High Core Issues – Threadripper 3960x ASRock sTR4 Creator
High Core Issues – Threadripper 3960x ASRock sTR4 Creator
I just completed my very first water cooling setup, after working with air cooled or overclocked rigs before. Since I work in VFX and need a modern workstation, I chose the Threadripper 3960x. Besides the unusual, exciting, and sometimes annoying water cooling experience, the build itself was straightforward. Right from the BIOS, I saw that both Ryzen Master and HWiNFO64 displayed unusually high vCore and peak core voltages. After adjusting the BIOS settings to 1.25 volts for the vCore, CPU-Z and ASRock A-Tuning reported a more accurate reading of 1.248v. The other tools still showed much higher values. I’m puzzled by this discrepancy and still learning how to manage PBO and overclocking with this new chip, especially after my old machine was a 7-year-old i7-3370K OC’d. Here’s a screenshot along with some photos of the finished build. Thanks in advance!
Normal operation under low loads with minimal current is typical, focus should be on adequate voltage when full load and high current is applied.
Really? That's quite a jump? In the end, the temps aren't the main concern. I'm testing Cinebench R20 from around 37-38°C idle up to about 63°C with PBO on, clocks near 4,150mhz and it's giving me roughly 13,500 points. It's a bit unsettling having such high voltage settings. I've adjusted the LLC features in the BIOS but keep making small changes.
I was monitoring my 3800x at idle and saw 1.4v when PBO was on; with PBO off I reached 1.5v (close to failure). The low idle temperatures were a concern, especially since the stock wraith prism cooler was still attached. It doesn’t seem connected to the LLC circuit. Switching from no load to full load (using Cinebench R20) caused a drop in core voltage by less than 0.01v. I’ve set up a static clock around 4.2/4.35ghz (CCX0/CCX1) at about 1.25v, which is within a small range. My idle and full load temperatures are now better than with PBO enabled. I’ve had to trade some single-threaded performance, but I’m satisfied. Theory suggests PB lets the CPU safely operate at these voltages during idle because cores aren’t fully powered for long. Once load comes on, voltages fall quickly. Why keep such high voltages? I’m not sure. The chip can easily hit its peak clock with far less than 1.5v, so I’m puzzled.
I initially thought a similar thing too, but ended up feeling uncertain about what was happening. The most unusual thing I observed was earning over 14k in R20 after resetting the BIOS to default DDR4 settings. No XMP applied, yet I saw around 13.5k; sometimes higher, other times lower. At 14,147 it surprised me and made me question the cause. In HWinfo64, with default settings and PBO enabled, I found that at least at that R20 level, seven cores reached 4.5GHz (one at 4.525GHz, one at 4.55GHz). Usually I see around 4.4GHz, 4.425GHz, or 4.475GHz. Once the R20 hit 14,147, it showed 1.3 to 1.319 volts on cores running between 4,150GHz and 4,175GHz. Typically I get 4.4GHz or slightly higher, but rarely 4.5GHz consistently. After that point, even with similar runs, the results didn’t match. Benchmarking memory in SiSoft showed XMP boosting speeds to 72GB/s compared to 50GB/s without it. This suggests XMP helps, and after trying water cooling adjustments, it remains a fascinating process.