All the AMD faggs in this comment section, Jesus Christ...
huur duuuur linus on intel payroll muh conspiracy muh unfair tactics muh healthy computition yuh intel would totaly do this and kill your mother too muh lisa su... go back to plebit gobbling bucket of 30 fps propeled AMD CEO cumbucket.
Why did they have to harvest tim from old cpus rather than just buying 2 new ones and delidding the other, applying the new paste and keeping the other intact?
WHY ... Why did they do it this way? they could have simply made a reading with the processor, then de-lid it , replace the TIM and .. yeah , you would have gotten much better data from just 1 .. 1 processor ..
That harvested TIM looked very much like the old, baked out MX-2 I had to clean out of a really hot-running laptop. (I know it was MX-2 because I applied it about 3 years prior.) Replacing that with fresh MX-2 (from the same tube the old stuff originally came from) produced a 15°C difference. And just like this one, the old MX-2 wouldn't come off with alcohol - it just retained the same rubbery consistency as before, not even discoloring the alcohol -, while fresh MX-2 would dissolve somewhat.
My guess is Intel's using the same binder material found in MX-2, and it polymerizes under excessive heat.
It may skew the test a little but if you put a drop of cheap thermal paste in with the intel tim paste it will soften it up so it spreads out as it would normally
Wouldn't it have been easier to get two of the same processors, test both thermals, de-lid one, put on new paste, re-lid it, then compare? Or can you not put a lid back on?
I want to know what happens if you delid then reapply the lid and test. I hadn't thought of the fact that the Intel TIM was chosen for longevity. That paste reminds me of dried paste found on dead(old) motherboards. To my question: I'm certain that it would result in overheat and death due to the dry paste not shaping into as good a thermal pathway as the solidified state of it, which may be engineering on Intel's part rather than neglect. It continues to be unacceptable that Intel stopped soldering, a longevity and performance solution, without publicly offering justification. We can only assume that stunted lifespan is good for business.
This test has so many problems, but the largest is that it misses the problem with Intel CPUs. The problem is not that Intel's TIM is bad, it's a combination of the use of TIM (any TIM, besides a metallic compound), and the large gap between die and heatspreader due to the adhesive. I've seen tests where the TIM was replaced with NT-H1 and the same gap was maintained between die and heatspreader, and NT-H1 performed *worse*.
Delidding and specifically using liquid metal compounds works well because: 1. The thermal environment between CPU die and heatspreader is hell; this is where the massive thermal conductivity advantage of metallic interfaces shows a significant benefit. 2. The heatspreader is usually brought closer to the die.
The best results are achieved when the adhesive is scraped off entirely and not replaced, i.e. leaving the heatspreader 'floating' on top. Sealing with adhesive consistently produces worse results because it increases the gap.
So no, Intel's TIM is fine. The problem is using TIM instead of solder.
Why not reach out to Intel and ask for a couple CC's of their paste? There are a bunch of Intel campuses a stones throw from where you guys set up shop :)
Chunk up some 6 year old used NT-H1, reconstitute it, and see how differently it performs from fresh Nt-H1. Otherwise... you know what you've done wrong.
so comparing a chip that had factory applied tim versus after delidding it and applying yours introduced too many new variables (such as???), but harvesting some dried up tim dust and dissolving it in alcohol to apply it didn't introduce new variables?
also a 25 degree difference isn't gonna happen from going from one paste to another. even if the stock tim was only rated at 1W/mK, going to the noctua paste isn't going to drop 25 degrees. going to liquid metal on that side of the ihs isn't going to get 25 degrees. delidding and replacing with liquid metal, then putting liquid metal on the outside of the ihs isn't going to get 25 degrees. I would say the results are largely due to poor applications of the stock dried up tim, since it can't spread you lost a huge amount of surface area and also the tim itself couldn't make good contact with the surface. I'm sure there were tons of air pockets in their wrecking the thermal conductivity.
would be a better test to just test the stock chip against a delidded chip.
Just a thought, shouldn't you find two identical CPU's with roughly the same thermal performance, test them, then delid one and apply standard thermal paste? The approach taken here adds way more variables for anything to be valid. And yeah... should have soaked the thermal gunk in silicon oil not alcohol.
Nice try guys but you can't conclude a result on this test. Unless you do a third test with good thermal paste that was once used and dried on a cpu, removed and reconstituted with alcohol the way that you did in this video. You have no control for that part of the test. All you did was compare good to bad and your results showed that but they aren't showing a control, the control would be what I had previously suggested. ''edit, I still gave a thumbs up and for what it's worth I still enjoyed the video''
I don't understand why the test would be done this way? If you had at least 4 old CPU's, why not just leave two untouched and install the high-performance paste on the other two so you could eliminate the wild variable of reapplying alcohol soaked old paste?
You should have compared reused Intel paste with original application Intel paste first to prove that it works the same when scraped off and re-applied.
If you want a sample of fresh Intel Thermal compound you can get 5 or 6 of the stock heat sinks and scrape the fresh compound off of them it comes stuck to the bottom of all of them . I'm sure Intel only stocks 1 kind of thermal paste corporations like to be efficient and save all the money they can
I just did this (thermal paste harvesting and re-use). Isopropyl alcohol doesn't work. Ethyl alcohol doesn't work. Acetone doesn't work. Beacon Solvent and Thinner 4 oz. (yes, NewEgg) works perfectly. I think it's Heptane based.
All the AMD faggs in this comment section, Jesus Christ...
ReplyDeletehuur duuuur linus on intel payroll muh conspiracy muh unfair tactics muh healthy computition yuh intel would totaly do this and kill your mother too muh lisa su... go back to plebit gobbling bucket of 30 fps propeled AMD CEO cumbucket.
That's like reusing a condom. I'm for the red team, but come on!
ReplyDeleteWhy not contacting with intel to send a simple ? .. pffff bs
ReplyDeleteWhy did they have to harvest tim from old cpus rather than just buying 2 new ones and delidding the other, applying the new paste and keeping the other intact?
ReplyDelete1:11 he's a dealer. A cpu dealer that is
ReplyDeleteThe worst kind of dealer on the corner of the block
WHY ... Why did they do it this way? they could have simply made a reading with the processor, then de-lid it , replace the TIM and .. yeah , you would have gotten much better data from just 1 .. 1 processor ..
ReplyDeleteThat harvested TIM looked very much like the old, baked out MX-2 I had to clean out of a really hot-running laptop. (I know it was MX-2 because I applied it about 3 years prior.) Replacing that with fresh MX-2 (from the same tube the old stuff originally came from) produced a 15°C difference. And just like this one, the old MX-2 wouldn't come off with alcohol - it just retained the same rubbery consistency as before, not even discoloring the alcohol -, while fresh MX-2 would dissolve somewhat.
ReplyDeleteMy guess is Intel's using the same binder material found in MX-2, and it polymerizes under excessive heat.
Why does Linus look like he has just eaten some Spaghetti Bolognese?
ReplyDeleteIt may skew the test a little but if you put a drop of cheap thermal paste in with the intel tim paste it will soften it up so it spreads out as it would normally
ReplyDeletePerhaps try again after you find an actual solvent for the thermal paste?
ReplyDeleteWouldn't it have been easier to get two of the same processors, test both thermals, de-lid one, put on new paste, re-lid it, then compare? Or can you not put a lid back on?
ReplyDelete0:48 Alex is a BADASS!!!
ReplyDeleteC'mon, just buy AMD chips which are soldered.
ReplyDeleteanyone remember dave chappel. i got the yellow cake right here..... dont touch that shit! @ 3:30
ReplyDeletethat paste was too thick. It didn't spread enough.
ReplyDeleteIt's toothpaste!!! J/k But, I've ponder this topic for sometime. I am glad you did this.
ReplyDeletethat's why you mix the flakes with tooth paste and not rubbing alcohol
ReplyDeleteWhat about Xeon ?
ReplyDeleteUsed Alcohol instead of heat for the application....
ReplyDeleteWhut.
1:35 Trigger warning.
ReplyDeleteI want to know what happens if you delid then reapply the lid and test.
ReplyDeleteI hadn't thought of the fact that the Intel TIM was chosen for longevity. That paste reminds me of dried paste found on dead(old) motherboards. To my question: I'm certain that it would result in overheat and death due to the dry paste not shaping into as good a thermal pathway as the solidified state of it, which may be engineering on Intel's part rather than neglect.
It continues to be unacceptable that Intel stopped soldering, a longevity and performance solution, without publicly offering justification. We can only assume that stunted lifespan is good for business.
This test has so many problems, but the largest is that it misses the problem with Intel CPUs. The problem is not that Intel's TIM is bad, it's a combination of the use of TIM (any TIM, besides a metallic compound), and the large gap between die and heatspreader due to the adhesive. I've seen tests where the TIM was replaced with NT-H1 and the same gap was maintained between die and heatspreader, and NT-H1 performed *worse*.
ReplyDeleteDelidding and specifically using liquid metal compounds works well because:
1. The thermal environment between CPU die and heatspreader is hell; this is where the massive thermal conductivity advantage of metallic interfaces shows a significant benefit.
2. The heatspreader is usually brought closer to the die.
The best results are achieved when the adhesive is scraped off entirely and not replaced, i.e. leaving the heatspreader 'floating' on top. Sealing with adhesive consistently produces worse results because it increases the gap.
So no, Intel's TIM is fine. The problem is using TIM instead of solder.
Ohh tunnelbear......
ReplyDeleteOh hey, a canadian jay bauman.
ReplyDeleteThe isopropyl alcohol route was the only way to go.
ReplyDeleteTrust me, I'm an honest to goodness person.
On reflection, white spirit might have been a better option.
ReplyDeleteGlycerin?...
ReplyDeleteWell, it is for sure better than toothpaste, This I can confirm.
ReplyDeleteI wonder if they would ever sell some of there CPUs that they use for testing
ReplyDeleteWhy not reach out to Intel and ask for a couple CC's of their paste? There are a bunch of Intel campuses a stones throw from where you guys set up shop :)
ReplyDeleteChunk up some 6 year old used NT-H1, reconstitute it, and see how differently it performs from fresh Nt-H1.
ReplyDeleteOtherwise... you know what you've done wrong.
Alcohol is the last thing you wanted to ues. that's just gonna dry it up more. should have tried a small amount of mineral oil
ReplyDeleteso comparing a chip that had factory applied tim versus after delidding it and applying yours introduced too many new variables (such as???), but harvesting some dried up tim dust and dissolving it in alcohol to apply it didn't introduce new variables?
ReplyDeletealso a 25 degree difference isn't gonna happen from going from one paste to another. even if the stock tim was only rated at 1W/mK, going to the noctua paste isn't going to drop 25 degrees. going to liquid metal on that side of the ihs isn't going to get 25 degrees. delidding and replacing with liquid metal, then putting liquid metal on the outside of the ihs isn't going to get 25 degrees. I would say the results are largely due to poor applications of the stock dried up tim, since it can't spread you lost a huge amount of surface area and also the tim itself couldn't make good contact with the surface. I'm sure there were tons of air pockets in their wrecking the thermal conductivity.
would be a better test to just test the stock chip against a delidded chip.
Just a thought, shouldn't you find two identical CPU's with roughly the same thermal performance, test them, then delid one and apply standard thermal paste?
ReplyDeleteThe approach taken here adds way more variables for anything to be valid.
And yeah... should have soaked the thermal gunk in silicon oil not alcohol.
Nice try guys but you can't conclude a result on this test. Unless you do a third test with good thermal paste that was once used and dried on a cpu, removed and reconstituted with alcohol the way that you did in this video. You have no control for that part of the test. All you did was compare good to bad and your results showed that but they aren't showing a control, the control would be what I had previously suggested. ''edit, I still gave a thumbs up and for what it's worth I still enjoyed the video''
ReplyDeletem/ Ryzen m/
ReplyDeleteThis is silly. Please make a contact and obtain some fresh Intel paste.
ReplyDeleteI don't understand why the test would be done this way? If you had at least 4 old CPU's, why not just leave two untouched and install the high-performance paste on the other two so you could eliminate the wild variable of reapplying alcohol soaked old paste?
ReplyDeleteYou should have compared reused Intel paste with original application Intel paste first to prove that it works the same when scraped off and re-applied.
ReplyDeleteLinus is an idiot. He should have used an OIL based solution NOT rubbing alcohol to dissolve the paste.
ReplyDeleteThought u gonna test the chips get average, delid change paste test again. but nope.exe this is aids rofl
ReplyDeleteThis is the dumbest video youve made you should be embarrassed
ReplyDeleteIts like comparin 2-3 years old scrapped thermal compound to 1 hours thermal compound, fucking stupid.
ReplyDeletewhich delid tool are you using?
ReplyDeleteIf you want a sample of fresh Intel Thermal compound you can get 5 or 6 of the stock heat sinks and scrape the fresh compound off of them it comes stuck to the bottom of all of them . I'm sure Intel only stocks 1 kind of thermal paste corporations like to be efficient and save all the money they can
ReplyDeleteI just did this (thermal paste harvesting and re-use). Isopropyl alcohol doesn't work. Ethyl alcohol doesn't work. Acetone doesn't work. Beacon Solvent and Thinner 4 oz. (yes, NewEgg) works perfectly. I think it's Heptane based.
ReplyDelete