Quiet Mini-ITX Gamer Build Guide

Do-It-Yourself Systems | Silent PC Build Guides
Viewing page 6 of 7 pages. Previous 1 2 3 4 5 6 7 Next

STRESS TESTING: CPU-CENTRIC

System Measurements
System State
Idle
x264 Playback
Video Encoding
Prime95x4
CPU Fan
720 RPM
CPU
27°C
28°C
39°C
44°C
MB
33°C
37°C
SSD
n/a
GPU
36°C
38°C
13 dBA
Power (AC)
35W
39W
81W
98W
Case fans at 500 RPM (low), GPU fans off (auto).
Ambient temperature: 21°C.

We began by testing the system with CPU-centric applications to see how it performs with non-gaming tasks. It's clear that despite its higher TDP (88W vs 84W) this Core i5-4690K is considerably less power hungry CPU than the i7-4770K we used in the Quiet ATX Gamer. Even when you consider the absence of a 3.5" HDD and the use of an 80+ Platinum efficiency PSU in this rig, the 131W AC with Prime95 (x8) in the ATX Gamer versus 98W in this system with Prime95 (x4) is striking. As you'll see in the later tests with the full load tests where the CPU and GPU had exactly the same load (Furmark + Prime95x2) as in the Quiet ATX Gaming system, AC power consumption remained lower by 15~20W throughout. This i5-4690K may be an exceptionally efficient sample, or perhaps it's the result of improving yields in Intel manufacturing.* Whatever the reasons, with this modest thermal load, the Silverstone AR03 CPU cooler just cruises at with its fan 700 RPM; there's no need to run the fan speed any higher, even if the ambient room temperature jumps to tropical levels.

*CA Steve, one of our trusty forum mods, comments: "The Devil's Canyon parts use a different TIM which should lower the temps a bit. The lower temps might lead to a couple of the watts saved vs the earlier i7. The majority of the savings is most likely yield and process improvements."

STRESS TESTING: GPU-CENTRIC

A. FAILED RUNS

We won't clutter this article with all the information about our initial failed trials. They were failures for one key reason: The GPU exceeded 90°C and its fans kicked into high gear, with SPL going over 30 [email protected] The GPU would then cool down over the next minute or two, and once temperatures dropped a few degrees below 90°C, the preset fan profile would take over (39% or ~990 RPM), bringing the noise down to under 20 dBA. But with continued Furmark load, the temperature would exceed 90°C again. As long as Furmark was running, the Strix GTX 980 fans cycled between pretty darn quiet to fail-safe loud over about a 3~4 minute period. Putting the case fans to high speed only stretched the cycling time, and it pushed the default SPL of the system to 25 dBA, so this was no solution, anyway.

The only thing that helped was to open up the case. Removing the left side panel kept GPU temperature under 80°C even at full load, with the GPU fans at just 38%. (All this was at 20~23°C room ambient.) The price? The noise of the fans, especially the rougher GPU cooler fans, became finely audible. The overall SPL was still modest, just around 20 [email protected], but the noise of the fans was a bit too richly detailed for us. This exercise told us that the only way to keep the GPU under 90°C with its fans running at <1000 RPM is to increase cool airflow to the GPU fans.

How could this be done without removing the side panel altogether or individually enlarging each of the small vent holes in the panel? The latter would be a messy, tedious job, and most user would shy away from this case if that kind of mod proved necessary.

There is one small detail mentioned in our look through the case: The side panel has a dust filter, a sheet of perforated plastic glued on the inside of the side panel vent holes. In case you've forgotten, here's that pic again.


Side panel with plastic film dust filter over VGA vent holes.


Side panel after "mod".

You knew this mod was coming, didn't you? A utility knife was used to pry under the plastic sheet to carefully peel it off. The adhesive is only under the non-perforated side strips. The end result is the bare unadored metal in the second photo above. Would this make a difference? As it turns out, yes, just enough.

B. SUCCESS

System Measurements
(After side dust filter removal)
System State
Prime95x4
Resident Evil 6 Benchmark
Prime95x2 + FurMark
GPU Fan RPM
off (auto)
870 (38%)
990 (39%)
1630 RPM (auto)
CPU
44°C
51°C
54°C
51°C
MB
37°C
44°C
46°C
44°C
SSD
n/a
GPU
38°C
79°C
86°C
72°C
13 dBA
18 dBA
19.6 dBA
25 dBA
System Power (AC)
98W
237 W
275W
272W
*Custom fan profile used on ASUS GPU Tweak: Fans turn on at 60°C at 35%, switch to 39% at 75°C then 40% at 87°C. These settings were established through a lengthy iterative process of tweaking. CPU fan constant at 700 RPM, system fans at 520 RPM (low). Ambient temperature: 22°C.


Quiet Mini-ITX GamerI at idle.


Quiet Mini-ITX GamerI at full load.

Once the GPU is fully loaded, it heats up the whole interior, causing a 7~10°C increase in CPU and motherboard temperatures. But with the slightly lowered airflow impedance of the side panel vents, our custom GPU Tweak fan profile kept the GPU from reaching 90°C. At this point in the fan curve, a 1% change in CPU Tweak's fan speed setting is equal to about 1 [email protected]

We were quite pleased to finally get the maximum load noise to under 20 [email protected] It took a full day of experimentation to achieve it. With the 13 [email protected] noise level at idle, the system is quiet enough to be barely audible in a quiet room when placed atop your desk close to the monitor.

If room ambient temperature goes up to, say, 27°C, these setting might not work; if the GPU exceeds 90°C then the cooler fans would ramp up. At that point, it would be quieter to just leave that side panel off, and position yourself to the right of the system. That would still be just 20 [email protected] with our settings.

On the other hand, Furmark + Prime95x2 is still at least 10% (more often 15~20%) higher power load than any game, and unlike games whose power varies dynamically, this stress load is constant. So there's a decent chance that even with a warmer ambient, this system as it is configured would remain under 20 [email protected]

AUDIO RECORDING

This recording was made with a high resolution, lab quality, digital recording system inside SPCR's own 11 dBA ambient anechoic chamber, then converted to LAME 128kbps encoded MP3s. We've listened long and hard to ensure there is no audible degradation from the original 24/88 WAV files to these MP3s. They represent a quick snapshot of what we heard during the review.

The recording starts with 7 seconds of ambient noise, then ~10 second segments of the system at various states. For the most realistic results, set the volume so that the starting ambient level is just barely audible, then don't change the volume setting again while listening.



Previous 1 2 3 4 5 6 7 Next

Silent PC Build Guides - Article Index
Help support this site, buy from one of our affiliate retailers!
Search: