[SOLVED] i9 9900k Air Cooling Help Please

[Axelrod]

I have a question here that I've kinda been pulling my hair out for, so first of all to understand my predicament, let me just tell you that Ebay and Amazon don't deliver in my country but I can get Aliexpress products delivered just fine.

I'm building an i9 9900k PC with a 2080 Ti that I won't be overclocking (CPU or GPU)

I want to go the air-cooling route but the problem is that Dark Pro or Be quiet or Noctua or any other mainstream air-cooling fan brands aren't available in my country.

The most expensive Air Cooler that I CAN get is the Cooler Master MasterAIR 620P.

I was thinking of doing a couple of modifications to it to get good temps that you'd generally get with on an i9 9900k with a Noctua NH D-15.

For starters, I was thinking that I'd replace the stock fans that the Master Air 620P comes with and put on some High RPM DC 12v Computer Server Fans or Corsair SP120/ML120 fans which are static pressure fans (I'll be testing both of them on the cooler to see which ones produce better temps and I'll be putting on 3 of them, the cooler only comes with 2 fans, so I'm betting that putting on 3 fans as opposed to 2 which are better than the stock fans will produce better temperature results.)

In all the benchmarks that I've seen online, the NH-D15 has performed better than the Cooler Master MasterAIR 620p and managed to provide an extra 7c - 10c of cooling.

I've provided ALL the benchmarks that I saw at the end of the post: (You don't need to read them, the only difference between cooling provided by MA620P and Noctua NH D-15 is a difference of 10c ONLY.)

Now physically comparing the coolers:

All three coolers come with 6 heatpipes.

MasterAIR 620P's Heatsink weight without fan = 850g
Noctua NH D-15's Heatsink weight without fan = 980g
Noctua's D14 Heatsink weight without fan = 900g

So Noctua's NH D-15 Heatsink is only 130g heavier than MasterAIR 620P.
And Noctua's D14 Heatsink is only 50g heavier than MasterAIR 620P.

I also watched a video by a youtuber where he used an i9 7900X paired with Coolermaster 212X and got 79c temps at 4ghz under an Aida64 Stress Test for 40 minutes.

And Coolermaster 212X cooler weighs a mere 492g, which is less than all the coolers mentioned here.

Now the i9 7900x is a 10 core 20 threads processor as opposed to the i9 9900k which is an 8 core 16 threads processor.

Now looking at the CFM numbers listed on the cooler's pages:

(The numbers listed on Noctua's website were cubic meter per hour so I converted them to cubic feet per minute using google.)

MasterAIR 620P CFM = 53.4
Noctua NH D-15 CFM = 67.6864
Noctua D14 CFM = 64.920129

Statistically speaking I think I'll be able to achieve the 65-67 CFM number by putting on any of the fans I mentioned above on the MasterAIR 620P cooler.

So do you guys think that the MasterAIR 620p cooler after these fan modifications will get some results very close to Noctua's NH D-15 and be able to support the CPU under 4k gaming load? and 1080p video editing/rendering?

P.S, Now I know that some of you are gonna suggest going with an AIO, I hate AIOs, I prefer air-cooling over liquid, I HAVE to go with air-cooling, AIOs are not an option.

Also, some of you might also suggest just buying a Noctua NH D15 Cooler from Aliexpress but the problem is that it being sold by 2 or 3 suppliers and neither listed product has a single review, so it might just be a cheap knockoff that doesn't produce any good results, so I'm going for a locally bought OEM product and doing fan mods on it instead.


Benchmark 1:

Benchmark 2:

Benchmark 3:

Benchmark 4:

 

[RealBeast]

I agree with your expectation that the MasterAIR 620P will perform pretty close to the other models by increasing the air flow. I also did not want an AIO.

While I did not keep track of the numbers, during some stress testing of my 9900KF with a beQuiet Dark Rock Pro 4, I put on some high speed but very loud fans that increased the airflow significantly (at the cost of a lot of noise) and I got 4-5 degree temperature drops from the already good level from that cooler. One pair were old high speed Delta 120mm at 58dBA!

 

[EndEffeKt_24]

Cant you get a Deepcool Assassin or Scythe Fuma 2?

 

[Axelrod]

I agree with your expectation that the MasterAIR 620P will perform pretty close to the other models by increasing the air flow. I also did not want an AIO.

While I did not keep track of the numbers, during some stress testing of my 9900KF with a beQuiet Dark Rock Pro 4, I put on some high speed but very loud fans that increased the airflow significantly (at the cost of a lot of noise) and I got 4-5 degree temperature drops from the already good level from that cooler. One pair were old high speed Delta 120mm at 58dBA!

Thankyou so much for the reply.
GOD, I've been waiting all day just to get at least one response from somebody xD

Thankyou for sharing your experience and your temperature numbers, that helps a lot.
Were you running those delta fans at 100%? What's the max RPM on those fans? How much RPM did you set your fans at? How many delta fans did you put on your cooler? And did you connect them to your motherboard fan headers or molex connectors?

The reason I don't want to go with AIOs is that there's a chance of failure, no matter how low, it's there, and when an AIO fails, the damage is far more catastrophic, one can potentially end up with a dead processor, dead motherboard and even some dead ram, that coupled with the fact that your AIO failed means that you've lost about 70% of your investment.

The most that can happen with an air-cooler is that the fan might fail and may need replacement, far easier and less messy, AIOs generally die within 3-4 years of use and Air-Coolers will probably last for 10 years plus without any problems, and may only need a fan replacement every 5 years, far less expensive than the AIO shenanigan.

So that means a person using AIO will be on his second AIO which will be close to dying while the other person will be comfortably using his Air Cooler.
Air coolers IMO is a far more safe and secure investment and future-proof tbh.

AIOs just have too many points of failure. (Radiator, Fan, Pump, Contact Plate, Liquid Degradation etc etc.)

One more idea that I did have was putting a 140mm fan in the middle of the dual tower heatsink with two 120mm fans on both the outsides, the 140mm fan in the middle as compared to the 120mm will obviously have a higher CFM further increasing the performance.

 

[Axelrod]

Cant you get a Deepcool Assassin or Scythe Fuma 2?

I cannot get ANY other brands except cooler master and the most expensive available cooler from Cooler Master is the MA620P sadly.

 

[Karadjgne]

The issue is capacity. Half of those tests above were done with 100w cpus that might be pushing 150w with OC. The i9 9900k series starts there and only goes upwards, a 5.1GHz all core boost has been clocked at 264w. That's well above what any aircooler can reasonably handle and keep under throttle temps.

As a cooler gets towards capacity, the temps go up logarithmically, they climb faster, so lower power usage isn't a problem, but higher power is, you can see the results with the above and see just how far the CM drops in standing when the thermal output is increased from the 6700k at 4.6/1.3v to 4.7/1.4v

 

[RealBeast]

The high CFM Delta (Delta FFB1212VHE 120mm x 38mm) runs at 3200 rpm I think but it is over 50dBA so it is quite obnoxious, and was only used a short time for testing due to the noise.

Make sure that your center fan would have enough space, the beQuiet only fits a 135mm center fan and it fits pretty tight, a 140mm would be questionable for it.

 

[Karadjgne]

Delta has 38mm fans that breach 8k rpm and 60db.

Sound pressure doubles every 3db, but it works logarithmically, so what ppl perceive in reality is sound volume doubling every 10db±. May not sound big, but if a Noctua fan is 20db, that Delta RealBeast tested is literally 8x as loud, the 8k Delta being 16x as loud. That's going from deep woods leaves rustling to sitting next to a normal dishwasher on rinse.

2 of those fans adds 3db, or 10db to your ears. Bring on the lawnmower.

At what point does the rule of diminishing returns override the ability of circum-aural noise canceling headphones?

 

[Karadjgne]

I want to run something by you, as I've been doing a lot of research on cooling systems lately, specifically radiators, for my upcoming build. Rads and heatsinks share almost identical thermal theory properties.

Tested was a koolance 240mm rad. Is 30mm thick and has 30 fins per inch (fpi). That's a high density rad, aios run @ 14fpi. Tested at 750/1300/1850 rpm.
Tested was a XSPC TX 240mm rad. Is 20mm thick with 22fpi. Same rpms.

At 750rpm, the koolance came dead last in cooling, the thinner TX was first, by a decent margin. At 1300, they were both middle of the pack. At 1850rpm the koolance was easy 10% ahead and the curve was still going up. The TX was dead last and going downhill in a hurry.

Fans have 2 properties, static pressure (the force of airflow) and cfm (amount of airflow). Fan stats don't show it, but sp rises in a curve, not a straight line. The higher the rpm, the faster the rise in sp.

So the thinner, lower fpi rad has less resistance to airflow at lower rpm so benefits more from what little sp the fans produce. The thicker, higher fpi rad benefits more from the higher sp-higher cfm produced from the higher rpm. The thin rad has air moving through it so fast its loosing efficiency and temps are higher overall.

All this is relative to the MasterAir 620p. When it was in design, CM fans were chosen for it. The designers/engineers know all the math about that fan, what it's sp/cfm is, and as a consequence designed the heatsink to take advantage of those properties. So fins were spaced a certain distance apart, heatpipes placed for maximum spread, the length and depth calculated to get the most performance at the fans rated speeds.

Now along comes you and want to add higher cfm fans, this may be a benefit or detriment, you are changing the properties. Higher cfm usually means lower sp. Higher cfm from higher rpm can mean increased sp. But like the TX, that may be a detriment or like the koolance a benefit. It'll all depend on the exact fan used. Increase the cfm/sp too much and you may blow air too fast and loose efficiency, blow too much cfm and not enough sp and loose efficiency. Too much sp and not enough cfm and loose efficiency.

It's a dance. You could try 10 different fans and none get the right results or get lucky and one specific fan gets incredible results. It's more than just rpm/sp/cfm, it's how it all combines with the design of the heatsink and its characteristics and properties. Noctua makes the best fans, but that doesn't necessarily mean a cheap Yate-loon might not actually out-perform it on the cooler itself.

So just wanted to say good luck with your testing, hope it works, but if you don't get the results you want, it may just be because it's the wrong fan for the application.

 

[Axelrod]

So just wanted to say good luck with your testing, hope it works, but if you don't get the results you want, it may just be because it's the wrong fan for the application.

Thankyou so much for the detailed reply, that is EXACTLY what I was looking for and I was thinking the same thing, at this point it's all just about testing different types of fans and finding one that hits the sweet spot.

When I do my experiments, I'll tag you here and let you know of the results because tbh I'm very curious myself to see if it works, and if it doesn't I guess I'll just take a chance at that Noctua Cooler that's available on Aliexpress and post temp results with different fans on it too.

 

[Axelrod]

So just wanted to say good luck with your testing, hope it works, but if you don't get the results you want, it may just be because it's the wrong fan for the application.

I have been doing some research on undervolting CPU, seen a few videos so far, I don't have a guru level grasp on that subject but I'm considering undervolting my CPU a bit to achieve better temps.

Do you have any experience in that arena?

 

[Karadjgne]

Every overclocker does. It's the premise behind it all. Every cpu is different, has to be, because no two bits of silicon are identical. So each and every cpu responds on a nano difference to voltages and current. As such, in order to guarantee stability, amd and Intel and nvidia all jack up stock voltages higher than any cpu could possibly need. Cover all bases. So your cpu could be barely under the stock voltages or way under. No telling exactly.

Overclockers start by bumping multipliers, it's the easiest way to use up what voltages are already applied and will then add as necessary. Or they'll jack up speeds and voltage to stability, then lower voltages to the lowest it'll handle and stay stable. The lower the voltages at the target speed, the lower the temps.

Ryzens are somewhat different in that they are dynamic cpus and will change voltages, currents and speeds as necessary for target temps. So fiddling with them gets trixy.

The 9900k is also dynamic in that it'll not allow thread usage if temps reach certain ranges, it'll protect itself to a certain extent unless you override that.

 

[Axelrod]

Every overclocker does. It's the premise behind it all. Every cpu is different, has to be, because no two bits of silicon are identical. So each and every cpu responds on a nano difference to voltages and current. As such, in order to guarantee stability, amd and Intel and nvidia all jack up stock voltages higher than any cpu could possibly need. Cover all bases. So your cpu could be barely under the stock voltages or way under. No telling exactly.

Overclockers start by bumping multipliers, it's the easiest way to use up what voltages are already applied and will then add as necessary. Or they'll jack up speeds and voltage to stability, then lower voltages to the lowest it'll handle and stay stable. The lower the voltages at the target speed, the lower the temps.

Ryzens are somewhat different in that they are dynamic cpus and will change voltages, currents and speeds as necessary for target temps. So fiddling with them gets trixy.

The 9900k is also dynamic in that it'll not allow thread usage if temps reach certain ranges, it'll protect itself to a certain extent unless you override that.

Do you recommend undervolting using the Intel Extreme Tuning Utility or should I directly change the vcore settings in the bios?

I've read about people not getting the chip to boot after they set the voltages too low in the vcore settings in the bios and now since the chip isn't booting up they can't even get into the bios to change the settings.

I was thinking I'd start undervolting at 0.025 mV and going down to see till the point it stays stable.

Is it better to undervolt using a software or directly? Can you recommend a good software to do it if so?
Also, does undervolting affect the life of the chip? Extend or reduce?

 

[Karadjgne]

No software. Even asus Turbo-V and other softwares designed to OC are done after boot/windows loads, it's not a system OC but a windows OC. So it's unreliable imho. There are also other settings in bios that you cannot see or access, they are deliberately hidden and those will often change slightly as they respond to other bios settings. With software, those are still in place, and different to what is really going on with the cpu.

Software is good for one thing, testing a cpu to its limits. Will it actually perform at 5.0GHz or 5.4GHz and what will the voltages look like. After that, it's go into bios, and manually set the pc up for a more permanent set of OC values. In bios there's also profiles, so you can save a 4.9GHz and a 5.1GHz and a default (always profile #1) etc so can always change how the pc is at boot.

Software will almost always leave behind remnants, orphan settings etc so can mess with stability, temps, voltages etc even after wiping everything you can. Some of those hidden settings. This creates issues later when/if you decide you don't get as much benefit from the high OC and drop it in favor of much better temps.

Unless you brick the bios during upgrades, you can always return it to defaults. Many overclockers will lock themselves out of bios with voltage settings, simply reset the bios/cmos by button or pins or pull the battery. It'll revert to factory defaults. Then recall your stable profile and keep tinkering. Bios will automatically revert to its last stable setting after (default 3x) attempts to start. Software will not.