I've done some undervolt/OC tinkering lately and i've used this Guide to setup the profiles to test – I've achieved the following settings on my 9800X3D:
CO Settings: -11, -16, -16, -15, -15, -9, -23, -19 from Core 0 to 7 respectively, Core 5 and 0 being the best.
Combined with a +150mhz Boost Clock OC these settings seem to clock to 5375mhz without clock stretching at around 1.28v VDDCR_VDD Voltage (SVI3 TFN) Average reading in HWiNFO while using StatusCore as shown in the guide for the Single Core Load.
Now I've tried to undervolt the SoC Voltage from 1.205v to 1.105v and noticed the Clocks are not hitting 5375mhz anymore under Single Core Load.
How does the SoC Voltage corelate to the clock speed at all?
As far as i thought the SoC is influencing the Memory Controller, PCIe lanes (and maybe Chipset?) Voltages.
As for memory, it runs on 6000CL30 (EXPO) and MCLK/FCLK/UCLK is at 3000/2000/3000 respectively which are the default settings. Nothing else has been changed in the BIOS.
This makes me think if i should've checked for 1, 2, 3 and 4 Core Loads aswell as these loads wouldn't be hitting thermal limits and are relevant for gaming.
Any Ideas on this are welcome, would love to know what the SoC Voltage is actually for or if im missing anything important.
EDIT:
-Forgot my System Details
————————————————–
System Specs:
CPU – Ryzen 7 9800X3D
GPU – ASUS TUF RTX 4070 Ti Super
RAM – Kingston 2x32GB 6000CL30
Mainboard – MSI B650 Tomahawk (running latest BIOS Version: 7D75v1M)
PSU – 1500W be quiet! Straight Power 12
2xNVMe SSDs, 1x SATA SSD & 2xHDDs
2 Comments
You’ve come a long way since your previous post.
I suggest you experiment a bit with different CO values where you put load on 2, 3 and 4 cores respectively and see if you notice any difference when you lower/increase Co.
Same with soc voltage. I’ve never seen anyone mention that before. Did you try to increase soc voltage again? Did you get the same clocks as before? Did you try between reboots?
There have been other users who get significantly different results in different loads just between reboots without changing anything.
I spoke to someone who’d get different aida64 read speed every third reboot. Still haven’t figured out why that was.
Did you test stability with 1.05 vsoc?
Probably 1.105 vsoc makes your uclk or fclk run unstable which probably makes your cpu itself unstable/run error correction.