The Intel i7 9700k failed all stress tests because of excessive heat when using Dark Rock Pro 4.
The Intel i7 9700k failed all stress tests because of excessive heat when using Dark Rock Pro 4.
Thank you for your response. I've observed that mx4 tends to dry out significantly. I recall using thermal paste regularly, but now it seems to dry out quickly. Your suggestions are appreciated.
I've never experienced MX-4 drying out before. AS5 is fine, but MX-4 doesn't hold up that well. This paste seems durable enough to remain effective for over six years before it begins to lose thermal conductivity due to chemical changes. The dried paste performs just as well when wet as when dry, provided it's applied while moist. The actual material isn't known for its thermal conductivity; instead, it mainly acts as a medium to hold the various components together—silicates, silver, carbon, diamonds, etc.—which are responsible for heat transfer. The moisture level of the paste doesn't affect how well these materials conduct heat.
The main problem with dried pastes (such as AS5) is when the cooler experiences a sudden temperature drop and the dried paste loses its surface adhesion. Once dried, it can't reattach, whereas wet paste can. A 9700k at stock performance isn't enough to overcome a DRP4. This issue often arises from the motherboard vendor's default settings being enabled by mistake. The 9700k runs at 125 watts with turbo activated under full core load. The DRP4 is a 250-watt cooler, and it requires overclocking beyond 5.0GHz to reach around 200 watts.
Stop experimenting with LLC. Stick with Auto if you're not overclocking. LLC adds extra voltage, which can cause a temporary drop in power delivery (VDROop). This compensates for the brief interruption, preventing crashes from voltage loss. However, this added voltage affects both peak and low draw—if LLC is set to 0.3V while your voltage core is configured for 1.3V, the CPU won't hit its full voltage if it drops to 1.0V, but at higher voltages (like 1.6V), the temperature spikes significantly.
My opinion on why temperatures rise is that the motherboard is set to PL2 on a permanent performance mode, combined with cooler mounting problems and high CPU voltages caused by LLC or other voltage adjustments.
The sole purpose of playing with the LLC is to lower voltage, not raise it, since on auto my CPU demands unusual voltages between 1.38 and 1.45 at standard settings. Appreciate the detailed response!
The voltage won't drop with LLC. To reduce voltage, use VID/offset. The range 1.38-1.45 is reasonable; single cores require higher voltages than multi-core ones when no static OC is active since the boost is greater. Silicon differences mean each CPU has its own voltage requirements—Intel and AMD tend to set defaults higher to ensure stability, so both are safe. You'll still see a 1.45v demand even if the CPU only needs 1.28v. Vcore is the actual voltage, while VID is the supply. Lowering VID/offset reduces supply and pushes the CPU to use less voltage. For example, with a -0.1v offset, you'd see 1.35v, but the CPU still uses 1.28v. Applying LLC with a -0.2v offset would push supply to 1.25v, triggering auto-adjustment to raise it above 1.28v. A low LLC setting might not bridge the gap and could cause crashes. It's complex and time-consuming, especially during testing, but understanding CPU-specific needs is essential.
Turn off AVX completely. AVX, AVX2 and AVX-512 are sophisticated vector analysis tools. There might be a limited use of AVX in some games, it performs better than regular instructions for particle calculations after explosions, but it's very demanding on the CPU. AVX2/AVX-512 is utilized by professionals such as LucasArts during game development, not within the games themselves. I haven't come across any games that employ AVX2.
So essentially, you're only simulating a load close to 135% while the system is at 100%. It's similar to lifting your arm: a full effort at 100% still requires 100% effort for the bicep, but an unnecessary increase just to make it look stronger. This is pointless.
You're pushing around 135% heat from a CPU that can only handle up to 100%. Occt also uses a different load variation, and examining the usage graph shows inconsistent results as the system changes speed. Temperatures will fluctuate as Occt shifts between 80% and 135% CPU usage.
Occt and Aida64 are good for stability testing but not suitable for thermal analysis. For thermal testing, use Prime95 Small FFT (without AVX enabled) as it provides a consistent 100% load, representing the worst-case scenario. For GPU testing, consider Firestrike or MSI Kombuster donut.
I understand, thank you.
The main concern is that with the past 4790k and DRP3 I managed to pass prime using AVX at normal temperatures, but with 9700k DRP4 I can't anymore.
Additionally, I’m a bit concerned about unstable performance in some games. It seems the game RUST drops every second, while others don’t.
I’ve also been dealing with issues where my PC stays off for hours—like overnight—and the first boot doesn’t pass it. The fans spin as if the system is working, but the CPU light turns red and there’s no signal on the monitor. Restarting fixes it, but other boots work fine now.
This has been happening consistently every morning; last week, even the first boot worked. I thought replacing the CMOS battery fixed it, but it came back quickly.
I’m wondering if the CMOS is stuck or if the CPU is malfunctioning.
My motherboard has already been sent to technicians and tested. I’ve replaced the PSU with a new platinum+ 750w Corsair unit.
So, I’m guessing the CPU or RAM might be the issue, or maybe my SSD has a 5-year-old M.2 slot with 65% health—it could be acting up sometimes but I don’t mind replacing it.
Replacing the CPU scares me.