• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

SOLVED Making Changes to my Rig - help sought to optimise cooling

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

Razour

Registered
Joined
Apr 13, 2016
Hello - this is my first post so before I get to the meat of things here's a quick summary of how I got here...

I've always been a bit of a gamer, starting out with dice and pencil games like D&D, over time this evolved into on-line gaming. In parallel I've also pursued a geeky career in Engineering in the embedded SW space.
Up until last year I'd mainly played games on a laptop but I became increasingly frustrated with the compromised performance of laptops so I decided to put some time and money into building a gaming rig.
The initial build was air cooled and even then largely beat my expectations in terms of what I could achieve.
BUT, what I quickly recognised was that I enjoyed the building and the tinkering with the rig more than I actually enjoyed playing games on it - So, it's largely been a constantly evolving beast since then!
Earlier this year I kind of ran out of things to do with the air-cooled setup so I decided to jump in and convert it to water cooled - which was a level of fun beyond the initial build and now I'm fully hooked.

So in summary I've been interested in water cooling for 2-3 months; I've got my rig to a point which is both (to me at least) aesthetically pleasing and somewhat performant; but based on what I've read I think my current setup isn't properly dimensioned for the heat being generated... so I'm here looking for a bit of advise from you seasoned experts and enthusiasts!

Let me start with some statements/assumptions to sketch out why I think the system isn't dimensioned properly and why I think the problem is with the dimensioning and not something like a botched application of heat paste:
1/ I believe that the key metric in understanding how effective the system is is dT at load = the delta between ambient temp and the temp of the system under load.
-- What isn't 100% clear to me is exactly which temperatures you measure to determine this - my *assumption* is that it should be the coolant in the system and that the ambient is the temperature with nothing running, i.e. the coolant is effectively at a temperature very similar to the air temp in the room.
-- Based on that my ambient is ~20degC and maxes out at ~40degC after a 2-3 hours of gaming on ultra graphics settings, making the dT 20degC.
2/ My *assumption* is that in order to raise the temperature in this way the transfer of heat between the different cores and the coolant must be reasonably efficient so my conclusion is that the issue is with how I dump that heat out of the coolant, hence why I believe that the system isn't properly dimensioned.

Some other observations:
3/ It seems to me that the biggest contributing factor to generating the heat is the GPU's (2 x GTX970's in parallel), at least when I stress the CPU (i5 6600K) with Prime95, even if I OC it to 4.7GHz, the dT is in the range of 5-7degC (I don't recall exactly), but when I run benchmarks that stress the GPU's the dT can run to 10-15degC (depending on the benchmark).
4/ Even when the coolant temp is at ~40degC I don't feel that the air being exhausted by the system is particularly warm which reinforces my feeling that the system isn't able to dump the heat effectively.
5/ My quick and dirty calculation is that the whole system must be generating somewhere in the region of 650TDP (~100 for the CPU + 2 x 250 for the GPU's + something for the pump + some mystery fudge factor)

Here are a couple of pictures so you get a feel for what it looks like today:
Detailed.JPG
Side shot.JPG

And here is a diagram to show the current configuration:
Gaming Rig Current Config.JPG
Please note:
- The use of the "industrial spec" Noctua fans internally is more for the asthetics - I much prefer the black over the default colour, especially with the addition of the red corner rubbers.
- The upper most of the two fans on the inner partition wall is almost purely there again for asthetics - there's no space behind it to mount a radiator but neither did I want to have an open gap exposing some of the ugliness behind. :)

Given the limited space that I have to play with the conclusions that I've drawn as to how I improve the capacity of the system to dump heat is as follows:
A. Swap out the Alphacool 8fpi rad with a Black Ice 16fpi fan (same-ish thickness).
B. Swap out the "standard" Noctua's on the top with "industrial" Noctua's for the extra airflow and extra static pressure.
C. Swap out the ST30 140 rad for an UT60 140 rad and add an extra fan for push-pull on that rad.

(worth noting that I'm prepared for the system being a little noiser - but at the same time I'm also assuming that the noise won't be too bad because the fans still push/pull a lot through even at low revs)

Hopefully all that makes sense - please tell me if it doesn't - but this is where I'm getting hung up... I know where I want to locate my fans but I can't make my mind up how to orientate them...the apparently age old conundrum of positive-v-negative pressure woven together with the conundrum of using rad-fans as intake-v-exhaust...

In short I've come up with two options as illustrated below:
Gaming Rig Option A.JPG
Pro: better balanced airflow (2 intakes and 3 exhausts)
Con: air through the top fans is already heated

Gaming Rig Option B.JPG
Pro: air through both rads is coming from outside the case so should be cooler
Con: airflow is badly balanced (4 intakes and 1 exhaust)

If you've got this far, thanks for sticking with me...
... now, what do you reckon:
- Do my assumptions make sense?
- Does my conclusion about the cause of the problem make sense?
- Will the changes to the fans and rads improve the situation in an appreciable way e.g. any change to get dT down to ~10degC rather than ~20degC? (assuming I'm even calculating this right!)
- What are your thought on Option A vs Option B?
- Other thoughts?

Thanks Again.

Raz.
 
Last edited:
Welcome!

But, none of your attachments work...

Please provide complete system specifications so we know what hardware you are working with.
EDIT: I see a 6600K amnd 2 970's now... :)
 
Very nice post!

Down and dirty, we recommend 120x1 radiator for 100 watts, we'll say 120 watts for the 140x1 radiators. Due to great 140mm fans are hard to come by, not as good as 120mm fans. So I'll fudge 120 watts. So you are cooling 650 watts under stress testing and only able to get a good DT when normal use when you are closer to the 360 watts you can keep cool.

A quick note from something I wrote:

DT is the foundation of your water cooling loop. The better your DT, the cooler your chips are. In water cooling, DT is simply the difference between the ambient air temperature and the water temperature on the outgoing side of the radiator. Room temperature vs. water temperature: that’s it. You can’t remove all the heat – no system is 100% efficient, nor can you go below ambient room temperature.


Source:
http://www.overclockers.com/guide-deltat-water-cooling/

DT means many things, but we use the room air air temp vs the water temp leaving the radiator. Ultimately all that matters is your temps are good enuff for you and the PC is quiet enough.

And ambient temp is the room air temp. It matters to some. Summer/winter temps and rooms warm up when they get an extra 650 watts of energy pumped into them.

If your temps are good at normal gaming etc loads, and only an issue when bench marking, then adding more cooling is up to you. Getting another bigger case is a pain, but sometimes that's part of the hobby!

 
He's cooling about 400w at worse/stock.

970s are 150w cards stock. And 6600k is 91W.
 
Ohh so he's good!

OP:

Please post full temps. Idle CPU/GPU. Room temps.

Then load up Heaven and Prime 95 at the same time, let it run 30 min, post room temps and CPU/GPU temps.

If temps look okay no worries.
 
Thanks for the engagement all - I'll provide the requested numbers later today when I can get back in front of my rig.

In the interim, I'm still interested to hear opinions on the two options that I'm considering for fan orientation - which should be the "least worse" ;)
Option A: has better balance in the airflow, but at the cost of using warmed air to cool the 420mm rad
Option B: has the advantage of drawing cooler air across both rads but has limited exhaust (albeit that the rear of the case is largely grills other than for the I/O plate)

View attachment 178065
View attachment 178066
 
Last edited:
Ok, so I ran Prime95 (Small FFT) and Heaven (Ultra and 256x1440) in parallel for 30 mins and recorded temperatures.

Some detail about the setup:

CPU i5 6600K OC'd at 4522MHz
GPU 2 x Gigabyte GTX970 in SLI both clocked at 1.328MHz

Core temps were taken using HWiNFO64 (v5.12-2740).

Water temp was taken at the output of the single 140mm rad using an XSPC inline sensor
(Loop: Res -> Pump -> GPU#2 -> GPU#1 -> 420mm Rad -> CPU -> 140mm Rad -> Retun to Res)

Air temp was taken using a "cheap" sensor situated to the side of the case outside of any direct airflow from the case fans - it seems to be reasonably accurate when compared against a couple of other less portable temp sensors.

In addition I also have an NZXT Sentry 3 in the system with the temp probe of that attached to the outside of the 140mm rad - I understand that the temps from this might not be so helpful, but as I have them I've also listed them.

I recorded the temps from the three temp sensors when I first booted up the machine.
After running both Prime95 and Heaven for 15 mins I made a note of the temps on all three sensors and for all of the cores and then again after 30 mins.
For the 30 min measurement I also noted the temps being reported for the Motherboard (ASUS z170 Pro Gaming) and the MB VRM.
For the core temps I noticed a bit of fluxuation through the test and across the four cores of the CPU so the number recorded is the maximum temp at the point of measurement and in the case of the CPU that means the max reached across all four.
All temps are in degrees centigrade.


Temp | Start | 15mins | 30mins Air |21.0°C|21.7°C|22.3°C
Water |20.8°C|36.0°C|38.8°C
dT |-|14.3°C|16.5°C
Radiator |19.4°C|35.0°C|37.8°C
CPU |-|77°C|79°C
GPU#1 |-|47°C|50°C
GPU#2 |-|47°C|50°C
MoBo |-|-|35°C
VRM |-|-|75°C

Note that this is all based on the current rad/fan setup:
View attachment 178064
 
Last edited:
What case is this all housed in? Forgive me if this it's mentioned anywhere but I cannot seem to find out
 
IMO, a 16.5C delta is not bad. It can be improved upon, sure, but still not bad. Your component temps also look good (you even have more room for overclocking, if you want).

As far as fan arrangement, I would leave the intakes/exhausts the directions they are in. If you want to invest in better fans, that's fine, but I don't think I'd even do that. I would, however, switch the fans to push on the top. The reason is it will provide more air for the fans, which may make them more efficient, or at least as efficient with less noise. Or it may make no difference at all.

Nice rig!
 
IMO, a 16.5C delta is not bad. It can be improved upon, sure, but still not bad. Your component temps also look good (you even have more room for overclocking, if you want).
I've taken the CPU to 4.7GHz but dialled it back because I was concerned about the temps - which kind of started this thread. ;)

As far as fan arrangement, I would leave the intakes/exhausts the directions they are in. If you want to invest in better fans, that's fine, but I don't think I'd even do that. I would, however, switch the fans to push on the top. The reason is it will provide more air for the fans, which may make them more efficient, or at least as efficient with less noise. Or it may make no difference at all.
I'd read that there are a couple of problems with push:
1/ You have a dead spot in the middle of the fan where the hub is (at least more of a dead spot than with pull).
2/ Dust can collect between the fan and the rad which is harder to clear than dust collected on the open side of the rad which is common with the pull config.

Nice rig!
Thanks :D
 
I've taken the CPU to 4.7GHz but dialled it back because I was concerned about the temps - which kind of started this thread. ;)

I'd push it back up to 4.7 and see what your temps are there. We generally recommend not letting the CPU get hotter than 90C, so you still have room.

I'd read that there are a couple of problems with push:
1/ You have a dead spot in the middle of the fan where the hub is (at least more of a dead spot than with pull).
2/ Dust can collect between the fan and the rad which is harder to clear than dust collected on the open side of the rad which is common with the pull config.

As far as the dead spot is concerned, you'll have that in either configuration. I've found that temps will stay similar at the same or lower fan speeds, but will almost always be quieter due to less air turbulence. The fins don't allow the air to move around the fins on the fan as easily in pull, is all I can figure.

The dust is a valid concern, but one that I would put up with. I mean, you'll have to tear the loop apart every 6 months or so anyway, right? I just dust them then.
 
I'd push it back up to 4.7 and see what your temps are there. We generally recommend not letting the CPU get hotter than 90C, so you still have room.



As far as the dead spot is concerned, you'll have that in either configuration. I've found that temps will stay similar at the same or lower fan speeds, but will almost always be quieter due to less air turbulence. The fins don't allow the air to move around the fins on the fan as easily in pull, is all I can figure.

The dust is a valid concern, but one that I would put up with. I mean, you'll have to tear the loop apart every 6 months or so anyway, right? I just dust them then.

Okay - all makes sense - thanks.
 
Ok, well I jumped in and made the changes and set the fans up as per Option B, i.e. with both rads as intakes with the compromise being the balance between intakes and exhausts.

I would have loved to have made the changes incrementally to see which elements brought me what but I think my wife would have killed me ;) - so I did it all in one big bang, here are the results:

Temp | 15mins | 30mins Air |21.6°C|21.6°C
Water |31.8°C|32.6°C
dT |10.2°C|11.0°C
Radiator |30.6°C|31.1°C
CPU |72°C|73°C
GPU#1 |44°C|44°C
GPU#2 |44°C|44°C
MoBo |-|33°C
VRM |-|72°C

Summary: dT has falled from 16.5°C to 11°C a change of 5.5°C, a change also reflected in the core temps with the CPU max falling from 79°C to 73°C, and the GPU's from 50°C to 44°C. I also note the max temp of the MoBo and VRM also fell by 3°C each.

Worth noting that this improvement has come at the cost of a little more noise but it's an acceptable compromise for me and it's given me the confidence to push the clocks a bit harder.
 
Airflow is more important than the temperature of the air entering the radiators. You want to have front, bottom and sides (if applicable) as intake and top and rear as exhaust. The way you have it set up now will likely cause "dead spots" in the case, and the air in these "dead spots" will continue to rise in temperature until forced out of the case.

I suggest that you run a 2 hour stress test at your current settings, and I think you'll find that temps continue to steadily rise due to lack of airflow.
 
Yep, it was a bit of a jump in to the unknown, and I had concerns about it, but I guess I rationalized it because the rear of the case is pretty much open and with all the fans pushing air down and to the back I was figuring that it would all get pushed out even if there is only one fan extracting at the back.

Anyway, here are the numbers over two hours of running Prime95 and Heaven in parallel.
Note that this isn't exactly an apples with apples comparison from the last run because in the interim I've upped the CPU clock to 4.7GHz and the GPU clock to 1.5GHz.

Temp | Start | 30mins | 60mins | 90mins | 120mins Air |21.5°C|25.3°C|26.7°C|26.8°C|27.9°C
Water |24.9°C|35.7°C|36.3°C|36.7°C|37.4°C
dT |3.4°C|10.4°C|9.6°C|9.9°C|9.5°C
CPU |27°C|81°C|81°C|82°C|82°C
GPU#1 |27°C|45°C|46°C|47°C|47°C
GPU#2 |27°C|45°C|46°C|46°C|47°C
MoBo |-|-|35°C|36°C|36°C
VRM |-|-|80°C|80°C|81°C

That suggests to me that the heat is being dumped into the room pretty efficiently and the system is sustaining a dT of ~10°C.

/goes to grab a cold beer :D
 
Last edited:
Cold beer is good! Cheers.

since you have plenty of radiator space, I'm pretty sure you're going to see temps like that no matter how your fans are configured. However, I'm also sure that you'll see at least these temps (if not a little better, though like I said, it won't be night and day), but at a lower fan speed and with less noise. I'm also sure that there's a hotspot in your case somewhere, but your cooling is combating that successfully as well.

Still a great build man! And like you said, it's always an ongoing project. I'm waiting on my 6 month tear down, and hopefully getting some sleeved PSU cables and mounting clamps for my res, and I'm redoing the lower half of my loop. It looks messier than I was hoping it would be, and I didn't have time or funds left before I needed to get the rig operational.

So, in short, it looks great, temps are good, I think they could be better and I wouldn't blame you at all if you left it as is or completely switched things around. Bottom line is that 4.7 is a beast of an overclock, and that rig will do everything you want without issues, I'm sure.
 
Back