Since the other two AMD CPUs have arrived and I was able to get a replacement CPU for the Ryzen 9 5900X (and buy it), I have revised the whole launch article and added the missing entries. This would give us the quartet of Ryzen 5 5600X, , Ryzen 7 5800X, Ryzen 9 5900X and Ryzen 9 5950X now complete and we can ask ourselves which CPU we really need to play and where does reason end, because then prosperity occurs and then the exuberance begins. Well, you want to have everything, but do you really need it? Answers? It’s there, don’t worry!
The architecture of the new CPUs including all slides I had already discussed in detail. A significant increase of the IPC, together with significantly more clock speed, should also be able to replace Intel as gaming champion. The proof is due today. The prices turn out as expected and of course one can be curious about the further availability in the end. Because one thing became clear: the parts were available in quantities and were also immediately gone again, as if a vanilla ice cream had been held in the sun at 12 noon in Dubai. Companies like Mindfactory wrote about the “best sales start ever, ever, ever”. There you go.
For all of you here again the most important slides of the presentation as PDF to browse through. The pictures speak for themselves, the rest will be proven by tests today and later. One way or another.
Ryzen 5000 Series Desktop ProcessorsPreliminary remark
Since Intel unfortunately can’t (or doesn’t want to) provide samples over a longer and reasonable period of time, but currently only sponsors the CPUs as giveaways in social networks and I don’t let myself be put under any conditions or time pressure, I’ll only use my self-acquired Core i9-10900K and Core i5-10600K in today’s test, which also happen to be the appropriate counterparts. This restricts the variety for the upcoming tests a bit, but since Intel doesn’t have anything to offer above the i9-10900K anyway, this is surely still bearable. And Intel can be politely told that the business is not only generated by influencing. It went bad enough on Twitch.
For this purpose, I am placing the Ryzen 9 3900XT and the Ryzen 3600X as direct counterparts from the generation that has now been replaced, so that in the end there are now eight CPUs, which I will analyze more closely in the course of these tests. In order to make the test more realistic and because the feedback of the readers is also important to me, I varied the test a bit this time and also worked with different settings. Today’s article will focus almost exclusively on gaming and the version used as the base for the graphical output, the detailed workstation and studio test will be given later, when the new test system with the Ryzen 9 5950X is finished and the remaining problems have been solved. Pre-work is already done, despite sickness related absence and delay.
I have decided not to benchmark all games, but to do so in a very target-oriented way and in much more detail than before, especially those games that (distributed in equal numbers) were best suited to the respective manufacturers. A total of 5 games, but in all four resolutions and with lots of details, should then complement (not replace) all the reviews of the esteemed colleagues, who were able to test myriads of games in the team, but can’t go into the whole background at all. What I will show today are the cases where the new Ryzen performs particularly well, where it performs as expected, and where there have been slumps in performance that I will explain in more detail.
Memory clock and Factory OC
In the past, it was hard to please anybody, some people wanted realistic tests with higher clocked memory (as many users do based on the XMP profiles of Intel CPUs), others swear by the specifications for the clock of the memory controllers. And are you allowed to enable Enhanced Turbo on Intel and what about PBO on AMD? Questions and questions and in the end you can’t please anybody.
That’s why I double test the two new Ryzen and also the two adequate Intel CPUs. One with the clock speed prescribed by the respective manufacturer for the memory controller (DDR4 3200 for AMD or 2933/2666 for Intel) and the other with the maximum stability of the new Ryzen CPUs. The two Ryzen 9 5900X and 5950X ran with an FCLK of 2000 MHz, so that these CPUs and the Intel Core i9-10900K could also be additionally tested with DDR4 4000. The Ryzen 5 5600X and Ryzen 7 5800X only managed one FLCK from 1933 to 1967 MHz stable, so these and the Core i5-10600K ran with DDR4 3934 (I call it simply DDR4-3900 in the charts). A pair of Corsair Vengeance RGB Pro DDR4 4000s with a matching profile is used here, but it also ran cuddly on the Intel platform.
The old Ryzen 9 3900XT and 3600X run on DDR4 3600, which is still a certain OC, but it’s almost negligible in many situations, because the actual game takes place somewhere else. As hard as it may sound, but more than just decoration, the Ryzens of the 3rd generation are the best. Generation now no longer, unless you play in Ultra-HD and then the graphics card screams for help. But I don’t want to anticipate the tests. What is more interesting is the maximum advantage a Ryzen 9 5900X can get out of what is often considered a simple memory-OC with stored profiles. The Ryzen 5 3600X then turns out to be very similar within its capabilities, but does not quite reach the peak values.
The new Ryzen achieve up to 7% performance increase for up to 800 MHz more clock speed. The power consumption of the CPU increases with less limitation by up to 23%, that of the GPU by just under 3%. I have achieved this in the ideal case and at 720p, normally this will have a much lower effect. But about 5% is actually always possible with Ryzen even on Full-HD. A benefit to take with you. Everything from WQHD onward is put into perspective, because the graphics card is becoming more and more the limiting link in the chain.
With Intel it looks a bit more extreme at first glance, but apart from the memory OC, the “Enhanced Turbo” is now added, which you can really see as a factory cheat without a guilty conscience and which goes far beyond PBO, as long as AMD doesn’t rely on predefined performance profiles. However, the almost 1000 MHz more memory clock and the Enhanced Turbo only result in a good 9% more performance in the best case, but instead with almost 30% more power consumption of the CPU in 720p, because the bottleneck widens a bit to the spout. In addition, the GPU consumes almost 6% more power. Also fits about the gain in AMD, now, with Zen3.
The measurement of power consumption and other things takes place here in the special laboratory on a redundant and identical in every detail test system then double-tracked using high-resolution oscillograph technology…
…and the self-created, MCU-based measurement setup for motherboards graphics cards (pictures below), where in the end the thermographic infrared images are also taken with a high-resolution industrial camera in an air-conditioned room. NVIDIA’s PCAT system is also used.
I have also summarized the individual components of the test system in tabular form:
Test System and Equipment |
|
---|---|
Hardware: |
MSI MEG X570 Godlike MSI MEG Z490 Unify 2x 16 GB Corsair Vengeance RGB Pro MSI GeForce RTX 3090 Gaming X |
Cooling: |
Alphacool Eisblock XPX Pro Alphacool Eiswolf (modified) Alphacool Subzero |
Case: |
Raijintek Paean |
Monitor: | BenQ PD3220U |
Power Consumption: |
Oscilloscope-based system: Non-contact direct current measurement on PCIe slot (riser card) Non-contact direct current measurement at the external PCIe power supply Direct voltage measurement at the respective connectors and at the power supply unit 2x Rohde & Schwarz HMO 3054, 500 MHz multichannel oscilloscope with memory function 4x Rohde & Schwarz HZO50, current clamp adapter (1 mA to 30 A, 100 KHz, DC) 4x Rohde & Schwarz HZ355, probe (10:1, 500 MHz) 1x Rohde & Schwarz HMC 8012, HiRes digital multimeter with memory function MCU-based shunt measuring (own build, Powenetics software) NVIDIA PCAT and FrameView 1.1 |
OS: | Windows 10 Pro (all updates, current certified drivers) |
Kommentieren