Interest in Ryzen 5800x3d appeared after the first tests of Microsoft Flight Simulator, where the novelty showed a striking performance. Then, at the forum of another air simulator, it was confirmed that the VR mode of FPS increases even more and significantly improves the smoothness of the image in scenes with a lot of units and when playing online. It was here that I sometimes lacked 5800X performance, so it was decided to buy it. At that time, 5800x3d was only available in some US stores, but I managed to find it in Eastern Europe, and thanks to a colleague (of which I am very grateful) who was there and returned to Ukraine, I got my processor.

Considering that only the CPU was changed (at the same time very similar), I had the opportunity to carry out quite demonstrative tests about the impact of L3 cache on speed for different situations.

I will not describe the theoretical part that addresses technologies and the situation on the market, but I have read a lot of reviews and most liked tom’sHARDWARE test, that describes many interesting moments about the features of this processor.

The main difference between 5800X3D from 5800x is L3 cache 96 MB against 32 MB and frequencies 3,4/4.5 GHz (base/boost) against 3.8/4.7 GHz. So, L3-cache increased by 67%, base frequency decreased by 12%, boosting by 4%.

Test stand configuration:

Processors: AMD Ryzen 7 5800X/5800X3D Def 8/16
Motherboard MSI X570 ACE BIOS I0 (AGESA1.2.0.7)
RAM CMK32GX4M2F4000C19 2X16Gb 3600 16-17-16-32 1T
Data storage Samsung SSD 960PRO (500Gb) + ADATA SX6000PNP (2Tb) 100/830Gb free
Video card PowerColor RedDevil 6900XTU Def SAM – on
Cooling system EK-XLC Predator 360 + EK-Quantum Vector RX 6900
Monitor Samsung C32HG70
VR headset HP Reverb G2
Power supply SeaSonic SSR-850PX
OS Windows 10 20H2

First 5800x3D VS 5800X tests in synthetic applications

5800X Δ Δ% 5800X3D
CPU-Z Single 668.1 -48 -7.74% 620.1
Multi 6624.1 -427.3 -6.90% 6196.8
CB20 Single 623 -43 -7.41% 580
Multi 5995 -451 -8.13% 5544
LinX Gflops 485.59 -18.88 -4.05% 466.71
Temp 81.3 9.7 10.66% 91
Vcore 1.46 -0,176 -13.71% 1.284
Power 138.6 1.99 1.42% 140.59
Aida64 MR 53390 108 0.20% 53498
MW 28812 -13 -0.05% 28799
MC 51573 -49 -0.10% 51524
ML 57.8 4 6.47% 61.8
L1R 2364.9 -199.4 -9.21% 2165.5
L1W 1180.4 -93.6 -8.61% 1086.8
L1C 2337.4 -161.7 -7.43% 2175.7
L1L 08 0.1 11.11% 0.9
L2R 1154.6 -79.3 -7.37% 1075.3
L2W 1074.4 -124.66 -13.13% 949.74
L2C 1112.7 -86.4 -8.42% 1026.3
L2L 2.5 0.2 7.41% 2.7
L3R 668.92 -83.35 -14.23% 585.57
L3W 605.32 -45.18 -8.07% 560.14
L3C 598.98 -41.89 -7.52% 557.09
L3L 10.8 1.5 12.20% 12.3
Queen 110729 -8838 -8.67% 101891
RW 27052 874 3.13% 27926
ZL 871.6 -16.6 -1.94% 855
AES 81560 -5695 -7.51% 75865
SHA3 2900 -201 -7.45% 2699
Julia 82741 -6261 -8.19% 76480
FPU M 43791 -3291 -8.13% 40500
FPU SJ 13097 -842 -6.87% 12255
FP32 RT 17964 -747 -4.34% 17217
FP64 RT 9307 -72 -0.78% 9235

AMD Ryzen 5800X3D vs 5800X testing


In all tests except one 5800X3D showed worse results than 5800X from 0 to 14%, which is quite expected to take into account the lower frequency. And here it is necessary to point out one feature – a boost of 5800X with good cooling was always not 4700 MHz, as stated in the specification, but 4850 MHz. In 5800X3D boost is 4550 MHz instead of 4500 MHz. Therefore, in low-flow loads, the difference in frequency is not 200 MHz, but 300 MHz, which is 6.6% less.

But in the LinX test there was already real trouble – 5800X3D reached 90° thermal limit, although this did not significantly affect the test result. In this case, the voltage of 5800X3D was much lower at the same power. Therefore, the problem is in the heat transfer between the crystal and the heat distribution lid. Wonders don’t exist – an extra layer of silicon has a negative impact on temperatures.

The following were tests in the famous benchmark 3DMark

Time Spy 5800X Δ Δ% 5800X3D
TS CPU 12389 -464 -3.89% 11925
FPSMin 39.93 -1.66 -4.34% 38.27
FPSMax 95.93 -0.42 -0.44% 95.51
TempMin 70 4 5.41% 74
TempMax 83 0 0.00% 83
FreqMin 4500 -325 -7.78% 4175
FreqMax 4600 -250 -5.75% 4350
TS GPU 21712 -20 -0.09% 21692
TempMin 54 0 0.00% 54
TempMax 73 -7 -10.61% 66
FreqMin 4550 -200 -4.60% 4350
FreqMax 4850 -400 -8.99% 4450

InTime Spy test the newcomer again loses 3.9%, although there are good news. The maximum temperature is still 83 °. There is no difference in the graphics card test, and the maximum temperature 5800x3d is even lower, which was expected.

Fire Strike 5800X Δ Δ% 5800X3D
FS CPU 30241 -1635 -5.72% 28606
FPSMin 90 -4 -4.65% 86
FPSMax 102 -4 -4.08% 98
TempMin 60 6 9.09% 66
TempMax 79 -4 -5.33% 75
FreqMin 4500 -325 -7.78% 4175
FreqMax 4650 -200 -4.49% 4450
FS COMB 15812 1305 7.62% 17117
TempMin 69 -3 -4.55% 66
TempMax 71 -1 -1.43% 70
FreqMin 4600 -350 -8.24% 4250
FreqMax 4725 -275 -6.18% 4450

In CPU Fire Strike test the novelty loses 5.7%, but the maximum temperature is even lower. And the good news is that in the combined test, the 5800X3D already wins by 7.6%.

So, there are signs of superiority that will be revealed only in the games.

Productivity in games

Actually, I wasn’t too worried about how the 5800X3D would perform in games that I’m not interested in, but I had access to the children’s Steam account, so I tested it in games that had a built-in test.

Metro Exodus

The game was tested on three graphics settings. Tests at resolutions 720 and 1080, unfortunately, gave incorrect results, so I will show only 1440.

AMD Ryzen 5800X3D vs 5800X testing

At low graphics settings, the 5800X3D has a 14% advantage, at high and extreme settings, everything depends on the video card.

Far Cry 6

The game was tested at close to maximum settings in 3 resolutions with Minimum, Average and Maximum FPS fixaction.

AMD Ryzen 5800X3D vs 5800X testing

The advantage of 5800X3D is felt in all modes, but it decreases with increasing load on the video card.

DiRT Rally

The game was tested at close to maximum settings in 3 resolutions with Minimum, Average and Maximum FPS fixation.

AMD Ryzen 5800X3D vs 5800X testing

It can be stated that the game almost did not react to the change of processor.

X-Plane 11

It’s probably the least optimized game I’ve come across, and it reacts very poorly to the hardware upgrade. And although it allows you to play more or less decently even on an ultrabook with integrated Vega graphics, there is no such thing as hardware to have a decent FPS in VR mode. The game does not have a built-in test, but a standard set of tests is provided on its forum, which allows you to compare the results by average FPS. As can be seen from the numbers, the iron load gradually increases from FPS1 to FPS3. FPS4 is a CPU test, FPS5 is a video card focused test, FPS54 is something in between.

AMD Ryzen 5800X3D vs 5800X testing

In more loaded scenes, the advantage of 5800X3D becomes more noticeable and, interestingly, it is most visible in the video card test.

But this game also has a VR mode, which actually interests me (tested only with high settings).

AMD Ryzen 5800X3D vs 5800X testing

In VR, the advantage of 5800X3D increases, but unfortunately it still does not allow you to “fly” comfortably in X-Plane 11.

The following games were tested only in VR mode using the fpsVR tool, which allows you to measure the more important value for VR geek – frame time, as well as calculate the average FT, maximum 99% and 99.9%. For the CPU and video card, these values are calculated separately, of course.

Headset displays run at 90Hz, so you need to have a FT of less than 11.1ms (ie > 90FPS) for a smooth picture. Of course, there is a reprojection, which modulates an additional frame based on 45 FPS, but this is a completely different feeling. Shortly speaking, those who have a VR headset will understand me, and it is very difficult to explain for those who do not have such experience.


This game was also tested on a monitor, but in all modes and settings the average FPS was higher than 387, which is more than enough, and the difference between processors was 0-7%. At the same time, the results did not have the required repeatability, and therefore I decided not to provide them.

In VR mode, the game was tested in the built-in benchmark at near-maximum settings with average FT, maximum 99%, 99.9% and percentage of frames with less than 11.1 FT (i.e. > 90 FPS).

AMD Ryzen 5800X3D vs 5800X testing

It can be seen that the results of the 5800X3D prevail for the most part, but in fact simracer will not notice this, because in both cases the number of frames with FT <11.1ms is 100%.

DiRT Rally 2.0

The game does not have a built-in test, so I had to go through a special section in Acropolis, and then watch a replay for the test. Only VR mode was tested.

AMD Ryzen 5800X3D vs 5800X testing

The conclusions are the same as for the previous game.

Universe SandBox

This, in general, is not a game, but rather a simulator of the universe. Allows you to observe and influence both real space systems and artificially created ones. It can load the CPU and video card in a way that no benchmarks have dreamed of. I highly recommend everyone try it (or show it to the kids), not just VR fans.

There is no built-in test, so a simulation of the collision of the Milky Way and Andromeda galaxies was launched.

AMD Ryzen 5800X3D vs 5800X testing

The 5800X3D has a slightly lower average FT, but the maximum 99% and especially the 99.9% are much lower, and the VR geek will immediately notice this, as the number of frames with FT <11.1ms is much higher for the 5800X3D. Naturally, the video card also “noticed” it.


And finally what it was all about. The only existing military aviation simulator that has “destroyed” any hardware all the time since its creation and continues to do so. The game is much better optimized than X-Plane 11, but it is also essentially single-threaded and requires a powerful CPU core. In most scenarios, the 5800X (and even much weaker processors) is sufficient, and the load on the most loaded core rarely exceeds 60% with any graphics card, which in VR mode is almost always loaded at 100%. But in missions with a large number of units, there are hard freezes, including online.

The game does not have a built-in test, and moreover, due to randomization, the recorded tracks almost never match. Therefore, the 5-minute test took place in the “aircraft carrier against aircraft carrier” mission, where the Nimitz-class nuclear aircraft carrier opposes Kuzya with the entire theoretical composition of its aircraft wings.

That is, 2 ships, 88 planes and 32 helicopters take part in the battle on both sides, which at some point begin to attack enemy ships and each other.

AMD Ryzen 5800X3D vs 5800X testing

The numbers speak for themselves, and the main thing is the number of frames with FT <11.1ms – 10% in 5800X vs. 80% in 5800X3D. But I felt it even before studying the results, everything was so obvious.

Of course, this was a specially created situation, but the sensations from real gameplay are similar – the improvements in sensations are significant in any conditions.

So, the Ryzen 5800X3D is not just a better gaming processor (as it has been proven in many tests), it is just a godsend for VR fans as well as aviation simulator players. Its advantage in some scenarios can be compared to the progress of the CPU for several generations. The ending 3D in the name of the model can be considered precisely as a gift to users of virtual reality, which provides real 3D sensations.

P.S. At the same time, the maximum/average temperatures of the CPU were measured during about 2.5 hours of active play, and they were not much higher in the new products (54/79 5800X vs. 64/82 degrees 5800X3D). Temperatures were cooler in other games. The problem is that the existing BIOS completely lacks an overclocking section, and that’s where the SMT parameter was, which is responsible for multithreading. Its deactivation gave not only some performance gain in low-current applications, but also allowed to reduce the temperature, which also had a positive effect on the frequency boost. In addition, AMD crystals already traditionally respond well to downvolting, because the temperature-voltage-frequency triangle is interdependent and key to the performance of their processors. The results of the first beta-BIOS tests are already available, and they allow us to hope for frequency parity with the predecessor.