Question i7-9700k overheats on 4600 MHz, but is fine at 3500-3600 MHz ?

[aplenite]

Hi guys, I'm a total noob in overclocking and I never really did that on purpose and the correct way I guess.

I have an i7-9700K processor and for more than 4 years of using it, I was pretty much running it as it was. It was always fine.
Somehow, not sure how it works, it was "boosting" itself to 4600 Mhz when under load while maintaining normal temps, like 60-70 degrees under load. So I was perfectly fine with that.

However, recently when playing Baldur's Gate 3 I noticed that my AIO is running like crazy, very loud. So I checked the temps and they turned to be around 100 degrees (voltage is around 105+ W)

So I was trying all possible things like re-applying thermal paste, reseating the cooler, and even going as far as updating the cold plate and gasket with a Liquid Freezer Service kit. (I got Liquid Freezer II 280) I checked and tried the basic things.

When I force it to work on 3500 MHz by setting Maximum Processor State to 99%, the temps go back to normal 60-65 under load and around 40-45W.
As soon as I allow it to run on 100%, 4600 MHz temps spike to 100 almost imeddiately.

There is one more thing, and you will probably think this makes no sense and I'm going crazy lol. So every time I was reapplying thermal paste / reseating the AIO pump, all temps were perfect on 4600 MHz under load... for like 2 weeks. And then they were going wild again. I don't understand what it is and I don't have any more ideas. If you do, please let me know.

 

[Misgar]

If you're seeing a large ridge of paste at the outer edge of the CPU, you've applied too much. The optimum amount of paste should just reach the edge of the CPU when the cooler is clamped down and it doesn't matter if the corners aren't entirely covered in compound. The heat is generated near or in the middle of the CPU.

Some time ago, I saw an article which said the top surface of a CPU was not perfectly flat, but concave for one manufacturer and convex for the other. I cannot remember if it was Intel or AMD who manufactured a concave IHS. The article went on to explain how you can "lap" or grind the CPU and heatsink surfaces perfectly flat, using a sheet of plate glass and very fine abrasive compound.

I'm not suggesting you do the same, but it's possible you've got the same problem, i.e. your CPU and heatsink aren't making good contact over the entire area. I'm used to seeing an even spread of paste on both CPU and cooler base whenever I disassemble any of my systems.

I doubt you're applying too little compound, but if there's nothing in the middle, perhaps you should apply a tiny fraction more? I normally squeeze a small garden pea sized blob into the middle of a square CPU, but if it's rectangular or a really large CPU, I place several smaller blobs in strategic positions.

If you have another cooler, air or liquid, try fitting it and then remove it immediately, to check the spread of compound. If it's different from your normal cooler, it implies the base of the two coolers have different "flatness".

It's unlikely you're applying too much mounting pressure, so I might be inclined to tighten down the screws a bit further to increase contact pressure. The compound is simply there to fill in the microscopic voids in the two metal surfaces. Most of the heat transfer is via metal to metal contact between CPU and heatsink. If you don't have sufficient pressure, heat transfer is reduced.

Even if you are applying too much pressure, I don't think it would squeeze all the compound to the edge of the CPU. There should always be a very thin coating left in the middle of the two mating surfaces.

I know some people experience bending on LGA1700 boards, leading to loss of contact in the middle of the CPU, so they fit a special frame around the CPU to counteract this problem, but I don't think it applies in your case.

 

[aplenite]

If you're seeing a large ridge of paste at the outer edge of the CPU, you've applied too much. The optimum amount of paste should just reach the edge of the CPU when the cooler is clamped down and it doesn't matter if the corners aren't entirely covered in compound. The heat is generated near or in the middle of the CPU.

Some time ago, I saw an article which said the top surface of a CPU was not perfectly flat, but concave for one manufacturer and convex for the other. I cannot remember if it was Intel or AMD who manufactured a concave IHS. The article went on to explain how you can "lap" or grind the CPU and heatsink surfaces perfectly flat, using a sheet of plate glass and very fine abrasive compound.

I'm not suggesting you do the same, but it's possible you've got the same problem, i.e. your CPU and heatsink aren't making good contact over the entire area. I'm used to seeing an even spread of paste on both CPU and cooler base whenever I disassemble any of my systems.

I doubt you're applying too little compound, but if there's nothing in the middle, perhaps you should apply a tiny fraction more? I normally squeeze a small garden pea sized blob into the middle of a square CPU, but if it's rectangular or a really large CPU, I place several smaller blobs in strategic positions.

If you have another cooler, air or liquid, try fitting it and then remove it immediately, to check the spread of compound. If it's different from your normal cooler, it implies the base of the two coolers have different "flatness".

It's unlikely you're applying too much mounting pressure, so I might be inclined to tighten down the screws a bit further to increase contact pressure. The compound is simply there to fill in the microscopic voids in the two metal surfaces. Most of the heat transfer is via metal to metal contact between CPU and heatsink. If you don't have sufficient pressure, heat transfer is reduced.

Even if you are applying too much pressure, I don't think it would squeeze all the compound to the edge of the CPU. There should always be a very thin coating left in the middle of the two mating surfaces.

I know some people experience bending on LGA1700 boards, leading to loss of contact in the middle of the CPU, so they fit a special frame around the CPU to counteract this problem, but I don't think it applies in your case.

Got you, thank you for the insights!
I'll try to tighten everything up before I do anything else.

As for the thermal paste amount, it's probably a bit above average. I was checking videos on how to apply thermal properly, so I have a sense of how much it should be, however, I probably apply a bit more than needed just to be sure I applied enough As I understand, not enough hurts temps, while too much doesn't, just creates a mess potentially. So that's why the thick outline.
I wish I had a photo to show you how "bad" it is, but I'll probably do one anyway soon.
Thank you for the help!

 

[Misgar]

Too much thermal compound can also be less than ideal, especially if most of it doesn't get squeezed out to the sides. The thermal conductivity of normal Silicone based thermal compound is not as good a the thermal conductivity of metal.

If the CPU is sitting under a really thick layer of thermal compound and not making direct metal-to-metal contact with the heatsink base, heat will be transferred more slowly.

Thermal compound is supposed to fill in the microscopic voids in the surface of the CPU's integrated heatsink (the metal cover over the Silicon die) and the tiny voids on the surface of the cooler heatsink.

To the naked eye, the CPU and cooler surfaces look smooth and flat, but examine them under a powerful microscope and you'll see an irregular surface and possibly a slightly distorted concave or convex shape.

If both the CPU and the cooler have slightly convex profiles, the cooler will only touch the middle of the CPU. If both the CPU and the cooler have slightly concave profiles, they will only touch around the edges.

Motherboard bending also comes into play when you tighten down the heatsink screws. Some LGA1700 boards bend quite badly, leading to poor heatsink contact and higher temperatures. You can buy an after-market frame to place over the CPU socket to counteract this bending.

There's a happy medium when applying compound. Go to extremes either way (far too little or far too much) and thermal transfer will suffer. It usually isn't catastrophic, just not ideal.

If after removing the heatsink you find the compound just reaches the edges of the CPU, but doesn't spill over, that's ideal.

Provided the heatsink makes good contact across the whole surface of the CPU, you're usually OK, but if you get uneven pressure, then only a small area will transfer heat and if this area isn't directly over the die underneath, overheating can occur.

You could just bolt the cooler to the CPU without any compound. The cooler would still work, but the CPU would run hotter.

Provided the heatsink is reasonably well attached, thermal throttling in the CPU should help to prevent damage, although the CPU may sit at 90/95/100C when throttling.

 

[aplenite]

@Misgar @thestryker @Phaaze88

Hi guys, I got a new update. I got some new information that doesn't give me any answers, only more questions... but perhaps your experience can help.

So on the 27th of October, I reapplied the thermal paste and reseated the AIO pump (12 days ago ). The temps normalized, and everything went back to normal.
I started doing an improvised test. I made a save in Baldurs Gate 3 that I loaded the first 7 days, every evening, and was checking the temps and wattage under the same conditions.

For 7 days straight my results were pretty much consistent. 1-2 degree difference in temps 1-5 difference in Wattage

This was the result I got each time - 58-63 degrees CPU, 79-85W.

I was doing it for 7 days straight, but unfortunately kind of stopped doing it at one point, since every time the result was the same, so I lost my enthusiasm.

Everything was fine until today's morning. While playing a different game I noticed that my coolers are going wild again and I physically can feel the heat since the PC is on a table close to me. I booted up the same save and here's what I got:

98-98 degrees, increased usage up to 100%, Wattage around 110-115 instead of 80-85. Idle temps also went a whole lot up and so it's everything all over again.
I even got less FPS (was capped at 72), while conditions are absolutely the same. CPU started struggling and probably throttling.
Unfortunately, I didn't test through the whole period, but I don't feel like the change was gradual. Temps didn't degrade during the first 7 days at all. It's rather as if they just changed in a click.

First of all, I unplugged the PC and tried shaking and turning it to see if that could help clear out the potential blockage in the tubes. I was doing my best, but it had 0 impact. So maybe we can exclude the blockage version.

So I took out the AIO pump to take photos of the thermal paste.

The consistency of the paste is pretty much the same as it was, the color changed from light-blue to dark-blue/grayish. Kind of looks like it's evenly spread, but let me know what you think. Probably applied too much.

So I reapplied the fresh paste and took a photo of it. I tried to apply a bit less than I usually would to not overdo it.

I tried to screw tighter everything I could, but the fastenings were pretty tight already, and had no room to screw them in even more. Everything nice and tight, no space for wobbling or anything like that.

This resulted in the following changes in the same scene:

So pretty much the same story all over again. The temps are like 3 degrees higher than before in a normal situation, but I assume this could be because I applied less paste, so maybe we can neglect that.
63 degrees, 86W, solid 72 FPS again.

I also did a test with my pump connector unplugged from the motherboard (while the coolers were running) to see if the pump was sort of idle and it wasn't. Temps went to 100 degrees real fast, plugged back in, all good.

--------

One more thing I thought about randomly but at this point I'm ready to believe in anything.. So a few days before things got worse my RAM's RGB started blinking uncontrollably (it never happened before). All 4 sticks just flickering as if it's a disco. I thought that it could be some software conflict. I had no RGB-controlling software at all, so I downloaded Trident's tool to control RGB and it seems to fix it, at least when the software is booted up.
But after that, I remembered that I have two 3000Mhz and two 3600Mhz sticks (XMP running them both at 3000). Both Trident Z DDR4, just different speeds.
So this brings two new questions - could having 4 RAM with 2 different speeds introduce some overheating/overload that would affect the CPU? Seems unlikely, but I can't tell.
And the second one is, is it possible that maybe my power supply isn't enough for all the RAM, GPU, CPU running on turbo, RGBs and all other stuff? Could this cause such issues?
My next tests would be checking the temps with only 1 RAM stick and also maybe even without a GPU, to see if it makes any difference. Unfortunately, I didn't think of it before and now I already "fixed" my CPU for the next 10-12 days.

I'm sorry for the wall of text, hope the screenshots and photos will make it a bit more engaging 😀

 

[Phaaze88]

For 7 days straight my results were pretty much consistent. 1-2 degree difference in temps 1-5 difference in Wattage

This was the result I got each time - 58-63 degrees CPU, 79-85W.

I was doing it for 7 days straight, but unfortunately kind of stopped doing it at one point, since every time the result was the same, so I lost my enthusiasm.

Everything was fine until today's morning. While playing a different game I noticed that my coolers are going wild again and I physically can feel the heat since the PC is on a table close to me. I booted up the same save and here's what I got:

98-98 degrees, increased usage up to 100%, Wattage around 110-115 instead of 80-85. Idle temps also went a whole lot up and so it's everything all over again.
I even got less FPS (was capped at 72), while conditions are absolutely the same. CPU started struggling and probably throttling.
Unfortunately, I didn't test through the whole period, but I don't feel like the change was gradual. Temps didn't degrade during the first 7 days at all. It's rather as if they just changed in a click.

For this part, I suspect something else was running during that last event. There's not much reason for power use to go up a whopping ~30w. What that is, is something you're going to have to narrow down from the apps you have installed on your PC. Perhaps something is updating or communicating with its servers at the time the game is running.

So I took out the AIO pump to take photos of the thermal paste.

The consistency of the paste is pretty much the same as it was, the color changed from light-blue to dark-blue/grayish. Kind of looks like it's evenly spread, but let me know what you think. Probably applied too much.

It is too much, but not a problem, per se. It's just messier for cleanup, is all. Mounting pressure is supposed to squeeze out excess. Paste should only fill in microscopic pits between the 2 plates.

Oh, and just a heads up, but MX-5 was discontinued because the company was having trouble keeping the quality consistent. The oils were separating from the rest of the compound(it's not supposed to do that). So use MX-6, if you haven't been already.

is it possible that maybe my power supply isn't enough for all the RAM, GPU, CPU running on turbo and stuff?

I've not read and seen any evidence suggesting so, but please provide the make and model of the psu, as well as how long it has been in use. The others may have more input on that.

I'm not sure about the other questions, so I had to skip them. Hopefully others have some insight on them.

 

[thestryker]

Which paste was used that changed colors? I have MX-5 from a good batch and it definitely doesn't change color so if it's that one toss it, and while I haven't used MX-6 the paste changing color still doesn't sound right.

Definitely too much paste in the picture, and while it does look like uneven pressure it looks good along the center which is where the die is.

When your chip is running less efficiently the power consumption will go up, but not quite that much by itself. The increased CPU usage seems like an entirely different issue as I can't think of anything that would link it to thermal issues.

 

[aplenite]

For this part, I suspect something else was running during that last event. There's not much reason for power use to go up a whopping ~30w. What that is, is something you're going to have to narrow down from the apps you have installed on your PC. Perhaps something is updating or communicating with its servers at the time the game is running.



It is too much, but not a problem, per se. It's just messier for cleanup, is all. Mounting pressure is supposed to squeeze out excess.

Oh, and just a heads up, but MX-5 was discontinued because the company was having trouble keeping the quality consistent. The oils were separating from the rest of the compound(it's not supposed to do that). So use MX-6, if you haven't been already.



I've not read and seen any evidence suggesting so, but please provide the make and model of the psu, as well as how long it has been in use. The others may have more input on that.

I'm not sure about the other questions, so I had to skip them. Hopefully others have some insight on them.

Yes, I also thought that was some sort of program or update. I even thought this could be a miner (checked with Malwarebytes and Windows Defender Offline scan, all clear), but I doubt a miner would stop after a cooler repasting, so yeah... It could be something that adds on top to the power usage, but why would it stop after repaste/reseat? That's the part that makes it all confusing.

Great info on the MX-5, thanks! I have both the MX-5 and MX-6 though, I think I got the MX-6 as a gift for buying something. So I tried both.

PSU - Seasonic Focus Plus Platinum 750PX 750W. Been in use for 5 years.

 

[aplenite]

Which paste was used that changed colors? I have MX-5 from a good batch and it definitely doesn't change color so if it's that one toss it, and while I haven't used MX-6 the paste changing color still doesn't sound right.

Definitely too much paste in the picture, and while it does look like uneven pressure it looks good along the center which is where the die is.

When your chip is running less efficiently the power consumption will go up, but not quite that much by itself. The increased CPU usage seems like an entirely different issue as I can't think of anything that would link it to thermal issues.

It was MX-5 I think, but honestly, I'm not 100% sure. This could be a faulty paste case, but I used both MX-5 and MX-6. I'm wondering if the paste could degrade in quality just by lying in the syringe it is shipped in.

Ok, good, at least I know that's not the issue.

What if my motherboard's bios sucks and it doesn't manage the CPU powers properly? But then again, why would it get back to normal after repasting..

 

[thestryker]

It was MX-5 I think, but honestly, I'm not 100% sure. This could be a faulty paste case, but I used both MX-5 and MX-6.

If you didn't use the MX-6 this go around then that's what I'd suggest doing should you run into this happening again.

I'm wondering if the paste could degrade in quality just by lying in the syringe it is shipped in.

It shouldn't so long as it's properly sealed back up when put away, especially the Arctic MX line as they're designed for long life.

Ok, good, at least I know that's not the issue.

Yeah it seems like it should be okay. I got my MX-5 with a spatula because Arctic was selling two versions and it was like 10 cents more so I figured I'd try it. Even though I've never had problems with paste application I really have liked using it as I know I've gotten a good thin even layer.

Anything else I think is more likely to be software side issues. Regarding mismatched DRAM while it's not a good thing what you'd experience if it was a problem would be stability related.

As for PSU unless you've got a video card pulling over 400W I highly doubt that would be related to any issues.

 

[aplenite]

If you didn't use the MX-6 this go around then that's what I'd suggest doing should you run into this happening again.

got you, will make sure to use MX-6 next time

Anything else I think is more likely to be software side issues.

yeah, looks like I'll need to wait until it happens again and do some more tests, but more software oriented now. Maybe running it in Safe Mode, should clear some of the possible causes.
thanks for your help!