The issue with MSI Z97 Gaming 3 overclocking has been resolved.
The issue with MSI Z97 Gaming 3 overclocking has been resolved.
Recently I experienced a CPU delid and used liquid metal, which significantly lowered my temperatures. This motivated me to attempt a higher overclock since previous temperature constraints limited me. For about two years, my i7 4770K operated perfectly at 1.25VID (1.264v Vcore peak under load) while playing games with a 980Ti, and even my Titan X performed without issues. However, after deciding to delid for the sake of it, I encountered problems when trying to reach above 4.6GHz. After several stress tests—Intel Burn Test 30 runs, AIDA64 stress CPU & FPU, Prime95 v26.6 with Small FFTs, and around 7 Cinebench R15 executions—I successfully achieved a stable overclock at 4.6GHz, 1.336v (peak voltage), and 1.900v VCCIN. When attempting to push above 4.7GHz, no matter the voltage applied, stability failed even at 1.5v. I didn’t adjust my ring ratio while trying to increase speed, but I managed a 4.5GHz frequency at 1.35v. It’s unusual for my CPU to hit such a ceiling suddenly, making it hard to rule out the CPU as the problem. My RAM passed stress tests and is functioning properly. The BIOS on my Z97 Gaming 3 is up to date (v2.10). Temperatures are stable, and my cooler is the NZXT Kraken X62 280mm AIO. Can someone help resolve this issue? It’s really frustrating.
It seems like adding another 100MHz would be quite a hassle, especially considering the increased power consumption and heat generation.
I would consider reducing your memory speeds back to 1600mhz and attempt to push the CPU further. With higher memory clocks, you might need to increase SA, Analog and Digital IO voltages. I've raised my SA voltage, Analog and Digital IO voltage to 1.3v on each side (with CPU IOA/IOD Voltage Boost set to +50mV) and increased my RAM voltage to 1.770v in the bios. I kept XMP Profile 1 (2400MHz 11-13-14-32) active, preferring to maintain RAM speeds at 2400MHz @4.6GHz rather than slowing it down by an additional 100MHz on the core. I doubt this improvement would be worth the effort.
In my BIOS, there are several options I'm unsure about, such as adjusting CPU Phase Control settings. I left it disabled. Research suggests CPU Vdroop Offset Control should reduce voltage under load, but it hasn't shown any effect even at a gentle +12.5% load. So I've kept it at the default +100.
For protection features, I set CPU Over Voltage Protection to 600mV, CPU Under Voltage Protection to 140%, and Phase Over Current Protection to disabled. CPU Switching Frequency is set to Auto, VRM Over Temperature Protection and Shutdown are both disabled, Digital Compensation is x1.2, and Transient Boost is enabled.
In CPU Specifications, I've defined long and short duration power limits at 4096, CPU current limit at 1023, and VR Iout Slope at 1023. Internal VR OVP OCP Protection is disabled, internal VR Efficiency Management is enabled, and internal VR switching frequency is set to +6%. I've configured VCCIN voltage to 2.240V and activated SVID Communication. Should I adjust any of these settings?
Well it seems like you know what you are doing, them settings look fine, most of them i don't really touch when overclocking unless i really need to.
1.3v on them sub voltages is a little to high, 1.25v is considered safe for testing and 1.15v is considered safe for 24/7 use.
Overclocking both the CPU and Memory at the same time is usually not a good idea, start with the CPU and find it's limits then do the Memory, that's how it's usually done.
You might be hitting a wall with that 4.7Ghz OC, but again, it could be the Memory OC holding it back.
Here's a graph of voltages:
https://imgur.com/a/vS6WMHj
I can't remember where i found the image, it's been so long, it could of been a guide from der8auer but i could be wrong.
ShadyHamster:
The settings look good overall. When overclocking, I usually don’t touch them unless necessary. For the 1.3v on sub voltages, it’s a bit too high; 1.25v is safe for testing and 1.15v is fine for continuous use. Overclocking CPU and Memory together isn’t ideal—start with the CPU to see its limits before moving to Memory. You might be hitting a ceiling with that 4.7Ghz boost, but it could also be the Memory setting limiting performance. Here’s a graph of voltages: https://imgur.com/a/vS6WMHj. I don’t remember where I found the image, maybe it was from der8auer, though I’m not sure. Thanks for the info, it helps have a baseline for safe max voltages.
I believe I’m nearing success. I managed to run 33 tests on my 4770K @4.7GHz with a 1.424v Vcore and 2.050v VCCIN without any crashes, except the last one. This is when things start to become unusual. Regardless of the voltage I set for Vcore or VCCIN, it always causes a blue screen, and this happens faster on both IBT and Cinebench R15 no matter what Vcore I choose.
Temperatures remain under 95°C, which isn’t too bad given that Intel Burn Test tends to make CPUs run hot.
When the Vcore exceeds 1.45v, other stress tests become necessary because thermal throttling occurs. I’ve tried Cinebench R15, which is good for quick stability checks, but around the 5th or 6th run with this CPU setup, it still causes a blue screen. I’m not sure if this improves things, but the error message keeps showing “CLOCK_WATCHDOG_TIMEOUT.”
I plan to tweak my BIOS settings further to see if I can get a stable OC at 4.7GHz. Here are the changes I made in my BIOS.
I think I've figured it out. I've successfully set up my 4770K stable at 4.7GHz with a 1.464V Vcore and 2.300V VCCIN. I can run 15 Cinebench R15 tests (5 real-time) without any crashes. My hottest core reached 83°C. Using 2.250V VCCIN helped prevent crashes, while enabling Internal VR Efficiency Management, Transient Boost, lowering IO Digital Voltage to 1.208V, and keeping ring voltage at 1.200V in Auto mode contributed to stability. I'm not sure if raising the ring voltage further helped, but it did. I'll keep monitoring with Prime95 v26.6, small FFTs, and a few games.
It seems like adding another 100MHz would be a significant hassle, especially considering the increased power consumption and heat generation.
It seems like overcomplicating things with an extra 100MHz would be too costly given the increased power consumption and heat generation. After some investigation, it appears the issue might lie with the motherboard itself. I discovered a discussion from two years ago where someone shared that they managed to reach 4.6GHz on their older Z87 Hero motherboard, which had better power phases and overall performance compared to their current board. They also found a related thread mentioning that budget boards typically have fewer power phases, making it harder to achieve higher overclocks. I suspect you might not realize the impact until you try it yourself. Eventually, upgrading to a new motherboard could help, especially if it supports lane splitting and better overclocking capabilities, allowing your components to run more efficiently without being limited by the current hardware. I won't regret trying, as my present board lacks features like lane splitting that would enhance performance further. I'll update once I have a new board.