Unable to cool 9900KS @5 Ghz on all cores without any limits

[John Naylor]

1. The Thermaltake 360 is not a great performer, As you can see here @ 4.4 Ghz (23:00 mark), it has to be almost twice as loud to get within 4 C the Swiftech H240X and more athan 4 times as loud as the D15 to beat it by 2C. And at 40 dbA, it ties the Noctua NH-D15 while the Thermaltake is 7 dbA (1.62 times) louder.

2. What do you mean pushing air thru the radiator ? For best performance, fans should be moving cooler outside ambient air thru the radiator into the case, not pushing preheated inside case air thru the rad. See Corsair's instructions below:

Lets look at this with your case situation ....

(3) 140s as Front Intake and let's say 20 % reduction in air flow w/ moderate dust buildup in filters so 3 x 0.8 = 2.40 EFs (equivalent fans)
(3) Exhaust fans on top, no filters = 3 x 1.0 = 3.00 EFs
(1) Exhaust fan on rear, no filters = 1 x 1.0 = 1.00 EFs

Even w/ filters removed, you still have a 4:3 deficit. If installed as was assumed above, flip the fans over for a 6 in / 1 out scenario ... the mindset that an intake fan "needs' help getting air out of a vented enclosure is non-scientific ... And attic fan in a building doesn't have a corresponding intake fan ... an exhaust fan in a kitchen does not need an intake fan ... a window fan in a bedroom window needs noting but another open window to work properly.

That's 4 EFs worth of air going out and just 2.4 EFs coming in leaving you with a deficit of 1.6 EFs. Where is that make-up air coming in ? It's coming in thru those veted slot covers and rear grille .... Is case installed on top of (or worse under) your desk against wall ? ... Isn't that the same space where your PSU and GPU exhaust their hot air ? Then that hot exhaust is likely being sucked back in to your PC. We use a "garage band' fog machine to test this and when discharging the fogger behind the PC, we often see the case fill up w/ fog when rad fans are installed contrary to the laws of thermodynamics and manufacturers recommendations. yes, we all learned in 8th grade earth Scince that hor air rises ... but when a fan is present, it goes the way a fan wants it to go.

3. CLCs are limited in what they can do due to teeny (0.11 gpm) pumps and low efficiency aluminum rads; they require extreme fan speeds to compensate. If ya plan on using an 360 AIO, would recommend the Swiftech all copper units (H360X) with 1.0+ gpm pumps and all copper componentry. Delta T (Ambient to Coolant) usually runs 8 - 12C

Swiftech Drive X3 AIO CPU Coolers

Drive X3 AIO, all-in-one CPU cooler with built-in addressable RGB lighting

www.swiftech.com

4. You should have the following fans as a minimum:

One (1) 1250 rpm, 140mm fan for each 75 - 100 watts of component max power draw
One (1) 1250 rpm, 120mm fan for each 50 - 75 watts of component max power draw

No. of Intake fans = 1.3 - 1.5 times No. of Exhaust fans to compensate for moderately dusty air inlet filters and avoid having hot GPU / PSU exhaust being sucked into case thru vented slot covers and rear case grilles.

5. What is the purpose of this build ?

a) Competitive Overclocking and earning place on web site leader boards
b) Running games and applications

If it is b), then I see no value in running synthetic stress tests. These utilities place unrealistic loads on the system. If its b), use a application based, mulithreaded stress test like RoG Real Bench which will place for heavier loads on the system then it will ever see again ... and in a multitasking environment.

 

[mrthanhnguyen]

Im selling some 360 aio in the selling thread. You may want to take a look

 

[EarthDog]

Everyone saying "don't focus on the 5Ghz" is out of their mind. The entire point of the chip is the 5Ghz all-core turbo. If his can't make it without throttling it's not getting it's rated speed, and should be warrantied.

But it's working in spec. 5 ghz turbo. When the boost time (tau) runs out, boost drops to spec.

Anyway, looks like it was resolved...that auto voltage will get you every time.

 

[TxGrin]

I would take the chip back and swap it out, Mine at 5.2Ghz right out of the box under 100% load and only gets to 70c. I paired mine with the Gigabyte Z390 AORUS MASTER motherboard and a
Hydro Series™ H115i RGB PLATINUM

 

[CopperBlockhead]

There is something else you might want to investigate: How your motherboard conforms to the Intel Spec regarding the TDP. The "throttling" phenomenon you mentioned appears to be a function of the motherboard manufacturer. For example, the rather pricey MSI "MEG Z390 GODLIKE" motherboard seems to hold overclocks higher and longer without throttling if you have the BIOS setting tweaked right and your cooling solution is managing the heat load. Not all motherboards treat the Intel spec the same. Some "chicken out" and throttle often. Others, like this MSI board, seem to say "OK, if you know what you're doing, here goes..." and you will live or die by your cooling solution.

Remember, TDP is not the full story. The technical details in the spec DO allow for 1.25 times to the TDP to be the "max" power for a duration not to exceed that "mu" factor.

 

[Vayra86]

There is something else you might want to investigate: How your motherboard conforms to the Intel Spec regarding the TDP. The "throttling" phenomenon you mentioned appears to be a function of the motherboard manufacturer. For example, the rather pricey MSI "MEG Z390 GODLIKE" motherboard seems to hold overclocks higher and longer without throttling if you have the BIOS setting tweaked right and your cooling solution is managing the heat load. Not all motherboards treat the Intel spec the same. Some "chicken out" and throttle often. Others, like this MSI board, seem to say "OK, if you know what you're doing, here goes..." and you will live or die by your cooling solution.

Remember, TDP is not the full story. The technical details in the spec DO allow for 1.25 times to the TDP to be the "max" power for a duration not to exceed that "mu" factor.

I can at least confirm that Asrock boards also go the full Tjmax, met that once already

 

[Okys]

You should definitely try to bench with an open case (both side panels off) and see what happens, just to get a feel of the impact of the case on your temps.

That way you can get a better judgment on what is happening with your other components. I doubt you have faulty gear, tbh. Judging from your last post you run a very comfy vcore now, and 82C is safe but it does also read like on the hot side given your cooling. 1.25V isn't a whole lot, and when you say Cinebench AVX load, is it really running 5 Ghz or are you using an offset?

That's what I already tried (mentioned in previous posts) but on Auto voltage it didn't give noticeable difference.
About the AVX offset I also mentioned before, it is set to 0.
The strange behavior that still doesn't have an explanation is, that only on Auto voltage (without changing anything else and keeping AVX offset at 0) I am able to achieve 5200 - 5220 pts in Cinebench R20, which seems an average for 9900K/9900KS on 5 Ghz AVX. As soon as I change voltage to manual or adaptive+ (and only that) the score drops by around 170-180 pts.

Just a quick time and my 2 cents. As long as you get the AIO pump speed sorted or have it sorted already, drop the vccio and vcsa down to 1.21 and 1.18 respectively. I'll have to check my bios for the same settings but i'm running roughly the same setup as you, board, ram, and just a 9900k at 5.1 stable, and i hit high 70's and low 80's on custom water. Dropping those two voltages from auto gave me a lot of headroom though.

Thanks, I will definately try that, when wife goes to work again and her PC becomes available.

Not sure if it was mentioned, but how are the fans on your AIO set up? Are they exhaust or intake?

They are exhausting air from the case through the 360mm radiator on top. It's just how wife wanted it to be. I will continue further experiments, when her PC becomes available, but it won't make a big difference anyway if i install 360 radiator on front. If she's rendering something, she aren't using both CPU and GPU, like AIDA64 stress test did. This test was the only way to throttle the CPU. For gaming, I ran ARK. CPU was around 60-62 degrees.

1. The Thermaltake 360 is not a great performer, As you can see here @ 4.4 Ghz (23:00 mark), it has to be almost twice as loud to get within 4 C the Swiftech H240X and more athan 4 times as loud as the D15 to beat it by 2C. And at 40 dbA, it ties the Noctua NH-D15 while the Thermaltake is 7 dbA (1.62 times) louder.

2. What do you mean pushing air thru the radiator ? For best performance, fans should be moving cooler outside ambient air thru the radiator into the case, not pushing preheated inside case air thru the rad. See Corsair's instructions below:

Lets look at this with your case situation ....

(3) 140s as Front Intake and let's say 20 % reduction in air flow w/ moderate dust buildup in filters so 3 x 0.8 = 2.40 EFs (equivalent fans)
(3) Exhaust fans on top, no filters = 3 x 1.0 = 3.00 EFs
(1) Exhaust fan on rear, no filters = 1 x 1.0 = 1.00 EFs

Even w/ filters removed, you still have a 4:3 deficit. If installed as was assumed above, flip the fans over for a 6 in / 1 out scenario ... the mindset that an intake fan "needs' help getting air out of a vented enclosure is non-scientific ... And attic fan in a building doesn't have a corresponding intake fan ... an exhaust fan in a kitchen does not need an intake fan ... a window fan in a bedroom window needs noting but another open window to work properly.

That's 4 EFs worth of air going out and just 2.4 EFs coming in leaving you with a deficit of 1.6 EFs. Where is that make-up air coming in ? It's coming in thru those veted slot covers and rear grille .... Is case installed on top of (or worse under) your desk against wall ? ... Isn't that the same space where your PSU and GPU exhaust their hot air ? Then that hot exhaust is likely being sucked back in to your PC. We use a "garage band' fog machine to test this and when discharging the fogger behind the PC, we often see the case fill up w/ fog when rad fans are installed contrary to the laws of thermodynamics and manufacturers recommendations. yes, we all learned in 8th grade earth Scince that hor air rises ... but when a fan is present, it goes the way a fan wants it to go.

3. CLCs are limited in what they can do due to teeny (0.11 gpm) pumps and low efficiency aluminum rads; they require extreme fan speeds to compensate. If ya plan on using an 360 AIO, would recommend the Swiftech all copper units (H360X) with 1.0+ gpm pumps and all copper componentry. Delta T (Ambient to Coolant) usually runs 8 - 12C

Swiftech Drive X3 AIO CPU Coolers

Drive X3 AIO, all-in-one CPU cooler with built-in addressable RGB lighting

www.swiftech.com

4. You should have the following fans as a minimum:

One (1) 1250 rpm, 140mm fan for each 75 - 100 watts of component max power draw
One (1) 1250 rpm, 120mm fan for each 50 - 75 watts of component max power draw

No. of Intake fans = 1.3 - 1.5 times No. of Exhaust fans to compensate for moderately dusty air inlet filters and avoid having hot GPU / PSU exhaust being sucked into case thru vented slot covers and rear case grilles.

5. What is the purpose of this build ?

a) Competitive Overclocking and earning place on web site leader boards
b) Running games and applications

If it is b), then I see no value in running synthetic stress tests. These utilities place unrealistic loads on the system. If its b), use a application based, mulithreaded stress test like RoG Real Bench which will place for heavier loads on the system then it will ever see again ... and in a multitasking environment.

There are some points on which i agree with you, but not all.
The top exhaust (through radiator) or front intake will, of course affect the CPU temps, but not as much as someone would think. Non-blower GPU is messing everything up anyway. With 3x140mm intake fans reacting to CPU temp, there's lots of fresh air getting into the case. It is not like everything is boiling inside without any cool air intake and only escaping through the top.
The only thing I would consider reversing is the single rear 140mm PWM fan, to get cool air in for VRM's. This will require dust filter, of course. At the moment it is just spinning at low rpm's. I think the air pressure inside the case is +/- balanced this way, but of course I will check it with time. If the pressure inside the case will turn out to be negative, I will reverse the rear fan, because I don't want dust entering the case through every possible opening.

About the other things.. PC sits on a desk. I always try to avoid putting them on the ground, or somewhere where airflow is restricted.
Thermaltake Water 3.0 ARGB is not a bad performer either. It's an average, Asetek type AIO, performing about the same as others. Was chosen by wife, because of the looks and availability in my country. For myself I'm using Noctua NH-D15S and would never choose an AIO, simply because of reliability and noise. I don't care the looks.
The Swiftech cooler you mentioned has limited ways for installing, because pump is located near the radiator and it's not available in my country anyway.
Purpose of the build is gaming, some video rendering and Photoshop.

But it's working in spec. 5 ghz turbo. When the boost time (tau) runs out, boost drops to spec.

Anyway, looks like it was resolved...that auto voltage will get you every time.

Im also using Auto voltage with best case scenarion on my Maximus X Hero and 8700K and it runs perfecty fine @ 4.8 Ghz on all cores with everything i feed to it.
For 9900KS on Maximus XI Hero it's a different story... It's a huge difference 1.33v auto voltage, or 1.25v manual (1.225v in HWmonitor). I haven't tried lowering the voltage more, 1.25v was my first try. When this PC becomes available again, I will continue experimenting.
Another person posted in the comments to try to get it to 5.2 Ghz and setting AVX offset to -2. This is also something I want to try.

 

[Agentbb007]

Okys it might be worth it to apply new high quality thermal compound paste to your cooler as well. I just applied ARCTIC MX-4 2019 Edition and my max temp went from 87C to 75C during a Cinebench R20 benchmark!

 

[mrthanhnguyen]

he's using Kryonaut which is one of the best thermal paste on the market now.

 

[EarthDog]

he's using Kryonaut which is one of the best thermal paste on the market now.

Not to mention no common thermal paste is bad enough to be 12C different... other factors are clearly involved there.

 

[mrthanhnguyen]

That temp is normal I think. For cinebench, Im also crazy about the score. I have not seen anyone hit 5200 at 5ghz (maybe with real time priority) I hit close to 5500 at 5.2ghz with real time priority.It stays around 60-70 during stress but some cores spike up to 83-84 within a second then drop back down.

 

[DeathtoGnomes]

1. The Thermaltake 360 is not a great performer, As you can see here @ 4.4 Ghz (23:00 mark), it has to be almost twice as loud to get within 4 C the Swiftech H240X and more athan 4 times as loud as the D15 to beat it by 2C. And at 40 dbA, it ties the Noctua NH-D15 while the Thermaltake is 7 dbA (1.62 times) louder.

2. What do you mean pushing air thru the radiator ? For best performance, fans should be moving cooler outside ambient air thru the radiator into the case, not pushing preheated inside case air thru the rad. See Corsair's instructions below:

View attachment 137500

Lets look at this with your case situation ....

(3) 140s as Front Intake and let's say 20 % reduction in air flow w/ moderate dust buildup in filters so 3 x 0.8 = 2.40 EFs (equivalent fans)
(3) Exhaust fans on top, no filters = 3 x 1.0 = 3.00 EFs
(1) Exhaust fan on rear, no filters = 1 x 1.0 = 1.00 EFs

Even w/ filters removed, you still have a 4:3 deficit. If installed as was assumed above, flip the fans over for a 6 in / 1 out scenario ... the mindset that an intake fan "needs' help getting air out of a vented enclosure is non-scientific ... And attic fan in a building doesn't have a corresponding intake fan ... an exhaust fan in a kitchen does not need an intake fan ... a window fan in a bedroom window needs noting but another open window to work properly.

That's 4 EFs worth of air going out and just 2.4 EFs coming in leaving you with a deficit of 1.6 EFs. Where is that make-up air coming in ? It's coming in thru those veted slot covers and rear grille .... Is case installed on top of (or worse under) your desk against wall ? ... Isn't that the same space where your PSU and GPU exhaust their hot air ? Then that hot exhaust is likely being sucked back in to your PC. We use a "garage band' fog machine to test this and when discharging the fogger behind the PC, we often see the case fill up w/ fog when rad fans are installed contrary to the laws of thermodynamics and manufacturers recommendations. yes, we all learned in 8th grade earth Scince that hor air rises ... but when a fan is present, it goes the way a fan wants it to go.

3. CLCs are limited in what they can do due to teeny (0.11 gpm) pumps and low efficiency aluminum rads; they require extreme fan speeds to compensate. If ya plan on using an 360 AIO, would recommend the Swiftech all copper units (H360X) with 1.0+ gpm pumps and all copper componentry. Delta T (Ambient to Coolant) usually runs 8 - 12C

Swiftech Drive X3 AIO CPU Coolers

Drive X3 AIO, all-in-one CPU cooler with built-in addressable RGB lighting

www.swiftech.com

4. You should have the following fans as a minimum:

One (1) 1250 rpm, 140mm fan for each 75 - 100 watts of component max power draw
One (1) 1250 rpm, 120mm fan for each 50 - 75 watts of component max power draw

No. of Intake fans = 1.3 - 1.5 times No. of Exhaust fans to compensate for moderately dusty air inlet filters and avoid having hot GPU / PSU exhaust being sucked into case thru vented slot covers and rear case grilles.

5. What is the purpose of this build ?

a) Competitive Overclocking and earning place on web site leader boards
b) Running games and applications

If it is b), then I see no value in running synthetic stress tests. These utilities place unrealistic loads on the system. If its b), use a application based, mulithreaded stress test like RoG Real Bench which will place for heavier loads on the system then it will ever see again ... and in a multitasking environment.

thats some crazy math, but not exactly wrong either, the typical measurement uses case pressure, which can use this math to get the same desired result. Meaning 3 intake fans and 3 exhaust fans typically relate to 0 (zero) case pressure. This math would result in a negative case pressure, unless you used better fans as suggest ( static pressure fans ). Tactical fan placement with quality fans can let you get away with an extra exhaust fan.

ideally, you want the 360 exhausting out the top, and the 120 out the back, the lower air flow thru the radiator and rear would make 3 high flow intake fans good enough to keep negative case pressure closer to 0. Mounting a 360 (3 fan) to the front as intake in a pull config i can see needing 4 exhaust fans just for the airflow, but this leads to a higher negative case pressure. The only main fault is that it would require higher than normal case maintenance.

 

[trog100]

"I just want to run a stable, advertised 5Ghz on all cores, in any situation, while keeping the temps below 90 degrees. Is that even possible "

to cut a very long story short.. when all cores are fully loaded and with normal type cooling the answer is no..

trog

 

[CopperBlockhead]

For my i9-9900KS build I am going to go with the a high performance two-phase thermosyphon coupled with a microchannel heat pipe. This should be able to cool 8-cores running 24x7 even at 5.2 GHz.

 

[freeagent]

Big heat sink, big fans, lots of air. Forget about AIO’s
High performance and silence do not go hand in hand, just sayin.

 

[EarthDog]

"I just want to run a stable, advertised 5Ghz on all cores, in any situation, while keeping the temps below 90 degrees. Is that even possible "

to cut a very long story short.. when all cores are fully loaded and with normal type cooling the answer is no..

trog

Depends on the test and what you consider normal cooling. High-end air or a AIO (2X120+) running AIDA64... more can than you may think.

If 95% of these cpus can run 1.25V and 5ghz with a -2 avx offset with a 240mm aio running tests generally more stressful than aida...

Intel Core i9 9900KS @ 5.0GHz Special Edition

Intel Core i9 9900KS Overclocked and Binned CPU.

siliconlottery.com

 

[CopperBlockhead]

High performance and silence do not go hand in hand, just sayin.

The sound of a 2-stage thermosyphon is the sound of a volatile drop of moisture turning to vapor
It removes about 260 Watts per square centimeter of surface area.

When my parts show up, I'll create a build log for you guys to demonstrate.

 

[EarthDog]

The sound of a 2-stage thermosyphon is the sound of a volatile drop of moisture turning to vapor
It removes about 260 Watts per square centimeter of surface area.

When my parts show up, I'll create a build log for you guys to demonstrate.

That's awesome... hope it isnt wasted on a 24/7 system and push that bad boy for real. That said...it isnt so much the cooler as it is actually getting the heat out of that concentrated place. Regardless, I cant wait to see the build log.

 

[CopperBlockhead]

That's awesome... hope it isnt wasted on a 24/7 system and push that bad boy for real. That said...it isnt so much the cooler as it is actually getting the heat out of that concentrated place. Regardless, I cant wait to see the build log.

Here is the video of one that is already built
link removed

 

[EarthDog]

Here is the video of one that is already built
link removed

Do you work for them or something and spamming links, lol? You are all up on this thing....I read the first one.

 

[CopperBlockhead]

Do you work for them or something and spamming links? You are all up on this thing....

No I don't work for them. And I am new here. I made a post and then I did not see where it went. So at most I did 2 posts about it. But I can only see the one I just made. I thought people would be interested in seeing something that looks like a real game-changer that isn't really widely known.

I'll just keep it to myself if that's what the board wants.

 

[EarthDog]

No I don't work for them. And I am new here. I made a post and then I did not see where it went. So at most I did 2 posts about it. But I can only see the one I just made. I thought people would be interested in seeing something that looks like a real game-changer that isn't really widely known.

I'll just keep it to myself if that's what the board wants.

lol, relax... there are two posts each with the link in it. We would have no idea you cant see your first post.

Anyway, looks cool, but is it ready/in production? The website is just that one page... you cant buy it, cant click on details of either product, nada. Just a video and that's it. Maybe I couldn't find it but how do you know it can mitigate 260w of heat in a cm? Did you talk with them?

Have you order it already? How much is it? Hiw much wattage can it cool at what delta? Info would be great...

Edit: maybe create a thread for it to discuss instead of off topic here? I'm really curious about this being an extreme overclocker (semi-retired).

 

[Deleted member 185158]

lol, relax... there are two posts each with the link in it. We would have no idea you cant see your first post.

Anyway, looks cool, but is it ready/in production? The website is just that one page... you cant buy it, cant click on details of either product, nada. Just a video and that's it. Maybe I couldn't find it but how do you know it can mitigate 260w of heat in a cm? Did you tall with them?

Have you order it already? How much is it? Hiw much wattage can it cool at what delta? Info would be great...

They just got their patent in may 2019. There's no specifications and most of the links are 404.

 

[EarthDog]

They just got their patent in may 2019. There's no specifications and most of the links are 404.

Exactly what I was getting at.

Wondering how this guy knows all this stuff about the product when it doesnt appear to be on the website. It looks cool as hell, but wondering how well it really works...cost...capacity...etc... feels like vaporware.....yet parts are on the way.

Please understand why I asked if you(copper) worked there.

 

[Deleted member 185158]

Exactly what I was getting at.

Wondering how this guy knows all this stuff about the product when it doesnt appear to be on the website. It looks cool as hell, but wondering how well it really works...cost...capacity...etc... feels like vaporware.....

It would be nice to have some specs. I mean I wouldn't be mad if he was advertising, it's not on my dime

But from the looks of it, it's alcohol based cooling system with a patent cold plate that may efficiently exchange heat well enough to remove from a processor.

Just exactly how many BTU is not clear, the video only said "hundreds" of BTU. This is not a chiller by any means.

No I don't work for them. And I am new here. I made a post and then I did not see where it went. So at most I did 2 posts about it. But I can only see the one I just made. I thought people would be interested in seeing something that looks like a real game-changer that isn't really widely known.

I'll just keep it to myself if that's what the board wants.

Am interested. I like chilling things, cooling things, removing heat from things. It's kinda what I do!

Please do share, we are curious and have almost no information to go by.