen 5 2600, will follow at a later date.
The Zen + Architecture
For the second processor generation, AMD uses the Zen architecture, which is now two years old, but has been revised in certain areas. Consequently, the company speaks of Zen+. In addition, there is a slightly smaller production process, which means that the second Ryzen generation is neither a real tick nor a real tock.
Since the launch of Raven Ridge, it was clear that AMD would focus on faster memory. Like the APUs, the new Ryzen processors support DDR4-2933, the predecessors were limited to DDDR4-2666. In practice, this is only of limited relevance, since faster memory can of course be used, which can have certain advantages – more about this later. It must be noted, however, that certain combinations are not or only partially possible. As usual, it is not only the set RAM clock, but also the structure of the memory bars (single or dual rank) and the number of bars that are used. No surprise: The dual channel connection is still in place.
AMD hardly reveals any details about the differences between Zen (Summit Ridge)/Zen 1.5 (Raven Ridge) and Zen+ (Pinnacle Ridge). There is only talk of “targeted improvements”, whose goal is however named: The reduction of latencies. The L1 cache latency is said to have been reduced by about 13%, the L2 cache by 34% and the L3 cache by 16%. In addition, the DRAM response time has been reduced by 11 %. A further measure is aimed at instructions per cycle (IPC). Here Zen+ is expected to perform about 3 % better.
The changeover in production from 14 to 12LP nm ensures higher transistor performance. AMD speaks of 10 to 15 % in this respect. It has therefore become possible to increase the peak clock rates by 300 MHz and at the same time reduce the voltage of the CPU cores by 50 mV. This should result in a 10 to 15% lower energy requirement for the same clock speed compared to Summit Ridge. The advertised structural reduction itself has hardly any effect – hardly surprising given the jump from 14 to 12 nm. Noticeable changes have been seen with Zen 2.
But AMD also promises something else, at least with regard to Ryzen 7 2700X and Ryzen 5 2600X: If all cores or all threads are used simultaneously, the clock rates are kept above 4 GHz at all clock rates. With the previous models, this was only possible on one core. In the test, the significantly finer clocking could be observed without any problems, even the 4 GHz mark, which was at least slightly torn. This alone already ensures that the computing power is higher than with the predecessors.
However, this is bought by the Ryzen 7 2700X with a higher TDP. This is now 105 W, the Ryzen 7 1800X was still limited to 95 W. The Ryzen 5 2600X remains at 95 W, while the Ryzen 7 2700 and Ryzen 5 2600 also remain stable at 65 W.
As with the first generation, the heatspreader is soldered on the new processors. AMD again uses a solder consisting of an indium alloy. According to the company, this would increase costs by an undisclosed amount, but would ultimately be the best method of dissipating the waste heat. Compared to other solutions, the die temperature should be about 10 °C lower.
The probably very cleverly chosen name Zen 5 as well as the statement that they were already working on this version of the processor architecture inevitably led to the question of how AMD numbers a few days ago. There is no clear answer to this question – among other things, because the company itself jumps back and forth in documentation provided. For example, the first generation of Ryzen is clearly based on Zen 1, Raven Ridge was referred to as Zen 1.5, and the second generation of Ryzen is now available as Zen+. It is conceivable that people are already talking about Zen 2 internally and simply orienting themselves on the model numbers. A Ryzen x 3xxx would then be based on Zen 3.
Completely detached from this discussion is the fact that the second Ryzen generation, apart from the fine tuning mentioned above, is based on the Zen architecture introduced two years ago. This is already shown by the number of transistors, which has remained the same at 4.8 billion.
This differs markedly from its predecessor bulldozer and represented a 180° turn back to the classic processor design. Each CPU core has four integer units with 168 registers each, which in turn can process 192 instructions “in flight”. Two load/store units are used to pass the data to the cache. There are also two floating point units, each with 128 floating point multiply accumulators. The associated instruction cache has a capacity of 64 KB, the data cache still has 32 KB. The former can be accessed four times, the latter eight times. The L2 cache has a capacity of 512 KB, and the L3 cache has a capacity of 512 KB.
However, this is not available in full size to all CPU cores, due to the structure called CPU Complex (CCX). In the case of the Ryzen 7 2700X and Ryzen 7 2700, each CCX has four CPU cores and 8 MB of L3 cache, which can only be used by these four cores. In the case of the Ryzen 5 2600X and Ryzen 5 2600, each of the two CCXs has only three CPU cores, but also 8 MB of L3 cache. The 8 MB are divided into four “slices” of 2 MB each, each slice being assigned to a CPU core. The partial L3 cache is based on the latter with regard to the clock rate and the general load. The latter ensures that not only an unused or only weakly used core is run at low voltage and thus ensures a lower energy requirement, but also the corresponding part of the cache.
Added to this is SMT, which means twice as many parallel executable threads as CPU cores. However, the two threads within the core do not necessarily have to be processed equally, this is where prioritization comes into play.
Infinity Fabric and SenseMI
Communication between the two CCXs, as well as with the main memory and various controllers, is once again handled by Interconnect, christened Infinity Fabric, due to the general retention of the Zen architecture. AMD has not made or documented any changes to this interconnect. In any case, it remains with a two-part structure. The Control Fabric controls the various Engine Hubs and is therefore responsible for Power Management, Security, Reset & Initialization and Test procedures.
The second part is taken over by the Data Fabric, which is primarily responsible for low latencies as well as a generally fast data exchange within the architecture and also takes over the communication with the main memory.
The CCX principle in conjunction with the Infinity Fabric should enable an almost linear increase in multi-thread performance.
A whole package of measures is to ensure the best possible mix of processor temperature, energy requirements and CPU clock speed. AMD calls this SenseMI Technology. It is based on a network of sensors that monitor current, voltage, power and temperature at millisecond intervals. All information is forwarded to the Infinity Fabric and evaluated there. This data is not only used to respond to current requirements, such as clock speed reduction due to excessive temperature, but also to prepare for requirements that will soon arise. According to AMD, the latter is based on machine learning.
For the second Ryzen generation, AMD has revised SenseMI in two areas that directly influence higher clock rates.
Precision Boost 2 und XFR 2
Precision Boost 2 is directly related to version 1 and still allows clock adjustments in 25 MHz steps. However, in contrast to its predecessor, Precision Boost 2 no longer only works when a maximum of two CPU cores are busy – from the third core onwards, the term all-core boost was previously used, only Precision Boost below that; we have already dealt with the concrete clocking behaviour. Instead, high clock rates are now no longer exclusively determined by the number of threads to be processed or the number of cores under load.
Precision Boost 2 uses three additional parameters for this: the currently available thermal budget, the currently available electrical budget and the absolute maximum clock rate. If none of the three limits is reached, Precision Boost 2 is used regardless of the number of threads and active cores. According to AMD, this can mean higher clock rates of up to 500 MHz in practice. This refers to the comparison between Ryzen 7 1800X and Ryzen 7 2700X.
Things get very confusing with XFR 2 – just like with the first generation. The idea behind Extended Frequency Range (XFR) is to reward, or rather to take advantage of, effective processor cooling. If the latter is powerful enough, the clock can be raised above the actual maximum. Decisive factors here are the Tcase value and the energy consumption of the processor. As with the first Ryzen generation, the limit was set at 60 °C and 95 A.
However, unlike XFR 1, XFR 2 cannot only be activated on two cores at the same time – if the budget is sufficient, the additional turbo is also ignited on all cores at the same time. However, this could not be proven beyond doubt in the test. Every now and then, individual cores reached 4.35 or 4.25 GHz (Ryzen 7 2700X/Ryzen 5 2600X) and thus just 50 MHz more than AMD calls for the maximum clock rate.
The fact that only 4.3 and 4.2 GHz are mentioned as limits in the external presentation is due to the conditions, the company says. The official values should be achieved with a cooling system that meets the minimum requirements – the rest is “on top”.
New Cooling Fans
AMD will sell all four new processors of the second Ryzen generation as boxed versions, including heat sink and fan. In terms of mounting, the spring-screw mechanism introduced two years ago remained the same – even with the new Wraith Prism.
This is the AMD flagship, but will only be included with the Ryzen 7 2700X. The cooling hardware is based on the Wraith Max, but for the lighting it now uses RGB LEDs, which are compatible with all relevant lighting systems such as ASUS Aura or Gigabyte RGB Fusion and are divided into three independent zones. The fan itself can be switched between two speeds, the volume indicates AMD with 39 dB(A). On the higher of the two speeds, the cooling capacity is supposed to be 140 W – enough to at least use the overclocking potential a bit.
The three other new processors have to make do with less powerful cooling solutions. AMD includes a Wraith Spire with RGB LEDs in the Ryzen 7 2700, the Ryzen 5 2600X has the same cooling system – but without LEDs. The Ryzen 5 2600X has to make do with a Wraith Stealth without LEDs.
The importance of high clock speeds for AMD is not only shown by functions such as XFR 2 or Precision Boost 2, because open multipliers once again offer the possibility to go beyond the standard speed. So it is possible with the simplest means to turn the clock screw. AMD’s own Ryzen Master software, which has been upgraded to version 1.3, provides the user with support. Among other things, CPU clock and voltage as well as RAM clock and voltage can be changed with a mouse click.
New features include several displays that show the current values for Package Power Tracking (PPT), Thermal Design Current (TDC), Electrical Design Current (EDC) and the fastest and second fastest core of a CCX. In addition, for the first time both CCX can be clocked independently of each other. On request, a simple diagram provides information about the current clock and the temperature. AMD points out, however, that not all values have to be exact. Because some data are called up from the BIOS, here the thoroughness of the mainboard manufacturer decides in the end.
In the test, using the Ryzen Master again turned out to be the easiest way to overclock the processors. However, newcomers may wish for a more in-depth explanation of some of the functions, and not everyone is familiar with details such as PPT or EDC. In addition, the grayed out field “Precision Boost Overdrive” might be irritating. AMD has already prepared this function fundamentally with version 1.3 of the software, but has not yet activated it. When this will be made up for was not yet known in mid-April.
Those who had hoped that the overclocking potential of the second Ryzen generation would be greater than that of the first should be disappointed. The Ryzen 5 2600X could be operated with water cooling at a stable 4.2 GHz on all cores at a voltage of 1.525 V, while individual cores could be clocked at 4.325 GHz. In Ryzen 7 2700X the end was reached at 4.25 GHz on all cores and at 4.475 GHz on one core. An increase in voltage above 1.45 V was not successful.
Similar values were achieved at an AMD event, also with water cooling. As with the Ryzen 7 1800X, this indicates that second generation selection is also carried out across all chips.
Performance and Verdict
If one orients oneself on the just two year old conclusion about the first Ryzen generation, Ryzen 5 2600X and Ryzen 7 2700X are somewhat disappointing. In March 2017, we wrote “Finally, a CPU test that was fun, as we were not only able to measure a 10% performance increase”, and this is exactly what happened. In productive use the performance increase is roughly 8%, in games about 4%.
But what looks like typical Intel mini-steps is due to the architecture change in 2017. This is now followed by the fine-tuning with Zen+, which explains the – in comparison to two years ago – smaller progress. It was not to be expected that AMD would again make great leaps. It was probably more important for the chip manufacturer to reduce or even close the gap to the Intel competition. According to the measurements, this has largely succeeded, even if old weaknesses are still present.
One such weakness is again the single thread, or single core performance. Thanks to higher clock speeds, the higher IPC and the faster RAM, Ryzen 7 2700X is about 8% and 19% ahead of its two predecessors Ryzen 7 1700X and Ryzen 7 1800X.
The Ryzen 5 2600X has an increase of between 5% and 10% compared to the Ryzen 5 1600X. Looking at the Intel stock, a Core i5-8400 or Core i5-7600K are on the same level. However, AMD itself positions the Ryzen 7 2700X against the Core i7-8700K, the Ryzen 5 2600X against the Core i5-8600K. But they still calculate up to 20% faster in such cases, and only in very few individual cases is something like a tie achieved.
This is not only noticeable when the user only uses a few threads or cores, but primarily in games. Most of them are still designed for operation with only a few cores – if at all. Intel is accordingly still ahead here, even if only in measurements. Because in practice, the difference, which on average is in the mid single-digit percentage range, should not really play a role. If a game runs smoothly on a Core i7-8700K, this will also be the case with a Ryzen 7 2700X.
Another problem that AMD can’t get rid of with the second Ryzen generation is the power consumption. Although one speaks of 10 to 15 % less energy being drawn off at the same rate due to the smaller structure and other architectural changes, this is not always noticeable in practice. Here the opposite is more likely to be the case. When Ryzen 7 2700X and Ryzen 5 2600X are required, they take a bigger sip than their predecessors. Although this also goes hand in hand with an increase in performance, the goal should actually have been parity with Intel.
In return, Intel might have to nibble at the multi-thread performance of the second Ryzen generation. The direct competitors leave Ryzen 7 2700X and Ryzen 5 2600X clearly behind, if all cores can be used. An increase of 25 and 32% in Cinebench 15 is a clear statement, even if this is only an isolated case. Usually the new AMD processors calculate 5 to 10 % faster here.
However, this is not only due to the higher number of cores or threads, but also to the clock performance under high total load. Precision Boost 2 comes into play here, if the cooling capacity is sufficiently high. Even when all cores are used, the clock rate now remains at a comparatively high level. XFR 2, on the other hand, like the first generation, is not important. Not only that the additional boost function could not be proven at all in many scenarios: A short-term plus of 50 MHz does not bring any significant additional power.
How well the second generation of Ryzen is suited for overclocking experiments depends, according to experience, strongly on the scope. The potential for increase with home remedies such as an easy-to-install AiO water cooling system was in the lower range of the test patterns.
Those who can and want to resort to more complex cooling solutions should be able to get much more out of the processors. However, overclocking newcomers can look forward to fast progress and a rather flat learning curve, open multipliers and the AMD tool Ryzen Master are very helpful at this point.
Whether in the end the grip to Ryzen 7 2700X and Ryzen 5 2600X or to Intel’s competitors is the better one depends strongly on the field of application. If the computer is to be used primarily for performance-hungry productive applications, the second Ryzen generation is the first choice.
However, the more important the topic of efficiency or single thread performance becomes, the more interesting the Core i7-8700K and Core i5-8600K become. If one assumes a mixed use, AMD and Intel are on the same level – perhaps the most important finding of the test. Whether one then chooses Ryzen 7 2700X or Ryzen 5 2600X depends on the actual requirements as well as the budget. However, there is a slight tendency towards the Ryzen 5 2600X, as it requires far less energy but still offers more than enough performance in many cases.
In terms of price, the two new AMD processors are in the same league as the competition from Intel. As of mid-April, only a few Dollars separated the processors, so in the end the total investment volume is decisive. If a complete platform change is pending, the second Ryzen generation should be cheaper in direct comparison. The same applies, of course, if an AM4 mainboard is already available. However, a change from the first to the second Ryzen generation is superfluous, the sense of a change from Intel to AMD strongly depends on the initial situation.