push push vs pull pull on Noctua NH-D15 + fan curve?

[majsta]

Hi guys, 

   

   I just bough Noctua NH-D15 too cool my i7 4770K and i mounted it in my case in pull pull configuration on both of the towers of the cooler, pulling the air to back of the case where i have one case fan as exhaust. Now the reason i did this is that the fans in push push configuration interfere with the high profile RAM (Kingston Beast in all 4 slots) and I have to put the first fan a bit up on the first tower above the RAM and it looks kinda ugly, as it is a windowed case (Define R4 from Fractal Design). 

So the first question is do I get anything performance wise if I run the fans in push other than pull? I kinda have the feeling i don't because the amount of air going trough both of the towers is the same in either config. Or am I in the wrong here. The temps are 40C on idle and around 65C @3900MHz when running stability test on AIDA 64. Bare in mind my room temperature is a bit high, around 30C. I didn't overclock the CPU, but I am planing to this weekend and I want to optimize the cooling as much as possible.  

 

The second question is for the ASUS ROG users. Can anyone help me with the fan curves in the AI 3 suite. I have two intake fans and one on the back as exhaust, all Fractal Design S2 fans, and now i have these two AF15 on the CPU cooler. So I would like to optimize the airflow vs the noise. The normal use of the CPU would be gaming and video rendering, so that is the biggest load on the CPU/GPU. What I am trying to ask is for some guidelines on how to setup the fan curves for best performance/low noise. 

 

Thank you all in the advance. 

[blu4]

No temp difference, 1-2C max. Do pull, it's easier to clean and you can hide the fans.

[SolarNova]

Push is generally considered better, BUT on a heatsink like the D15 ur talking 1c ...maybe.

 

Pull is easier to keep clean on RAD and Heatsink as the dust builds up on the rad or heatsink side without the fan on it, so its easy to get to. In push config the dust builds up on the fan and behind the fan on the rad and heatsink, requiring the removal of said fan to clean properly.

 

So what ever u prefer is fine.

[paddy-stone]

I have a d-15 on my i5 4670k in push and get approx 2C over ambient temp at lowest fan speeds and idle CPU, this also includes lowest case fan speeds also... on load (aida 64 stress) I get approx 20c above ambient @3.8Ghz no O/C.. and when I O/C to 4.4Ghz I get around 40C above ambient which still gives me a bit of room temperature wise as I prefer stability at all times of the year over a slighly better overclock.

What is more important IMO than puch/pull config is a good case with an excellent intake for fresh cooler air to enter the case and a well positioned exhaust fan setup.

[majsta]

First of all thank you all fro replaying so quickly.
@blu4 and @SolarNova, thank you for confirming that the temp difference is not that big of a deal for different configs, IMO 1C or 2C improvement is not that important, so I think I will stay with the pull pull.
@paddy-stone, I couldn't agree more, the stability is far more important than the overclock, 10 mins more or less on a render is not that important, and in gaming the CPU will never get that hot, and does extra 100MHz will not make a difference when we talk about 4GHz speeds. 
So I think I will go for a slight overclock for now, just to break the 4GHz barrier, and keep the temps as low as possible, maybe change the lowest fan speed on CPU cooler.

 

Still I am open for some opinions for the curve on the AI 3 suite. Trying to get as low noise as possible while keeping the temps in normal ranges. 

[paddy-stone]

I haven't used the Ai suite at all, I don't have an ASUS board. I have an MSI gaming 5 z97, not impressed with the fan control options to be honest, and might break my fan controller out of storage again soon... basically  with this board if you set on auto and select your fan curve for scale up of fan speed, you set the fan speed you would like for under 40c, for instance mine is like 12.5% I think, then when it hits 40C I scale up pretty quickly as the 4670k doesn't like too hot temps at all. This board is however very easy to overclock, started at 4Ghz @40x 100 and personally I used the MSI command centre to see where my limits were first as that saved a bit of time, just upping the multiplier x1 each time and seeing if it's stable, then again.. until it bluescreens, then you know your basic parameters without having to adjust core voltages, then take it back down to the previous multiplier setting eg 44x 100 and give it a proper stress test to check stability. Then if all is well, you can either stick with that stable overclock, or carry on upping voltage and multiplier and even the base clock to get a higher OC as long as you're within thermal limits. I know some will probably disagree with my method, but it works for me and I generally stick with just upping the multiplier, as you have a similar CPU gen, I would say you are probably good to go around 4.4Ghz same as me and still have a little head room for different seasons being warmer etc.

 

Oh, forgot to say that when I found my theoretical limit OC wise, I then went into the BIOS and manually entered the settings for the OC and then tested again just to be sure... as you can tell I am not a big overclocker expert by any means, LOL

[SolarNova]

I use AI suit, but because im on a custom water loop I don't really have a curve, just a flat line at the speed I want with an emergency bump to max speed if temps ever go crazy for what ever reason. Its more an audible alert (sound of fans ramping up to max) than use for cooling since any drastic change in temps like that will likely indicated a cooling failure rather than a need for more airflow.

 

I used to use a NH-D14, and I had a similar setup then as well, set the fan speed to the maximum speed before noise became an issue, then let it stay at that speed until temps hit like 75c or 80c, which it never did unless I was benching. Thus I ended up with a constant fan speed , constant noise level (more or less silent) no matter the 'normal' usage CPU load.

 

that is what I recommend.

[majsta]

Once again thank you for the quick replay.

I did the overclock yesterday and a day before and I can say that i had a couple of surprises, unfortunately bad ones.
So I vent for the manual overclock, 4.5GHz @ 1.2V which is some sort of a standard as I could find on the internet for my i7 4770K, and it booted up. I ran the AIDA64 stability test (maybe like 15 - 20 mins just to see the temps) and the temperatures seamed fine (82C max and around 70-72C average), except on the one core, which was maybe around 4-5C higher in average. I ran the test on intel OC tool as well and it seamed fine. So I was happy with it, and went for the 4.6GHz 1.2v, which didn't even boot up (the voltage was to low), so I went for the 4.5GHz again and upped the cache speed to 4.5GHz @ 1.2V, didn't boot. So I up the voltage just a little bit more 1.25V no boot. So I downscale the speed 4.4Ghz @ 1.2V and up the ram speed with XMP, it boots up... Great I was happy, ran the stability (15 - 20 min), temps were fine, everything looked fine. I turned on the power saving mode and set the voltage to adaptive with 1.25V and done.
So I went for some rendering after that and my CPU crashed with same error, not enough voltage . So disappointed.
So I reverted everything to stock, and went for the AI suite overclock. The ASUS AI suite is not working on windows 10 properly, but I managed to install it somehow and went for the OC. It go to 4.4GHz @ 1.264V adaptive mode, and with XMP on my RAM 2400MHz, but the cache speed is only 3.9GHz. So disappointed, it looks like i got not so great CPU a bad luck of a draw. On the plus side the fans can not be even heard, so the curve with AI suite is great even under load. 

I will probably try again for the manual OC, 4.5GHz @ 1.275V, with XMP and cache at 4.5GHz, and if it boots up i will try the stability test for the whole night long, maybe try and mount the cooler once again, as the tamp on that one core wasn't that great (but this is probably due to the internal paste on the chip, bad CPU, and not the Noctua). 

Once again thank you for the advice.