[Experiment] Household materials as thermal compound extreme

[Lord Nicoll]

In this test, I really wanted to test the boundaries, and I did. Even the good actual thermal compound failed, although it lasted longer. This test was performed on an i5 3570k at a very high 1.4 Volts on a stock Skylake intel cooler (there was no way anything could cool this). While actual thermal compound seemed to be managing, after 2 minutes (most failed within 30 seconds) it too succumbed, which might have been the result of the cooler rather than the Thermal Compound. The results are below. Nothing I tested actually worked, but again, 1.4 volts was extreme, on a stock cooler was never going to work. I wanted to see could it last. I also tested single core stress tests and some of them lasted 10 minutes, Thermal compound, Tooth paste, WD40 and washing up liquid all managed to keep 1 core at a time under control, so maybe if you're stuck they could be used. The nice formatting text file is included.  If you think I should retest fairly or with a better cooler just flame away. The system that was used was
Intel Core i5 3570K stock clocks @ 1.400 Volts
8GB Crucial DDR3 1600
ASUS ROG Maximus V Formula
Intel stock cooler (for Skylake, it's a lot smaller)

AMD FirePro V5900
Some random Dell 350 watt PSU
Some Random 320GB Toshiba Drive

i5 3570k @ 1.400 volts stock clocks, stock cooler. Prime 95 small FFT for 10 minutes OR until 25% thermal trottling OR all cores at one point log.

	thermal paste		temp				status		notes
	No Thermal Paste	H:105°C (all cores) L:53°C	fail		all cores logged within 1 minute, test stopped. 
	model polly cement	H:105°C L:47°C			fail		strong smell of glue, not very effective, +25% throttling, difficult to clean. only core 1 didn't throttle, at 99°C
	WD40			H:105°C  L:59°C			fail		instant failure and trottling
	petrolium Jelly		H:105°C L:53°C			fail		Throttling over 50% and 3 cores reached TJ max
	Strontium Toothpaste	H:105°C L:47°C			fail		works well compared to others, but still failed, test stoped at 50% thermal throttling, tested lasted almost a minute 
	Magnesium Sulphate	H:105°C (all cores) L:50°C	fail		Test aborted after 1 minute, all cores throttled to 1.6GHz
	washing up liquid	H:105°C L:45°C			fail		At first did good, but quickly throttles and fails, but did better than some
	Metalic Acrylic paint	H:105°C L:52°C			fail		All cores reached over 100, two throlled before falling to just 1.4GHz, a pain to clean
	Plasterine		H:105°C L:54°C			fail		failure within seconds as three cores log, and trottles back over 50%
	aluminium foil pieces	H:105°C L:53°C			fail		instant failure with multiple cores throttling and over 50% clock throttle
	Gel Shaving foam	H:105°C L:46°C			fail		test quickly failed, thermal throttling caused clocks to fall 50%
	Actual thermal paste	H:105°C L:46°C 			fail		failure was much more controlled, fan kept up for 56 seconds before throttling began, test ran for 2 minutes and 13 seconds before throttling reached 50%, no more than 1 core reached TJ max at the same time.

 

 

temps.txt

[Guest]

Wow this is amazing. Though I would suggest changing format to

H:105 C L: 40 C

So that people doesn't get confused and thought you got negative temps

[Lord Nicoll]

1 minute ago, deXxterlab97 said:

Wow this is amazing. Though I would suggest changing format to

H:105 C L: 40 C

So that people doesn't get confused and thought you got negative temps

Possibly, that although to go to minus temps with a stock cooler would put all heatsink makers out of money. 

[Oshino Shinobu]

No mayo? It's the best of the TIM. 

[Lord Nicoll]

3 minutes ago, Oshino Shinobu said:

No mayo? It's the best of the TIM. 

I wanted to avoid food products for this one, there'll be a food and body fluids one later, maybe at a lower voltage, while a few runs at 1.4 won't do much, regularly doing it might.

[Guest]

5 hours ago, Lord Nicoll said:

I wanted to avoid food products for this one, there'll be a food and body fluids one later, maybe at a lower voltage, while a few runs at 1.4 won't do much, regularly doing it might.

Put mustard and ketchup later then

[Lord Nicoll]

Just now, deXxterlab97 said:

Put mustard and ketchup later then

Oh boy do I hate those condiments, I might need my respirator for that (I avoided it when dealing with the noxious poly cement that actually made be pretty high and dizzy and gave me a splitting headache) 

[Lord Nicoll]

If anyone is interested at stock volts (in a room that is now 5-10°C colder) the max temps are under 75°C with Prime95 small FFT with actual thermal paste

[Energycore]

I'm interested in tests with a cooler that can actually keep up, or with a lowered heat load on the stock cooler

[SCHISCHKA]

I think all your experiment shows is that a CPU consistently fails at 105C regardless of thermal compound. A more scientific test: put a meth burner under a sheet of wood or non-flammable product with a hole cut for a CPU lid. Measure the temperature difference over time.

[aisle9]

The fact that actual thermal paste failed suggests that maybe you should have used a bigger cooler.

[gtx1060=value]

13 minutes ago, Lord Nicoll said:

Oh boy do I hate those condiments, I might need my respirator for that (I avoided it when dealing with the noxious poly cement that actually made be pretty high and dizzy and gave me a splitting headache) 

homemade heatsink? or could you try out the heatsinks you get with stuff like rasberry pi? the metal ones with fins

[vanished]

6 minutes ago, aisle9 said:

The fact that actual thermal paste failed suggests that maybe you should have used a bigger cooler.

The results are "clipped" and so it's not possible to compare performance on the high end where it really matters.

[Lord Nicoll]

6 minutes ago, Energycore said:

I'm interested in tests with a cooler that can actually keep up, or with a lowered heat load on the stock cooler

Well, the cooler couldn't keep up but only actual thermal paste lasted more than 1 minute and 30 seconds, but none of them thermally shutdown (I was watching the whole time)

5 minutes ago, SCHISCHKA said:

I think all your experiment shows is that a CPU consistently fails at 105C regardless of thermal compound. A more scientific test: put a meth burner under a sheet of wood or non-flammable product with a hole cut for a CPU lid. Measure the temperature difference over time.

That is the TJ Max, but not the point of failure, a fail is when three or four cores all reach or exceed 105°C, clock speed fell below in this case 2.65 (some did this even though the core temps where at 80-90, that is why the clocks went down). I could use the stock cooler with a precision heat source and just change up the TIM but that is a lot of work and has other variables. 

4 minutes ago, aisle9 said:

The fact that actual thermal paste failed suggests that maybe you should have used a bigger cooler.

It failed because 1.4 volts is a lot and even a good cooler would have struggled, but the stock cooler held out for a little bit when it actually has good transfer, Thermal compound and tooth paste to name a few lasted longer and didn't throttle within a 10 seconds.

 

9 minutes ago, Ryan_Vickers said:

The results are "clipped" and so it's not possible to compare performance on the high end where it really matters.

Yes, an info dumb or graph would have been a little difficult to display and process on a PC that is already being tortured. No core at any time exceeded 105°C though, so it stopped a complete thermal event shutdown. 

 

9 minutes ago, wrathoftheturkey said:

lol. Even at stock, it'll throttle almost any mainstream processor in something like AIDA

Actually the i5 3470k at stock volts leveled at 75°C for all 4 threads in Prime95, maybe if you throw in memory and all the cache it might hit 80°C but it held just.

[Clanscorpia]

Could you give the recipe? 

[Lord Nicoll]

1 minute ago, Clanscorpia said:

Could you give the recipe? 

recipe for what?

 

[J. Hammond]

38 minutes ago, Lord Nicoll said:

I wanted to avoid food products for this one, there'll be a food and body fluids one later, maybe at a lower voltage, while a few runs at 1.4 won't do much, regularly doing it might.

Quote

there'll be a food and body fluids one later

Quote

body fluids

Oh god please no.

[Clanscorpia]

5 minutes ago, Lord Nicoll said:

recipe for what?

 

Fior the thermal paste. Like ratios and stuff like that 

[Lord Nicoll]

5 minutes ago, J. Hammond said:

Oh god please no.

yes, that was a suggestion, I'm em, not too keep to do that one actually.

[Lord Nicoll]

2 minutes ago, Clanscorpia said:

Fior the thermal paste. Like ratios and stuff like that 

OH, the procedure wasn't all that uniform, as some of the materials where more vicious, but on average, a pea of the globey stuff and usually about .25ml of the more liquid ones

[Lord Nicoll]

On 1/10/2017 at 2:32 AM, Energycore said:

I'm interested in tests with a cooler that can actually keep up, or with a lowered heat load on the stock cooler

 

 

[Lord Nicoll]

On 1/10/2017 at 2:41 AM, Ryan_Vickers said:

The results are "clipped" and so it's not possible to compare performance on the high end where it really matters.

Perhaps these results are better?

[Lord Nicoll]

On 1/10/2017 at 3:00 AM, J. Hammond said:

Oh god please no.

Yes, I did them, they where bad and actually killed that very system basically.

 

[vanished]

9 minutes ago, Lord Nicoll said:

Perhaps these results are better?

Yes, that's far more useful, if a bit disturbing.