TIM Testing Going Forward

FrgMstr

FrgMstr

Just Plain Mean
Staff member
Joined
May 18, 1997
Messages
55,602
One of the projects that I never got time to start, was a TIM test. As you might guess, this type of testing is extremely time consuming.

Today I am shipping rgMekanic the 20 TIMs I purchased for the test.

When I closed HardOCP, I left Rich with all the testing equipment (as I did with most editors), so he has the hardware needed to pull off this huge project. The ball is now in your court, Rich! Godspeed!

Below is a pic of what is coming your way this week. Also, there is a package of 24K gold leaf as well. I thought it would be fun to throw into the mix. Just don't pawn it, is all I ask. ;) It is edible however should you need a snack.

IMG_20190416_112338.jpg
 
Here's something for The Director of Enthusiast Engagement to ponder:

Seems like every time Intel releases a new CPU gen, there are howls of derision from the enthusiast crowd that the TIM under the IHS ain't up to par. While some gens of the X series Enthusiast-grade processors are (were) soldered, it still doesn't always seem to be enough.

Take the i9-7980XE (or any of the 79XX series). They weren't soldered and de-lidding/applying liquid metal TIM seemed to really help the thermal performance. Then the i9-9980XE comes along, soldered - but even with solder, it seems like a properly de-lidded/LM TIM/re-lidded 7980XE will run cooler/OC higher on cooling systems less than LN2. Seems like I also remember something about Intel not being too keen on soldering these huge dies, different thermal expansion between the die and the solder might lead to cracking the die IIRC.

I'm sure I'm not the first to bring this to their attention, and I realize that it just won't be feasible to explore on the consumer grade chips, but seems like for the $2K enthusiast chips they could find something that would perform as well as a delidded CPU with LM TIM where customers don't have to resort to delidding/voiding warranty on these multi-thousand dollar CPUs. I admit there's a bit of adrenaline rush associated with popping the lid on a $2K chip, but I guess part of that is because I didn't ruin the thing.
 
Dullard said:
Here's something for The Director of Enthusiast Engagement to ponder:

Seems like every time Intel releases a new CPU gen, there are howls of derision from the enthusiast crowd that the TIM under the IHS ain't up to par. While some gens of the X series Enthusiast-grade processors are (were) soldered, it still doesn't always seem to be enough.

Take the i9-7980XE (or any of the 79XX series). They weren't soldered and de-lidding/applying liquid metal TIM seemed to really help the thermal performance. Then the i9-9980XE comes along, soldered - but even with solder, it seems like a properly de-lidded/LM TIM/re-lidded 7980XE will run cooler/OC higher on cooling systems less than LN2. Seems like I also remember something about Intel not being too keen on soldering these huge dies, different thermal expansion between the die and the solder might lead to cracking the die IIRC.

I'm sure I'm not the first to bring this to their attention, and I realize that it just won't be feasible to explore on the consumer grade chips, but seems like for the $2K enthusiast chips they could find something that would perform as well as a delidded CPU with LM TIM where customers don't have to resort to delidding/voiding warranty on these multi-thousand dollar CPUs. I admit there's a bit of adrenaline rush associated with popping the lid on a $2K chip, but I guess part of that is because I didn't ruin the thing.
Click to expand...

I've made the point several times that Intel's overclocking headroom was far ahead of AMD's despite the inferior TIM. Intel was never concerned about giving you even more overclocking headroom on CPU's that could already go a 800MHz-1GHz higher than their stock settings allow for. The TIM as they saw it saved them some money and it was good enough. Intel doesn't seem to fully understand the enthusiast market, much less know how to cater to it. It also seems unlikely from a business standpoint that they would cater (IE spend more money) to what amounts to a fraction of that market.
 
Dan_D said:
I've made the point several times that Intel's overclocking headroom was far ahead of AMD's despite the inferior TIM. Intel was never concerned about giving you even more overclocking headroom on CPU's that could already go a 800MHz-1GHz higher than their stock settings allow for. The TIM as they saw it saved them some money and it was good enough. Intel doesn't seem to fully understand the enthusiast market, much less know how to cater to it. It also seems unlikely from a business standpoint that they would cater (IE spend more money) to what amounts to a fraction of that market.
Click to expand...
Maybe ASUS can buy Intel's dies, fit a tiny AIO on them, and resell them for $750 more, with integrated RGB.
 
Honestly, properly delidding, cleaning and applying liquid metal TIM will make ANY CPU with a heat spreader have better thermal performance.

This is because the thermal performance of fresh liquid metal is superior to the indium solder AMD and (sometimes) Intel use. The problem is that liquid metal dries and needs to be periodically cleaned and re-applied to maintain that advantage. The indium solder is apply-once-at-factory-and-forget-about-it, and still performs better than any paste (liquid metal not being paste).

Edit: Edited to say I hope Rich runs with it and am looking forward to the results should he do so. A good TIM roundup has been sorely lacking. For the vast majority of my enthusiast career I have been using Arctic Silver 5. When I went looking last year to see how much better newer TIMs may be, I saw a lot of recommendations, but no real solid data showing the actual differences (initial and over time) between the various pastes out there.

If he doesn't run the testing, well, he'll at least have enough TIM to last a few years :)
 
Last edited:
Dullard said:
Here's something for The Director of Enthusiast Engagement to ponder:

Seems like every time Intel releases a new CPU gen, there are howls of derision from the enthusiast crowd that the TIM under the IHS ain't up to par. While some gens of the X series Enthusiast-grade processors are (were) soldered, it still doesn't always seem to be enough.

Take the i9-7980XE (or any of the 79XX series). They weren't soldered and de-lidding/applying liquid metal TIM seemed to really help the thermal performance. Then the i9-9980XE comes along, soldered - but even with solder, it seems like a properly de-lidded/LM TIM/re-lidded 7980XE will run cooler/OC higher on cooling systems less than LN2. Seems like I also remember something about Intel not being too keen on soldering these huge dies, different thermal expansion between the die and the solder might lead to cracking the die IIRC.

I'm sure I'm not the first to bring this to their attention, and I realize that it just won't be feasible to explore on the consumer grade chips, but seems like for the $2K enthusiast chips they could find something that would perform as well as a delidded CPU with LM TIM where customers don't have to resort to delidding/voiding warranty on these multi-thousand dollar CPUs. I admit there's a bit of adrenaline rush associated with popping the lid on a $2K chip, but I guess part of that is because I didn't ruin the thing.
Click to expand...
The factory TIM will likely never be as good as what you can get from delidding, and that is just a fact. There are all kinds of factors at play there.

That said, what exactly do you want? I am listening.
 
travm said:
Possibly ignorant comment, but, why do we need these heat spreaders? I don't recall ever having an issue before ihs was a thing. I get it for OEMs and such, but I'm never going to damage a cpu die installing a heatsink. / Looks around, taps wood.
Std practice was always remove hsf pre-installed Tim, watch the verge for how much Tim to apply, apply quality paste Tim and hsf.
I guess power levels have increased by a lot...
Click to expand...
I'd bet it has more to do with avoiding confusion with socket specs and heatsinks - If they offered an "enthusiast" option without an IHS it wouldn't be compatible with motherboards that could fit cpus with an IHS (unless the clamp were removed) and any heatsink compatible with both would need to have two sets of mounting hardware to account for the height difference of the IHS. Probably more complexity and potential for confusion than Intel is interested in.

While I'm sort of with you - I de-lidded my 7700k and thought long and hard about running it bare under water - I'd settle for a move back to solder for all i5 skus and up. I ran my Q6600 bare-die (butane torch, double-edged razor blades, and two coffee cups de-lidding method ;) ) and iirc I only saw a few degrees difference in load temps - less than the difference between crap TIM and AS5 (the "best" I was aware of at the time) and nothing like the difference replacing the crap TIM on my 7700k made.
 
capt_cope said:
I'd bet it has more to do with avoiding confusion with socket specs and heatsinks - If they offered an "enthusiast" option without an IHS it wouldn't be compatible with motherboards that could fit cpus with an IHS (unless the clamp were removed) and any heatsink compatible with both would need to have two sets of mounting hardware to account for the height difference of the IHS. Probably more complexity and potential for confusion than Intel is interested in.

While I'm sort of with you - I de-lidded my 7700k and thought long and hard about running it bare under water - I'd settle for a move back to solder for all i5 skus and up. I ran my Q6600 bare-die (butane torch, double-edged razor blades, and two coffee cups de-lidding method ;) ) and iirc I only saw a few degrees difference in load temps - less than the difference between crap TIM and AS5 (the "best" I was aware of at the time) and nothing like the difference replacing the crap TIM on my 7700k made.
Click to expand...
Oh right, intel.
My AMD kit has always clamped the pins ;)
I never thought about Intel clamps.

Hell, why dont they just build the cpu into the heatsink and solder the damn die right to the copper. Or... Liquid cooled interposer? I'll stop now, I think someone with a more realistic idea should chime in.
 
If there is gonna be funky stuff like gold leaf or cheese slice or mayo, then i think it would be good to have a "no Tim" comparison point ;)
 
You must log in or register to reply here.
Back
Top