For Science!: FLIR Imagery of PC Watercooling Loops

[Skripka]

So I know someone who recently got their hands on a $3000 FLIR (Infrared for those not hip) camera in their department. We wanted to play around and see what we could see…so the idea of scoping a PC waterloop came up to see what it looks like in the infrared.
Instrument is one of these FLIR E-Series: http://www.flir.com/cs/apac/en/view/?id=41372

So first we need to calibrate the scope, and what better way than with ONE Standard Cat. Because if anything shows anything in calibration, it is using a $3000 FLIR instrument sensitive to 0.1*C to look at my house cat Reilly

*So the “Spot ____*C” reading is the temperature in the middle of the reticle. The scale on the right side recalibrates from time to time with the high/low being the maximum/minimum temps in the frame.


Without further ado, the PC waterloop is my sig rig system as pictured below in visible light with a nice DSLR camera for clarity:

Loop order is:
D5 pump->GPU->240mm Radiator #1 with push/pull AP-15s (lefthand radiator in all pictures)->CPU-> 240mm Radiator #2 with push AP-15s (in the right hand of all pictures)->D5 Pump
*NOTE Ambient temperature in my room was measured to be ~23-24*C for comparison

1) First we need a baseline. So I left my machine powered off overnight. So this is as cold as my machine gets without flipping the switch on the PSU

As you can see…Most everything in frame is dark. The only warmspots aren’t that warm peaking at all of 27C or +3C over ambient…those hotspots are the backlit power buttons on the motherboard. Funny note, the acrylic GPU waterblock top is an IR reflector. You can see the IR reflection of the Z68 motherboard chip heatsink bounced off the top of the GPU block.

Now let us boot my machine and see what we can see. Until load tests, all fans were UNDERVOLTED via fan controller to 5V, for silence.

FLIR CONTINUED IN POST #2

 

[Skripka]

2) This is the IR after POST and during windows loading, with all fans undervolted. Remember the entire loop was cold, so the water has not equalized.

Already the motherboard is glowing in the IR, causing all the PC components and cables to stick out visually as they are cold. The radiators themselves are starting to glow mildly as demonstrated by Radiator #2, however note that the radiator housing itself is cooler than the channels and fins. Water hose has warmed somewhat. Remember that our scale changes, whereas the prior photo had white peak temp at 27*C it is now 34*C.

FLIR CONTINUED IN POST #3

 

[Skripka]

3) This is the system after equalizing at Win desktop and CPU and GPU idle, with fans still undervolted.

Note that even with our moving color-temperature scale that all components including the hose are now glowing much more above the low-end of the scale. The crosshairs here are just on the water fittings on the CPU block. The glowing area around the block is the CPU VRM. Now equalized at idle, we see the heat the waterpump generates in the glow around the circumference around the D5 which is running at full throttle. Another interesting bit I spotted was the ATX 24-pin cable having a few glowing wires on it on the right. Note how uniform Radiator #2 is in color.

*For perspective per AIDA64, my CPU temp under the IHS is 38C according to the thermostats.

**Something to NOTE-I found in outside trials is that watercooling hose is opaque in the IR. So all watercooling hose glowing is due to the hose itself heating up from the coolant…the glow is not from the coolant itself.

FLIR CONTINUED IN POST #4

 

[Skripka]

4) CPU Load=100% stressing CPU/cache/FPU/System Memory, GPU idle, Radiator #2 fans at 12V, Radiator #1 push/pull fans at 5V. Note this is an e3-1230 part, which is only rated for 80W TDP. This cooling configuration is my common one as it gives decent CPU cooling but is still fairly quiet. Here we should see things really start glowing.

Now here we go with getting things glowing. The entire motherboard is glowing. Note how the RAM appears hot on the sides but cooler on the tops, probably ascribable to the heatspreader designs on these Ripjaws. Also Radiator#2 has lost some uniformity, I can spot the AP-15 fan bullets wind shadow blocking the air movement throught eh heat exchanger. The CPU waterblock itself is warmer now, with the CPU as measured by AIDA64 hitting 46*C.

FLIR CONTINUED IN POST #5

 

[Skripka]

5) Having loaded the CPU…It is now time to load the GPU and see what we can see. GPUs being high power draw and high heat devices…this should be fun. GPU was loaded with FurMark 1.9.2 doing burn in for 15+minutes to make sure things were stable. All radiator AP-15 fans are now running at 12V (1850RPM), as this is how I normally game with headphones on.

Note the coolant temps as indicated by the heating of the hoses are warmer. Also note the bright glow line around the GPU waterblock itself. That is the PCB of the VGA card getting hot.

Temps as measured by AIDA64 during GPU load:
GPU DispIO=42*C
GPU MemIO=50*C
GPU Shader=50*C
GPU VRM1=36*C
GPU VRM2=26*C
CPU temp=40-41*C

This test being about the GPU…let us focus ONLY on the VGA card itself in the IR rather than the whole loop. These images are all of the back of the GPU PCB, as this is the only area visible to IR during operation. First we need a baseline-the GPU at idle load:

In this photo we can see the GPU VRM area being warmer than the backside of the GPU itself. We also see a hotspot near the PCIe backplate-right wear the GPU sends data over DVI/HDMI/etc.


Now at FurMark 100% load:

Note the new scale…the VRM instead of reading 35C according to the IR now hits almost 60C. Further the hotspot near the VGA ports is still there…but the GPU itself is now warming a bit but still nowhere near as warm as the rest of the card. Why? My guess is not only the IHS-less GPU getting cooled by the waterblock, it is also the VRAM chips surrounding the GPU getting watercooled as well.