Test Setup
Please read our test setup information carefully below, we are using Windows 11 Pro, and each testbed is customized for each CPU platform being compared. There are common components such as cooling and power.
System Config
Take note that the same G.SKILL DDR5-6000 memory kit is being used on the Intel Alder Lake platform and AMD AM5 platform. We are using the same memory profile on both systems at the same frequency and timings. Naturally, for the AMD AM4 system, we are using a fast DDR4-3600 kit. We have updated the BIOSs on every motherboard to the absolute latest BIOS versions. We are using the same MSI MEG CORELIQUID S360 AIO Cooler on all CPUs. We are using the same ASUS ROG STRIX 1000W Gold PSU on all CPUs. We are also using the same GeForce RTX 3090 Ti Founders Edition video card on all systems. We have updated the chipset drivers to the latest versions and using the AMD-supplied drivers for the AM5 system.
Note that we installed a fresh copy of Windows 11 Pro and applications and games on each platform tested (AM4, Alder Lake, AM5). With each new platform, we set up, a new fresh copy of Windows and games was put on. Note that we are using Windows 11 Pro 21H2 on all platforms, the minor version number did change on the AM5 platform because of a new update, but it only changed the minor number, not the primary 21H2 version. We are still using 21H2 NOT the newer 22H2 (2022 Update) that has just been released on September 20th, the minor number update did not in any way change performance, as the same major version of Windows 11 is being used on all platforms.
Also, take note of the motherboard settings we have disabled. Firstly, we did set each motherboard to its “Optimized Defaults” before testing each CPU. We utilize the default motherboard settings across all CPUs. All of our motherboards are ASUS motherboards, so they have similar defaults in regards to power and multi-core enhancement.
However, we did specifically disable PCI-Express Resizable BAR in the BIOS. The reason for this is so that we can isolate CPU performance only when it comes to gaming performance differences. We did not want Resizable BAR to add to, or interfere with the CPU performance in games. We wanted to isolate the CPU differences so you see what the real CPU performance is bringing to the table, not what PCIe Resizable BAR is adding to it.
The same reasoning is also why we disabled Hardware Accelerated GPU Scheduling on the GeForce RTX 3090 Ti FE in Windows. In addition, Virtual Based Security is also disabled so it doesn’t interfere with performance. We are using the “High Performance” power profile on all CPUs.
Again, we want to isolate CPU performance, and this was the goal of turning these features off. In our GPU reviews, we ALWAYS have Resizable BAR enabled, so if you want to see what performance in games is with that feature enabled, all of our GPU reviews have it enabled and will continue to in the future.
AMD AM4 Comparison Platform
For all of our AMD AM4 platform comparisons, we are using a Ryzen 9 5950X, Ryzen 9 5900X, Ryzen 7 5800X3D, Ryzen 7 5800X, and Ryzen 5 5600X CPUs in all of our comparison graphs. The motherboard being used is an ASUS ROG CROSSHAIR VIII HERO Wi-Fi X570 motherboard. RAM is Crucial Ballistix DDR4 BL2K8G36C16U4B (8GBx2) at 3600MHz with timings of CL16-18-18-36. The same MSI MEG CORELIQUID S360 AIO is being used.
Intel Alder Lake Comparison Platform
For all of our Intel Alder Lake platform comparisons, we are using Intel Core i9-12900K, Core i7-12700K, and Core i5-12600K CPUs in all of our comparison graphs. The motherboard being used is an ASUS ROG MAXIMUS Z690 APEX motherboard. We are using the same DDR5 memory on the Intel platform that we used on the AMD AM4 platform. In the BIOS the EXPO timings are revealed and labeled as D.O.C.P. profiles, so we can set that profile and it will be exactly the same as we have it set on the new AMD AM5 platform. This means frequency and timings are the same between the Alder Lake and AM5 platforms. The same MSI MEG CORELIQUID S360 AIO is being used.
Discussion (13 replies)
Join Discussion →My opinion. If you're currently on Zen2+ or Zen3 on anything less than a 5800X3D, and all you do is game, then the 5800X3D is the best bang for the buck upgrade. That CPU is a gaming powerhouse.
I kind of have mixed feelings about these mid-tier Zen4 CPU's. At least from a gaming perspective. It would cost significantly more to build a 7600X system, for inarguably less gaming performance, than it would be to build a cheaper 5800X3D system. Especially with AMD saying they were going to continue development of X3D AM4 parts.
They kind of put themselves in a tough situation with their product stack.
Yes, for gaming, I agree.
The question one has to ask is what is their intention and use case. What workloads will they need? If all you do is gaming on the PC, then you have your answer. However, if you do much more, or throw other certain kinds of workloads at the CPU, then perhaps there is value. It depends on your needs.
Yar, tough sell to migrate from AM4 to AM5 currently (though if I was building new I'd for sure go AM5).
Looking forward to see how the 7000X3D's perform, but think I won't be migrating until the second revisions come out.
It's nice to see that these lower cost 6C/12T chips can still keep up in gaming, at least if you aren't trying to do TOO much in the background (like encode streams)
I bet these will find their way into a lot of decent gaming machines.
I'm betting they won't find their way in to much of anything until cheaper X670 boards come out, or B-series is released. The value isn't quite there yet. Especially with DDR5 costs, PSU and cooling requirements.
Yep, I think the single chip B650/E will be the way to go. All those extra USB and PCIe connections on the expensive X670/E will be limited by the bridge between the 2 chipsets.
B650 + 7_00X3D will be the best early AM5 for gaming.
If you game at 4k you are better off getting a 5600X.
BTW thanks a lot for using settings one would actually use for gaming. Not Low/mid 720p/1080p settings like many reviews out there.
I think both have value.
Using minimum settings and low resolution you see the potential performance a CPU might have if you have a really fast GPU.
This is valuable for theoretical comparisons and to get an idea how "future proof" the CPU is.
Benchmarking a CPU at real world settings is easier to understand for those who don't get the fact that they will in all likelihood be GPU limited, and helps minimize the arguing over irrelevant performance differences no one will ever actually experience in their CPU due to this GPU limitation.
Honestly, if I had to pick just one, I'd pick the CPU benchmark at minimum settings and low resolution. Using that and GPU benchmarks I can project what my performance will be in almost any combination of hardware. if I only have the latter, I can only really predict how it will perform in that one CPU/GPU combination.
So yeah. I like to see both, but if I had to pick just one, I'd pick the "everything graphical at minimum settings and resolution using an overkill top end GPU" just to try to isolate the CPU performance, because that is honestly more useful.
If I have that CPU isolated data I can look at a CPU benchmark and at a GPU benchmark separately and just pick the minimum framerate of the two, and know that is likely a reasonable prediction of how they will perform together. This is of much greater value to me than real world benchmarks on a system with a GPU I may or may not ever use.
I also see value in both results, low rez, and high rez, for both reasons. That's why I include 1080p, 1440p, and 4K, I think it's all relevant, and important to see scaling between resolutions, comparing CPUs. I'm rather excited about the fact that we represent that, that way anyone reading can hone in on the resolution they take value from, and make up their own mind. I think it gives an overall bigger picture of CPU performance, and I'm very happy people get value out of that in our reviews.
More than the resolution, my issue is low settings. Normally I see cpu benches even at 720p low and those exacerbate the impact of the mobo/cpu.
I recall asking Kyle at several times to do hi-res/high settings on the CPU/Mobo reviews as it would reflect a more real world scenario (which was policy). Never happened.
Thankfully you got all my bases covered.
That's why you look at both the GPU and CPU reviews, and - if you have CPU reviews run at low settings, and GPU reviews with overkill CPU/Platform to avoid it being the limiting factor - you can just take the minimum framerates of each review, and use that as a pretty reasonable predictor of what the combination of that CPU and GPU will look like.
The theory is to make the component under review the bottleneck to see what it is truly capable of so you can understand that particular component, rather than having the results confounded with the system as a whole.
Or just give the reader all the info you can in each review and let them decide what's important to them :p
I'll never argue against more information :p