High VCore Concerns - Threadripper 3960x ASRock sTR4 Creator

[Greywolf411]

I just recently finished my first ever water cooling build, previously only built air cooled/overclocked machines.

 

As I'm in the VFX industry and wanting a new home workstation I went with the Threadripper 3960x, and apart from the strange, magical, and frustrating world of water cooling it has been an easy build. However, initially straight out of the BIOS gate I noticed that in Ryzen Master as well as HWiNFO64 that my vCore / peak core voltages were seemingly way high. I have since set the vCore to be 1.25 volts in BIOS which made it show up in CPU-Z and the ASRock A-Tuning software as 1.248v, however, the others still show really much higher vCores.

 

I'm wondering what the heck is going on, and also still trying to wrap myself around dealing with PBO and overclocking with this new chip as I've been out of practice as my old computer was 7 year old i7-3370k oc'ed.

 

Here's a screen capture as well as some finished build shots. Thanks in advance.

[Jurrunio]

Use Hwinfo

 

high voltage at low loads (low current) is normal, what matters is reasonable voltage during full load (high current)

[Greywolf411]

Really? That high? I mean, ultimately the temperatures aren't an issue. Cinebench R20 I'm going from ~37/38 C idle to ~63 C with PBO enabled, seeing clocks around 4,150mhz and scoring ~13,500pts. Just unnerving to me to have the voltage that high. I tried fiddling with the LLC properties in BIOS and still tweaking.

[Meat_PoPsiclez]

Bone stock my 3800x was hitting 1.4v at idle, and with pbo enabled I saw 1.5v (I nearly s*at myself), idle temps were obviously poor (especially given I still have the stock wraith prism cooler on it).  I don't think it's related to llc. Going from no to full load (cinebench r20) with my static overclock sees a change of core voltage <0.01v
I've dialed in a 4.2/4.35ghz (ccx0/ccx1) static clock with ~1.25v (can't recall the exact voltage but within a step of that) and my idle and full load temps are lower than what I was getting with pb/pbo. Yes I've sacrificed some single threaded performance, but I'm content for now.

Theory time: PB can allow the cpu to use these high voltages safely because at idle no core is actually receiving full voltage for more than a few moments due to core parking. As soon as any load is present the voltages drop significantly. Why run these high voltages, I have no idea. The chip can clearly hit it's peak clock with much less than 1.5v, so I'm at a loss.

[Greywolf411]

That's what I started to assume as well, but came to the same conclusion of being at a loss. My newest oddity is getting over 14k in R20 when I had my Corsair Vengeance Pro RGB set to default ddr4 settings from a reset of the BIOS. No xmp. If I put in xmp I get around 13.5k, sometimes a bit higher, sometimes lower. But when I hit 14,147 I was surprised and wondering why that's happening. I also noticed in HWinfo64, that the Maximum clock on my cores with the default settings and PBO enabled in bios yielded, at least at the time of the 14k R20, 7 cores hitting 4.5ghz, 1 at 4.525ghz, and 1 at 4.55 ghz. I haven't been able to get that many cores showing the ability to hit 4.5ghz on any runs. Usually smatterings of 4.4ghz, 4.425ghz, some 4.475ghz.

 

When R20 hit the 14,147 number, it was showing 1.3 to 1.319 volts on the cores with speeds from 4,150ghz to 4,175ghz. In times after I've had it hit those same areas but not get the same results, ie...~13.5k

 

I did benchmark the memory in SiSoft and the xmp profile I was able to get 72 GB/s vs 50 GB/s with the non-XMP settings. So, at least there the XMP is doing some good.

 

That along with working through my first water cooling tweaks has and continues to be an interesting adventure.