8700k overclocking: Which Vcores are secure? How extended should stress tests last?
8700k overclocking: Which Vcores are secure? How extended should stress tests last?
Hello, I’m currently working on overclocking my 8700k. My aim is to achieve a highly stable 24/7 overclock at 4.8Ghz and a “stable enough” configuration for gaming (if higher FPS is needed), which should remain idle at 5.0Ghz—possibly with an AVX offset of -2. I’m using the following stress tests: prime95 current version, prime95 26.6, Intel Extreme Tuning Utility, and Cinebench.
My questions are:
1. What core counts would be regarded as good and safe for continuous 24/7 overclock?
2. What core counts would you consider safe for occasional overclocking?
3. Which core counts would be deemed unsafe and should be avoided during gaming sessions?
4. How long should I run the stress tests? a) For initial stability check b) For the final test before using it as a 24/7 setup?
5. How do different LLC settings influence stability? Should I experiment with more or less?
6. What temperature should I keep an eye on during the stress test in HWmonitor?
What I’ve discovered so far is that my system remains stable at 4.8Ghz across all cores with a Vcore of 1.27 and LLC 4. It can handle all benchmarks for about 30 minutes without issues. At lower core counts below 1.27, some workers stop in the latest prime95 small FFP test. QUESTION 1: Would tweaking the LLC settings affect that stability?
I can run Cinebench and IET at 5.0Ghz with Vcore 1.35 and LLC 4 (no AVX offset) without any problems. In fact, I can run both tests simultaneously. Prime95 26.6 also works well for at least 30 minutes. However, the current version of prime95 halts workers, and temperatures spike to mid to high 90s. Prime95 26.6 performs better, staying in the mid to high 80s. Adding an AVX offset of -2 might improve those settings to be considered “game stable”?
Thank you!
I believe the chip performed very well.
Here are some statistics from the silicon lottery:
As of 11/11/2017
What percentage of I7-8700k chips can operate at an aggressive vcore of 1.4 or higher and remain delidded?
- 4.9 100%
- 5.0 81%
- 5.1 58%
- 5.2 30%
- 5.3 6%
I would keep your OC unchanged. 1.27 is great; you won’t feel a noticeable extra multiplier or two.
If you apply speedstep, the vcore and multiplier will decrease when the CPU is idle—this is beneficial.
I believe your chip performed very well. Here are some statistics from the silicon lottery:
as of 11/11/2017
percentage of I7-8700k chips that can handle an aggressive vcore of 1.4 or higher and remain delidded at 4.9 100%
81% at 5.0, 58% at 5.1, 30% at 5.2, and 6% at 5.3
I would keep your OC as it is. 1.27 is great; you won’t feel a noticeable extra multiplier or two.
If you enable speedstep, the vcore and multiplier will decrease when the CPU is idle or under light load – which is beneficial.
Thank you. What MSI BIOS/UEFI setting would trigger speedstep? Right now my clock slows down during idle and low loads, but the vcore stays at 1.27. I’d prefer it to drop further on lighter usage.
Also, where did you obtain these numbers/percentages from?
CompuTronix: Read this: Intel Temperature Guide - Appreciate the helpful article. However, it doesn't address questions 4 and 5. The guide explicitly mentions that the tests provided are for temperature measurement, not stress testing. Still, it's a very informative piece.
Silicon lottery is a company that groups processors and offers the better ones at a higher price. That’s how I learned it. Keep in mind these figures apply to delidded chips. Removing them from your system ends the warranty but helps keep the chips cooler. I believe you’re checking the BIOS for adaptive voltage settings—it might be labeled differently on your motherboard.
For stability i prefer performing an evaluation using a single Cinebench run.
Regarding the final check: one program can be completely stable while another may crash after just 5 seconds.
I recommend running Cinebench a few times, doing a short CPU stress test for a few minutes, and then a similar test with Aida 64. If both are stable, you can try games; if no crashes occur within two days, you have a fairly stable OC.
Mjbn1977,
Stability assessments have consistently sparked intense discussion and disagreement, with perspectives differing widely depending on the specific metric being evaluated. Silicon Lottery, referenced by geofelt, employs Asus RealBench for these evaluations. Their previous statements indicated they conduct tests for a minimum of one hour, yet any claims of shorter durations are met with counterarguments advocating for extended periods—potentially up to 24 hours—or alternative methodologies. I wholeheartedly support and strongly suggest using Asus RealBench:
• Asus RealBench - http://rog.asus.com/rog-pro/realbench-v2-leaderboard/
Finstar,
Below is an example of test tools displayed based on TDP under default BIOS configurations:
TDP …
Thermal
Test -
Steady
Workload
129% … Prime95 v29.3 - Small FFT’s (AVX, No Offset)
101
%
Prime95 v26.6 - Small FFT’s
89% … HeavyLoad v3.4.0.234 - Stress CPU
87% … FurMark v1.19.1.0 - CPU Burner
78% … CPU-Z v1.82.0 - Bench - Stress CPU
66% … AIDA64 v5.95.4500 - System Stability Test - Stress CPU
55% … Intel Processor Diagnostic Tool v4.0 - CPU Load
TDP …
Stability
Test -
Fluctuating
Workload (Peak)
123% … OCCT v4.5.1 - CPU: OCCT (AVX, No Offset)
118% … LinX v0.6.5 - Default
116% … IntelBurn Test v2.54 - High
113% … OCCT v4.5.1 - CPU: Linpack (AVX, No Offset)
110% … AIDA64 v5.95.4500 - System Stability Test - Stress FPU
99
%
Asus RealBench v2.56 - Stress Test (AVX, No Offset)
94% … Sandra 2017.09.24.41 - Burn in - Processor Tests
92% … CineBench v15.0 - CPU - Render Test
79% … Intel Extreme Tuning Utility v6.4.1.15 - CPU Stress Test
Each evaluation indicates full CPU utilization as shown in Windows Task Manager, irrespective of the actual workload. Elevated TDP tends to correlate with higher core temperatures. Outcomes will differ based on various factors such as microarchitecture, core count, cache, speed, Turbo Boost, voltage, hyperthreading, instruction sets, BIOS versions, and CPU microcode.
Despite these variations, it becomes clear that CineBench’s CPU Render Test requires a lower TDP workload compared to Asus RealBench, and the Bench Stress CPU Test in CPU-Z is only around 78%. This lower demand can create a misleading impression of stability due to reduced core temperatures. Prime95 version 26.6 Small FFT’s stands out as an accurate representation of Intel’s proprietary conditions. It is also the tool utilized by Real Temp for monitoring core temperatures.
All assessments confirm 100% CPU utilization in Windows Task Manager, regardless of workload actuality. Greater TDP generally leads to increased core temperatures. Results are influenced by numerous parameters including architecture, core count, cache, speed, boost settings, voltage, hyperthreading, instruction sets, BIOS updates, and microcode.
CompuTronix :
mjbn1977,
Stability assessments have consistently sparked intense discussion and disagreement. Perspectives differ widely depending on the specific purpose. Silicon Lottery, referenced by geofelt, employs Asus RealBench for such evaluations. Their site indicates testing sessions last a minimum of one hour, yet any assertion of shorter duration is met with rebuttal using reasoning that supports extended testing or alternative methods. I strongly endorse Asus RealBench.
Finstar,
Here’s an example of test tools displayed based on TDP at default BIOS configurations:
TDP …
Thermal
Test -
Steady
Workload
129% … Prime95 v29.3 - Small FFT’s (AVX, No Offset)
101
%
Prime95 v26.6 - Small FFT’s
89% … HeavyLoad v3.4.0.234 - Stress CPU
87% … FurMark v1.19.1.0 - CPU Burner
78% … CPU-Z v1.82.0 - Bench - Stress CPU
66% … AIDA64 v5.95.4500 - System Stability Test - Stress CPU
55% … Intel Processor Diagnostic Tool v4.0 - CPU Load
TDP …
Stability
Test -
Fluctuating
Workload (Peak)
123% … OCCT v4.5.1 - CPU: OCCT (AVX, No Offset)
118% … LinX v0.6.5 - Default
116% … IntelBurn Test v2.54 - High
113% … OCCT v4.5.1 - CPU: Linpack (AVX, No Offset)
110% … AIDA64 v5.95.4500 - System Stability Test - Stress FPU
99
%
Asus RealBench v2.56 - Stress Test (AVX, No Offset)
94% … Sandra 2017.09.24.41 - Burn in - Processor Tests
92% … CineBench v15.0 - CPU - Render Test
79% … Intel Extreme Tuning Utility v6.4.1.15 - CPU Stress Test
All evaluations indicate 100% CPU utilization in Windows Task Manager, irrespective of actual workload. Elevated TDP correlates with increased core temperatures. Outcomes depend on various factors such as microarchitecture, core count, cache, speed, turbo settings, voltage, hyperthreading, instruction sets, BIOS versions, and CPU microcode.
Nonetheless, the CineBench CPU Render Test demonstrates a lower TDP workload compared to Asus RealBench, while the CPU-Z Bench Stress CPU test shows only about 78% load—relatively light. This may create a misleading impression of stability due to low core temperatures. Prime95 version 26.6 Small FFT’s is considered ideal for thermal assessments. It closely mimics Intel’s proprietary conditions and is also the tool used by Real Temp for Core temperature sensor testing.
And still, an OC that Prime95 26.6 showed as fully stable collapsed within five minutes of AIDA64 2017.09.24.41 reporting a burn-in test result, even on an i5 6600k with only 60% CPU load. The same chip also hit higher temperatures during a single Cinebench run than Real Temp’s stress test after ten minutes. Yet, this might simply be a coincidence—let’s examine my older i5 4690k.
I completed a full hour of Prime95 on that processor without any crashes, yet it would still fail in Dota 2 every few hours at around 60% CPU usage. You claim expertise, but you overlook how different architectures expose unique vulnerabilities. Sure, Prime95 is excellent for measuring maximum temperatures and board power delivery, but it falls short when assessing whether a CPU is hitting its physical limits. It’s wise to employ a wide range of stress tests to uncover as many weaknesses as possible.
CompuTronix :
mjbn1977,
Stability checks have consistently sparked strong disagreement. Perspectives differ widely depending on the purpose of the test. Silicon Lottery, referenced by geofelt, employs Asus RealBench and claims to conduct tests for a minimum of one hour. However, any assertion of shorter duration is met with counterarguments advocating for extended periods or alternative methods. I strongly endorse Asus RealBench.
Finstar,
Here are examples of test tools displayed based on TDP at default BIOS configurations:
TDP …
Thermal
Test -
Steady
Workload
129% … Prime95 v29.3 - Small FFT’s (AVX, No Offset)
101
%
Prime95 v26.6 - Small FFT’s
89% … HeavyLoad v3.4.0.234 - Stress CPU
87% … FurMark v1.19.1.0 - CPU Burner
78% … CPU-Z v1.82.0 - Bench - Stress CPU
66% … AIDA64 v5.95.4500 - System Stability Test - Stress CPU
55% … Intel Processor Diagnostic Tool v4.0 - CPU Load
TDP …
Stability
Test -
Fluctuating
Workload (Peak)
123% … OCCT v4.5.1 - CPU: OCCT (AVX, No Offset)
118% … LinX v0.6.5 - Default
116% … IntelBurn Test v2.54 - High
113% … OCCT v4.5.1 - CPU: Linpack (AVX, No Offset)
110% … AIDA64 v5.95.4500 - System Stability Test - Stress FPU
99
%
Asus RealBench v2.56 - Stress Test (AVX, No Offset)
94% … Sandra 2017.09.24.41 - Burn in - Processor Tests
92% … CineBench v15.0 - CPU - Render Test
79% … Intel Extreme Tuning Utility v6.4.1.15 - CPU Stress Test
All evaluations indicate 100% CPU utilization in Windows Task Manager, irrespective of actual workload. Elevated TDP correlates with increased core temperatures. Outcomes depend on various factors such as microarchitecture, core count, cache, speed, turbo settings, voltage, hyperthreading, instruction sets, BIOS versions, and CPU microcode.
Nonetheless, the CineBench stress test demonstrates a lower TDP workload compared to Asus RealBench, and the CPU-Z benchmark stress results are modest (~78%), potentially misleading users into thinking performance is secure due to low temperatures. Prime95 v26.6 Small FFT’s suits thermal evaluation well, while Real Temp relies on it for core temperature checks.
And still an OCCT version that 1 hour of Prime95 showed stable performance collapsed within minutes when running AIDA64? (i5 6600k)
The same chip also hit higher temperatures during a single Cinebench run than 10 minutes of Real Temp stress. Yet, this might simply be a coincidence—let’s examine my own i5 4690k.
I performed a full hour of Prime95 on that processor without any crashes, yet it would still crash every few hours under only 60% load. You label yourself an expert, but you overlook how different architectures expose unique vulnerabilities. Sure, Prime95 is excellent for thermal profiling, and Asus RealBench excels in stability checks.
It’s worth noting that Prime95 v26.6 Small FFT’s isn’t the best fit for CPU stress evaluation, nor is it ideal for benchmarking under heavy loads. The CineBench CPU render test offers a lighter workload, which can give a false impression of reliability.
Please note that I’ve emphasized the points clearly. My experience spans nearly 11 years, during which I’ve dedicated over 5,500 hours to thorough research and hands-on testing to back my assertions. I’m also the author of the Intel Temperature Guide, which stands as a verified resource—not just a claim.
Our Forum Rules encourage members to review the key points before commenting. If you’re unfamiliar with the guide, take the time to read it; it may contain valuable insights. A link is available in my signature.
Respectfully,
Your feedback raises concerns. I’ve consistently worked on this subject for almost a decade, investing significant effort into research and testing. My claim is supported by extensive data and personal experience. I assure you that the Intel Temperature Guide is a reliable reference, not merely an assertion. It’s prominently featured in the CPU forum and indexed by Google. If you haven’t encountered it yet, consider reading it for useful information. A link is provided in my signature.