Dual Stacked Intel Stock Cooler Experiment on Bed-Station PC

Calibos

I took apart the Bed-Station PC the other night to take some photo's and took the opportunity to replace the stock Intel Heatsink with a clean one I had lying around rather than clean the dust and grime out of the current one. Applied some new Thermal Grizzly Kryonaut and put the little PC back together. Re-installed Core-temp to make sure everything was OK and was disappointed to see the 4 cores of this low powered i5 4570S were idling at 33-38ºc.

I was pretty sure the last time I monitored temperatures of this PC a few years ago it used to idle in the mid 20's. I did have some trouble with the plastic retention pins of the stock cooler and figured that maybe the heatsink wasn't seated right. I opened it up again last night and removed the heatsink. Sure enough, it was obvious from looking at the underside of the heatsink that it hadn't been mounted right as half the thermal compound on the CPU heatspreader was untouched.

Was about to re-apply more compound and try again but had a brainwave. I dug out one of the other Stock Intel coolers and figured out a way to mount both. They're both those 1.5cm thick ones, however one is all aluminium and the other had a copper core. The All aluminium one is flat on top so I used that one on the bottom of the stack on the CPU heat-spreader and I mounted the copper cored one on top. I modded the fan retention pins by removing the black twisting/locking part and reversed the white plastic part that was a wide flange at one end and which are hollow plastic tubes that its possible to screw into. I added some white plastic around the bottom heatsink as a shroud to ensure the fan air travels through both set of heatsink fins and reused the stiff black plastic shroud for the fan. Couldn't use the cases Mesh top panel for this experiment.

Put it back together and.....disappointment.

It was still idling at 33-38ºc

However, tonight I decided to run a stress test for sh!ts and goggles. I ran Intel Burn Test (IBT) with the 'High' Stress profile. The i5 4570S held at 3.2ghz at 100% load and maxed out at 76ºc

Looks like I have nothing to worry about then really if even IBT on High doesn't send the CPU into the dangerzone.

Maybe the current 'high' idling temps are simply due to a current higher house ambient temperature as compared to when I last monitored temps on this PC a few years ago.

Was going to pick up a Noctua NHL9i that fits this case but tbh I'm not sure I'll bother now. Doesn't matter if my Twin Stacked Intel OEM Heatsinks are ugly AF given that the mod can't be seen anymore anyway what with the Bed-Station PC now mounted behind the bedside locker.

I don't know if the fan was running louder or if it was in my head seeing as I'd being paying attention to it while doing this. Anyway, I installed and configured speedfan, found the fan speed that becomes inaudible which turned out to be about 22% and set a fan profile to keep it at 22% at all temperatures up to 45ºc with a smooth ramp up from there. This should mean that normal daily usage of this PC (Browsing and youtube) doesn't cause the fans to audibly speed up and slow down anymore. Silence!! I setup a scheduled task to start Speedfan minimized with windows. Rebooted to ensure the scheduled task was working and then ran Intel Burn Test again to ensure the fan profile was working. Sure enough the fans were at 22% till I started the test and it started to ramp up in speed as soon as the CPU core temps went above 45º. The cores maxed out at 75ºc for this test and when the test was over the fan slowed back down in steps till the temp went below 45ºc whereupon the fan settled back at a steady 22% speed. Jobs a Goodin' !

Interestingly, it turns out that before I set the new fan profile in speedfan, the fan was running at 40% at idle temperatures. However now with the fan running at a silent 22% thanks to speedfan, the idle temperatures are exactly the same as before. ie. I didn't seem to lose anything by running the fan slower.

ED E

33 isn't a problem.

Thermal transfer is based on the delta between hot and cold, if a room is 20*c an item at 30*c sheds heat a lot slower than one at 50*c. Once you're down close to ambient its harder to get further. Wait until december leave the heating off and you'll bring it down

Calibos

ED E wrote: »

33 isn't a problem.

Thermal transfer is based on the delta between hot and cold, if a room is 20*c an item at 30*c sheds heat a lot slower than one at 50*c. Once you're down close to ambient its harder to get further. Wait until december leave the heating off and you'll bring it down

I blame the hour I posted this at.

I realised later that I'm in an attic bedroom now whereas back then I was in a first floor bedroom. I'm testing now at the end of a still warm and muggy Irish Summer with night time outside ambient temperatures in the 15ºc+ range this weekend. No Hive Thermostat reading for the attic but I do have a Bord Gais cardboard one that after taking out of the drawer is showing a temp between 24-27ºc. Previous readings might have been in December as you say and I may have even purposely biased the test with the window open back then for all I know. Bedstation Mini-PC is now mounted on the back of a bedside locker with less air circulation whereas it was on a table top or 1.5m high wallmount back then.

Given all that, you are correct ED. Its not actually that surprising there is an 11ºc delta from a current ambient in the room of 24-27ºc.

The interesting thing though now that we realise all is actually well with the CPU cooler and its ambient temps causing the current higher readings is that the CPU was still idling at these same temperatures even when I hadn't mounted the stock cooler correctly and it wasn't clamped down on one side evidenced by zero thermal compound transfer to the underside of the Heatsink on one side of the CPU heat-spreader. ie the CPU heatspreader was still able to shed the same amount of heat even with only half its surface in contact with the stock cooler. Maybe the Package consumption of only 10-15 watts at near idle is the reason.

(ED-E. Still rocking a 140mm fan running at 5v off the USB port via a modded cable sandwiched between my old CISCO EPC3925 Virginmedia Bridged Cable Modem. Less than 24 hours UPC/Virginmedia downtime in 8 years. LOL. still getting the full 250/25mbit 24/7/365 with it. Theres only one thing thats going to take me away from Virgin cable Broadband....and thats the ESB SIRO Fibre thats now hanging from the lamp post outside the bedroom window as off last week )