Review: ROG Crosshair X670E Extreme

0

CPU overclocking:

Now that we have our memory profile at 6400 MT/s CL30, let’s move on to the CPU overclocking. So, considering the temperatures obtained at the original frequency under Cinebench R23, 84°C, while we are under a custom watercooling, the overclocking margin may be quite limited. At the same time, this is good news since our Ryzen 9 7950X will offer excellent performance from the start. Currently, to overclock our processor, we must either do it directly via the BIOS or via the AMD Ryzen Master software. It is this last solution that we have chosen for these first tests. A small remark before starting, by default our ROG Crosshair X670E Gene offers less voltage to the 7950X (1.21v) which allows us to have lower temperatures, 84°C against 91°C on the X670E Master (1.36v) on a benchmark of Cinebench R23 in multi.

Protocol:

Before starting, it is necessary to have one or more reference scores with your processors at stock frequencies, that is to say, without having made any modification other than having activated the XMP/EXPO profile. It is also very important to have a sheet of paper at hand that will allow you to note down all the tests that you will perform as well as the results or failures. Personally, I have hundreds of sheets of my tests on which I sometimes return to know how the processor behaved on such motherboard or in such test conditions. My own little bible. For H24 overclocking, it will of course be necessary to take more time to fine-tune the voltage in order to provide it with the minimum necessary but also to adjust each core independently if you want to take advantage of the maximum performance.

Load-Line Calibration:

The idea of modifying the Load Line Calibration is that the voltage chosen should be as close as possible to the load voltage. Asus offers several levels of “calibration” and recommends a Level 5 for overclocking . Personally, we opted for a Level 7 in order to have as little VDroop as possible. Basically, I want the voltage chosen to be as close as possible to the load voltage. Here is an example of two different readings I took for testing.

The readings are made with a multimeter at the PCB level of the motherboard. It is very important to know the real voltage brought to the different components but especially to your processor.

Ryzen 9 7950X Operation:

Once again, it is via the OS that we will start our overclocking tests using AMD Ryzen Master. It is a tool that can be used very easily and is therefore very intuitive provided that you have a good understanding of the structure of your processor and here more particularly that of the Ryzen 9 7950X. The Ryzen 9 7950X has two CCDs. Each CCD has a single CCX which has 8 cores. In the case of our 7950X, none of the CCX cores are disabled so that there are 8 cores per CCD and therefore 16 in total. The idea is to see how the CCDs behave when all the cores are used. We saw on page 5 that the CCD0 was more efficient than the CCD1. It will therefore be possible via Ryzen Master to try to push this one higher than the other.

Finally, you can also decide to vary the frequency of each core. The one with a star is the one that performs best of the 8 and the second best is marked with a disk. To overclock, go to profile 1 and select manual overclocking. You can then control either each CCX or each core independently.

Overclocking in different stages:

We will first start the tests at 5000 MHz on the two CCX with a voltage in Ryzen Master of 1.15 volt which when Cinebench R23 multi is launched is fixed at 1.10 volt. The idea is then to test the stability on several runs of Cinebench R23 multithread. If it’s stable, I increase the frequency by 100 MHz while checking the stability again. If the benchmark crashes, I increase the voltage slightly in order to regain a stability allowing to run the benchmark.

The objective is to get a “quick” idea of the maximum benchmarkable frequency. First result with a frequency of 5400 MHz on the 16 cores for a voltage of 1.19 volt and a score of 40766. Things get more complicated after that because the CCD1 seems to be really less performing and therefore limiting. But thanks to AMD Ryzen Master, you can overclock the CCDs independently. So we will continue to push the CCD0. In the end, we were able to benchmark at 5600/5400 MHz (CCD0/CCD1) and reach a score of 41634 points in Cinebench R23. We also took the opportunity to quickly switch to Geekbench3 at 5550/5400 MHz.

We get an excellent score of 41499 points under CinebenchR23 with a temperature of 85°C while at stock we had a score of 39223 for 95°C. As for Geekbench 3, the single core score is lower since our cores are at 5.6 GHz against 5.85 GHz at stock but the multi core score goes from 112664 to 119688 points.

Finally, and this is excellent news, it is possible to deactivate the synchronization between the BCLK frequency and the PCIe frequency via the BIOS. It will therefore be possible to overclock via the BCLK frequency. We will be back very soon with more cold tests with this GENE.