Are NZXT krakens really this bad??

slim01 · 12 Apr 2019 at 22:14

Hey guys.

Just wondering your thoughts on what should be expected of a 280mm aio on a delidded/liquid metal 4770k, and some overclocking tips.

I got 2 Kraken x61 AIOs new for £35 each (crazy deals in China) and am a little underwhelmed tbh.
Am I right in thinking a 280mm aio (kraken x61) should waaaaay outperform a 120mm (corsair h80i)??

Corsair H80i (120mm) 4.5ghz @ 1.32v = 81c max (25c ambient)
Kraken x61 (280mm) 4.5ghz @ 1.32v = 75c max (25c ambient)

Max overclock with x61...
4.6 @ 1.38v = high 80's (25 ambient)

I was expecting a LOT more than a 6 degree drop from max... And more than a 100mhz bump (maybe bad silicone lottery to blame).

Its also louder, and idle temps are about the same.
I even dremelled out the fan guard of case beneath the Rad for better airflow!

If I'd have paid the full price for them (well over £100) I'd be pretty damn annoyed.

Things to note...
-installed horizontally as an exhaust at top of case
-the case is open (both tests)

I'm just overclocking by adjusting the core clock multiplier and bumping the vcore voltage... Is there something I should know about ocing haswell? I was expecting at least 4.8 with this "beast" of a high end cooler.

Adexus · 13 Apr 2019 at 16:13

I used to have an NZXT Kraken x61 and x62, the x61 I remember performing pretty wellon my old 4770k, I can't remember what the temps were or anything but I didn't have reason to change it, in fact it's still running as my mate has that PC now, but then I upgraded to the x62 and thought it was awful temps wise, thought it might have been mounted wrong or something but nope, nothing helped it, eventually returned it for the Corsair H150i which is a great cooler.

slim01 · 13 Apr 2019 at 19:08

Adexus said:
I used to have an NZXT Kraken x61 and x62, the x61 I remember performing pretty wellon my old 4770k, I can't remember what the temps were or anything but I didn't have reason to change it, in fact it's still running as my mate has that PC now, but then I upgraded to the x62 and thought it was awful temps wise, thought it might have been mounted wrong or something but nope, nothing helped it, eventually returned it for the Corsair H150i which is a great cooler.

Ahhhh someone who's had both! Thanks.

Awful? Really? Well the x62 is supposed to be a few degrees cooler. But it was the x62 for £135 or the x61 for £40... No brainer really, just thought a 280mm compared to a 120mm would be a considerable upgrade.

Can you remember overclocking on the x61? Or has your mate oc?
I'd push mine further but don't really wanna go past 1.38v... Lots of people saying even 1.35 is bad for the chip.

I might try and re-mount it then, thought I did a good job. Might even chuck some liquid metal on there.

Danny75 · 13 Apr 2019 at 21:48

1.4v is the official max safe voltage for Haswell according to Intel, far as I know (haven't read that from Intel themselves only from others saying so).

So above 1.3v is when you have to watch out for AVX instructions spiking it to around 1.4v. Unless running with Fixed/non-Adaptive which prevents that.

My own 4770K needs 1.35-1.375v for 4.5GHz stability. Most the time I just run with 1.224v at 4.2GHz as I don't really see a difference other than benchmark numbers and the fps counter. But the power draw and heat are considerably less.

slim01 · 14 Apr 2019 at 01:51

Yeah maybe a silicon lottery thing then. But there's no way a 4770k could be ran at 1.4v without being delidded... Mine used to hit 100oc within seconds of prime 95 starting at auto volts (and a h80i) @ 4.2ghz.

What cooling did you use Danny? You think a 280mm should perform better?
Not sure I can be arsed doing another day of tests and rebuilds if it's normal Temps, even after a delid.

Adexus · 14 Apr 2019 at 01:55

slim01 said:
Ahhhh someone who's had both! Thanks.

Awful? Really? Well the x62 is supposed to be a few degrees cooler. But it was the x62 for £135 or the x61 for £40... No brainer really, just thought a 280mm compared to a 120mm would be a considerable upgrade.

Can you remember overclocking on the x61? Or has your mate oc?
I'd push mine further but don't really wanna go past 1.38v... Lots of people saying even 1.35 is bad for the chip.

I might try and re-mount it then, thought I did a good job. Might even chuck some liquid metal on there.

Yeah I heard nothing but good things about the x62 but mine didn't perform well at all, and I couldn't get my chip to clock that highly, I think it was at 4.5 or 4.6, somewhere round there.

He's got the same OC on as when I did it as I forgot to default it but still running fine as is on the x61.

So I'm not sure if I just got a bad x62 or not :\

Rroff · 14 Apr 2019 at 03:01

4770K tended to run warm even delidded in my experience also AIOs I've always found a bit hit and miss for noise with 2 identical models side by aide potentially very different noise wise.

I got good temps on my 4820K at stock didn't go over 48C max with an Antec 1250 and 4.6GHz stayed around 60C max but couldn't recommend it as one of the fans packed in, which was quite a common fault, and the fans are integral to the pumps.

One thing that can easily add 10C to the max load temps of those CPUs is VTT/System Agent voltage it also potentially killed them quicker than VCore could - I know a guy that used a whole load overclocked in render farms and in some cases VTT over 1.3V IIRC was killing some of them in weeks or months while most were taking crazy VCore indefinitely if cooled sufficiently.

EDIT: Ambient temperatures in my operating environment are a good bit lower than 25C though - usually more like 17C.

slim01 · 14 Apr 2019 at 04:37

Yeah same, I heard amazing things about all the krakens tbh.

Yeah it's pretty toasty in China. And am I right in thinking a rise of say... 10c in ambient Temps should actually cause a lot more than a rise of 10c in max/average TEMPS?

Regardless, I'm just so confused as to how a 280mm Rad surface area can only perform slightly better than 120mm... It's more than double the surface area and liquid.
If a bigger temp decrease should not be expected with bigger rads, I'm glad I didn't get the £150 x72 (I think, the one with a360mm Rad).

Noise wise, I'm not expecting anything to be quiet at 2000rpm, but *normal* operation should be quieter right? Isn't that the point of bigger fans/rads?

I mean... A 14cm fan SHOULD at the same rpm as a 12cm fan have a higher cfm, and bigger Rad, therefore reduce Temps more, therfore, not need to spin as fast, therfore... be quieter.

I've read the x61 fans are great (100s of 5* reviews on amerzon) but I can clearly hear the rattling of the bearings on one fan (when over like 70% rpm) and high pitched whine from motor.
This stops if I cover the exit of it, or press down slightly on the Rad it's behind.

That might be to do with the fact I dremeled out the fan guard, and potentially an uneven amount of metal/obstruction is causing it to rattle.
I will change the fans around, and see if the noise is mirrored. Maybe use a fan from the other x61 I have.

slim01 · 14 Apr 2019 at 05:06

Rroff said:
4770K tended to run warm even delidded in my experience also AIOs I've always found a bit hit and miss for noise with 2 identical models side by aide potentially very different noise wise.

I got good temps on my 4820K at stock didn't go over 48C max with an Antec 1250 and 4.6GHz stayed around 60C max but couldn't recommend it as one of the fans packed in, which was quite a common fault, and the fans are integral to the pumps.

One thing that can easily add 10C to the max load temps of those CPUs is VTT/System Agent voltage it also potentially killed them quicker than VCore could - I know a guy that used a whole load overclocked in render farms and in some cases VTT over 1.3V IIRC was killing some of them in weeks or months while most were taking crazy VCore indefinitely if cooled sufficiently.

EDIT: Ambient temperatures in my operating environment are a good bit lower than 25C though - usually more like 17C.

Wow so I really need to check this VVT isn't over 1.3v then... Thanks for info.
Yesterday, I settled on 1.38v, changed voltage to adaptive mode then "additional turbo mode vcore" to 1.38v (I think it was called) but then noticed VERY high temps...
Then I noticed each core was NOT drawing the 1.38v I set, it was more like 1.45v.
*then pc powered off. I thought I'd killed her *
I then tested and found "additional turbo mode vcore" @ 1.33v makes the cores take about 1.38v.

I THINK there was a number in hw manager and cpu-z that WAS what I set the voltage to (1.33v)... Maybe this is the VVT?

So Danny and Rroff, how do you go about overclocking?

-I'm just adjusting the multiplier, and bumping cpu with manual voltage.
-Then stress testing with prime95 (fft first for 10 mins then blend). If it doesn't bsod, core 2 usually fails on the blend.

Everything else I've left the same. Should I be overclocking differently with haswell? Like play with cache speed/volts? Use blk + memory instead?

slim01 · 14 Apr 2019 at 05:10

Should i be using offset mode instead of adaptive?
I don't really understand offset mode... Well... Apparently not adaptive mode either. I thought by setting adaptive maximum additional turbo mode to 1.38v it would take a maximum of... Well... 1.38v

bimbleuk · 14 Apr 2019 at 06:14

The thermal limit may be the interface at the CPU so increasing the radiator area/volume will have a smaller affect. Yes I've gone from a single 120mm to 280mm Kraken and the change wasn't that impressive though the increased water volume means it takes longer to reach thermal stability and then stays very constant.

From idle when you start a stress test what does the CPU temp jump to as that's probably only going to creep up after that as the water temp increases?

Danny75 · 14 Apr 2019 at 08:48

slim01 said:
Everything else I've left the same. Should I be overclocking differently with haswell? Like play with cache speed/volts? Use blk + memory instead?

I wouldn't use BCLK because on Haswell that affects PCIe and drives which is not a good thing.

As for Cache leave that to the end, find the CPU OC you want, then start raising Cache and increasing Cache voltage as needed. Same for Memory overclocking. With Haswell Cache the advice (although this may differ) is to stick to around 1.2-1.25v. That translates to 4.2GHz on mine.

Your CPU Vcore might be overshooting to 1.45v because of using Adaptive instead of Fixed/Override, and programs with AVX instructions (like Prime95 with AVX). On Fixed/Override, AVX won't make Vcore overshoot.

Personally I've found tweaking the Vcore to make way more difference than anything else. I only had to mess with other things as a challenge to get the 4.5GHz OC stable. Because at that point I didn't want to raise Vcore further. To this end I also tweaked the following (marked in green):

Upped DRAM voltage from 1.62v to 1.65v.

CPU Phase Control from Auto to Normal (which was the best option, the others were more focused on power-saving)
CPU Vdroop Offset Control - this is called LoadLineCalibration/LLC on other boards. Some people warn that too high LLC is worse than too high Vcore when it comes to reducing component lifespan. 25% did the trick there so left it at that.
CPU Switching Frequency - if motherboard VRMs are good enough, around 400 KHz is supposed to be okay for a 24/7 overclock.
CPU Power Duty Control - by default it was on Thermal Balance. In theory, the change to Current Balance would provide more current stability/less ripple. This is not something I tested much for effect but thought why not.

So if messing with any of that helps you, great. But mileage varies from chip to chip and board to board and memory. Your scenario is pretty similar though - you're just under 1.4v and presumably don't wish to exceed that. So could look into this other stuff. And also power and current limiting if needed, to act sort of similar to an AVX Offset on newer motherboards. See this thread: https://forums.overclockers.co.uk/threads/avx-power-limit-for-stability.18846605/

Adexus · 14 Apr 2019 at 16:06

slim01 said:
I've read the x61 fans are great (100s of 5* reviews on amerzon) but I can clearly hear the rattling of the bearings on one fan (when over like 70% rpm) and high pitched whine from motor.
This stops if I cover the exit of it, or press down slightly on the Rad it's behind.

That might be to do with the fact I dremeled out the fan guard, and potentially an uneven amount of metal/obstruction is causing it to rattle.
I will change the fans around, and see if the noise is mirrored. Maybe use a fan from the other x61 I have.

Never had any luck with the NZXT fans, they always break sooner or later or the bearings go on them.