Jump to content
Corsair Community

Upgrade my Corsair Hydro H115i (i9-9900K OC 5,1Ghz)


tigerblue77

Recommended Posts

Hi, guys! ;):

 

I built the war machine of my dreams earlier this year and bought a Corsair Hydro H115i (1st generation) watercooling to cool my Intel Core i9-9900K overclocked at 5.1Ghz on all its cores.

 

Now, this cooling is a bit noisy in games and with the overclock the chip reaches 90/100°C on load and the water about 45/50°C...

Could you advise me a better Corsair watercooling adapted for my processor ?

Will the gains be significant ?

Do you advise a delid or direct die cooling ? (seems very cool but touchy !)

I don't care about RGB, my only criteria are the cooling capacity, the noise generated and... the price... a little bit... anyway...

 

My case is a Corsair 780T, so it accepts 360mm watercooling... I ask a question here about them : https://forum.corsair.com/v3/showthread.php?p=1041975#post1041975

 

Thank you per advance!

 

Here's my configuration:

 

- Intel Core i9-9900K CPU overclocked at 5,1 Ghz on all cores

- Asus Strix Z390-E motherboard

- Corsair RGB PRO DDR4 2x8 GB 3600Mhz RAM/memory

- Corsair Hydro H115i (1st version) top mounted

- Asus Dual OC Nvidia RTX 2080

- Intel Optane M10 32GB

- Corsair AX760i power supply

- LG Bluray disc reader/writer

- a few drives... 5 actually

Edited by tigerblue77
Link to comment
Share on other sites

You can probably search up the same delid results as me, but from what I've seen it usually is a small gain. Not like the prior 8700 series where you could steal back 20C by doing this. Still, if -5C keeps you out of the trouble zone, it could be worth it.

 

Unfortunately, the above is likely the only real options, if everything is working correctly. I am concerned about your coolant temperatures. You should never get to 45-50C coolant temp unless the GPU is feeding its waste heat into it as well or your room is super hot to begin with (30-35C). Given this is 780T, I am assuming you did the logical placement and put the H115i as top exhaust like most people. These are not normal results.

 

That specific series of cooler seems to be vulnerable to a building flow restriction or blockage over time. One way to analyze this is to watch the behavior of the H115i Temp (coolant temp) at cold boot. When you first power on, the coolant temp should be only 1-2C above room temp and will warm to +4-7C above room temp over the first 5-10 minutes. If you see the coolant temp methodically tick upwards 25,26,27...35, 36... etc. in front of your eyes, there likely is an issue.

 

However, since idle wattage is low, this may not identify the problem unless it is severe. Take a mild CPU test like the "Bench Stress Test" in CPU-Z. This is an easy and linear 100% load with no complex instructions. I would expect a +6-7C rise for your combination for a 10 minute test. Run it for 10 and then stop. Obviously if the coolant temp went up +15C, that is a problem. If it is more like 8-10C, watch the coolant for a few minutes. It should shed 3-4C within 1-2 minutes after the load stops. If the temperature seems to hang at this elevated level, that is another sign the heat isn't getting to the fins as fast as it should.

Link to comment
Share on other sites

5°C would be a good beginning to cool this x)

 

I saw a video on youtube where the guy is hand-sanding his 9900K chip before putting the lid back on... If I understood correctly (because English is not my main language), he explains that the 9900K is much thicker than the 8700K and that this excess of material "suffocates" the processor a bit by preventing a good heat propagation. I was rather surprised... I don't see what interest Intel would have in putting "too much matter"... Here is the link if you can give me your expert opinion :) [ame]

[/ame]

 

About configuration, yes everything is working fine but my core voltage is very high I think... It's my first OC so it is probably very unperfect... I made 3 OC modes :

- 5,0 Ghz on up to 3 active cores and 4,8 Ghz for 4 to 8 active cores => 1,215 V on core voltage

- 5,0 Ghz on all cores & 4,7 Ghz cache => 1,45 V on core and cache voltages

- 5,1 Ghz on all cores & 4,7 Ghz cache => 1,48 V on core and cache voltages

 

And the default one :

- 5,0 Ghz on up to 3 active cores and 4,2 Ghz for 4 to 8 active cores => 1,175 V on core voltage

 

My cooler, is actually as you said : in push mode, on top of the case as an exhaust :)

I have :

- 2x140mm input fans in front

- 1x140mm input fan at bottom

- 1x140mm exhaust fan at back

My desktop is not super hot, about 21°C before the PC starts and my GPU is at 31°C on average, I configured it to be cooled actively (thanks to its 2 fans) regardless of its temperature instead of being passively as in the factory settings.

 

Ok I'll do the test with CPU-Z, in which profile listed below do you want me to run it ? 5,1Ghz ?

Actually I'm on default mode (4,2Ghz), CPU usage about 10% and average CPU temp is 48°C, average water temp is 36,8°C and room about 22°C so my water seems hot...

 

Thanks for your answers and your help !! :)

I'll wait this topic to be solved before responding the second one : https://forum.corsair.com/v3/showthread.php?t=195621

Edited by tigerblue77
Adding infos and correcting current mode
Link to comment
Share on other sites

I think your voltage for the 5 and 5.1 GHz settings are too high. Intel did something different with the 9900 family and the voltage is comparatively lower than prior models by 0.05 - 0.10v. That means your 1.4x voltage is really high.

 

The other things to look at is your per core overclocking, although this is not related to the temperatures. I used to do this in years past, but starting a few years ago the ability of the CPU to confine its work to a select group of cores disappeared. In general, the newer CPUs will try to shift the work around as much as possible and in particularly Coffee Lake CPUs are highly tuned and very snappy. They will try to maximize every task. In the end, this means you may never get your 5.1 because it is constantly using 6-8 cores. To that end, I prefer to set an All Core overclock for the entire package at the voltage I need. You can still use adaptive voltage so it shifts down at idle and lighter loads.

 

As for what you need now, I can make an educated guess and you can compare that with other 9900K owners and results around the web. Most users will need 1.25v for 5.0 on very good CPUs and closer to 1.35 on poor performer. That is a pretty decent range, so perhaps start at 1.30v and work either up or down. Either way, that is a huge difference between there and 1.48v. Actually, I am impressed it didn't shut you down at that level, so perhaps you do have a good CPU.

 

Don't stress test until you get a fix on the voltage. No need to test the upper temp range threshold. However, I do think the coolant is too warm so there may be two issues here.

Link to comment
Share on other sites

My rig began life with a h115i cooling my 9900K and to be honest about 3 months i came to the conclusion a 280mm RAD was just not enough to cool the 9900k when overclocking. It just isnt.

 

I sold the h115i and stepped up to a h150i Pro 360mm AIO. I also relocated the RAD to the front of my case and using the Corsair ML 120 RED LED Fans in push/pull config (2400 RPM MAX)(custom fan curve) and a fresh application of thermal grizzly paste i dropped my idle temps on my 9900K to 26c from 30-31c.

 

The voltages you mentioned are oddly high to achieve those overclocks. On Cinebench R20 i get 77c at 5.2GHz OC.

 

Every chip is different yes, but here are my OC profiles i have saved in my BIOS for my OC's - These are my settings after HOURS and DAYS of testing i want to point out. I started high and worked it down till i got a BSOD. I can run every game i play and benchmark software & video software with no BSOD's.

 

Only test i have not ran because it doesn't matter to me is AVX instructions. Its just not a real world scenario and therefor, i couldn't care less if i can run AVX for 45683 hours or not.

 

5GHz - 1.21V

5.1GHz - 1.27v

5.2GHz - 1.29V

 

5.2GHz is as far as my CPU will go being able to run anything. I can get my PC to post at 5.3GHz put as soon as i try to put any load i get a BSOD. i tried as much as 1.32V at 5.3GHz and still couldn't run Cinebench R20. I stopped there and said its good enough being able to reach 5.2GHz i'm happy with that.

Link to comment
Share on other sites

I think your voltage for the 5 and 5.1 GHz settings are too high. Intel did something different with the 9900 family and the voltage is comparatively lower than prior models by 0.05 - 0.10v. That means your 1.4x voltage is really high.

 

The other things to look at is your per core overclocking, although this is not related to the temperatures. I used to do this in years past, but starting a few years ago the ability of the CPU to confine its work to a select group of cores disappeared. In general, the newer CPUs will try to shift the work around as much as possible and in particularly Coffee Lake CPUs are highly tuned and very snappy. They will try to maximize every task. In the end, this means you may never get your 5.1 because it is constantly using 6-8 cores. To that end, I prefer to set an All Core overclock for the entire package at the voltage I need. You can still use adaptive voltage so it shifts down at idle and lighter loads.

 

As for what you need now, I can make an educated guess and you can compare that with other 9900K owners and results around the web. Most users will need 1.25v for 5.0 on very good CPUs and closer to 1.35 on poor performer. That is a pretty decent range, so perhaps start at 1.30v and work either up or down. Either way, that is a huge difference between there and 1.48v. Actually, I am impressed it didn't shut you down at that level, so perhaps you do have a good CPU.

 

Don't stress test until you get a fix on the voltage. No need to test the upper temp range threshold. However, I do think the coolant is too warm so there may be two issues here.

 

Thank you for confirming that my voltages were too high, unfortunately lowering them causes a restart.

 

I think you misunderstood my previous message, as I said my two profiles (at 5.0 and 5.1Ghz) are overclocked at these frequencies on their 8 cores and I use adaptive voltage too. However, there is a core base voltage that can be varied and I had to increase it to reach such a frequency (I overclock using Intel XTU (Extreme Tuning Utility)).

 

I really hope I have a good CPU because I had a few dozen that I quickly compared to keep the one that seemed to be the best and then I sold the others haha! Honestly I'd really like to improve my overclock skills, I made mine by following the videos and advices of Wizerty, quite famous apparently.

 

Anyway, back to my problem if you don't mind, I think I'll have to fix my cooling problem first before overclocking it?

 

Since my overclock was not good, I just put back the factory BIOS settings and run the test you explained to me earlier, exactly in the requested conditions. So : overvoltage motherboard's switch turned off, closed PC case (as in real use conditions), a minimum of applications running, the mode of my watercooling on "silent" and I waited for my temperatures to drop to a minimum before launching the test.

 

Initial state :

Ambient temperature: 21°C

Initial CPU temperature: 40°C

Initial water temperature: 33.2°C

 

3 minutes after the stress test started:

CPU temperature: 100°C

Water temperature: 48.5°C

 

Five minutes after the stress test started:

CPU temperature: 100 degrees Celsius.

Water temperature: 51.4°C

 

10 minutes after the stress test started:

CPU temperature: 100 degrees Celsius.

Water temperature: 53.2°C

 

1 minute after stopping the stress test:

CPU temperature: 45°C

Water temperature: 47.2°C

 

two minutes after you stopped the stress test:

CPU temperature: 45°C

Water temperature: 42.6°C

 

3 minutes after stopping the stress test:

CPU temperature: 43°C

Water temperature: 40.1°C

 

The graphics card remained at 31°C for the duration of the test.

During the test, the red LED on my Asus motherboard lit up until I stopped the test.

 

https://imgur.com/a/pupgKHr

 

- According to the manual, this indicates a CPU problem, but it doesn't give more details... Probably because of temperature error ?

- Would you like me to replace the thermal paste for testing ? (it was put in place just a few months ago but I have some left over if I need it, it doesn't cost anything to try)

 

I'm still listening to you to try to solve this thermal problem and thank you again for your help!

Edited by tigerblue77
Link to comment
Share on other sites

You likely hit the top warning threshold at those coolant temperatures. However, it is clear to me the H115i is on its way out. The idle coolant temp is high and the +15C rise during a mild test that should have only generated about 130-165W is out of range too. Unfortunately, this seems to be a common end for the GTX/v2 series coolers. I never liked that series and I have spend a lot of time diagnosing issues with them. The Pro series that replaced it is massively better in every aspect.

 

Well, I guess we are back to the 360 vs 280 question for the 780T. As far as I know, you can do a 360mm up top by removing the 5.25 drive bays. This use to come up often and I will see if I can find some older threads with pictures. In the same 280mm size, you have choices between the H115i Pro, H115i Platinum, and H115i XT. The Pro is Asetek OEM, the other two are CoolIT if that matters to you. The Pro offers an ultra low quiet pump speed option for those that need a really quiet desktop environment and then two more high speed setting . The Platinum is the full tilt D-RGB pump ring plus lighting controller, plus RGB ML fans. The XT is the same cooler, but striped of the RGB fans and lighting controller to bring the price down. It is also black versus silver/grey.

 

And then there is the 280 vs 360 question for performance. You can run 140mm fans fast enough for a 280mm to keep up with a 360mm. However, it will obviously be louder in doing so. A 360mm can do it with less fan fan speed. An overclocked 9900K can pull a good 225W and that is just enough to create differences. I have been a 140mm fan lover for a long time. I preferred the sound. However, I think I have finally been moved over to the 360 size and just replaced my dual 280mm build with push-pull radiators to a 2x360 with very pretty non-radiator specific fans. That is the difference. I can run the 360s with any fan but need the highest performance options for the 280mm to match it. The only 360mm offering from Corsair right now is the H150i. This is a very good cooler and I prefer the Pro series over the Platinum. At some point, there will be a H150i XT black, but it appears to have been delayed by the current world state.

 

We can probably help you set up a manual overclock as well. I am on a Asus Z390 and the settings are very familiar.

Link to comment
Share on other sites

My rig began life with a h115i cooling my 9900K and to be honest about 3 months i came to the conclusion a 280mm RAD was just not enough to cool the 9900k when overclocking. It just isnt.

 

I sold the h115i and stepped up to a h150i Pro 360mm AIO. I also relocated the RAD to the front of my case and using the Corsair ML 120 RED LED Fans in push/pull config (2400 RPM MAX)(custom fan curve) and a fresh application of thermal grizzly paste i dropped my idle temps on my 9900K to 26c from 30-31c.

 

The voltages you mentioned are oddly high to achieve those overclocks. On Cinebench R20 i get 77c at 5.2GHz OC.

 

Every chip is different yes, but here are my OC profiles i have saved in my BIOS for my OC's - These are my settings after HOURS and DAYS of testing i want to point out. I started high and worked it down till i got a BSOD. I can run every game i play and benchmark software & video software with no BSOD's.

 

Only test i have not ran because it doesn't matter to me is AVX instructions. Its just not a real world scenario and therefor, i couldn't care less if i can run AVX for 45683 hours or not.

 

5GHz - 1.21V

5.1GHz - 1.27v

5.2GHz - 1.29V

 

5.2GHz is as far as my CPU will go being able to run anything. I can get my PC to post at 5.3GHz put as soon as i try to put any load i get a BSOD. i tried as much as 1.32V at 5.3GHz and still couldn't run Cinebench R20. I stopped there and said its good enough being able to reach 5.2GHz i'm happy with that.

 

I assume you're talking about a rig for mining cryptos?

If so, why would you put an i9-9900K in there? Did you have too much money, man?

 

Ok thanks for your feedback, I love to tinker with my tower and I can't do without my hard drive supports.

Link to comment
Share on other sites

You likely hit the top warning threshold at those coolant temperatures. However, it is clear to me the H115i is on its way out. The idle coolant temp is high and the +15C rise during a mild test that should have only generated about 130-165W is out of range too. Unfortunately, this seems to be a common end for the GTX/v2 series coolers. I never liked that series and I have spend a lot of time diagnosing issues with them. The Pro series that replaced it is massively better in every aspect.

 

Well, I guess we are back to the 360 vs 280 question for the 780T. As far as I know, you can do a 360mm up top by removing the 5.25 drive bays. This use to come up often and I will see if I can find some older threads with pictures. In the same 280mm size, you have choices between the H115i Pro, H115i Platinum, and H115i XT. The Pro is Asetek OEM, the other two are CoolIT if that matters to you. The Pro offers an ultra low quiet pump speed option for those that need a really quiet desktop environment and then two more high speed setting . The Platinum is the full tilt D-RGB pump ring plus lighting controller, plus RGB ML fans. The XT is the same cooler, but striped of the RGB fans and lighting controller to bring the price down. It is also black versus silver/grey.

 

And then there is the 280 vs 360 question for performance. You can run 140mm fans fast enough for a 280mm to keep up with a 360mm. However, it will obviously be louder in doing so. A 360mm can do it with less fan fan speed. An overclocked 9900K can pull a good 225W and that is just enough to create differences. I have been a 140mm fan lover for a long time. I preferred the sound. However, I think I have finally been moved over to the 360 size and just replaced my dual 280mm build with push-pull radiators to a 2x360 with very pretty non-radiator specific fans. That is the difference. I can run the 360s with any fan but need the highest performance options for the 280mm to match it. The only 360mm offering from Corsair right now is the H150i. This is a very good cooler and I prefer the Pro series over the Platinum. At some point, there will be a H150i XT black, but it appears to have been delayed by the current world state.

 

We can probably help you set up a manual overclock as well. I am on a Asus Z390 and the settings are very familiar.

 

Indeed, the consumption of my CPU during the test was well in the range of 130 to 165W as you said. I admit that these temperatures seem high to me because I've always been praised for the merits of watercooling but as this is my first one, they didn't particularly alarm me.

I'm sorry that you have to diagnose this kind of problem with me again, so... I didn't think there'd be any big differences between the watercoolings when I chose it. I took a Corsair for reliability and in 280mm dimensions to cool down the CPU properly. The fact that it is an old model didn't bother me because the RGB doesn't matter to me but I see now that the RGB is not the only improvement made by Corsair! :)

 

Thanks for summarizing all the differences between the 360mm and the 280mm! In fact I've read a lot of topics before posting here (but unfortunately after I bought the H115i) so I've already read them but it could be useful for others ;)

 

So if I understand well your two answers, I can't improve anything ? The best solution is to change of watercooling ? If it's the case I'll do exactly like you, prefer silence for the same performances !

 

Your answer is very technical, what's the difference between asetek and coolit pumps ? As I'm only looking for performance & silence and the noise of my current Hydro H115i pump (1st generation so has an asetek pump if I'm right) is ok for me as my drives are louder haha ! :)

 

And then I'll be able to work on the overclock ! Your help will be welcome, thanks a lot ! :D

Edited by tigerblue77
Link to comment
Share on other sites

First and immediate improvement should come with ANY new cooler. In water cooling, adding a different set of fans, going push-pull, faster pump speeds, usually result in 1-2C differences at most. It is small gain. Right now you are losing 5-6C at idle and 10-15C at load because the cooler is not able to get the heat to the radiator in the manner it is supposed to. That is a huge difference. A normal H115i anything should be +4-7C coolant over room temp at idle and have a load increase of +6-8C for the stress test. If you are still under warranty (5 years), you can use that to replace the H115i (v2). However, I would consider one of the newer models.

 

Asetek vs CoolIT. No real technical differences, but it is more of brand war thing (Coke v Pepsi, whatever). People who have one or the other die on them express hate at that one and go to the other side and then maybe back again later. The CoolIT version during your generation of H115i was the H110i (GT). It was clearly the better model. I prefer the Asetek this time around, but again it is really about how your own experience goes. Only the Asetek's have the super low 1100 rpm pump speed. I never used it because you need to manually switch back to the medium speed before doing any kind of extended load. However, I do that all the time now on my custom loops, so perhaps not really an obstacle. Supposedly the CoolIT has a slightly improved cold plate to take back a few degrees in CPU temp. I can't say if that is true without doing my own side by side test. Only other thing of note is pump tone. The Asetek tends to be lower, more "grumbly". The CoolIT higher in pitch. Most people will never hear the difference or care. If you are sensitive to some higher frequency sounds, it might be something to consider especially since you can turn the Asetek to low at the desktop. With any moderate fan speed, the pump sound will be buried.

 

The H115i Pro (280mm) and H150i Pro (360mm) are Asetek. The H115i Platinum and XT are CoolIT. I don't see any reason for you to get the Platinum unless you want to start a small RGB cluster. That model is good for those with no other Corsair RGB products, but want to add some light. The XT would be the one to get if CoolIT and no need for RGB.

Link to comment
Share on other sites

Alright I see, thanks for your advice, you've got experience, I'm a bit short of it ^^

I've just opened a ticket at Corsair's to give it a try, but it's not won... I don't know if it's still under warranty.

 

Anyway, I will consider newer models, I will go for a Corsair H150i XT since you seem to be very satisfied with it and it seems to cool better than the H115i for the same noise. Thanks for your advice, I'll come back if needed or if Corsair warranty works :)

Link to comment
Share on other sites

I assume you're talking about a rig for mining cryptos?

If so, why would you put an i9-9900K in there? Did you have too much money, man?

 

Ok thanks for your feedback, I love to tinker with my tower and I can't do without my hard drive supports.

 

I do not mine with my rig, never have, never will as mining is dead.

 

I use my rig strictly for gaming and benchmarking only. if not overclocking a 280mm RAD will do you just fine. If you're trying to overclock that chip to 5.1GHz and beyond a 280mm RAD is not enough my friend especially at the voltages you're using.

 

have you tried to reinstall your pump onto the CPU with a fresh coat of thermal paste? I recommend Grizzly Kryonaut

Link to comment
Share on other sites

Haha okayy :P yes I was probably a bit crazy with tensions but on Intel's website they say i9-9900K can accept up to 1,52v ! Not my fault, here is the proof :

 

https://www.intel.com/content/www/us/en/support/articles/000032392/processors/intel-core-processors.html

 

No I haven't tried yet, do you want me to do it ? :)

I use server's thermal paste after cleaning with alcohol wipes.

Link to comment
Share on other sites

Since the coolant is hitting 30C over ambient, we know heat is getting from CPU into the coolant stream. The problem is it can't get out at the expected rate. With a contact problem, you would get the terrible CPU temps, but the coolant would be barely off room temp. 100C CPU to 22C coolant. Amazingly, your CPU-Coolant differential at that crazy voltage is only +50C, which is about what we expect from most Intel CPUs at high overclock. I am surprised it isn't more at that level. With the coolant 20C lower, your CPU temps will be as well. A voltage reduction will take even more off. You can save the next layer of TIM for the new cooler.
Link to comment
Share on other sites

Since the coolant is hitting 30C over ambient, we know heat is getting from CPU into the coolant stream. The problem is it can't get out at the expected rate. With a contact problem, you would get the terrible CPU temps, but the coolant would be barely off room temp. 100C CPU to 22C coolant. Amazingly, your CPU-Coolant differential at that crazy voltage is only +50C, which is about what we expect from most Intel CPUs at high overclock. I am surprised it isn't more at that level. With the coolant 20C lower, your CPU temps will be as well. A voltage reduction will take even more off. You can save the next layer of TIM for the new cooler.

 

I totally agree with you but I don't understand my cooler because... The water temperature never exceeds 55°C, so it cools down... Even when I touch the hot hoses we see a big temperature difference between the start and the finish of the radiator but it is obviously not efficient enough for my i9...

 

So I'll not replace the thermal paste, not usefull as you said :)

Link to comment
Share on other sites

It does cool down, so it's not totally blocked/stopped yet. If it was, it would be essentially unusable. A typical 280-360mm radiator can only reduce coolant temp by 1-2C going through the radiator. When you have one "hot tube" and one cool one, that usually indicates a problem since 1-2C surface temp should be imperceptible to human touch.
Link to comment
Share on other sites

It does cool down, so it's not totally blocked/stopped yet. If it was, it would be essentially unusable. A typical 280-360mm radiator can only reduce coolant temp by 1-2C going through the radiator. When you have one "hot tube" and one cool one, that usually indicates a problem since 1-2C surface temp should be imperceptible to human touch.

 

Hum are you sure about your 1-2 degrees reduction going through the radiator ?

I certainly have a problem but which kind of problem could conduct to have a big difference between the two tubes :confused: (if you know, if not, doesn't matter)

A pump problem ?

 

EDIT : Well, you're absolutely right, it seems that I have a pump problem as it doesn't speeds when water becomes hot... (>40°C) I just made the test

Edited by tigerblue77
Link to comment
Share on other sites

1-2C per pass through the radiator. So if the coolant goes into the radiator at 32C, it will come out at 30-31C. It's a balancing game of heat in vs heat out. You can start doing the math of Watts/mL of water per second to calculate expected heat increase to the coolant temp, but that gets messy. At X watts you increase the coolant by 1C, less the heat released through the radiator. With the AIO you only get one coolant reading (after the heat pickup) so you only see the hot end. With custom loops you can put sensors on the in/out ports and see exactly where the reductions happen and if one radiator or one fan set does better than another.
Link to comment
Share on other sites

1-2C per pass through the radiator. So if the coolant goes into the radiator at 32C, it will come out at 30-31C. It's a balancing game of heat in vs heat out. You can start doing the math of Watts/mL of water per second to calculate expected heat increase to the coolant temp, but that gets messy. At X watts you increase the coolant by 1C, less the heat released through the radiator. With the AIO you only get one coolant reading (after the heat pickup) so you only see the hot end. With custom loops you can put sensors on the in/out ports and see exactly where the reductions happen and if one radiator or one fan set does better than another.

 

Hum ! You know so much about cooling ! That's soo cool :)

Okay I'm pretty sure them problem is the pump because on "silence mode" it stays arount 2000 RPM even if I launch a stress mode.

 

Can you confirm where do I have to plug the Corsair pump pin on my motherboard ? "CPU_FAN" or "W_PUMP" ? Actually it is on "CPU_FAN"

 

My Asus Motherboard manual say that "W_PUMP" fan pins give about 3V if I remember but some websites advice to use those pins... I'm lossed, maybe the problem is stupid in fact...

 

I read here

 

https://linustechtips.com/main/topic/845982-h115i-pump-not-working-causing-overheating-cpu/

 

that the pump could "get stuck in a position where the electric motor is not able to start."... damn it... How can I be sure it is running...

Link to comment
Share on other sites

Look at it... I did overclock the CPU to 5,0 Ghz on all cores and cache to 4,7Ghz, I kept voltage to "default"

 

On the first half of each graph shown on the below screen, I was on "silence" mode for the water pump and on the second half I set it to "extreme" mode (we see it pass from 2000 RPM to 3000 RPM)

 

https://imgur.com/a/4PD6pSp

 

And the CPU temp never go above 90°C ! And 38°C for the water :D

I stopped the test after 2/3 minutes just to make the capture but we clearly see that it was working at this moment, the question is why ^^...

Link to comment
Share on other sites

On your H115i (v2), the location of the MB lead does not matter. It can go on any header or none at all. I believe the GTX/v2 series reports 1/2 the pump speed, but the actual power comes from the SATA cable. You can't "throttle it down" by cutting voltage to it like some DC fan. We generally recommend it goes on CPU fan for two reasons. 1) Something has to be on CPU fan to boot, although that can be disabled in the BIOS. 2) CPU is always tied to CPU temperature and was designed for a CPU air cooler. It will react quickly and aggressively to any CPU temp change and has safety measures in place that override your settings. That makes it terrible for standard case fan control. Might as well put the AIO lead that does nothing on there, solve the boot problem, and save the CHA fan header for a chassis fan.

 

I am sorry. Should have suggested to bump the pump speed as a temporary measure after we determined there is some type of flow problem. Normally and in most any functional cooler, you will get 0-1C difference between the low and high pump speeds. An AIO has short total flow length and the CPU and radiator are not overly restrictive. Since there is not a lot of flow resistance, the pump speed isn't as important. Pump speed or cycle rate is part of the calculation, but the dispersion rate at the radiator and fans is the heavier factor. Faster cycle rate somewhat balances with the amount of time in the radiator channel for dispersion.

 

However, we did talk about it being possible to be too slow and getting inadequate cycle rate. That was in relation to the slower 1100 rpm pump speed on the Pro series, but what is happening to you is similar. Most likely something is blocking or restricting the flow into one section of the cooler (radiator inlet, CPU outlet, etc). This slows down the flow to the point where the coolant at the CPU block does not leave fast enough and that small section of coolant heats up more than it should. That heat is then passed back to the CPU. If you switch between 2000 and 3000 rpm at the pump and see an immediate coolant drop, something isn't right. This is further evidence you have a flow restriction and need to move toward a replacement. The sudden decrease in CPU temp should match the sudden decrease in coolant temp.

 

Looks like the Auto voltage is still putting you at 1.43v. That is still too high and what happens to everyone using Auto voltage and running stress tests these days. You likely are looking for something around 1.30v (to start). In the Advanced BIOS, change CPU voltage from Auto to Adaptive. It will create 2 drop down boxes below. Set "additional turbo voltage" to 1.30. This will make the CPU follow the Intel specified voltage curve anytime you are below 4.4 GHz (or whatever the baseline is for 9900K) and then it will apply the 1.30 when you go higher. This keeps it lower at idle and full on when working. However, you may want to postpone any further stress testing until the cooling situation is resolved.

Link to comment
Share on other sites

Okay, thanks, so I keep it on the "CPU_FAN" header.

 

Don't worry you didn't do anything wrong, my pump was already in "extreme" mode (full speed) since the beginning of this topic and during the tests, I tried to change it only for testing last night and put it back in "extreme" mode as a temporary solution.

 

In fact I don't see what would block or slow down the flow of the circuit... I wonder if the pump is not just broken (even partially) because sometimes it seems not to turn (cases where I have a big temperature difference between the two hoses) and sometimes it turns but I have the impression that its rotation is not regular (as if a fin was broken or I don't know ...)

Anyway, I'll wait for the answer from the Corsair support and I'll come back to you.

 

Your explanation seems quite logical, we can see on my previous screenshot that the temperature decreases linearly from the moment I go to 3000 rpm.

 

Indeed the automatic voltage puts me at 1.43V on average at 5.0Ghz, nevertheless I was happy to see that the CPU did not reach 100°C. Yes this time I stop all overclock until my cooling problem is solved. My goal was simply to test my pump and seeing that it was cooling better last night I overclocked a bit but that was not the main goal.

Link to comment
Share on other sites

  • 1 month later...

So I got my replacement watercooling from Corsair. As the H115i v2 is no longer manufactured I received a H115i Pro RGB, that's great!

 

I assembled it the day before yesterday and frankly... It's night and day... It's incredibly quiet even in stress test with CPU overclocked at 5.1 Ghz on all cores and cache overclocked at 4.7 Ghz ! The processor has reached a maximum of 90°C and 39.2°C for the water temperature, compared to a processor temperature of 36°C and 31.3°C in normal/daily use. It's magical and incredibly powerful !

 

Here are different screenshots taken before, during and after the stress test with the overclocked processor : https://imgur.com/a/Gx8V9p0

You can see that my vCore frequency is about 1.41V (I didn't make any adjustment, it's in "auto") but it looks good to me (c-attack can you confirm ?)

 

So finally I'm going to keep this Corsair H115i Pro RGB and not switch to an H150i, it's really excellent and meets my needs!

Thanks a lot to c-attack for his great advice here and in private messages, thanks for your time, without you I would never have been sure that my previous watercooling was faulty! But now I can see the difference, for example, the old radiator was always cold even with the processor on load while the new model heats up evenly when my processor is on load, that's perfect !

 

Good continuation !

Edited by tigerblue77
Link to comment
Share on other sites

×
×
  • Create New...