[Asmodian]

That is just too much. 24Hrs of Aida is just torturing your CPU.
You can loop Cinebech for 30 min, do a couple of hours of Realbench and you're good to go.
Mine never failed in anything so far, games or otherwise. Can't go lower on voltages either, as it gets unstable. Currently about 4 voltage steps above that point.

LOL One example of "I did it for 30 min and I didn't have any problems". Yeah, super convincing.

How many systems have you overclocked?

I still wouldn't consider it stable after only 24h of Aida64, based on my experience over the decades. I have had many systems pass 24h of Aida64, Realbench, and many Cinebench runs that was still unstable in a game, x265, or something else. Both Cinebench and Realbench seems fairly pointless for stability testing to me, it is easy to pass either and have other loads crash the system right away, not that Aida64 by itself is a reliable test either.

Stability testing is hard but assuming a system is stable after only 30 min of Cinebench and a couple hours of Realbench is simply laughable.

 

[Techie007]

Agreed. I would add that Aida64 really isn't that great at testing stability either. For that, use OCCT and Linpack; they are far faster and will flag systems that can pass Aida64. And yes, 30 minutes is great for dialing in an overclock, but definitely not enough to be declaring the system "stable".

 

[Minioverclocker]

That is just too much. 24Hrs of Aida is just torturing your CPU.
You can loop Cinebech for 30 min, do a couple of hours of Realbench and you're good to go.
Mine never failed in anything so far, games or otherwise. Can't go lower on voltages either, as it gets unstable. Currently about 4 voltage steps above that point.

@Minioverclocker
Those are some damn nice temperatures you're getting on that VR OUT voltage!
With only 72c during Cinebech, you can definitely try for 5.2Ghz.

I managed to get lower temps to 67c max @5.1 ghz, the thing is i barely find time for it haha , i would try this out soon.

 

[GeneO]

I have HTT enabled , speedshift enable, C-states down to C7 enabledturbo boost enable, fixed 5.1 GHz

I do run CineBench at elevated priority (High) because it is sensitive to background processes. That eliminates interference pretty well.

You can also run Cinebench at real-time priority.

A word of warning though. If you have Intel XTU installed (you don't have to be running the XTU utility) and run a benchmark at real-time priority, your computer may abruptly crash - just a click and boom, black screen crash with no indication of why. I finally tracked this to Intel Xtreme Tuning Utility . - there must be a bug in one of its drivers or services that cause a hard crash when running a compute intensive programs in real-time. I uninstalled it and can run CineBench, IBT in real-time once again.

Others have reported this (crashing) with real-time while others have had no issue. Now I know why

 

[Cryptedvick]

LOL One example of "I did it for 30 min and I didn't have any problems". Yeah, super convincing.

How many systems have you overclocked?

I've been overclocking for the better part of the last 15 years. See my sig for my CPU history. Each and every one of those CPUs has been running as much as I could get out of them.
I used to be the same kind of crazy testing my overclock to death with Prime95, IBT, OCCT, Cinebench and y cruncher, until I blew my former 8700k while pushing crazy AVX loads through it. Made it have a nice little belly down below the middle, under the resistors.

I still wouldn't consider it stable after only 24h of Aida64, based on my experience over the decades. I have had many systems pass 24h of Aida64, Realbench, and many Cinebench runs that was still unstable in a game, x265, or something else.

If after all of that bashing on the CPU it still failed some tests or in games, then it might indicate that other things are unstable, such as RAM.

bold: I call that straight up BS

Stability testing is hard but assuming a system is stable after only 30 min of Cinebench and a couple hours of Realbench is simply laughable.

Laugh as much as you want. Fact of the matter is, CPU overclocking has been far simplified these days and all you need to adjust now is vcore and maybe VCCSA and VCCIO if you OC your RAM too. IDK if you remeber the C2Q days of overclocking, but I do!
With the aforementioned test, I have not had a single BSOD, crash of freeze due to unstable CPU overclock.
I do these tests and after I find the minimum voltage I need to pass this, I up it 4 more steps. This gives it slightly more voltage than it needs to be stable, very little increase in heat, but rock solid.
You can be my guest and stress test your CPU as much as you want, just don't go around advising people to destroy their CPU with 24hrs of crazy unrealistic AVX loads. And I speak this from my personal experience, again, blowing my 8700k while running P95 AVX stress testing.

 

[zayd]

I've been overclocking for the better part of the last 15 years. See my sig for my CPU history. Each and every one of those CPUs has been running as much as I could get out of them.
I used to be the same kind of crazy testing my overclock to death with Prime95, IBT, OCCT, Cinebench and y cruncher, until I blew my former 8700k while pushing crazy AVX loads through it. Made it have a nice little belly down below the middle, under the resistors.



If after all of that bashing on the CPU it still failed some tests or in games, then it might indicate that other things are unstable, such as RAM.

bold: I call that straight up BS



Laugh as much as you want. Fact of the matter is, CPU overclocking has been far simplified these days and all you need to adjust now is vcore and maybe VCCSA and VCCIO if you OC your RAM too. IDK if you remeber the C2Q days of overclocking, but I do!
With the aforementioned test, I have not had a single BSOD, crash of freeze due to unstable CPU overclock.
I do these tests and after I find the minimum voltage I need to pass this, I up it 4 more steps. This gives it slightly more voltage than it needs to be stable, very little increase in heat, but rock solid.
You can be my guest and stress test your CPU as much as you want, just don't go around advising people to destroy their CPU with 24hrs of crazy unrealistic AVX loads. And I speak this from my personal experience, again, blowing my 8700k while running P95 AVX stress testing.

I concur with your sentiments. I recently overclocked my i9-9900k to 5Ghz, 4,7Ghz ring ratio, 0 AVX offset at 1.32 Vcore, LLC at Turbo and all power saving options disabled. This was an upgrade from my trusty i7-9700k which also ran rock solid for 17 months on a similar bios setup but it needed 1.370 Vcore. I use my PC primarily for gaming and at present, I'm playing Warzone. This is my stability test, in which I monitor temps, CPU usage and power draw mostly. So far its been stable and the CPU usage has gone from an average of 70% with the 9700k to 30% with the i9 9900k, hence my upgrade. My new GPU was just to taxing for the younger brother of this intel family.

 

[Asmodian]

I used to be the same kind of crazy testing my overclock to death with Prime95, IBT, OCCT, Cinebench and y cruncher, until I blew my former 8700k while pushing crazy AVX loads through it. Made it have a nice little belly down below the middle, under the resistors.

I totally agree on your point here. Don't run power virus loads on new architectures. It is dangerous for the hardware and does not help prove stability since the architectures react differently to power virus loads anyway.

I also notice Aida64 is not one of the items in your list.

If after all of that bashing on the CPU it still failed some tests or in games, then it might indicate that other things are unstable, such as RAM.

Or it could be the CPU and the stability test does not reproduce it. This is also a reason only using 30min of Cinebench and two hours of Realbench to decide the system is stable is laughable.

bold: I call that straight up BS

Believe what you will but it is absolutely true. Many times. x264 and x265 has crashed many CPUs that can pass 24h of multiple artificial stability tests.

Games have also been pretty common. With your amount of experience I am surprised you doubt this. A game crashing due to a bad CPU overclock immediately after 24h of Prime95 seems like a meme to me or something. Isn't it common knowledge?

IDK if you remeber the C2Q days of overclocking, but I do!

Yes, I do. It still sounds like a pretty modern CPU to me. I was overclocking my 486.

In some ways it is much easier to overclock today but is some ways stability testing is also trickier. Today we need to worry about stability during light or medium loads too, or while changing power states. Passing a nasty power virus load for 24h isn't particularly convincing anymore.

We are well past the good ol' days of simply changing the jumpers and having the CPU always run at that speed and voltage.

With the aforementioned test, I have not had a single BSOD, crash of freeze due to unstable CPU overclock.

Great for you. Others will not have the same experience.

It is pretty easy to get an apparently stable overclock today if you aren't pushing to the absolute limit but it is also very hard to be sure an overclock is stable without simply using it for a long time.

 

[Cryptedvick]

I also notice Aida64 is not one of the items in your list.

I really dislike Aida64, which is why I don't use it.
I also use P95 in place 1344FFTs. This seems to filter out WHEA errors or possible freezes.

Believe what you will but it is absolutely true. Many times. x264 and x265 has crashed many CPUs that can pass 24h of multiple artificial stability tests.

Games have also been pretty common. With your amount of experience I am surprised you doubt this. A game crashing due to a bad CPU overclock immediately after 24h of Prime95 seems like a meme to me or something. Isn't it common knowledge?

Which is why you test with several things, not a 24hr constant, balanced load doing one thing. That is just a waste of time.

Yes, I do. It still sounds like a pretty modern CPU to me. I was overclocking my 486.

Thats nice, I also remember jumper overclocking one of my first CPUs.
Reason I was referring to the C2Q, was because this was the hardest (IMO) CPU to overclock and test for stability. This is the CPU I spent the most time overclocking and stability testing. You had a lot of voltages to fiddle with and had to find the perfect balance of all of them to find proper stability. You had vcore, CPU termination, PLL, Reference. Then you moved to MCH core, MCH reference, ICH IO and Core voltages, etc.
Now its only Vcore and if you need faster RAM, simply up the VCCIO and VCCSA and thats it.

edit: just found this old pic in my phone with the stable voltages for my particular setup from back then, with the Q9550. Old times. THIS took a lot to properly stabilize and verify.

In some ways it is much easier to overclock today but is some ways stability testing is also trickier. Today we need to worry about stability during light or medium loads too, or while changing power states. Passing a nasty power virus load for 24h isn't particularly convincing anymore.

I disagree. I think its actually easier to stress test today and I'm basing that on my own experience. If you really want to make sure your PC is stable, a few hours of combined tests will be more than fine for that, assuming you're testing your CPU only, and your RAM is stock. If you put RAM into the equation, things are a whoooole lot different and thats when you really need to get into many, specific stress tests, with a lot more time dedicated for this, as RAMs are far trickier to stabilize than CPU, let alone stabilize together with the CPU.
Also, when overclocking, its highly advisable to disable power states, as this can cause instability. I have never overclocked any CPU and have power states enabled. Its just a bad approach and the only reason power states are used is for low idle power usage. Well, my 9900k is only taking about 25-28W right now on idle, full power 5Ghz and 1.273v idle VR OUT. It really doesn't make any difference.

Putting any kind of power virus on the CPU is just asking for trouble and should be avoided as stress testing, let alone a 24hr run.

It is pretty easy to get an apparently stable overclock today if you aren't pushing to the absolute limit but it is also very hard to be sure an overclock is stable without simply using it for a long time.

As I said, pusing to the absolute limits is a really bad idea and has high chances of degrading your CPU or even burn it.
You're loading the CPUs cache that feeds the cores workload in such a way that can never ever be recreated by any regular software, game or whatever. Thats why its called a power virus.
You can get perfectly reasonable real world scenario stress testing software like Realbench, which really pushes the CPU to max with AVX loads. You can also run some in place 1344FFTs.
As a closing note, please remember the ONE POINT that I made which is very important to all of this. You pass all these tests and if you up the voltage by 3-4 steps, you are pretty much guaranteed that you wont have any issue and it will spare you of unreasonable testing of 24hrs or more and will also spare your CPU unnecessary load for such a long time.

 

[Techie007]

I used to be the same kind of crazy testing my overclock to death with Prime95, IBT, OCCT, Cinebench and y cruncher, until I blew my former 8700k while pushing crazy AVX loads through it. Made it have a nice little belly down below the middle, under the resistors.
<snip>
I do these tests and after I find the minimum voltage I need to pass this, I up it 4 more steps. This gives it slightly more voltage than it needs to be stable, very little increase in heat, but rock solid.
You can be my guest and stress test your CPU as much as you want, just don't go around advising people to destroy their CPU with 24hrs of crazy unrealistic AVX loads. And I speak this from my personal experience, again, blowing my 8700k while running P95 AVX stress testing.

This person does not know the definition of a "stable overclock". If your CPU blows up under extended load—regardless of whether it's real or artificial—you did not have a "stable" overclock. Straight up fail! So now you go around telling people to push their systems and not fully test it because "it might blow". So what if the system is actually loaded like that for some task and it blows? You clearly did not set your TDP limits correctly. This is exactly what they're there to prevent!

 

[zayd]

I really dislike Aida64, which is why I don't use it.
I also use P95 in place 1344FFTs. This seems to filter out WHEA errors or possible freezes.



Which is why you test with several things, not a 24hr constant, balanced load doing one thing. That is just a waste of time.



Thats nice, I also remember jumper overclocking one of my first CPUs.
Reason I was referring to the C2Q, was because this was the hardest (IMO) CPU to overclock and test for stability. This is the CPU I spent the most time overclocking and stability testing. You had a lot of voltages to fiddle with and had to find the perfect balance of all of them to find proper stability. You had vcore, CPU termination, PLL, Reference. Then you moved to MCH core, MCH reference, ICH IO and Core voltages, etc.
Now its only Vcore and if you need faster RAM, simply up the VCCIO and VCCSA and thats it.

edit: just found this old pic in my phone with the stable voltages for my particular setup from back then, with the Q9550. Old times. THIS took a lot to properly stabilize and verify.
View attachment 2480276


I disagree. I think its actually easier to stress test today and I'm basing that on my own experience. If you really want to make sure your PC is stable, a few hours of combined tests will be more than fine for that, assuming you're testing your CPU only, and your RAM is stock. If you put RAM into the equation, things are a whoooole lot different and thats when you really need to get into many, specific stress tests, with a lot more time dedicated for this, as RAMs are far trickier to stabilize than CPU, let alone stabilize together with the CPU.
Also, when overclocking, its highly advisable to disable power states, as this can cause instability. I have never overclocked any CPU and have power states enabled. Its just a bad approach and the only reason power states are used is for low idle power usage. Well, my 9900k is only taking about 25-28W right now on idle, full power 5Ghz and 1.273v idle VR OUT. It really doesn't make any difference.

Putting any kind of power virus on the CPU is just asking for trouble and should be avoided as stress testing, let alone a 24hr run.



As I said, pusing to the absolute limits is a really bad idea and has high chances of degrading your CPU or even burn it.
You're loading the CPUs cache that feeds the cores workload in such a way that can never ever be recreated by any regular software, game or whatever. Thats why its called a power virus.
You can get perfectly reasonable real world scenario stress testing software like Realbench, which really pushes the CPU to max with AVX loads. You can also run some in place 1344FFTs.
As a closing note, please remember the ONE POINT that I made which is very important to all of this. You pass all these tests and if you up the voltage by 3-4 steps, you are pretty much guaranteed that you wont have any issue and it will spare you of unreasonable testing of 24hrs or more and will also spare your CPU unnecessary load for such a long time.

Would you mind telling me what Vcore you are using for your CPU and what is your ring ratio too. Thank you.

 

[Asmodian]

Which is why you test with several things, not a 24hr constant, balanced load doing one thing. That is just a waste of time.

And you don't call it stable after only two hours of Realbench.

 

[Cryptedvick]

And you don't call it stable after only two hours of Realbench.

Cherry-picking only one part of my entire post/reasoning does not help your argument.
Please read my above closing note again.

This person does not know the definition of a "stable overclock". If your CPU blows up under extended load—regardless of whether it's real or artificial—you did not have a "stable" overclock. Straight up fail! So now you go around telling people to push their systems and not fully test it because "it might blow". So what if the system is actually loaded like that for some task and it blows? You clearly did not set your TDP limits correctly. This is exactly what they're there to prevent!

You're joking right? My entire point is related to excessive stress testing that could cause degrading or CPU blowing up, which is what I had happened.
While running AVX stress testing, with excessive loads for stress testing purposes (exactly like Asmodian is mentioning that it should be done), the CPU blew.
The WHOLE point of running stress testing is to test stability.... what kind of argument is "If your CPU blows up under extended load—regardless of whether it's real or artificial—you did not have a "stable" overclock" This is straight up laughable.
Whats the point in overclocking if you use TDP limits, which will just reduce multiplier and voltage to stay under TDP defined? You'll just hit the TDP limit, multiplier will be lowered, vcore lowered and your overclock stability testing will be meaningless.

Would you mind telling me what Vcore you are using for your CPU and what is your ring ratio too. Thank you.

Righ now I am at 1.36v in Bios, LLC lvl3, 1.32v vcore Idle, 1.30v VR OUT idle, 1.209v VR OUT under stress testing.
x50 CPU multi
x45 cache multi

 

[Asmodian]

Cherry-picking only one part of my entire post/reasoning does not help your argument.

My argument was with your "You can loop Cinebech for 30 min, do a couple of hours of Realbench and you're good to go." That is not enough to know your settings are stable.

Aida64 isn't a power virus so I don't see why your point about not running a power virus is relevant to my original post.

You seem to have your technique which works for you. You run higher voltages than needed, don't care about idle power use, and don't overclock the cache or memory. I guess that is fine, but it isn't the only way to overclock.

 

[Cryptedvick]

My argument was with your "You can loop Cinebech for 30 min, do a couple of hours of Realbench and you're good to go." That is not enough to know your settings are stable.

Aida64 isn't a power virus so I don't see why your point about not running a power virus is relevant to my original post.

You seem to have your technique which works for you. You run higher voltages than needed, don't care about idle power use, and don't overclock the cache or memory. I guess that is fine, but it isn't the only way to overclock.

AIDA64 FPU test ran alone is most definitely a power virus.

Indeed, I do have this technique, which has proven to work just fine, as a circumvention to countless hours of stress testing.I was just too sick and tired of overnight stress tests and after the CPU blowing, I realised a method change was in order.
A stock CPU will run a stock voltage much higher than what it needs to be stable, I am applying the same principle, but not overdoing it so that heat doesn't become too much of an issue.

When we overclock, we tend to try and find the least amount of needed voltage for obvious reasons and that implies proper and longer stress tests. If you find a voltage that passes 30 min of cinebench and 2 hours of Realbech, then you are, I'd say, 98-99% confident of stability. Just upping the voltage then 3-4 more steps, just does what one would do with constant stress testing ending with maybe 2 extra steps of vcore to pass 24hrs and many other tests.

You are also correct about the last part. I'm only talking about CPU overclocking. Cache overclocking yields close to zero net benefit (tried from 4000Mhz to my max, 4800Mhz, settled on a round 4500Mhz) and RAM overclocking gives very diminishing results past 3200Mhz. (not that I wouldn't, but these ****ty Corsairs won't move past 3200Mhz no matter what )

Try to see the other guys POV as well, it may not be as bad as you think. Cheers!

 

[Techie007]

You're joking right? My entire point is related to excessive stress testing that could cause degrading or CPU blowing up, which is what I had happened.
While running AVX stress testing, with excessive loads for stress testing purposes (exactly like Asmodian is mentioning that it should be done), the CPU blew.
The WHOLE point of running stress testing is to test stability.... what kind of argument is "If your CPU blows up under extended load—regardless of whether it's real or artificial—you did not have a "stable" overclock" This is straight up laughable.
Whats the point in overclocking if you use TDP limits, which will just reduce multiplier and voltage to stay under TDP defined? You'll just hit the TDP limit, multiplier will be lowered, vcore lowered and your overclock stability testing will be meaningless.

No; I'm dead serious. With each post, you are clearly revealing that you know very little about proper overclocking. The whole purpose of the TDP limit is to make the overclock stable and safe. In fact, it is Intel's intention and design that the TDP limit will be the primary method of regulating the top clock speeds of the CPU, and it is by far the most reliable and scalable limiter built-in to their CPUs. By limiting the maximum instantaneous power the CPU can draw, both it and the motherboard are protected from damage in the event of extreme load with boosted clocks. There's software stability (no BSODs, WHEA errors, or crashes), and then there's hardware stability (no powering down, hardware degradation, or smoke). If you're having ANY of those problems under ANY load scenario, your overclock is not stable. By eliminating the TDP limits, you are asking for hardware instability. The hardware cannot handle the extra load and power dissipation and damage eventually results—overclock fail. Who cares if you managed to boost the voltages enough that there was no software instability. Your computer is literally a ticking time bomb, waiting for that accidental load that will make it blow.

 

[Arctucas]

DELETED

Double post.

 

[Arctucas]

<SNIP>Just upping the voltage then 3-4 more steps, just does what one would do with constant stress testing ending with maybe 2 extra steps of vcore to pass 24hrs and many other tests.
<SNIP>

When you say 'steps', you mean millivolts, correct?

No; I'm dead serious. With each post, you are clearly revealing that you know very little about proper overclocking. The whole purpose of the TDP limit is to make the overclock stable and safe. In fact, it is Intel's intention and design that the TDP limit will be the primary method of regulating the top clock speeds of the CPU, and it is by far the most reliable and scalable limiter built-in to their CPUs. By limiting the maximum instantaneous power the CPU can draw, both it and the motherboard are protected from damage in the event of extreme load with boosted clocks. There's software stability (no BSODs, WHEA errors, or crashes), and then there's hardware stability (no powering down, hardware degradation, or smoke). If you're having ANY of those problems under ANY load scenario, your overclock is not stable. By eliminating the TDP limits, you are asking for hardware instability. The hardware cannot handle the extra load and power dissipation and damage eventually results—overclock fail. Who cares if you managed to boost the voltages enough that there was no software instability. Your computer is literally a ticking time bomb, waiting for that accidental load that will make it blow.

TDP

Thermal Design Power (TDP) represents the average power, in watts, the processor dissipates when operating at Base Frequency with all cores active under an Intel-defined, high-complexity workload. Refer to Datasheet for thermal solution requirements.

So, according to the above definition, (copy/paste from Intel website), there is no overclocking that will meet the definition; because it clearly states Base Frequency.

 

[Cryptedvick]

No; I'm dead serious. With each post, you are clearly revealing that you know very little about proper overclocking. The whole purpose of the TDP limit is to make the overclock stable and safe. In fact, it is Intel's intention and design that the TDP limit will be the primary method of regulating the top clock speeds of the CPU, and it is by far the most reliable and scalable limiter built-in to their CPUs. By limiting the maximum instantaneous power the CPU can draw, both it and the motherboard are protected from damage in the event of extreme load with boosted clocks. There's software stability (no BSODs, WHEA errors, or crashes), and then there's hardware stability (no powering down, hardware degradation, or smoke). If you're having ANY of those problems under ANY load scenario, your overclock is not stable. By eliminating the TDP limits, you are asking for hardware instability. The hardware cannot handle the extra load and power dissipation and damage eventually results—overclock fail. Who cares if you managed to boost the voltages enough that there was no software instability. Your computer is literally a ticking time bomb, waiting for that accidental load that will make it blow.

You have some serious logic issues and actually just proved how poor your knowledge of proper overclocking and testing is.
Please tell me, how is properly testing stability of an overclocked CPU done, by adding power PL1 and PL2 targets, which reduces your multiplier when hit? Or do you think that when a reasonable person overclocks his/her CPU, they're pushing 1.6v through their CPU and drawing 300W? (ignoring the fact that you need to cool that first). Adding PL1 and PL2 targets completely invalidates the stability testing attempt as soon as the limits are reached and multiplier reduced.

You know, there's also a max safe amperage a CPU is designed to have flowing through. The 8700k has 138A at its max VID and default LLC. While regular stress testing AVX apps such as Cinebech and Realbech or other of this kind will not hit that target, even at 5Ghz and 1.3-1.35v VROUT (assuming you are delidded), a power virus such as P95 Small FFT or even AIDA64 FPU test will go past that rating and endanger the CPU, which is precisely why I advise against. No real world scenario will push the CPU to that power draw and while it might pass and be stable and even have decent temps if you have excellent cooling, its only asking for trouble. And trouble I found, blowing up going over 138A, even tho my temps were very decent, having it delidded and with liquid metal applied.

With my current 9900k, which has a max amperage draw of 193A using default LLC, I am drawing no more than 125-130A before I am limited by heat so tell me, how exactly will PL1 and PL2 help me in this case, to avoid the hardware instability you speak of? At 195-200W I'm already maxing out my temps with a rather low VR OUT voltage.

This is my last reply for you, as I really dislike losing time like this. Use google and OCN for further info if you want. Search for Falkentyne's posts on overclocking, LLC, amperage, calculating formulas and so on. You might learn a thing or two.

When you say 'steps', you mean millivolts, correct?

I mean an extra step when you select your voltage in BIOS. say it goes in 0.005 or 0.01 increments. For me it goes in 0.01 increments, so if I need 1.31v in BIOS to pass my tests, I then add another 3-4 steps, to 1.34-1.35v and I'm good to go.

 

[Techie007]

Thermal Design Power (TDP) represents the average power, in watts, the processor dissipates when operating at Base Frequency with all cores active under an Intel-defined, high-complexity workload. Refer to Datasheet for thermal solution requirements.

So, according to the above definition, (copy/paste from Intel website), there is no overclocking that will meet the definition; because it clearly states Base Frequency.

Yes, that's how Intel rates their CPUs. However, you can put a "better than stock" heatsink on the CPU and use a gaming motherboard with enhanced power delivery, which will increase the TDP (Total Design Power) limit of the whole system. Then you can go into your BIOS and increase the Short Term TDP and Long Term TDP figures to increase the CPU's power limits accordingly. These would be the "first steps" of overclocking. But you have to keep in mind that the motherboard can only safely supply so much power, and that there's thermal resistance in the CPU that will result in differential thermal stress if you've got the CPU heavily cooled but putting out several times its originally rated power—even if kept cool. 2x seems to be the maximum "safe limit", but it would be very foolish to up those protective numbers to 4096w (which many ill-advised overclockers do)—this allows the CPU to draw huge power spikes under certain loads that will cause degradation or damage given enough load and time. Since the CPU in question (i7-8700k) is originally rated for a TDP of 95w, I would recommend setting TDP limits no higher than 190w Long Term (2x) and 230w Short Term (2.5x) as an absolute max (assuming adequate cooling; otherwise Long Term TDP needs to be lowered further to mirror what your thermal solution can actually dissipate on a continuous basis without the CPU thermal throttling). Then use Linpack (which has a pulsed load) and watch the thermals to determine if you need to drop the Short Term TDP further and/or reduce the switchover timeframe.

You have some serious logic issues and actually just proved how poor your knowledge of proper overclocking and testing is.
Please tell me, how is properly testing stability of an overclocked CPU done, by adding power PL1 and PL2 targets, which reduces your multiplier when hit? Or do you think that when a reasonable person overclocks his/her CPU, they're pushing 1.6v through their CPU and drawing 300W? (ignoring the fact that you need to cool that first). Adding PL1 and PL2 targets completely invalidates the stability testing attempt as soon as the limits are reached and multiplier reduced.

You know, there's also a max safe amperage a CPU is designed to have flowing through. The 8700k has 138A at its max VID and default LLC. While regular stress testing AVX apps such as Cinebech and Realbech or other of this kind will not hit that target, even at 5Ghz and 1.3-1.35v VROUT (assuming you are delidded), a power virus such as P95 Small FFT or even AIDA64 FPU test will go past that rating and endanger the CPU, which is precisely why I advise against. No real world scenario will push the CPU to that power draw and while it might pass and be stable and even have decent temps if you have excellent cooling, its only asking for trouble. And trouble I found, blowing up going over 138A, even tho my temps were very decent, having it delidded and with liquid metal applied.

With my current 9900k, which has a max amperage draw of 193A using default LLC, I am drawing no more than 125-130A before I am limited by head so tell me, how exactly will PL1 and PL2 help me in this case, to avoid the hardware instability you speak of? At 195-200W I'm already maxing out my temps with a rather low VR OUT voltage.

This is my last reply for you, as I really dislike losing time like this. Use google and OCN for further info if you want. Search for Falkentyne's posts on overclocking, LLC, amperage, calculating formulas and so on. You might learn a thing or two.

I can't even... This is a hilarious response. You're so far off base that I'm not even going to waste my time formulating much of an intellectual response. Ohm's law itself disproves your logic above. Just remember that you're the one who blew a CPU stress testing! Also, keep in mind that Intel's own Turbo Boost feature relies heavily on those limits and that maybe—maybe—they might know something that you do not. 👋👋

 

[Cryptedvick]

I can't even... This is a hilarious response. You're so far off base that I'm not even going to waste my time formulating much of an intellectual response. Ohm's law itself disproves your logic above. Just remember that you're the one who blew a CPU stress testing! And keep in mind that Intel's own Turbo Boost feature relies heavily on those limits and that maybe—maybe—they might know something that you do not. 👋👋

1.Just as hilarious as your so called overclocking knowledge, oh wise one.
2. I am well aware on how Intel's Turbo Boost works and how it relies on TDP to boost cores, but if you manually overclock, you disable all of these features.
And yes, my CPU blew during excessive stress testing, due to overclocking, but you know what they say, you're not a racer if you haven't crashed at least once