[SOLVED] Nothing Can Keep my Overclocked i7 9700K Cool

[Panzerbjorne39]

I’ve gone through three coolers for my overclocked 9700k. Looks like no off the shelf coolers either air or water can keep it under 90. I tried two air and one aio. One four heat pipes single tower air cooler, one six pipes twin tower, and lastly a 360mm AIO cooler. The 140mm fan options will not fit in my case and do not outperform my 360 AIO so that’s not it.
I’m leaving brands out of this because I want to avoid “try the brand I like” or “that brand sux” response which aren’t helpful. All three of the coolers were good, proven coolers in the price and performance brackets.
I’m using an 5.1 All core overclock on a ASUS Z-390 Prime using dynamic Vcore that maxes out 1.38. I’ve tried stepping down to 5.0 and it made almost no difference. Maybe 1-2 degrees at best, but still Over the acceptable thresholds. I’ve also ruled out the paste. I’ve tried three different pastes that are good quality.
Any bright ideas or suggestions?

 

[blazerboy]

Wel guys. Thanks especially to @blazerboy and @Phaaze88 for their assistance with this. The primary issue with the temp was the fans! Nothing to do with my overclock settings. I swapped the Arctic fans out for the stock fans and it was an immediately noticeable temp reduction.
When I originally considered using the artic pressure fans I was focused on the RPM difference and thought it would be ok. I had no idea how big of a difference the fans could make in this application. My PC is now sitting at 5.1 All core @1.385vcore that adapts up to 1.403 max. High temp under load is now 78. Excellent!

Glad you fixed your issue, that seems like a reasonable temperature. You may also be able to configure your settings a bit to further improve Temps by lowering voltage if it stays stable among other improvements @Karadjgne added. If you don't care about noise, there are fans out there that are higher rated for rads.

Noctua NF-F12 industrialPPC-3000 PWM is one that comes to mind

 

[Phaaze88]

Noctua NF-F12 industrialPPC-3000 PWM is one that comes to mind

Uhh... as someone who exclusively uses those and the 140mm version, I would not advise using those in a typical PC.
These fans get - obviously - ridiculously loud if one does not set a fan curve, or does not connect them to PWM headers.

I normally run them at 50%, where they're just as loud/quiet as the regular NF-A12 and NF-A14 fans, but they have a different pitch to them due to the 3 phase motor.
When stress testing - 9 of these things at 100%... you'll swear you're at an airfield!

If one never runs them at 100%, they're essentially wasting money. There are good rad fans that can be had for cheaper and don't run quite as loud.

 

[Karadjgne]

Noctua A12x25 are the best rad fans (arguably) to be had. Phanteks PH-F120MP are also excellent, as are the EK Vardars. The Noctua NF-F12 are good for intakes, but the P12 are better overall if used as exhaust.

Glad you did get temps sorted out by swapping fans, seriously though, 1.385 topping 1.4+ is pushing it for voltage, although Alot depends on what the actual vcore is as measured after the LLC is involved. If vcore + LLC is hitting 1.5v ish, your cpu will be short lived, electromigration will burn it up. If that's the voltage after LLC is applied, you can probably lower it some still. There's many people pushing that OC under 1.35v in bios, topping out under 1.37v. So I'd not stop tweaking yet. Even that little of a drop could easily drop 5ish °C or better from your load temps.

 

[Panzerbjorne39]

Noctua A12x25 are the best rad fans (arguably) to be had. Phanteks PH-F120MP are also excellent, as are the EK Vardars. The Noctua NF-F12 are good for intakes, but the P12 are better overall if used as exhaust.

Glad you did get temps sorted out by swapping fans, seriously though, 1.385 topping 1.4+ is pushing it for voltage, although Alot depends on what the actual vcore is as measured after the LLC is involved. If vcore + LLC is hitting 1.5v ish, your cpu will be short lived, electromigration will burn it up. If that's the voltage after LLC is applied, you can probably lower it some still. There's many people pushing that OC under 1.35v in bios, topping out under 1.37v. So I'd not stop tweaking yet. Even that little of a drop could easily drop 5ish °C or better from your load temps.

Thanks. Using cpu-z to watch vcore and clock speed under load, that voltage is after LLC is applied, correct? I will try to get it a little lower. The ASUS rep said anything under 1.425 and under 90 degrees is acceptable so that’s what I’ve been shooting for. I do understand that the lower, the better though.

 

[Karadjgne]

Yes, according to what I can find, the vcore on cpu-z is After LLC. Just make sure it's vcore, not VID. I've got 2 versions of cpu-z, the Asus ROG and the MSI, and for some odd reason, the Asus has it listed as VID, but the MSI has it listed as vcore. Not sure what's up with that, I also had a default version that said VID and that was ugly until I realized what I was looking at wasn't what I thought it was. Sat at 1.114v for months, turned on the pc, did the update and running it right after I was suddenly at 1.208v....

 

[blazerboy]

Uhh... as someone who exclusively uses those and the 140mm version, I would not advise using those in a typical PC.
These fans get - obviously - ridiculously loud if one does not set a fan curve, or does not connect them to PWM headers.

I normally run them at 50%, where they're just as loud/quiet as the regular NF-A12 and NF-A14 fans, but they have a different pitch to them due to the 3 phase motor.
When stress testing - 9 of these things at 100%... you'll swear you're at an airfield!

If one never runs them at 100%, they're essentially wasting money. There are good rad fans that can be had for cheaper and don't run quite as loud.

I use the 2k rpm variant of these fans, and in hindsight I wish I had gotten the 3k and simply built a fan curve. By far the best rad fans iv used, including a comparison with the a12 variant. It definitely has far more cfm and static pressure. 2k rpms is far more tolerable if the sounds are identical on the 3k @ 2k. I personally have a thermaltake f suppressor case and while I can hear the fans, I don't find them bothersome. When I am away and the system is simply acting as a recording server with no one home, the added jump to 3k would have been a nice feature to increase cooling without worrying about noise whatsoever when ambient is over 100 (southern California heat) and no one is home. I'm pulling 56c on full load Intelburntest in 77F ambient , so pushing ambient to 100 or beyond (highest iv seen where it sits is 108), I'm sure you can understand why higher speeds would be beneficial. To help keep my chip cooler (marginally so).

All I'm saying is that for the price difference, I'd get the 3k and have that as an added bonus in potential even if you don't need to use them at 100%. Overdo it now so you don't have a need for it later. A 360mm rad is overkill. Why not go all the way on it

 

[Phaaze88]

@blazerboy
Yeah... when I was first contemplating whether to get the 2k or 3k model, I checked the specs and pricing, and was like, "Why did Noctua even bother with the 2k model, it's redundant..?
Just get the 3k model and set a fan curve... how hard is that?" So I bought several 3k 140 and 120mm fans.

 

[Karadjgne]

Ambient temps don't affect load temps much (if any) because the cpu has a far greater output of heat than the case temps. Ambient heat mostly affects idle temps because the cpu output is far less, so case ambient sets the starting temp of the Delta-T.

With 100°F, that's only 40ish °C, case ambient being @ 6-12°C higher on average, so you'd be looking at idle temps of @ 48-54°C ish. (assuming normal 23°C room ambient). Your loads would still be @ 60°C ish or under.

Higher airflow fans won't change that. Point of Diminishing Return vs physical impossibility to cool a cpu below ambient temps by mechanical means (liquid and air are both mechanical, you'd need chemical like peltier or ln2 or ac compressor)

 

[blazerboy]

@Karadjgne
Ambient heat will cause the Temps to rise even under load. By logic, the air going over the fins on the rad will be 30 degrees higher, therefore raising the water temp by 30 degrees roughly. Higher fans would move greater volumes of warmer air, however may be able to move it closer to ambient due to increased cfm. Now there are a few other factors at place like if my radiator is even getting saturated with heat, higher cfm won't do any good if my radiator is sized adequately and my water in my loop after the rad is near ambient. The higher fans will however make a difference if my radiator isn't large enough or my fans aren't good enough currently, and my water is well above ambient once leaving the rad.

Currently I am running at 24c on my intake and my cpu waterblock is running at 28c. There isn't a whole lot of improvement I can really do on that front on idle. Now under stress testing, I have seen the waterblock push to 32c when left for hours on end. That leads me to believe that I can achieve even cooler Temps by making sure the loop is fully dissipating the heat. We are talking about marginal returns here, but in theory, I could drop my load Temps another 8c under load by increasing the heat dissipation, by adding higher cfm fans, more rad space, and / or a stronger pump. Is it worth it, to drop my Temps from a working temp of 50c to 42, no it isn't. But it can be done.

 

[Karadjgne]

No no no lol, doesn't work that way. You assume it's a 1:1 linear ratio, it isn't. Not even close.

You coolant temp and ambient temp will be about the same. The radiator itself has a certain efficiency curve whereby air flowing through it at particular rates will dissipate a certain amount of energy. The cpu dumps energy (watts) into the coolant at the cpu block, that charged liquid hits the rad a second later (if that) and dumps the wattage to the fins and moves on. It has little to no time to effectively change the actual temp of the coolant and there's enough dissipation that even at 70°C on a 200w cpu ouput, it takes roughly half an hour to raise the coolant temp from 32°C to 40°C, and thats as high as the coolant goes.

With a rad, you aren't cooling off the coolant, you cool off the fins that have had the energy transfered to them from the coolant.

Put a pan of water on the stove on high (that's about 1500w or more) it takes several minutes before you can tell the temp difference in the water and a much longer time before it comes to a boil. Put that same pan on 'warm' (about 150w - 200w) and come back tomorrow and it's still not boiling because the surface area of the liquid can transfer to the air more than the burner can transfer to the pan.

Same as a fin can only absorb 'X' amount of energy from the coolant and transfer that to the airflow, once that limit is overcome, further gains in airflow don't mean anything.

Nzxt x61 280mm aio on Silent mode, fans at 700rpm got my i7-3770K @4.9GHz Prime95 small fft (@ 200w OC) 70°C. Same exact setp on Performance mode, fans @ 1600rpm, 68°C. With the Noctua 3k ippc I'd have been looking at maybe 67°C. More than likely the same 68°C. The fans at low rpm were effective enough to be almost at the limit, the fans at 1600rpm being at or past the limit.

Point of Diminishing returns, even a Delta 8k rpm fan wouldn't have made much, if any at all, difference in cpu temps.

 

[blazerboy]

Right, fin efficiency plays a roll, but as you stated the coolant goes to the fins, and depending on the design and fin structure, can absorb more or less heat from said coolant. No matter what you have, you will have a weak link unless everything is perfectly specced but it isn't. I feel like that was the goal of AIO units, simplicity and engineering where they could achieve close to maximum performance for the given components when assembled together. That doesn't mean maximum potential of a water cooling loop, that simply means maximum potential for any one given component. To reduce diminishing returns. They don't need a 1400L/hr pump on a cpu block AIO cooler. They get by with what is on it.

The greater the amount of fin density and on your rad, thickness, and size goes, the closer to ambient you can go, and having strong enough fans for that rad are essential for the largest heat dissipation.

Now again, the coolant will heat up, based upon how hot your cpu is running, and therefore you have a greater return by adding size or cfm to your fins if you haven't yet achieved their maximum potential. Adding a larger fan wouldn't necessarily drop my Temps at all, if the maximum coolant>fin>dissipation had been achieved. It would however decrease Temps if the maximum hasn't been achieved. If there is still leftover energy in those fins that the fan isn't removing, I'll call it, the thermal dissipation threshold for lack of knowing the proper term, then yes the fans would help move it. If not, you would simply be adding noise and unnecessary dust to your system.

The real true problem I have in my system, which isn't really a problem, more like a brain teaser is how to get better transfer from my ihs to the coolant. If I could improve heat transfer from my ihs to my coolant, then the water Temps would rise (albeit not much due to large volume in my loop), but then bigger fans may or may not play a role in removing that added heat / energy.

 

[Panzerbjorne39]

I like this discussion with that in mind @Karadjgne how can I stop my fans from needlessly hitting 90-100% RPM? With Corsair, their software run the pump and fan off the coolant temp which is much quieter than what I’ve got going which is controlled by the UEFI, based on CPU temps, which are always wildly fluctuating under load or no load.

 

[Gmoney06ss]

With the corsair software(using icue correct?), you should be able to tell it to use either cpu temp or coolant temp. Iirc. If icue is giving you a much quieter pc, and temps are still manageable, I'd use that. I would set the pump to extreme and leave it. It should always be at 100%, then adjust the fans in icue via cpu temp, if possible, to a level that keeps good temps and lowest possible noise levels.

I could be off on how icue works with the aio, as it has been a bit since I've used it for cooling. Now I just use it for keyboard and mouse lights. But above is how I had my h100i setup.

 

[Panzerbjorne39]

With the corsair software(using icue correct?), you should be able to tell it to use either cpu temp or coolant temp. Iirc. If icue is giving you a much quieter pc, and temps are still manageable, I'd use that. I would set the pump to extreme and leave it. It should always be at 100%, then adjust the fans in icue via cpu temp, if possible, to a level that keeps good temps and lowest possible noise levels.

I could be off on how icue works with the aio, as it has been a bit since I've used it for cooling. Now I just use it for keyboard and mouse lights. But above is how I had my h100i setup.

I’m not using Corsair. I was just making a statement and asking a question about taming AIO fans that are running off UEFI settings.

 

[Karadjgne]

Well that's the funky thing. Aio rads are pretty much all the same, very few variations. Custom rads are quite different. If you look at the xspc TX series super slim, they run 20mm thick but have very high density fpi, and very low flow resistance. As a consequence, they do awesome at low rpm, 700ish, but faster speeds with higher sp and their ability declines rapidly. The 30mm BlackIce Nemesis GTS slim is just the opposite, even having almost the same fin density, but flow restriction is higher and higher speeds are a serious benefit. All for a 10mm difference. And yet other 30mm like the AlphaCools run right down the middle, not as good at either end.

So rad design has a significant impact on cooling, but only when paired with the right fan. Pair it with the wrong fan and that ability goes in the toilet. A 3k ippc on a Tx240 would get horrific temps, same as a 1000rpm NF-F12 would on the GTS. Swap fans and Bingo.

Gotta get past the sales hype and see exactly where the components sit, strengths and weaknesses, what pairs good and what doesn't. Corsair was all about speed, sacrificing volume, even going as far as using 2800rpm SP120 fans on their rads, and here comes nzxt with a 1300rpm fan that beats them. Scythe Gentle Typhoons were the best rad fans for a long time, but looking at the stats was useless as they were not that good, mediocre sp, mediocre cfm, perfect ratio combination for an aio.

 

[blazerboy]

Well that's the funky thing. Aio rads are pretty much all the same, very few variations. Custom rads are quite different. If you look at the xspc TX series super slim, they run 20mm thick but have very high density fpi, and very low flow resistance. As a consequence, they do awesome at low rpm, 700ish, but faster speeds with higher sp and their ability declines rapidly. The 30mm BlackIce Nemesis GTS slim is just the opposite, even having almost the same fin density, but flow restriction is higher and higher speeds are a serious benefit. All for a 10mm difference. And yet other 30mm like the AlphaCools run right down the middle, not as good at either end.

So rad design has a significant impact on cooling, but only when paired with the right fan. Pair it with the wrong fan and that ability goes in the toilet. A 3k ippc on a Tx240 would get horrific temps, same as a 1000rpm NF-F12 would on the GTS. Swap fans and Bingo.

Gotta get past the sales hype and see exactly where the components sit, strengths and weaknesses, what pairs good and what doesn't. Corsair was all about speed, sacrificing volume, even going as far as using 2800rpm SP120 fans on their rads, and here comes nzxt with a 1300rpm fan that beats them. Scythe Gentle Typhoons were the best rad fans for a long time, but looking at the stats was useless as they were not that good, mediocre sp, mediocre cfm, perfect ratio combination for an aio.

This actually makes alot of sense, I suppose design of the rad would definitely dictate that. Interesting enough, I felt that increased air flow would always be a good thing, which it is... But there may be a point at which there is too much pressure or movement causing turbulence and fighting against itself. It would all boil down to rad and fan cohesion.

I have an EK Coolstream SE 240 Rad, I know EKWB likes to say that their vardar fans are purpose built for their rads, but at least on paper my noctua industrials are far better in both cfm and pressure. Interesting though. I'll tweak my setup and see how my Temps do

 

[Karadjgne]

These guys are serious about rads.

EK CoolStream SE 360mm Radiator Review - ExtremeRigs.net

EK CoolStream SE 360mm Radiator Review - SE compared to HardwareLabs Nemesis GTS, Magicool G2 Slim, XSPC EX 360, Alphacool ST30

www.xtremerigs.net

Hope it sheds some light 🕯️

 

[blazerboy]

These guys are serious about rads.

EK CoolStream SE 360mm Radiator Review - ExtremeRigs.net

EK CoolStream SE 360mm Radiator Review - SE compared to HardwareLabs Nemesis GTS, Magicool G2 Slim, XSPC EX 360, Alphacool ST30

www.xtremerigs.net

Hope it sheds some light 🕯

Yep iv read their xe 240 testing as well as the 360 se. The 360 faired pretty poorly but the 240 was quite good. Granted we are comparing the fat boy vs the slim versions but one thing was rad restrictions among other things that were likely in favor of the 240 outperforming (watts dissipated /size) the 360. No tests on the 240 but it works in my application and is largely oversized even if the sentiment of the 360 carried over to the 240. Cooling a 65w tdp chip is within reach of just about any 240mm rad iv come across.

 

[Karadjgne]

True, I'd be ashamed to market a 240mm if it couldn't.