Fractal Design S24 Kelvin - FAULTY?

MicroAlex

New member
Hello,

Decided to give the latest Kelvin series cooling solutions a try... but sadly the impressions are not to expectation. The looks are there, but performance issues make it almost comical for my first personal H2O setup. I did not have too much exposure to AIO's yet, however I do not remember anything similar on the systems observed or worked on. Please read below.
And before I start I would like to say THANKS to OC3D team for doing a great job on reviews and support.
Also, I have seen the review on S24 and S36.

Setup:
Intel i7-4790k (stock)
Asus Impact VII
Corsair Dominator 16GB Kit 2400mhz
Samsung EVO 840 250GB
WD Black 2.5 inch 750GB
Corsair 380T
Corsair AX860i
S24 Kelvin (stock fans)

Case fan layout:
Front - 2x 120mm Corsair SP fans
Back - 1x 120mm Corsair stock fan
Side - 2x 120mm Fractal fans + S24 Kelvin

Testing:
The system was tested under OCCT Linpack mode with 90% used memory and AVX enabled (also with AVX disabled) stressing all logical cores.
With AVX the system would get over 85c on the cores sometimes in 2 minutes and stop the test. Fans were connected directly in to Asus MOBO, fans and pump calibrated and set to a NORMAL profile. Under more aggressive fan profiles the temperatures do not change much but does help to keep it at high 70s. However since the fan curve on all profiles at the end ramps up to the same speed - the noise coming both from the fans and the pump is very loud.
AVX disabled results in temperatures of 60c-70c with everything being almost whisper quiet compared to the previous results observed with AVX enabled.
First testing sessions were done on stock Fractal paste, secondary attempts were made on MX4 to double-check the old spread and exclude a possible problem.

Problem:
After mounting the new cooling solution and powering the system I did notice that there is a significant amount of noise coming from the air bubbles in the loop (not sure if this is acceptable or normal with AIO's as I have heard Swiftek had similar problems with their first revision).
Attempts to fix the issue were made by leaning the system side to side in order for the bubbles to settle somewhere away from the pump.
The pump is only silent at its minimal speed, going up another 200-300rpm makes the pump buzz quite a significant amount.
Temperatures are obviously high under a decent stress while running stock clock speeds and voltages.
Attempts to increase airflow were made as well by ramping up the front fans and reversing the back fan to act as an intake. In result the temps did drop a bit, but nothing that would live up to the "Silent" operation under load.

Please advise plan of action or a possible resolution.

Thanks you for your time guys.
 
Last edited:
Hello,

Setup:
Intel i7-4790k (stock)

Testing:
The system was tested under OCCT Linpack mode with 90% used memory and AVX enabled (also with AVX disabled) stressing all logical cores.

With AVX the system would get over 85c on the cores sometimes in 2 minutes and stop the test.

Think this sums it up -"Stock"-. I bet if you changed this to a manual OC it wouldn't get hot. Takes away nearly the entire problem. Why? Well i bet the4790k is jumping up to 1.3-4+ on the volts. That'll kick an intel CPUs ass. Overclock it correctly dude. Tend to get more(ghz) for less(volts) if you do so.

Edit: Might have come off as a prick. Sorry about that, wasn't the intention.
 
Last edited:
Above comment is spot on in regards to stock

As for airbubbles it will calm down, you may need to follow a 'bleed' process like you would with a normal WC loop. Easiest way is leave the CPU block attached and hold the rad up vertical, hoses at the bottom and the rest of the rad above it.
 
Think this sums it up -"Stock"-. I bet if you changed this to a manual OC it wouldn't get hot. Takes away nearly the entire problem. Why? Well i bet the4790k is jumping up to 1.3-4+ on the volts. That'll kick an intel CPUs ass. Overclock it correctly dude. Tend to get more(ghz) for less(volts) if you do so.

Edit: Might have come off as a prick. Sorry about that, wasn't the intention.

Confirmed!

I recently hooked up my 4790k and it is still set to auto. Ran a prime95 and the volts ramped up with 1.38V!!!!

Just waiting for a few other key components to come in before I start doing my manual adjustments. I suggest you follow the overclocking guide on this forum to achieve the best result for yourself.
 
If it's stock clocked with auto voltage the board shouldn't go 'too' mental, check the cpu core voltage with CPU-Z when under full load. I would imagine it's more likely a combination of a bad mount, poor pressure or still having some air in the loop.

Ultimately what are your fan speeds, pump speed (you really shouldn't have this varying! set it at a static speed somewhere between 60% and 100% DC Voltage controlled) and what's the ambient temperature like.

JR
 
Think this sums it up -"Stock"-. I bet if you changed this to a manual OC it wouldn't get hot. Takes away nearly the entire problem. Why? Well i bet the4790k is jumping up to 1.3-4+ on the volts. That'll kick an intel CPUs ass. Overclock it correctly dude. Tend to get more(ghz) for less(volts) if you do so.

Edit: Might have come off as a prick. Sorry about that, wasn't the intention.


No worries mate, as long as I am getting useful information :)
Now, I am getting 1.27V on stock/auto settings under AVX stress. This is at least what OCCT, CPU-Z from Asus and Aida64 version of it tells me.
Forgot to mention - I also did run a test with Core Voltage set to 1.25v in BIOS, results were a bit better.


Above comment is spot on in regards to stock

As for airbubbles it will calm down, you may need to follow a 'bleed' process like you would with a normal WC loop. Easiest way is leave the CPU block attached and hold the rad up vertical, hoses at the bottom and the rest of the rad above it.

Hey Tom, long time no see. Many thanks for your input, I know how busy you are.
Yeah, bubbles sorted! That is exactly what I have done to get them damn bubbles out of the way, I guess its a logical solution. :) Now it sounds alright, although the pump still whines once it goes from 1000rpm (as shown in the Asus fan calibration software) to something like 1300rpm... definitely noticeable. Not sure how it compares to your favorite H100 and H100i.
It is one of the annoying high frequency sounds, so it does get on your nerves especially if the pump is pushed even higher.


Confirmed!

I recently hooked up my 4790k and it is still set to auto. Ran a prime95 and the volts ramped up with 1.38V!!!!

Just waiting for a few other key components to come in before I start doing my manual adjustments. I suggest you follow the overclocking guide on this forum to achieve the best result for yourself.

WOW! I have not seen my go that high O.o
My first reaction to the problem actually was to double checking Tom's underclocking guide. But my CPU did not really like it that. I will give it a shot tonight again.

If it's stock clocked with auto voltage the board shouldn't go 'too' mental, check the cpu core voltage with CPU-Z when under full load. I would imagine it's more likely a combination of a bad mount, poor pressure or still having some air in the loop.

Ultimately what are your fan speeds, pump speed (you really shouldn't have this varying! set it at a static speed somewhere between 60% and 100% DC Voltage controlled) and what's the ambient temperature like.

JR

My thoughts exactly. Cant imagine the manufacturers to cock it up that much for the whole platform. A little bit high on the volts - possible, but that mental?!
Well the air is definitely still there. But the mount was re-seated just to exclude that option. Judging by the spread it was good the first time (I am judging this as I have probably took apart 100 laptops by now, and yes I know that they are different). The screws + mount are tight as a nuns ... Did not go mental, just screwdriver/finger tight.
I could tear it apart again and take a picture and some point.

And YEP, the pump was set up for DC voltage in BIOS, as after a reset it jumped back to PWM.

I will try getting the voltage down for now by underclocking it again, maybe this time it will go better.
One thing I have noticed, even though I understand the fact that my fans and pump were set up to ramp up with the load, the temperatures were jumping around a lot. Lets say I would reach a peak of 80-85c at the Max Temp section of OCCT - while the RPM is reliability stable, the temps jump around from 80c to high 60s and back to mid 80s

Stress

I am also adding a screenshot of a quick and dirt stress test Asus fan profile for everything was set to FULL SPEED. In other words they were running at their full RPM.


Idle temps with fan speed set to Standard (was getting def by now). You can judge the room temps from this.




I did build this system to eventually include a GPU after some planning and modding, but mainly I wanted a silent and compact-ish/portable rig to work by. I have two Alienware M17X R4 sitting next to me, but I am tired of the noise especially since my work revolves around them even at the main daytime job. Hence the build...

Hope you can shine some more light guys! Many thanks!
 
Last edited:
Just run the CPU at stock clocks and find its lowest volt it'll run at stable.

Also about the temps in OCCT. That is normal for them to drop and rise. What the test is doing is stressing it and then does a lighter load then stresses again. This simulates a more real life and avg usage of the computer.

http://forum.overclock3d.net/showthread.php?t=45604
That should give some insight on OC'ing. Not a devils canyon chip but basically same rules apply

http://overclocking.guide/haswell-and-devils-canyon-socket-1150/
This has a lot of terms you should get to know and learn. It'll help you understand how it all works better.
 
That pump is running way to slow dude. Something is amok there.


Don't copy my settings or expect my temps because i'm running a different chip and my Kelvin has everything changed on it but this is as mine looks.

The pump is specified to run from 7-12v DC and as such I set my fan curve up (CPU fan you can see) to be flat at 60% (12x60%=~7) and the pump runs at 1700 RPM. Now I know the radiator is different so the restriction will vary but it's not THAT different. I've found plugging fans into the CPU_OPT header screws with the CPU_FAN control and as such have stuck all of my fans onto CHA_FAN headers.

kelvin.jpg


JR
 
Just run the CPU at stock clocks and find its lowest volt it'll run at stable.

Also about the temps in OCCT. That is normal for them to drop and rise. What the test is doing is stressing it and then does a lighter load then stresses again. This simulates a more real life and avg usage of the computer.

http://forum.overclock3d.net/showthread.php?t=45604
That should give some insight on OC'ing. Not a devils canyon chip but basically same rules apply

http://overclocking.guide/haswell-and-devils-canyon-socket-1150/
This has a lot of terms you should get to know and learn. It'll help you understand how it all works better.

Oh thanks mate, something useful to read before bed :) Lets see how much I actually dont know about the dark side!


That pump is running way to slow dude. Something is amok there.


Don't copy my settings or expect my temps because i'm running a different chip and my Kelvin has everything changed on it but this is as mine looks.

The pump is specified to run from 7-12v DC and as such I set my fan curve up (CPU fan you can see) to be flat at 60% (12x60%=~7) and the pump runs at 1700 RPM. Now I know the radiator is different so the restriction will vary but it's not THAT different. I've found plugging fans into the CPU_OPT header screws with the CPU_FAN control and as such have stuck all of my fans onto CHA_FAN headers.

kelvin.jpg


JR

Pump is as Asus calibration has set it (under Standard mode). Since temps are low at idle to low usage, the RPMs as you can see are very low. On the bright side if it is under 1200RPM I cannot hear the pump.
The Pump is obviously in my situation connected via CPU fan header.
The fans on Kelvin and the back fan are connected via special mini fan board from Asus that sits on the MOBO and has two fan headers sticking out.
Front intake fans are set to run through Corsair 380T fan controller ATM.



The test below is once again a quick and dirty run with all fans in the case maxed! Thank god for in-ear earphones from Sennheiser :(
CPU Core Voltage set to 1.2v, dialed in whatever I thought should work. Please note that all other settings are stock/auto, even the ram runs at default speeds, not 2400mhz.

PS - it was a 3rd 5-10minute run.



Mm... reminds me of the days when i was so bored, I took apart an Acer Ferrari One 200 and replaced the CPU with a proper 2ghz AMD part. Had to undervolt and step down the cpu multiplier by one to get that little thing cool enough. What a beast it was after that, was playing Warhammer SoulStorm on it :D Not many knew the fact that this netbook was a unicorn with a proper socket for a CPU, not soldered.

Anyway, back to the present :)
Question, probably silly one - is there a point when the higher pump RPM is worse or no better than a slower RPM? What I mean by it is that heat dissipation has a proportion at which it grows with flow. But if you go past a specific point, would it actually reduce the efficiency as the mass of the source doesn't have enough time to transfer the energy from one body to another (in this case to a radiator)?
 
Last edited:
Update

Hey guys,

Once again thanks for your help with the issue.

Fractal in their email have recommended to run the tump at a minimum of 1800rpm stating that it might overheat and break...
Also they have stated that if there is a lot of bubbles I will have to get in and refill it.
Dunno what I was expecting to be honest as they have skipped over other issues that I have mentioned in the email.


Now, UNDERCLOCKING! I went down as far as 1.175v on the core, however there are some issues.
The test AVX test runs with no issues to be seen. On the other hand I do get a BSOD if I run a CPU: OCCT test on all cores and a Large Data Set.
This issue continues across to higher voltages as well... 1.175v, 1.185v, 1.200v. I am yet to test something closer to the default AUTO voltage on the CPU.

The temperatures are a bit better tho :)

Please once again note that everything else was on AUTO, maybe I have missed something? RAM was running at 1333mhz for the time being.

Hope you can help out.
 
Ram wont auto run at 2400 you have to set it up / use XMP fella.... I say this in every single one of these threads......
 
If you get a BSOD on an OC or underclock then it's unstable and volts should be lowered/raise till stabiltity happens.

In the Bios what have you left on auto?

For reference, M&P on this forum has done a guide. While it is Ivy Bridge, just go down to the terminology part and read up on a few of them to see if they aid you in temps.
http://forum.overclock3d.net/showthread.php?t=45604

Ram wont auto run at 2400 you have to set it up / use XMP fella.... I say this in every single one of these threads......

He knows Tom. He said he left it at 1333 for the time being.
 
Ram wont auto run at 2400 you have to set it up / use XMP fella.... I say this in every single one of these threads......

Hey TOM,

I dont want them to run at 2400mhz, I am excluding this option for testing purposes.
I was running them at the specified speeds via XMP before we started troubleshooting on this forum. :)
 
Back
Top