The overclocking process using a Ryzen 9 7950X is exactly the same as the previous Ryzen generations. However AMD has already tuned the turbo algorithm to the ceiling, very little room is left for manual overclocking. This is a good thing for those who have no desire to overclock, but also leaves those who like the hobby somewhat unsatisfied. Gone are the days of heavy all-core overclocks being beneficial, what is left is AMD Ryzen Precision Boost Overdrive. When configured correctly, you can have similar performance to a stock CPU setup, but with a lower power draw. In some cases, upwards of 5% increases in CPU performance are achievable as well. It is worth taking the time to go through the settings and set it up correctly according to your hardware and use case.
If you are not keen on using the BIOS for CPU overclocking, AMD provides a piece of software called Ryzen Master. This is a all-in-one tool for overclocking on Ryzen platforms. You can also have the software try to auto optimize PBO, or use the AMD Eco mode preset to limit the AMD 7950X to 90 Watts. Multi-threaded scores are reasonable considering the lower power target with just over 30,000 pt in Cinebench R23.
The next step was a brute force overclock, after trying out PBO first. AMD says any, and all overclocking (including memory) will void the warranty. Lots of warnings in software are given as well. In any case, if you do plan on overclocking, a media-only overclocking guide suggested not to exceed 1.3 Vcore, and AMD says Tj.Max is 95 °C for the cores. This temperature threshold is for the "lifetime" of the CPU, where you can exceed this up to 115 °C before the system shuts off. Doing so will degrade the CPU over a shorter period of time, but 95 °C is perfectly acceptable, the CPU being designed to operate at that temperature. The good news is that more than 1.3 Vcore will never be practical without first delidding the CPU, or going with an exotic, sub-ambient cooling solution. The bad news is, as you may have guessed, is that the CPU temperatures can easily reach 95 °C with just 1.25 static Vcore, under load. A primitive overclocking session ended with a all-core overclock of 5.3 GHz at 1.25 V. Not very exciting overclocking using this method, but its possible none the less.
HWInfo Sensors
HWInfo Sensors have been requested. Here you go folks!
Memory Overclock
Memory Support (BIOS F2c)
Speed
Ranks / DIMMs
Boots
Stable(?)
Auto Voltages
MCLK Gear
FCLK
DDR5-6000
Single-Rank / 2x
Yes
Yes
Yes
2:1:1
2000 MHz
DDR5-6200
Single-Rank / 2x
Yes
Yes
Yes
2:1:1
2000 MHz
DDR5-6400
Single-Rank / 2x
Sometimes
No
No
2:2:1
2000 MHz
DDR5-6600
Single-Rank / 2x
No
No
No
2:2:1
2000 MHz
DDR5-6200
Single-Rank / 4x
Yes
Yes
No
2:1:1
2000 MHz
DDR5-6000
Dual-Rank / 2x
Yes
Yes
Yes
2:1:1
2000 MHz
DDR5 support for the AM5 platform is somewhat sporadic. With every new platform, there's always growing pains and re-education for consumers. After testing a number of X670 motherboards, it is clear the BIOS and subsequently AMD AGESA version implemented plays a vital role in memory compatibility. The motherboard manufacturers also have a part to play in overall memory support as the PCB quality, PCB layer count, memory training algorithms and auto voltages can all contribute to what works out of the box. Gigabyte B650E AORUS Master memory QVL lists go to DDR5-6600, but so far DDR5-6200 is the highest stable frequency reached in this review (and others). It is expected that memory compatibility will improve with future BIOS updates.
Trying different configurations, it becomes clear that compatibility is one thing to be concerned about early in the AM5 lifecycle. Similar to other vendors, Gigabyte is also at the mercy of AMD AGESA updates. Right now with the current F2c BIOS, it is highly suggested to buy memory DDR5-6000 or below and avoid using four DIMMs. To be fair, Gigabyte is doing much better compared to other vendors, where dual-rank memory didn't work at all and four DIMMs was questionable in a previous motherboard review.
The good news here is when using four DIMMs, a maximum frequency of DDR5-6200 was stabilized. Using that as a baseline, the primay timings were brought down to the lowest possible with 1.5 V (28-38-38-50). By doing so this benchmark time is now within 1 second behind the top Ryzen 7950X CPU y-cruncher 10b score. I'm sure it will be will be crushed by those XOC players soon enough (if not already), but this feels like an accomplishment using ambient cooling. Some more attempts were made after, but without successfully lowering the time.