News How to Overclock an Intel CPU: Get the Most MHz from Your Processor

[Admin]

If you have the right Intel processor and motherboard, you can overclock and get more performance from your chip.

How to Overclock an Intel CPU: Get the Most MHz from Your Processor : Read more

 

[Gurg]

Curious why you don't use Intel Extreme Tuning Utility to set your overclock and then the included stress test rather than going through the bios and using non-Intel tests? I've gone the bios route before and its more of a pain to continually tinker with changing the rates by restarting to get to the bios. Once I set my overclock I like occasionally leaving the Intel Extreme system monitor screen open while running benchmarks to see how my PC is running. My 9600K runs at 5.05 all-core. I left my core and cache voltages in the adaptive mode rather than setting a value.

 

[Unolocogringo]

Curious why you don't use Intel Extreme Tuning Utility to set your overclock and then the included stress test rather than going through the bios and using non-Intel tests? I've gone the bios route before and its more of a pain to continually tinker with changing the rates by restarting to get to the bios. Once I set my overclock I like occasionally leaving the Intel Extreme system monitor screen open while running benchmarks to see how my PC is running. My 9600K runs at 5.05 all-core. I left my core and cache voltages in the adaptive mode rather than setting a value.

I'm not the author of this simple piece of information, but I can answer your question.
Almost all overclocking software produces terrible overclocks from an experienced overclocker"s point of view.
Many are just plain unstable at the software's setting.
Almost all apply too much voltage, which is counterproductive for really high desktop overclocks.
Their 10,15,30 minute stress tests are a joke for true stability testing.
Most software overclocks are unstable and will slowly corrupt the hard drive over time.

If he/she had suggested using software overclocking he would of lost all credibility as a author and most comments would be complaints.
Most serious overclocking enthusiast would call him/her out for giving bad advice.

 

[mdd1963]

None of the above list of alleged grievous 'software-overclocker deadly sins' listed above remotely or even applies to Intel's XTU, ergo, listing what alleged issues some other software OC sources (assorted OC Genies/Masters, etc) had or might have caused some 10-15 years ago seems ... irrelevant, at best.

Too high a voltage at default? - a core voltage value or offset, either plus or minus, is/can be applied within XTU, identically as if it had been set in the BIOS; ergo there is no default voltage increase or decrease applied. (THis fault would actually readily apply to any BIOS within any mainboard manufacturer, but, can be quickly overridden within the XTU)

The ability to manipulate base clock speed of 100 MHz is there, but, I don't know many that advocate this on today's mainboards.

The normal Intel power limits may also be tweaked here, as well as applying AVX offsets. and desired clockspeed/multiplier values entered for specific number of cores active. User may adjust Turbo Boost features (on/off, short boost power max, max power enable, boost duration, etc) a lower cache clock speed may be specified if desired.
stress Test - although the XTU does include both CPU/RAM stress test options (which have fluctuating workloads that induce temperature peaks approximately equal to those of CPU-Z's included stress routines), those are to be run manually, and are not used in some 'auto-tuning'/stability test. No one included test is really a test of stability anyway, as what good does it do to pass Prime 95/small FFTs, but, fail in CInebench/Blender, or hard reset in gaming, etc..? With SpeedStep still enabled and Balanced still selected in power plans, the rig still turbos to the high specified maximums, but, downclocks to 800-1200 MHz when loafing at the desktop, etc..

Last but not least, an XTU config file is referenced/accessed very early in each bootup; if the file was not accessed/correctly appended during a clean shutdown, all values manipulated in XTU are set aside/returned to default, quite useful if one has overstepped the bounds of stability in the search for that last 100 MHz, etc...

Intel's XTU has been a delight to work with, a very quick way to access ~90% of what most folks tweak in the BIOS. Are there a few more options in the BIOS? Of course.

 

[jpe1701]

Just a note to the author. In the paragraph talking about setting voltage you say it won't hurt to add more volts. You should word it a little differently I think so that it's not taken literally. Lol

 

[Deicidium369]

AMD Overclocking Guide - long version

You can not overclock an AMD CPU - AMD already overclocked it.
==
AMD Overclocking Guide - short version

You can't.

 

[Unolocogringo]

None of the above list of alleged grievous 'software-overclocker deadly sins' listed above remotely or even applies to Intel's XTU, ergo, listing what alleged issues some other software OC sources (assorted OC Genies/Masters, etc) had or might have caused some 10-15 years ago seems ... irrelevant, at best.

Too high a voltage at default? - a core voltage value or offset, either plus or minus, is/can be applied within XTU, identically as if it had been set in the BIOS; ergo there is no default voltage increase or decrease applied. (THis fault would actually readily apply to any BIOS within any mainboard manufacturer, but, can be quickly overridden within the XTU)

The ability to manipulate base clock speed of 100 MHz is there, but, I don't know many that advocate this on today's mainboards.

The normal Intel power limits may also be tweaked here, as well as applying AVX offsets. and desired clockspeed/multiplier values entered for specific number of cores active. User may adjust Turbo Boost features (on/off, short boost power max, max power enable, boost duration, etc) a lower cache clock speed may be specified if desired.
stress Test - although the XTU does include both CPU/RAM stress test options (which have fluctuating workloads that induce temperature peaks approximately equal to those of CPU-Z's included stress routines), those are to be run manually, and are not used in some 'auto-tuning'/stability test. No one included test is really a test of stability anyway, as what good does it do to pass Prime 95/small FFTs, but, fail in CInebench/Blender, or hard reset in gaming, etc..? With SpeedStep still enabled and Balanced still selected in power plans, the rig still turbos to the high specified maximums, but, downclocks to 800-1200 MHz when loafing at the desktop, etc..

Last but not least, an XTU config file is referenced/accessed very early in each bootup; if the file was not accessed/correctly appended during a clean shutdown, all values manipulated in XTU are set aside/returned to default, quite useful if one has overstepped the bounds of stability in the search for that last 100 MHz, etc...

Intel's XTU has been a delight to work with, a very quick way to access ~90% of what most folks tweak in the BIOS. Are there a few more options in the BIOS? Of course.

Their are still lots of overclocking software available today.
Most published by motherboard makers themselves.Built into the bios and software for windows. Not 10-15 years old.
Brand new in boxes waiting to be purchased.
Those of us who must have 100% stability can not rely on them. My folding comruters are overclocked to stability limits and run 3 months between reboots.
I install updates. shut down for cleaning. take them outside and blow them out with an air compressor. Bring it back in and fire it back up for another 3 months.
They run almost 100% load for three months and repeat.

Never used XTU but it sounds like most other overclocking software that has a manual mode or an automatic OC mode.
Never had good luck with any of them so manual it is for me.Gave up on trying software and went back to bios overclocking many years ago.
Other people might want to learn how and this is a decent beginners guide.

Otherwise it would be a very short article.
start overclocking software.
press start.
put check mark in box to start with windows.
press accept.
done.
end of article.
If the article title was "how to overclock with XTU" that would be different.

 

[mamasan2000]

Number 1 rule when it comes to overclocking: Every silicon is different. Don't you think Intel built the software with that in my mind? There is variation in OC you can reach and therefor Intel probably went with the safer bet, lower clocks and voltage.
They don't want support tickets or returned CPUs, that hurts the bottomline.
You are leaving perf on the table. If you are going to OC, do it right. Manually. Test with multiple tests. If any of them crash, it is not stable.

If you are going to overclock, learn what each setting does.

And if you've noticed, none of the professional overclockers are out of a job. That should tell you something.
Software is not the way to go for OC. No one half-serious uses it.

 

[Gurg]

Number 1 rule when it comes to overclocking: Every silicon is different. Don't you think Intel built the software with that in my mind? There is variation in OC you can reach and therefor Intel probably went with the safer bet, lower clocks and voltage.
They don't want support tickets or returned CPUs, that hurts the bottomline.
You are leaving perf on the table. If you are going to OC, do it right. Manually. Test with multiple tests. If any of them crash, it is not stable.

If you are going to overclock, learn what each setting does.

And if you've noticed, none of the professional overclockers are out of a job. That should tell you something.
Software is not the way to go for OC. No one half-serious uses it.

Intel XTU overclocks the cpu in steps until it fails so as to find the highest possible overclock. Mine would do 5.0 fine, but not 5.1. When I used the advanced tuning I set it to run at 5.0 and so it does all day at max temps generally in seventies. When I tried 5.1 it failed. When reviewers overclock 9600K, most everyone gets to 5.0. I saw just one that made 5.2 and one that could only get 4.9.

TH: "We tapped Corsair's H115i v2 to test our Core i5-9600K, which gave us enough headroom to run at 5 GHz with 1.36V Vcore and Auto Load Line Calibration settings. An AVX offset wasn't needed; our sample maintained ~80°C during AVX workloads. The temperature only reached ~64°C during non-AVX workloads. Although some Core i5-9600K CPUs reportedly run stable at up to 5.2 GHz, we aren't comfortable pushing our chip beyond the "safe" 1.35V limit."

Time Spy Advanced which allows you to search by CPU and GPU shows a handful of 9600K that evidently won the silicon and probably the Z390 motherboard lottery at 5.3. The motherboard factors in because when I tried to double my memory to 32gb with the same identical RAM my system would not run at same settings as before. It shouldn't be my PSU as it is a 1000 watt PSU. I admit I only went with a $120 Z390.

 

[PCWarrior]

With regards to XTU, it does have direct control of BIOS settings –in other words you can directly modify the BIOS settings. It is not a mere software overclock applied at windows startup like the various gpu software OCing tools such as the MSI afterburner. You do not need to run XTU each time you boot your computer. That being said it depends on the motherboard manufacturer or OEM to expose these BIOS controls to XTU. All DIY desktop boards from the big 4 (ASUS, Gigabyte, MSI, ASRock) do so. For OEMs and laptops where they may not expose these controls, XTU can act like traditional software OC tools. Also, some people seem to confuse XTU with auto-overclocking tools like the Intel Performance Maximiser or ASUS’ AI overlocking tool.

 

[PCWarrior]

My Overclocking stress testing procedure

Phase 1
For a quick stability test I use 3-5 consecutive runs of Cinebench R15 and/or Cinebench R20. If the system is unstable it will be revealed though this method about 90-95% of the time. I also check whether the score has increased nearly linearly with the frequency – if it does not it is almost always a sign of an underlying instability. At this stage I perform the tests with fixed voltage. Once I find an overclock that is stable and with predicted score increase in Cinebench I switch to adaptive voltage and rerun the tests.

Phase 2
If phase 1 is successful I pass to the second phase of testing the system further.

1. I first perform a thermal test to check what are the worst-case, steady-state temps and see if my cooling solution is adequate. For that I use the IntelBurnTest and small FFT prime 95 for 15-20 minutes each. These tests will absolutely pummel the cpu and produce maximum heat and highest temps. Best way to monitor temps and power is HWinfo and Intel’s XTU. The latter also shows you whether the cpu is throttling.
2. If I am happy with the temps, I move to further test stability. Intel’s XTU benchmark for starters followed by running the build-in stress tests (CPU and memory) for 15minutes each.
3. Then Realbench as it also includes Blender. I do a couple of runs.
4. Then I test with OCCT were I can I set a temperature limit for the stress test (e.g. 84C) above which the test is automatically terminated. I test for 1hour with the OCCT test and for another 30minutes-45minutes with the Linpack including full memory test.
5. Then I test with the 3D Mark Time Spy by running the benchmark a couple of times followed by a couple of runs of Fire Strike and Fire Strike Extreme. When I run the benchmarks, I look (obviously for crushing but also) for any signs of stuttering and whether there is the expected score increase in the cpu score with the increase in frequency.
6. Subsequently I use Memtest64 for almost the entire RAM capacity (e.g. 30.5GB for 32GB of installed RAM). I do 3-4 full passes.
7. As a final confirmation I may run Prime95 for 4-5 hours, but I find this practice unnecessary as it never produces any error after all the previous stages had successfully passed.
8. Just to guarantee 100% stability, especially if I am aiming for my “mild 365/24/7 Overclock” I may sometimes decide to dial back 100MHz while keeping the voltage the same or set an AVX offset.

Phase 3
The third phase is to try to do per-core overclocking and try to basically ‘rewrite’ Intel’s turbo 2.0 table with higher frequencies. You can do this easily through XTU but I prefer to do it the old fashioned way through the BIOS. So, for example, for an i9 9900K, if my all-core OCed turbo was 4.9GHz I would try:

1-2 active cores - 5.2GHz
3-4 active cores - 5.1GHz
5-6 active cores - 5.0GHz
7-8 active cores - 4.9GHz

Then repeat phase 1 and 2 above all over again…

Phase 4
Then I just use the PC but I keep looking for possible odd behaviour while using various intensive apps or games.

 

[tom111111]

I don't think, that it is a good idea to overclock an Intel cpu.
Even at stock clocks. they are running so hot, that is difficult to cool them.

 

[samopa]

I don't think, that it is a good idea to overclock an Intel cpu.
Even at stock clocks. they are running so hot, that is difficult to cool them.

My 12900KF is happily running in core clock of 5252 MHz (52 x 101 MHz), All P-Core running 5252 MHz and all E-core running 3939 MHz.
Temperature range 41-45 Celsius in idle and 65-85 Celsius on all full load. I'm using Noctua NHD15S as a cooler.