S4 Mini w/ Ryzen 1700 and 1080 TI

MarcParis

Spatial Philosopher
Apr 1, 2016
3,616
2,705
Stupid question : why not orientate all your fans as exhaust instead of intake? That's generally the best setup in sff in order to avoid to heat other components?
 

Reglar

SFF Lingo Aficionado
Original poster
Jul 16, 2017
111
129
Amazing dedication @Reglar. I would've shoved the entire system underneath my desk or plan an lesser specced one.

Thanks, I hate leaving things unfinished.

Stupid question : why not orientate all your fans as exhaust instead of intake? That's generally the best setup in sff in order to avoid to heat other components?

I was worried about air restrictions in the case. I'll definitely try it though on both CPU and GPU (need to cut off the top corners though for the CPU test).
 

Reglar

SFF Lingo Aficionado
Original poster
Jul 16, 2017
111
129
OK, mini update. Completed test 4 with just the Noctua, no dice, card throttled all the way down to min and was above 84C.

I then tried the Noctua in exhaust and it was worse, throttled to min and hit 90C.

Moving onto test 5 which will be the NZXT fan, it won on CPU let's see how it goes on GPU. For now the test will just be it, I'll then see if adding some 40mm fans help at all.
 

Reglar

SFF Lingo Aficionado
Original poster
Jul 16, 2017
111
129
I think I can get a winning combination, just need to work through all the permutations and combinations.
 

Reglar

SFF Lingo Aficionado
Original poster
Jul 16, 2017
111
129
Small update, I have now completed testing 13 combinations of fans on the GPU, and now I'm parsing all the data. In addition to the noise and temp measurements that I did on the CPU tests, I'll be including benchmark scores for each, as well as stats on GPU clock speed.

I'll post data as soon as I have it complete.
 

Reglar

SFF Lingo Aficionado
Original poster
Jul 16, 2017
111
129
First up, some graphs, followed by some comments at the bottom.

Noise level



Noise to performance - synthetic tests



Noise to performance - Metro benchmark



Benchmark (score and median frequency) - Heaven



Benchmark (score and median frequency) - Time Spy



Benchmark (score and median frequency) - Fire Strike Ultra\




You can clearly see a relationship between frequency to score, and a pretty significant relationship at that. You can also see that Noctua by itself is definitely the worse performing fan combination.

Score wise most of the rest of the fan combinations will do the job, so now it comes down to personal preference on noise level. Note in the discussion below I refer to the main fan, but there are usually 2 50mm Scythe fans paired with it.

Looking at the noise to temp graph on synthetics, the Cougar looks VERY attractive, with the Gentle Typhoon (69 CFM model) right behind it. If I then look at the noise to temp graph on Metro benchmark, the Phanteks looks good (I'm ignoring the standalone Noctua given the horrible benchmark scores).

The Gentle Typhoon does better in the benchmarks than the Cougar and the Phanteks, so I'm going to go with the Gentle Typhoon.

Next steps are to now try different thermal paste on the GPU. I have 3 to try out:
- Noctua NT-H1
- IC Diamond
- Themal Grizzley Kyronaut

My prediction is that the Kryonaut will work best, so I am saving that for last.

The testing for the thermal paste will probably start Sunday, although I may take a break Sunday and actually just play some games :)

** Edited to fix some labels on the graphs
 
Last edited:
  • Like
Reactions: Josh | NFC

MarcParis

Spatial Philosopher
Apr 1, 2016
3,616
2,705
Hourrah!!!
Gentle typhoon has won another fight...just 8years after its release...;) what is making competition?..;)

Ek vardar will perform similarly but they are by far louder
 

Reglar

SFF Lingo Aficionado
Original poster
Jul 16, 2017
111
129
OK well I did the GPU TIM testing today to get it out of the way. Bottom line is that the commercial TIMs performed better, but only slightly. I'm not sure there was enough of a change to warrant doing it yourself.

Of the TIM's I'd say the Kryonaut provided the best temps, but only by 2-3 degrees. Certainly no significant boost in scores or median frequency with the change.



Next step is I think I'll replace the CPU TIM, just to get a few degrees improvement. I won't be doing a ton of testing for this, I'll just use the Kryonaut and report the temps.

The final step will be to create the video with all the audio from all the fan testing. That will hopefully get done this week and then I'm done with this build.
 

Siba

Chassis Packer
May 19, 2017
19
9
is this accurate to sum things up?
  • 1 gentle typhoon + 2x 50mm scythe
  • kyronaut paste
is the best noise + performance combination? can you provide pictures of the orientation and setup? does it require the 3d bezel or can it work with the stock aluminum faceplate?
 

Reglar

SFF Lingo Aficionado
Original poster
Jul 16, 2017
111
129
is this accurate to sum things up?
  • 1 gentle typhoon + 2x 50mm scythe
  • kyronaut paste
is the best noise + performance combination? can you provide pictures of the orientation and setup? does it require the 3d bezel or can it work with the stock aluminum faceplate?

Siba, let me preface this that my choices are a bit subjective. From a performance/noise perspective on the Zotac GTX 1080 TI I found that the Gentle Typhoon (63 CFM variant, there are 3 types for GT) + 2 50mm Scythe fans worked the best and gave me the "gentlest" noise. The thermal compound change didn't really do anything but lower my temps 2-3 degrees. There was no material change to the scores for the tests I did.

The Zotac GTX 1080 TI will NOT fit unless you have the 3d Bezel sold by Josh (NFC systems) specifically for the Zotac 1080/1080 TI.

Here's a picture of the card before I started doing the fan mod. You can see how the card extends through the opening in the chassis.



Some other points from earlier in the thread:

  • I'm using a cardboard baffle to create separate chambers for the GPU and the motherboard; that had the most appreciable heat improvement prior to me playing with the fans on both CPU and GPU
  • You have to be willing to FLEX the GTX 1080 enough so that the 25mm tall fan will fit on the end; if your card is particularly rigid this could cause an issue for you so please make sure you understand the risk before you try this
 

Siba

Chassis Packer
May 19, 2017
19
9
How far does the heatsink stick out past the frame? I was trying to find out if milling the existing aluminum bezel would allow for enough clearance to fit it, and if using the non-stock fan setup would allow for everything to fit after milling (in case the stock plastic doesn't fit w/ milling)

Can you take a quick profile pic or vid to show it? Something like this
Also a pic of the finished setup so we can see how the GT63 + 2x 50mm looks
 

Reglar

SFF Lingo Aficionado
Original poster
Jul 16, 2017
111
129
Heads up everyone, I realized I have a mistake in the name of the 50mm fan, the fan is NOT the Scythe 50mm but the Fractal Design Silent Series R3 50mm. Apologies for mislabeling that fan, it looks similar to the Scythe and I got confused.
 

Reglar

SFF Lingo Aficionado
Original poster
Jul 16, 2017
111
129
Quick weekend update, I have started putting together the video on the build. I hope to have it posted by Monday or Tuesday.
 
  • Like
Reactions: Choidebu