Motherboard and impact on CTR
Motherboard and impact on CTR
Finally, we get to my favorite topic. VRM. The very foundation that is most frequently underestimated. That's the foundation for high acceleration/optimization results. Since CTR fault-tolerance was a priority, I used the ASUS ROG STRIX B550-E GAMING for most tests.
The representative of the middle segment on the chipset B550, which has everything to become the basis of the PC for many years. Why so? Let's look at an example of overclocking the same AMD Ryzen 9 3900X processor, with a fixed voltage value (1225 mV) on two different boards from ASUS.
As you can see the boards based on the new B550 chipset have shown different CTR results. Why did this happen? Firstly, the boards had a different number of phases, in the case of ASUS TUF GAMING B550M-PLUS - the board has 4 strengthened phases, and ASUS ROG STRIX B550-E GAMING as much as 7. The following picture just shows the "naked" VRM ASUS ROG STRIX B550-E GAMING.
The second is the difference between VCC MOSFETs: Vishay SiC639 (50A) and Intelli-Phase MP86992 (70A). That is, at the moment we already have a significant difference in the power subsystems. It would seem that this is it, but no. Every representative of VCC MOSFET has as significant characteristics that will affect the overclocking of the processor. Since there is no documentation yet on Intelli-Phase MP86992, which are used in ASUS ROG STRIX B550-E GAMING, let's consider VCC MOSFET of similar class - Infineon PowIRstage IR3555. In the next picture, 2 graphs are combined, the first one is the efficiency relative to the currently used (red line), and the second one is the power loss relative to the current (black line).
At the red cross, I noted the most optimal operation mode of VCC MOSFET. That is, from the maximum 60A load of 15-17A per 1 VCC MOSFET will be the most energy-efficient mode, but with a VCC MOSFET temperature of 25 degrees. Further heating of VCC MOSFET will have a negative impact on the efficiency and power of VRM as a whole. To sum up, the value that is documented can be safely divided by 2 - 2.5 to get the real value. In the case of ASUS ROG STRIX B550-E GAMING we'll get nominal (14*16) = ~220A (recall that 3950X eats 105-140A in the stock), and in the case of ASUS TUF GAMING B550M-PLUS about 80A (8*10). Of course, both motherboards will cope with the high-end processors, but due to a number of physical properties on one of the boards, the potential of the high-end processor will be impossible to unlock. In the future, when you are going to choose a motherboard, pay attention to the VRM temperature of the board, the lower it is, the greater the overclocking potential can be.
Another important element is that the motherboard should have is a wide range of phases and LLC options.
ASUS boards have a bonus in this regard, which allows the user to customize VRM individually for their tasks. During internal testing, in addition to ASUS motherboards, the CTR project demonstrated excellent MSI compatibility. Gigabyte was slightly behind because in Auto mode Vdroop was far from the recommended range, which is 1-2.3%. Despite the excellent VRM components, ASRock boards in most cases failed to meet the challenge set by CTR. Vdroop "floated" within the range of 6 - 3%. Of course, this can be fixed in microcode, but it's up to ASRock to meet the users. Unfortunately, I did not have a BIOSTAR board.