AMD Ryzen 9 3900X Review (English) – Dethroning the Intel Core i9-9900K? AMD Ryzen 9 3900X Review (English) – Dethroning the Intel Core i9-9900K?
(Última actualización: September 11, 2019)AMD Ryzen 9 3900X Review (English) – Dethroning the Intel Core i9-9900K? Disclaimer: Doing a double launch review (also in... AMD Ryzen 9 3900X Review (English) – Dethroning the Intel Core i9-9900K?
(Última actualización: September 11, 2019)

AMD Ryzen 9 3900X Review (English) – Dethroning the Intel Core i9-9900K?

Disclaimer: Doing a double launch review (also in Spanish). Using at the moment a quick translator (I’m a fluent speaker) but due to time restrictions and lack of sleep, probably won’t be editing for quite some hours (done a bit). Hope it is still comprehensible since we have some important data related to this launch. 6AM, going for a few hours sleep.

Update 1: After reading this review, please check here on first update about frequencies.

Update 2: Then check this update. (PBO working, Ryzen 5 3600 MHz working at 4325 MHz) and NVIDIA looking on WHEA errors.

Update 3: AGESA 1003 ABBA Preliminary testing. Link here.

We are pleased to release the review of the processor, AMD Ryzen 9 3900X, AMD’s current top-of-the-range flagship CPU (Ryzen 9 3950X will be released during September) using the new enhancements/changes in its architecture, Zen 2. The “hype” is quite high with this release and we will see if the processor reflects the expectation of what is expected. At the time of this release, there is controversy and it’s probably the most difficult and critical review we’ve had to date, so get ready for a roller coaster of information and excitement.

The processors based on the Zen 2 architecture, brings a lot of changes and we will try to talk briefly about them. We’ll probably have content for several weeks, just talking about Zen 2, but let’s start with the Ryzen 9 3900X review, changes in Zen 2, specifications, then go to the respective benchmarks, some problems with this release, our final analysis and recommendation.

Ryzen 9 3900X Review

Zen 2 Core, Matisse and Chipset X570

With this new launch of Zen 2 proccesors, there is a fundamental change in architecture and design. Each Zen 2 chip (7nm TSMC) comes with a maximum of 8 cores, divided into 2 separate complexes (4+4) also known as Core Complex. Each Ryzen consumer chip can come up to 2 Zen 2 chips (up to 16 cores) that interconnects with a I/O die (12nm). The link between the I/O chip and the Zen 2 Cores is known as “Infinity Fabric”.

The I/O chip (or I/O die) is responsible for housing several parts outside the Zen 2 cores, such as PCI Express lanes, memory channels and I/O operations (USB, etc.).

The change with Ryzen Zen 2 is that its chiplet philosophy (in this case Ryzen 9 3900X) helps to reduce the intercom latency that can exist between different cores, since all this is intercommunicated through a single module (the I/O die).

Ryzen Zen 2 Chiplet

Although this is quite technical, it is also quite intelligent from AMD, since instead of manufacturing a single die, it uses three separate parts that are manufactured and then interconnected. This allows for less loss of silicon due to manufacturing defects and gain more profit margin for the company.

With the introduction of the new Ryzen 3000 Matisse series processors, there is a new interface support, PCI Express Gen 4 (24 lanes CPU side) that is retro compatible with earlier versions. PCI Gen 4 offers twice the bandwidth offered by Gen 3 and is a solution that can currently attract many professional and enthusiastic consumers.

With the TSMC 7nm manufacturing process improvements, AMD has managed not to lose frequencies in its cores (there is usually a reduction) and in the design and test process, get more frequency compared to Zen+ and Zen. In addition, the official RAM support (JEDEC standard) has risen to 3200 MHz.

Using 3600 MHz memory with the new Ryzen processors will be quite simple using a 1:1 speed with Infinity Fabric. According to AMD Tech Day, they suggest working at 3600 MHz and adjusting the memory timings, to get the best performance inside the platform.

Another thing we want to “underline”, is that the L1 cache has decreased, while the L3 has increased. The L3 cache increment brings latency penalty, but according to AMD, in games, this increased L3 cache should help in the performance of games (they call it Game Cache).

AMD Ryzen 9 3900X – Specifications

The AMD Ryzen 9 3900X processor comes with 12 cores and 24 threads. This means that the processor has two Zen 2 cores (7nm) and they are interconnected by a die (12nm) I/O. The processor has a base frequency of 3.8 GHz and a Boost frequency up to 4.6 GHz. In addition, it has a maximum TDP of 105W at the factory and the product has an included heatsink (Wraith Prism RGB).

Now let’s look at the specifications of the Ryzen 9 3900X processor in addition to the other processors released (Ryzen 9 3950X will be available in September 2019):

Ryzen Serie 3000 Zen 2 Especificaciones

Unboxing AMD Ryzen 9 3900X and Ryzen 7 3700X – Press Kit

First of all, don’t worry, the box of the Ryzen 9 3900X includes processor (we were in full testing with the processor at the time of the photo session) but we have the photos of the press kit, as well as the heatsink that comes included in Ryzen 9 3900X and the Ryzen 7 3700X CPUs.

The photos on our portal is not our strong suit, but as it is a important release for us, we asked for external help. Special thanks to Istav Nile for the hard work and you can follow him on Facebook.

Productivity Benchmarks

We have some productivity benchmarks and will compare it to several Intel processors. For AMD Ryzen, we have used Windows 10 version 1903 at their suggestion and some technical improvements for AMD Ryzen (CPPC2 support and the new scheduler for Ryzen on Windows 1903). As for Intel, we opted to use Windows 10 version 1809 since there’s kernel problems last time we checked. Both Windows 1903/1809 versions have the vulnerability exploitation mitigations that have been released by Microsoft for Intel processors.

The specifications of the test benches are as follows:

AMD test system

CPU: Ryzen 9 3900X
RAM: G.Skill Trident Z RGB 2x8GB 3200 MHz CL14
Cooling system: EVGA CLC 280 (AIO with 280mm radiator, 2800 RPM Pump, 1500 RPM Fans)
Video card: GIGABYTE GeForce RTX 2080 Ti GAMING OC (
Primary SSD: Kingston 120GB UV300
SSD Test: Corsair MP600 2TB (Phison E-16 PCIe 4.0)
Gaming SSD: Silicon Power P34A80 1TB (SSD PCIe 3.0 NVMe) (
Operating system: Windows 10 Version 1903 (includes security mitigations)

Intel® Test System

CPU: Various processors (i9, i7 and i5)
Motherboard: GIGABYTE Z390 AORUS Ultra (F8)
RAM: G.Skill Trident Z RGB 2x8GB 3200 MHz CL14
Cooling system: EVGA CLC 280 (AIO with 280mm radiator, 2800 RPM Pump, 1500 RPM Fans)
Video card: GIGABYTE GeForce RTX 2080 Ti GAMING OC
Primary SSD: Crucial BX300 120GB
SSD Test: Corsair MP600 2TB (Phison E-16 PCIe 4.0)
Gaming SSD: Silicon Power P34A80 1TB (SSD PCIe 3.0 NVMe)
Operating system: Windows 10 Version 1809 (with security mitigations)

You help us by using our Affiliate links.

An additional note with our sample of Intel Core-i9 9900K. Our sample doesn’t overclock very well and works at 5 GHz stable on all cores. We had a very good 9900K press sample that reaches 5.3 GHz in all cores, but we didn’t get access to that chip for these tests.

The productivity benchmarks we have are the following:

-Cinebench R20 (Single-core/Multi-core)
-Blender 2.79
-Corona Benchmark
-WPrime 1024
-VRAY Benchmark
-Geekbench 4 (Single-core)
-AIDA64 RAM transfer rate
-AIDA64 RAM latency

Cinebench R20

Cinebench R20 is a benchmark based on Cinema 4D, which is a 3D animation software of different uses, developed by MAXON. Cinema 4D also has variants and is also available as a plugin for other programs, as it is optimized to use the various processor cores optimally. CBR20 measures the performance of the hardware configuration based on this software/plugin.

The software has been used for several multi-million dollar Hollywood movies, in addition to professional use for smaller scale 3D modeling.

The test can measure multi-core/thread and single-core performance.

Ryzen 9 3900X Cinebench R20 Multi Ryzen 9 3900X Cinebench R20 Single

The Ryzen 9 3900X (499 USD MSRP USA) simply crushes in performance an Intel Core-i9 9900K (485 USD Amazon) in multi-core, with a brutal difference of 47.28% in factory configuration. Even with the Core i9-9900K overclocked at 5 GHz on all cores, the 3900X has a 38.94% advantage.

This doesn’t surprise us so much, since AMD Ryzen in several price ranges, offers more cores and threads at the same cost and that’s very good in software where they scale well with more cores/threads.

As for the single-core performance in this test, in the X570 AORUS XTREME motherboard in factory configuration and with the liquid cooling we have (EVGA CLC280), the single-core frequency reaches up to 4.65 GHz with the AGESA 1002 code (BIOS F1) in several cores, except for a few (the rest at 4.6 GHz).

The single-core performance we obtained with the AMD Ryzen 9 3900X was on par with the Intel Core-i9 9900K, i.e. a technical tie (within the margin of error).

This is very good news for the redesign and the manufacturing process of the processor (TSMC 7nm) that AMD has bet.

The mono-core performance measurement of the processor is quite good in this test, but we would advise you to also review the results in Geekbench 4, which we will see for a while.

Blender 2.79

Blender is a 3D animation software that is getting more and more adoption in our country, besides it’s free. It has different types of use, since it has enough versatility in terms of animation. It also has professional use in movies, for pre-production, but is popular among enthusiastic 3D animators.

Ryzen 9 3900X Blender

The Ryzen 9 3900X with its twelve cores, has a render time less than 26.25% compared to its counterpart Intel Core-i9 9900K in manufacturing configuration. Compared to the i9 9900K at 5 GHz in all cores, the Ryzen 9 3900X renders with a 21.85% advantage.

In rendering, the less time, the better.

Corona Benchmark

Corona Benchmark is a special renderer for some 3D modeling programs, such as Cinema 4D and 3DS MAX. If you want to see the complete options offered by this renderer for 3DS MAX, you can follow this link. For Cinema 4D, use this.

The benchmark is something old, as there are more current versions than Corona Benchmark version 1.3, BUT, according to its developers, have not updated it, since in the new versions of Corona Plugin for C4D and 3DS Max, have not seen performance changes.

3900X Corona Benchmark

Corona shows a similar performance to Blender, since the advantage of the Ryzen 9 3900X over the Intel Core-i9 9990K is 26.80% and 22.82% when the 9900K is in OC at 5GHz in all its cores.

VRAY Benchmark

VRAY is an add-on or plugin for different softwares, such as 3DS Max, Maya, Rhino, Cinema 4D among others. Within its capabilities, outside render, it offers a better workflow as well as tools for the 3D animator. The plugin is developed by CHAOSGROUP and also has the option of hybrid rendering (using the processor and the video card at the same time) but because we are only seeing the processor, we are using the CPU-only test.

VRAY has released its new version (software upgrade), VRAY NEXT and its new benchmark measures this update to the software. Instead of using a measurement like minutes and seconds, use KSAMPLES. The higher the KSAMPLES, the better the processor performance.

3900X VRay Benchmark

VRAY NEXT Benchmark shows us an advantage of the AMD Ryzen 9 3900X over the 37.89% Intel Core i9-9900K when the 9900K is in factory configuration. With the 9900K at 5 GHz on all cores, the advantage is reduced to 34.29%. Both differences are quite substantial.

Geekbench 4

Geekbench 4 is a benchmark that measures processors and video cards, in different types of tasks quickly. It is a performance indicator and includes memory tests. This benchmark includes the multi-core theoretical performance, as well as the processor mono-core.

3900X Geekbench 4 Multi 3900X Geekbench 4 Single

Again, the Ryzen 9 3900X shows a multi-core performance far superior to the Intel Core-i9 9900K processor in this test, with a 35.55% advantage. In single-core, Ryzen 9 3900X is still behind the Intel processors in this test. Compared to the Ryzen 7 2700X in single-core, there is a 16.71% improvement.

WPrime 1024

WPrime is a benchmark that uses mathematical calculations that are processed by the CPU and shows some performance in operations of this type, using all available threads. The results are in seconds and the less, the better.

3900X WPrime 1024

AIDA64 (RAM transfer rate and latency)

AIDA64 is a synthetic benchmark that tests RAM memory that interacts and communicates DIMMs with the processor. We will see two aspects of the benchmark:

-RAM transfer rate

We start with the first capture, the transfer rates:

3900X AIDA64 RAM Transferencia

The results in this benchmark are mostly similar and as it is synthetic, they do not necessarily reflect the performance for most users, but it is a reference to how RAM scales with processors. Within the information provided by this test, the most relevant is the latency of the processor with RAM and this is the table.

3900X AIDA64 Latencias RAM

Intel continues to have a very significant advantage in latencies between the processor and RAM modules. In addition, the latency of Ryzen Zen 2 (3900X) is even higher than its previous generation, 2000 Zen+ series (the less latency, the better). Although these are the results, there is a logical explanation for this.

Let’s see the following screenshot and I suggest you read the analysis, to understand the architecture changes that AMD Ryzen Zen 2 (Matisse) has.

Additional: Capture and analysis of the AIDA64 benchmark

First let’s look at the capture and pay attention to the transfer rate for the L3 Cache.

AMD Ryzen 9 3900X – Zen 2

AIDA64 Ryzen 7 2700X

AMD Ryzen 7 2700X – Zen+ (From our review in spanish)

Do you see the difference compared to Ryzen Zen 2 and Zen+ (1000/2000 Series)? Yes, Ryzen 3000 series (Zen 2, Matisse) has a larger amount of L3 cache and when one increases more L3 cache, one generates more latency between the processor and the RAM. That explains why chiplets using Zen 2 Cores + the I/O chip has more latencies in AIDA64 compared to Intel and previous generations of Ryzen.

AMD believes and is promoting that the L3 cache will benefit games, increasing performance. They are even using the name “Game Cache”. It’s difficult to measure how much impact the cache has on games, but we’ll see if the gaming results reflect these changes in any way.

Benchmarks (Gaming)

================= IMPORTANT TO READ THIS=================

And where are the game benchmarks? We will dedicate an entire section, of why I haven’t tested games yet, as there are problems with the latest AGESA code and X570 motherboards (X470 also happens). It has taken us up to four days of testing, debugging bugs and using different types of configurations.

You can click here to skip directly the problems we have experienced with the AMD Ryzen processors we received, plus this problem has not been foreign to us (other media have reported problems).

================= IMPORTANT TO READ THIS=================

Testing PCI Express 4.0: Corsair Force Series PCIe Gen 4 MP600 2TB NVMe M.2 SSD

We did some quick tests with the new Corsair storage, the 2TB MP600 that AMD gave us to test the new features offered by PCIe Gen 4, as it is one of the new features offered exclusively in AMD Ryzen processors and boards that have X570 chipset.

The 2TB Corsair MP600 uses the PHISON E16 controller and Toshiba BiCS4 3D NAND 96 layer memory. If you want to compare it with storage like we did in other brands AORUS (Review in Spanish) and Silicon Power (Review in Spanish), the mentioned M.2 PCIe 3 SSD use PHISON E12 and Toshiba BICS3 3D NAND 64 layer memory.

Let’s see the performance that SSDs PCIe 4.0 offers in the M.2 format.

CrystalDiskMark 6

ATTO Disk Benchmark

Corsair MP600 ATTO Disk Benchmark

AS SSD Benchmark

Anvil Storage Utilities

Corsair MP600 Anvil Storage Utilities

The performance of solid state drives (SSDs) using the new PCI Express 4.0 protocol is impressive. These new products are not really focused on gamers, as the load improvement is not going to be noticeable in this type of tasks.

The people or users who will really appreciate this kind of storage performance are enthusiastic and professional users (digital production) who use video files 4K or higher (6K RED, 8K RED) as the bottleneck for them is storage. However, seeing this performance on a consumer platform is welcome.

Users with quite stringent professional demands will be looking at this type of RAID storage, as AMD Ryzen Threadripper Zen 2 will have more PCI Express 4.0 lanes (this speculation is quite likely).

Seeing this, we will see the apple of discord… problems related to the launch of the new Ryzen 3000 series processors (Zen 2 Matisse).

Product launch problems – AGESA Code and WHEA PCI Express errors on NVIDIA GeForce cards

First I begin by underlining all the time invested to obtain the following results. It has taken me three whole days trying out different configurations for this analysis, so let’s get started:

The problems shown with the two processors I have tested, the Ryzen 9 3900X and Ryzen 7 3700X (which are press samples) are as follows:

-Problems with code AGESA 1002 NPRP (BIOS for press) and later BIOS
-Problems with PCI Express and NVIDIA video cards (WHEA error)

These are all the motherboards that were tested:


The whole story…

During the first three hours of testing of the AMD Ryzen 9 3900X processor, using the X570 AORUS XTREME board, I noticed the problem when PCMark 8 did not pass the first test after 40 minutes (this is a total of ten tests). I noticed WHEA error (Windows Hardware Error Architecture) in HWInfo64 (se this software for PC telemetry, highly suggested).

From there I also decided to pay more attention to HWInfo64 and also checked that the BOOST frequencies of the processor had problems, since it didn’t get to “boost” all its cores to the maximum that it should, which is 4.6 GHz. It reached 4.5 GHz to 4.575 GHz in a pair of cores and the rest of cores to 4.3-4.4 GHz… We used manufacturers chipset driver, we have used press chipsets, as more current chipset driver version, same results.

X570 AORUS MASTER BIOS N11 NRPR, as well as BIOS F5c and F5e

It seemed strange to me, so I first decided to write to my contact with GIGABYTE USA (Matthew Hurwitz, I thank him for all the time he has put in to find a solution) and showed him the WHEA (PCI Express) errors, as well as the rare behavior of the 3900X boost frequencies.

To spoil you a bit of what you are about to read… this is not a problem of GIGABYTE (it was also corroborated in MSI board that I tested, we have capture in GODLIKE and we know another media got the same results with ASUS board).

At the time we did not know what was going on and tried to scale the problem to Taiwan. Until I received some answer from Taiwan (it was still hours in the afternoon, in Taiwan, early morning) disassemble plate, I tried in the MASTER, same results with the latest BIOS available.

Test, disassemble, test, disassemble…

Midnight (Wednesday) GBT HQ gives us news and according to their tests, the new AGESA code, including NPRP BIOS (BIOS for press) replicated our results in single-core frequencies, BUT, the original BIOS (AGESA 1002, without code introduced NPRP) turbo boost was working well.

With this information, I decided to flash BIOS, the first BIOS released for the X570 AORUS MASTER board and surprise, the boost frequencies were working as they should, even beyond the processor at 4.65 GHz. The WHEA error problem in the PCI Express was still going on, so I kept pressing and trying if the problem was maybe the chipset driver.


X570 AORUS XTREME BIOS F1 (Boost works well, you can see on capture)

X570 AORUS XTREME BIOS F1 WHEA PCI Express Errors (only on the morning of July 5th could we start preparing machine for review after we found a BIOS that works)

From there I tried to test an X470 board since the first problem was “solved” (Precision Boost Overdrive does not work with AGESA 1002, but factory performance has no problems) since a board with X470 chipset, doesn’t have PCI Express 4.0, it may not have  WHEA errors in PCI Express.

We got blue screenshots in Windows 10 when we did a new Windows installation (you can’t imagine how many times we’ve installed Windows in the last three days) on the X470 and Windows 1903 boards. We were able to mitigate the blue screens by first installing a 2700X, installing chipset driver and then switching to 3900X.

The boost frequencies on the X470 AORUS Gaming 7 WIFI board work well (AGESA 1002, BIOS F41a) with the 3900X we have as a sample, but still, WHEA errors of the most current NVIDIA driver (we use a GeForce RTX 2080 Ti).

We tried another additional driver, same results. Why do we use a GeForce RTX 2080 Ti? To remove 1080p bottleneck limiter.

Let’s continue… After several days of testing, I concluded that the reasons for PCI Express (WHEA) errors can be two:

-NVIDIA Driver
-PCI Express Driver

If it’s NVIDIA driver (since all tech websites and media use for processor testing) we hope there will be a solution soon. The other reason may be that there is some problem with the PCI Express driver for the Ryzen processors with GeForce cards, BUT, I don’t think it goes that way… (although there is some possibility, we will eventually know).

PCI Express (WHEA) errors don’t appear with the Radeon VII (which I also tested) or with Vega 56 (someone tested with that card and passed me the information corroborating the data). Unfortunately, to benchmark game processors and avoid processor bottlenecks, we can’t use any of these alternatives (besides our database uses GeForce RTX 2080 Ti).

Well, we had access to press drivers (chipset), press BIOS and got the same results. Even GBT Taiwan were kind enough to provide us with more current drivers, but without any change (to solve the WHEA with NVIDIA cards). Also the new drivers do not solve the problems of factory boost frequencies (it’s AGESA code problem).

As I mentioned before, it’s not a GIGABYTE problem. Luckily in our press kit, we were given a MSI MEG X570 GODLIKE, so we also took the time to try GODLIKE, with the BIOS that came installed (old one) with the other Ryzen 7 3700X processor that came in the press kit, while we were testing the 3900X trying to fix the WHEA problem and it turned out that the processor didn’t boot more than 4.3 GHz with the exception of a pair of cores (it should be 4.4 GHz), with the NPRP BIOS (for press) as well as the most current BIOS of its web page (AGESA 1003AB).

All with the same boost problem, here the corresponding screenshots (using X570 GODLIKE).


X570 GODLIKE BIOS 1013. Processor Ryzen 7 3700X, problems with Boost and PCIe WHEA errors

X570 GODLIKE BIOS 1015 Boost Frequency

X570 GODLIKE BIOS 1015 (BIOS Press NPRP). Processor Ryzen 7 3700X, problems with Boost and PCIe WHEA errors

In addition, WHEA errors were also presented in MSI MEG X570 GODLIKE. I tried to communicate with a MSI rep, but in the end I was not successful, to get a more current BIOS (although they uploaded a more recent one to their website, as previously mentioned). Also the same problems in three different BIOSes.

I also found out, that another international media had the same behavior with ASUS Crosshair VIII Hero board.

So with ALL these serious problems, how do you possibly launch a review on Sunday?

That’s what I decided and that’s what you’ve seen with productivity numbers and the NVMe PCIe 4 SSD storage numbers:

-Launch the review with the X570 AORUS XTREME (BIOS F1) motherboard as it shows turbo frequencies working normally. I think the benchmark numbers they show will be quite real when there is new AGESA micro-code with corrections.

-To obviate Precision Boost Overdrive 2 for the moment, since in all the BIOS, with the two chips that I have, the 3700X and the 3900X in the two boards of different marks do not work correctly.

-To avoid gaming tests, because the frametimes (FPS) we see an erratic behavior and until not having a stable driver, I prefer not to launch tests that may not show the true power of the processor and I have to retract, in addition to double expenditure of time. (This review has already taken me three to four times the time I normally need for processors reviews).

-Consumption numbers may not come out in time, for all the time I’ve invested to show all this data while “debugging”.

-When the corrections are out, we will test gaming using our best skill that always characterizes us.

We have done some internal gaming tests, but we prefer not to comment or show the numbers for the WHEA errors shown and the frametimes we have observed internally. I don’t know if all Ryzen processors are affected by premature AGESA code and there are problems with NVIDIA drivers, but we will see what other media observed. I rather skip until my own criteria is met.

Let’s jump to the final analysis of this review, which as you can see, is going to be incomplete. If these problems had not occurred, we would probably have been able to review the Radeon RX 5700 XT and show you numbers on gaming we feel confident to release.

Final Analysis (Incomplete) – Promising, but still incomplete

If you have not read the previous section, we would suggest that you read it, AS IT IS VERY IMPORTANT and therefore “the incomplete” title of this section. I want to emphasize quickly, that despite the problems shown, at the level of architecture and chip design (two Zen 2 chiplets and an I/O chip) we find impressive at first sight, but not having the full image (gaming) we can not give a final verdict of these changes.

We will divide our final analysis into four sections:

-Gaming (Incomplete)
-Our recommendation
-Extra criticism/opinion

Productivity and X570 chipset – Very promising…

We haven’t tested all the benchmarks we would have liked (because of time issues) and also because we don’t have a 4x16GB kit needed to test Adobe Photoshop, After Effects and AutoCAD in a standardized way (we’re trying to get kits).

However, AMD Ryzen 9 3900X simply dethrones in any application where the software uses all cores and threads optimally and to the maximum the Intel Core i9-9900K processor, which is in its price range.

Despite the problems we showed, with the release of Ryzen 9 3900X, we have the confidence of the numbers shown in productivity for the most part.

We do not believe that when the new AGESA (BIOS) update comes out with the necessary corrections, we will see a lot of difference in the benchmarks we have tested.

At the price of 499 USD, Intel is in serious trouble for these types of use; as well as in its number of cores and threads with the i9 9900K processor (483 USD in Amazon at the moment). I will have a better picture of the situation once we can do additional testing.

You may not be informed, but not all software is optimized to use more cores and threads. Some use few cores and/or favor mono core and IPC performance. You really need a software by software analysis, as you do in games.

The benefits that PCI Express 4.0 (chipset X570) brings to the gaming world is limited at the moment.

This option is probably useful when video cards demand more bandwidth than is available in PCI Express 3.0 (maybe in a couple of generations).

However, the professional use of new PCI Express 4.0 storages is something that is going to be quite attractive to those who really need it and now they can also find it on a consumer platform.

No doubt there will be some gamer, who wants to buy a board with X570 chipset and wants the best of the best in storage (we see that a lot in Intel users) and well, is included in the package because Ryzen 9 3900X has support for the new transfer rate.

Now, the cost of a X570 + 3900X motherboard (in this case, comparing the i9 9900K) is more expensive than a Z390 + i9 9900K motherboard, if we compare motherboards of the same range… but again, the performance in productivity software that scales with cores, justifies it.

Gaming (Incomplete) and consumption

Well, there are problems as shown from WHEA errors with NVIDIA drivers as shown. We prefer not to comment on the few tests we have in gaming, but we have seen erratic behavior.

The moment this is fixed, we will “benchmark” immediately when we see that the platform is stable and error-free.

In terms of consumption, we could not measure for lack of time (that after the errors demonstrated) we are lucky to have collected all this data. We will add the same in the following days.

Our sincere recommendation – aimed at consumers

It’s our first review in which AMD has given us the sample directly. We appreciate the gesture honestly, but our responsibility, first and foremost, is to the public. Please, once again (I think I am repeating about five times…) read the evidence we have shown of some serious anomalies with the AMD Ryzen 3000 series processors, which we hope will be resolved as soon as possible.

Perhaps the AMD commercial and marketing sector will not agree and in the worst case they will ask for samples ahead of time (I sincerely hope not), but I suggest to the public, to wait for some update as to the problems we have shown before making a purchase.

The problem is not with the X570 chipset (we have also tested on X470) and we expect it to be solved in less than a week. We are aware of getting some official response from AMD/NVIDIA regarding WHEA errors soon with GeForce cards and/or a new driver.

Remember it’s a new technology, as well as a new architecture being released. The original release of Ryzen 1000 series had many problems (Ryzen 7 1800X) and this release has not been executed in the most successful way.

I understand that the date, 7/7 has a very important value for the company, but I think there were quality control problems and AGESA code. AMD is responsible for the main code (AGESA) that gives partner companies (ASUS, ASRock, GIGABYTE, MSI) in the case of processors.

Maybe it was a lot of things in such a short time (remember that Radeon RX 5700 XT is also released among other products today).

It’s a pity, because AMD, from a marketing point of view, has done a tremendous job of changing the perception of its products from a few to the end user and admit that its direct competition Intel recognizes it as a “formidable competitor”.

In productivity, it looks very promising and certainly in software that scales well in cores and threads, simply a better choice than the i9 9900K.

We would not be surprised if Intel offers a price reduction when all this anomaly has been corrected.

In gaming, we can not say anything yet for the problems shown …

If we still have the samples when we see that the BIOS and WHEA bugs related to NVIDIA are corrected, we will do the respective tests and give our final conclusion.

Well, that’s all I have to tell you, but I will give a little critique to the end user; if you who are reading this, always make a responsible purchase.

Extra criticism/opinion – Review portals and hype is bad

I know some other media have noticed the Boost Clock’s unusual behavior. I don’t know if all the samples sent are demonstrating this same behavior, nor to what magnitude. Hopefully they will report it as well. We were among the first to notice this on Wednesday July 3rd and the errors related to WHEA PCIe.

If in the case that ALL the samples sent from the press present these problems, all or most of the reviews could show data with not due performance.

So take this into consideration when analyzing the review of another media and I hope you have reviewed the behavior of the boost frequencies with the press BIOS. We used initial AGESA 1002, as we saw that it was the most stable, at least for productivity.

So I would take with tweezers the reviews of other media portals today.

Now, the end user… Every hype is bad, because maybe after reading this, several people are disappointed (at least as far as the release is concerned). In the worst case, we’ll see memes, because the Internet is the Internet…

We believe that eventually the problems will be fixed (we hope) and the faster it is, the better… But “hype” and fanboyism of all kinds is bad. It’s not bad to have a preferred brand (for example, I have one for mice) but it’s you, the users who buy the products and you always have to be careful with “the marketing machinery”, as it can be something double-edged. (Extreme example, not related or compared to this review: pre-orders of No Man’s Sky).

It’s good to make purchasing decisions and make things ahead of time with a cool head. Pre-sales are a double-edged sword, like today.

We’ll see what happens over the next few days and while we inform the public through our social networks (like Facebook) when a product is launched, please take the time to choose your review portals seriously and be well informed. Also, try to realize that media has independent behavior and is impartial.

It’s been the toughest review I’ve had to do to date (it’s the first time AMD gives us a sample for launch day directly) and we were really excited about the launch, but these things can happen.

We will continue to report in the best way we can, as before, today and in the future, as long as we can… As mentioned earlier, when the behavioral problems have been corrected, we will update this review if we still have the samples, it’s a promise.


Alex Bustinza (better known as Xanxo)

If you liked this article, you help us a lot by doing like, share. If you have any questions, you can follow us on our social networks:



Overclocker entusiasta, reviewer y actualmente rank 1 en HWBOT en Perú. Gamer desde niño y ahora reviewer/media tiempo completo.