Saturday , August 13 2022

Review AMD Ryzen 9 3900X, Ryzen 7 3700X and Ryzen 5 3600 |


It was the day of one of the most anticipated editions of the second half of the year, the AMD Matisse or the 3000 series of AMD or Zen 2, no doubt regardless of the name you knew about this edition, the important thing is that the first processors produced in 7nm would be part of the market, and what a better day than 7/7. Will they cope? Let's see!

This new generation of Ryzen processors is produced not only with a smaller size (7nm vs 12nm), but also the change comes from how they are designed in such a way that more cores can be included without increasing the size of the matrix would lead to a more expensive process and more likely to fail.

That is why they chose a chip, ie. each specialized unit is on a piece of silicon, regardless of the other, so "enough" to add another chip and increase the cores, or add more features.

In this way, the processor was modulated and we saw it with the Ryzen Threadripper and the Epyc servers for servers, using the high-speed path called Infinity Fabric 2.0 that will communicate both CCDs via cIOD.

With these new processors and new support for PCIe 4.0, it has become necessary to update the chipset. This is why the new X570 delivers more SATA and NVMe ports, all via PCIe 4.0 interfaces, in addition to 10Gbps USB ports (USB 3.2 Gen2). Obviously, CPU communication continues through the PCIe 4.0 x4 connection.

Everything is on the route

To be able to communicate between the kernels, all route routes pass through the IO die, which is finally responsible for CCD communication with each other, and the CPU with the memory and other components. To achieve this, we had to increase the amount of substrate layers to 12.

Approach from one of the layers from CPU, Infinity Fabric lines ranging from CCD to IO can be seen at a glance. There is no direct communication between the two CCDs.

New Topology in Design

With this new design challenge to include more cores, in one socket size, the clearest move was to move to a chip topology. In this way, they can add more and more cores simply by adding or removing chips from the packaging.

The basic unit of Zen2 is a CCD, which consists of 2 CCX units of 4 cores and 16MB of L3 cache (double before). Therefore, each Zen2 CCD module has a maximum of 8 cores / 16Threads and 32MB L3 cache.

This topology is what we see in the third-generation Ryzen 5 and Ryzen 7 processors. As the maximum of the cores is 8. Remember that IO communication is through Infinity Fabric, and that chipset is responsible for communicating with the other components.

When you want to go beyond the eight cores, you need to add another CCD, the maximum being 16 cores / 32 threads and 64MB L3 cache. That's why a processor like Ryzen 9 3950X (16C / 32T) is possible.

Improvements to the internal level (Better Cash and New Stages)

At a high level, the kernel is very similar to what we already know about Zen. The basic elements of the Zen 2 design include a different predictor L2, known as TAGE predictor, micro-op cache mirror, L3 cache duplication, , increasing the load / storage resources and maintaining the AVX-256 from one operation (or AVX2).

Despite all these changes, AMD said there was no frequency punishment when using AVX2, depending on its power recognition frequency platform.

AMD has also made adjustments to the caching system, with the most noteworthy being the L1 instruction cache, which is reduced by half to 32K, but the association has doubled. L1 data cache and L2 cache have not changed, however, translation buffers (TLBs) have increased their compatibility. AMD also claims to have added more security virtualization support that helps to activate features in the future.

Looking fast, it is easy to say that duplication of micro-operational cache will offer significant IPC improvements in a number of scenarios, and combining this with increasing the boot / storage resources will help with more instructions. Dual L3 cache will help with specific workloads as well as AVX2 support for a single operation, but the improved predictor will also show increased raw results. In summary, for paper analysis, the improvement of AMD + 15% in IPC seems to be a very reasonable number for promotion.

Security: something left with Zen 2

One of the important issues last year, backed up in 2019, is the security of processors, especially those based on the X86 architecture. From its design and microcode, it allows malicious people to perform tasks remotely or execute malicious code to retrieve data without noticing one.

Like all Zen-based processors, these new Zen 2s have hardware-level mitigation as well as the operating system level already implemented by Microsoft. Therefore, they will not be affected by the recent and more known security issues in processors that Intel has become a major headache.

Memory support and overclocking

Another moment that enhances the new Ryzen family of processors is maintenance and overclocking in memories. Something that has caused a lot of headaches in today's generations of Ryzen processors. Well, with this new Zen 2 and I / O chip enhancements, the use of high-frequency memories will not be a problem and it's no problem to take them to the highest frequencies.

In fact, they promise that frequencies up to 5100MHz can be achieved in the memories (which allow it, clearly) and all by air. This can be achieved, partly by the above-mentioned, by splitting the chips into a CPU and providing a better memory controller in the IO chipset in conjunction with the X570 chipset. In addition to this with Infinity Fabric 2.0, some separators are automatically executed to maintain the stability of the system as a whole.

Every time we install fast memories in our new system, Infinity Fabric will work at a 1: 1 ratio with the memory frequency until a frequency of 3733MHz is reached. When we exceed this frequency, if it automatically enters the 2: 1 ratio mode, it means it will operate at half the memory frequency.

This obviously means that latency will increase slightly, and productivity will be affected by about 1-2%, but they have to do to maintain the stability of IF and in the process to achieve higher frequencies in memories.

The family of processors Ryzen 3000

Let's learn a little about these new third-generation processors.

Prices remain very attractive, something that we like and attracts attention, as this is a new way of production and production, must reduce costs, but by including more cores they are increasing. With this, AMD continues to offer more cores at an attractive price for the consumer.

Windows 10 and its enhancements

Along with all the new features of these processors, AMD together with Microsoft are implementing operating system enhancements, thus correcting a problem that arises in some current scenarios where the Windows scheduler does not know how to redirect. the priorities for performing the tasks that lead to higher latency and hence reduced overall work.

With so-called topology awareness in Windows and CPPC2 in BIOS, AMD along with Microsoft improved the response time of Windows frequency changes as well as the load distribution of each CCX. With this corner of improvements, the CCX now has to be full of tasks, just skipping instructions to another CCX.

This can generate up to 15% improvement in performance and 6% improvement in application response time. This is available with the Windows 10 update of May 10, but is only enabled with X570 boards.

The whole family

Finally, the new Ryzen family of 3rd generation processors remains from the "modest" Ryzen with Radeon Vega Graphics to the top of the Ryzen 9 3900X line.

A whole range of options, from embedded graphics processors to up to 12 cores and 24 threads at frequencies up to 4600MHz.


glasses AMD Ryzen 9 3900X AMD Ryzen 7 3700X AMD Ryzen 5 3600 AMD Ryzen 7 2700X Intel Core i9 9900K Intel Core i7 9700K
Manufacturing process 7 nm 7 nm 7 nm 12 nm 14nm 14nm
cores 12 8 6 8 8 8
threads 24 16 12 16 16 8
Freq. basis 3.8 GHz 3.6 GHz 3.6 GHz 3.7 GHz 3.6 GHz 3.6 GHz
Freq. max 4.6 GHz 4.4 GHz 4.2 GHz 4.3 GHz 5.0 GHz 4.9 GHz
Cache L2 6MB 4MB 3MB 4MB 2MB 2MB
Cache L3 64MB 32MB 32MB 16MB 16MB 12MB
TDP 105W 65W 65W 105W 95W 95W
Syst. cooling Prism with RGB LEDs Prism with RGB LEDs Wraith Stealth Prism with RGB LEDs N / A N / A
Price USD $ 499 $ 329 $ 199 $ 254 $ 485 $ 365
CLP cost $ 445,000 $ 295,000 $ 175,000 $ 250,000 $ 439,900 $ 349,900

The protagonists

A small look from CPUz to the main characters of this article.

Ryzen 9 3900X

We see that 4 CCXs are listed in the L3 cache, perhaps CPUz error, or this model has 4 active CCXs.

Ryzen 7 3700X

Finally, we have the R7 3700X where, unlike the previous one, only 2 CCXs are used because the design is only 1 CCD in the package.

First look.

As usual, AMD lets us test a complete set of everything we need to demonstrate the full potential of our processors

Packaging of Ryzen 7 3700X remains the same, except for aesthetics, but Ryzen 9 3900X we see a different package, more attractive and better coverage.

At the top of this processor we find that the processor is completely thrust to avoid hitting.

The content of. T Ryzen 9 3900X shows a two-piece package with a strong lining of foam and underneath this cooling system.

AMD included the best Wraith Prism cooler in both processors. Obviously with RGB.

Platform for testing and methodology.

Test platform
processor – AMD Ryzen 9 3900X
– AMD Ryzen 7 3700X
– AMD Ryzen 5 3600
motherboard – GIGABYTE X570 AORUS MASTER
memoirs – G.SKILL TridentZ Royal 3600MHz 2x8GB
cooling – AMD Wraith Prism RGB LED
Video card – NVIDIA GTX 1070 Ti 8GB
Power source – Corsair RM1000X
storage – SAMSUNG 960PRO 512GB M.2
monitor – ASUS MG28UQ

  • Windows 10 Pro x64 operating system [Update 1903],
  • The tests were carried out in a medium having a temperature of approximately 25 ° C.
  • The platform was used without a cabinet.
  • Drivers used for NVIDIA graphics cards were: ForceWare 430.86 WHQL
  • The resolutions of the synthetic tests are predetermined by each of the benchmarks.
  • The actual test solutions are 1920 × 1080 and 3820 × 2160 with all the maximum graphics in each game.

Picture of the pedestrian platform.

2D tests.

3D tests.

Consumption and temperature.

With the new change in the production process, this should affect the customer who will benefit directly from consumption and temperature generation, and in the following charts we can see if the theory of these benefits will really be effective.

As far as consumption is concerned, only the R9 3900X comes out of others and over its declared TDP.

As far as temperatures are concerned, we see Cooler "Stock" doing good work, but again the 12 cores and 24 threads of the R9 3900X make them rising to 80ºC. When we apply overclocking, we need to look differently and look for a better cooler.

PCIe 4.0

Another issue that AMD refers to these new processors is the inclusion and support of the new PCI Express standard. Where the bandwidth is quietly doubled compared to current PCIe 3.0. But will we really see a real change?

A test carried out on the NVMe PCIe 4.0 SSD gives us the result above, apparently not at the top of the technology as it is limited by the controller (Phison PS5016-E16). Well, day by day, you will not notice a big change of ten, what we will notice when you work with very large files or when we transfer a huge amount of information.

The NVMe PCIe 4.0 SSD, which will go on sale today or within the next few weeks, will all have the same controller as it is the only one currently available. So its performance will be very similar to the one shown.


As it did to previous generations, AMD Ryzen is a bit limited, in order to increase performance by manual overclocking, mainly because they are already coming at a frequency that is quite similar to what the processor can develop so restrictive As the temperature prevents us from increasing the frequency additionally through conventional cooling solutions, however, not all cores manage to develop the maximum frequency in XFR technology. After what was discussed, it was decided to leave the frequency for all 4.2 GHz kernels at 1.4 V at the 3 CPUs and result in the following.

As you can see, there was a slight improvement in the values ​​one can get with the default processors, and it is clear that With Cooler Stock it is not possible to achieve a good OC levelAs mentioned in the temperature range, as peaks with 80 ° C and 90 ° C are reached when we apply some OC.


The technological leap that AMD delivers with this new generation is enormous, obviously leaving behind the key factors that previous generations have not been able to overcome, is performance in games and high-frequency memory maintenance.

In the three processors we had the opportunity to review, we can conclude that AMD really is generating a real change in the way we see the processor industry, it is clear that AMD's prices and performance gives a 180 degree turn around , which has been seen earlier, and leaves the competition very bad, which we have undoubtedly regarded as desperately trying to respond mainly to the only thing left to lower its prices.

In general, AMD takes advantage of the advantage and takes over the crown, more cores and threads, which adds significantly to multitasking, if we add lower purchase costs, the balance will focus on the green side. In any case, it should be noted that when we talk about pure power, Intel still benefits in this respect, although with Zen2, the difference was significantly shortened.

AMD should continue to work to improve the temperature, especially now that the number of cores under the same piece of silicon is increasing, we can see in the numbers of the R9 3900X. It is appreciated that a good cooler is included, but if someone wants to achieve even higher performance, it is also highly recommended to replace the cooler cooler or coolant with liquid.

Many of us know that true competition between the two major processors is beneficial to both the price and the tireless need to have the best product that leads competitors to develop technologies that are unlikely to take several years without pressure. ,

Source link