OC issues i7-4790K on MSI Z97 PC Mate – Weird Issues with VCore, C-States, and Performance Drops

[Patikekfad]

Hey everyone, need help understanding some weird overclocking behavior!

I'm trying to overclock the legendary i7-4790K on an MSI Z97 PC Mate motherboard, and while I’ve had some success, I keep running into weird issues that I can’t fully explain. I am not an expert, so I’ll go step by step so you can see where things get strange and hopefully help me out. I've been able to overclock this motherboard with a G3258 Pentium in the past quite well, from 3.2Ghz to 4.0Ghz, if I remember correctly.

1. The CPU VCore and VID (Max) voltage values don't seem to match. My VCore value is seemingly stuck around 0.9v's (idle or load), even though I set it to 1.25v for example (VCore sensor from the motherboard doesn't seem to change at all, even when I do change the CPU Core voltage in the BIOS). Pre-OC my Cinebench R23 score was 4865, at 4.0Ghz. I managed to OC it (somehow, we'll get to that later), to 4.5Ghz, giving me a score of 5250 in R23. Even in this scenario, the VCore value in HWMonitor, given by the motherboard, remains around ~0.9v. The VID value the CPU reports is (in this 4.5Ghz OC-scenario) 1.20V, which is what I set it to in the BIOS. So here's the question: I know VID is what the CPU requests, but why isn’t the motherboard actually delivering it?

• This motherboard does not have LLC, so I can't really fix potential VDroop that way. I could try my luck with adaptive+offset mode instead of override, on the CPU Core Voltage setting in the BIOS maybe?
• I've tried to increase the VCCIN voltage from 1.80v to 1.90v, but also that didn't seem to help.
• Disabling Intel C-states (Package C State Limit) actually made my computer crash within 5 seconds of the R23 benchmark test.
• I do have the latest drivers installed, V4.11

About that 4.5Ghz OC, I couldn't really get it to work until I changed the VCCIN and CPU Ring voltage to 1.8v and 1.15v respectively, and especially changing that Ring voltage value seemed to make it stable (AIDA64), for at least 45 minutes until I decided to up the clock speed to 4.6Ghz, but I wasn't able to get that stable at any point, even though i feel it should (kept crashing within seconds of putting load on it through R23, but am able to get into windows). Now the weird thing about that 4.5 OC is that, I accidently had enabled Intel C-state and put the Package C State Limit to C6, which should be a deep power saving mode. If I disable or put it on C0 (meaning the CPU runs all the time, I guess?) it actually crashes my 4.5 AIDA64 tested OC within 5 seconds of putting a 100% load on it.


Now there is an built-in OC feature called OC Genie (OG), that overclocked the CPU to 4.4Ghz and disabled the Intel C-states, whilst running the VCore at 1.20v, and actually, it seems pretty damn stable! So what is going on here?

To summarize:

1. Why is my CPU VCore still so much lower than what I set in BIOS?
2. Why does disabling C-States break my overclock, even though Overclock Genie does it just fine?

System Specs for Context:
CPU: Intel i7-4790K
MOBO: MSI Z97 PC Mate (No LLC option, 10 years old)
RAM: 16GB (2x8GB Kingston KHX1866C10D3/8G @ 1.5V)
GPU: MSI GTX 1650 Super
PSU: 650W (10+ years old, might be weak?)
Cooling: Cryorig H7
Monitor: 1080p 60Hz
I also increased the Long and Short Duration Power Limit's to 255 (just as the OG) and the CPU Current limit (in Amperes) at 256A.
I added 2 screenshot of my HWMonitor under load in AIDA64 in the stable 4.5Ghz OC setup, which actually managed to run for 8 hours now!

Thanks a lot, and please let me know if you'd like to know some extra information!

 

[ShrimpBrime]

VID is only relevant to a stock system.

Use the CPU v-core option only, leave VID on auto. I dont remember the values for my 4690K, but I do know they like v-core. And as long as you have the cooling, more v-core is ok.

4.5ghz, maybe 1.3v to 1.35v could be stable 4.5ghz, C-states disabled.