[POLL] EPIC SHOOTOUT MATCH!!!! AS5 vs NT-H1 on Thermalright Ultra-120 vs. EVGA ACX!!!

[soccerballtux]

The total amperage on a pair of fans connected to a single fan port cannot exceed what the port is spec'd to provide, but you can wire two identical fans in parallel and monitor one of them. You can also purchase a $5 3-pin fan-splitter that does the same thing without the tedium of soldering.

I've done this before, and it's sucked. Badly I'm not sure why, because electrically the load should just look cut in half :hmm:

 

[DrMrLordX]

yes, I've seen that before and did that for the nth1.

also, holy server fan noise batman! the included fan is terribad worse than the one I already have!!!

Good thing you have the Ultra Kaze then . . .

is there any way to tell how static pressure will be by the design?

Generally speaking, 38mm-thick fans will produce more static pressure than 25mm-thick fans. Finding 25mm fans that produce good static pressure is really hard. The best one I found years ago when I got fans for my nh-d14 was a Delta that's ~100 cfm and only puts up maybe 6.5mmH2O. It makes about as much noise as an Ultra Kaze 133cfm fan @ 12v (Ultra Kaze = ~8mmH2O static pressure).

I had to get at least one 25mm fan, since that's the only thickness that fits between the towers. But I digress.

It usually comes down to thickness, blade design, and RPMs.

 

[soccerballtux]

hm, this is frustrating. It's been sorta rough to get this NT-H1 to consistently perform better than the AS5. Right now AS5 is doing just as well, and I was confident in the NT-H1 application.

It could be because the AC is actively running. Tough to get all the variables. I'll be re-testing NT-H1 again as I move some fans around.

another variable: previously applied AS5 that was doing 10C worse than NT-H1 was probably an 8 year old tube. New stuff that's doing great is ... new

and one more variable: I cleaned up a bit of cable routing, but it's hard for me to accept that as being ~2.5C improvement (which my most recent NT-H1 vs AS5 run approximately showed).

 

[soccerballtux]

Good thing you have the Ultra Kaze then . . .



Generally speaking, 38mm-thick fans will produce more static pressure than 25mm-thick fans. Finding 25mm fans that produce good static pressure is really hard. The best one I found years ago when I got fans for my nh-d14 was a Delta that's ~100 cfm and only puts up maybe 6.5mmH2O. It makes about as much noise as an Ultra Kaze 133cfm fan @ 12v (Ultra Kaze = ~8mmH2O static pressure).

I had to get at least one 25mm fan, since that's the only thickness that fits between the towers. But I digress.

It usually comes down to thickness, blade design, and RPMs.

what I mean is I think blade pitch is going to have a sizable impact, at least in pushing. In free air all you have to consider is CFM, but at pressure some gulp and swallow better-- the sickle shaped blades in particular I think might be better at this

 

[ClockHound]

and one more variable: I cleaned up a bit of cable routing, but it's hard for me to accept that as being ~2.5C improvement (which my most recent NT-H1 vs AS5 run approximately showed).

I can. Intake airflow is probably as critical as the cooler and its fans. In many cases (pun alert) the cooler intake temp is 10 degrees higher than ambient.

With some tweaking, cable wrangling and optimal intake fan placement, can achieve a 2-3 degree gradient between ambient and cooler intake temp.

A cheap indoor/outdoor thermometer with a probe is a great test device.

But, yeah...sometimes optimizing air cooling is a big bag of variables mixed with unrealized expectations, capped with diminishing returns.

 

[BonzaiDuck]

I can. Intake airflow is probably as critical as the cooler and its fans. In many cases (pun alert) the cooler intake temp is 10 degrees higher than ambient.

With some tweaking, cable wrangling and optimal intake fan placement, can achieve a 2-3 degree gradient between ambient and cooler intake temp.

A cheap indoor/outdoor thermometer with a probe is a great test device.

But, yeah...sometimes optimizing air cooling is a big bag of variables mixed with unrealized expectations, capped with diminishing returns.

I can be corrected, but I think he was disappointed in the relative performance of the TIMs. But I'm not surprised: the temperature spread is consistent with published reviews.

As to issues about ambient, cases with poor ventilation or other shortcomings, those issues wouldn't affect the temperature difference of two TIMs using the same processor and cooler. Only their absolute temperature values would be different from a case with better airflow, etc.

On the topic of CPU and/or ported exhaust fans: I've used a 140mm PWM fan between the back side of a heatpipe tower and a 120mm case vent with no modification but use of a clear-blue plastic "fan-adapter." Even if the rated static pressure of this or that fan is lower, you can still achieve the same airflow at lower RPMs with the proper attention. If there's a decent gap between the fan on both sides, you alleviate the infamous "radiator-dead-spot" or limits to venting air because some holes are blocked.

These things are all tedious pursuits, but they are options. My 140mm Akasa Viper did an equal job without the potential noise problem of my AP-30. Being curious about the AP-30, I also rid myself of a clumsy ducting construction that was difficult to remove. But those problems would have been easy to eliminate, even so.

 

[soccerballtux]

time of day makes a significant impact on temperature.

ICD7 is no better than AS5. NTH1 MIGHT be 0.5-1degree better than AS5, but it might have just been time of day and wall-HVAC-thermometer-drift in this hot summer sun.

NTH1 _does_ seem to perform better with lapped surfaces. I do not think the difference can be chalked up to a room temp.

I need to decide if I want to lap the cooler. It's got a very rough, grainy surface that I think might be holding back NT-H1's performance (ICD7 pictured):

 

[DrMrLordX]

You could probably apply less paste and get better performance by using little lines running along the aluminum separating the heatpipes instead of running one big line perpendicular to the heatpipes.

 

[BonzaiDuck]

time of day makes a significant impact on temperature.

ICD7 is no better than AS5. NTH1 MIGHT be 0.5-1degree better than AS5, but it might have just been time of day and wall-HVAC-thermometer-drift in this hot summer sun.

NTH1 _does_ seem to perform better with lapped surfaces. I do not think the difference can be chalked up to a room temp.

I need to decide if I want to lap the cooler. It's got a very rough, grainy surface that I think might be holding back NT-H1's performance (ICD7 pictured):

1st off -- lap the freakin' cooler base. You don't need to make it "shiny." I usually start with 240-grit wet-or-dry, with the grit becoming "more refined" with use. Then I use 400-grit. For some TIMs, scratches in the surface may actually improve the performance, but you likely couldn't measure it. Lapping the base on that cooler shouldn't void the warranty on it -- as if it would ever matter with heatpipe-MTBF = 1 million years.

2nd. If you're showing no difference in the ICD and NT-H1 or AS5 performance, the inconsistency of your result with published or first-hand observations is easily explained by your very sharp photo (a thank you follows later.)

I don't care what Noctua says, nor can I remember. But even their directions -- no less those of folks who've tried it and think it's too unwieldy -- erase any doubt that just putting a blob or a line of the paste on the center of the cooler won't spread it in any optimal way. Micronized diamond is an abrasive; it will form its own "locked matrix" even with additional layers of diamond piled on a hypothetical single layer, but that's not so much the problem in your case. You may not have used an appropriate amount of the paste; what I see there seems grossly inadequate for the surface area. Whatever amount is used, and howsoever you take pains to make a thin application on a nice flat surface, you want as much as possible of the IHS copper and heatpipe-base to make contact with diamond particles in between like an Oreo-filling.

A credit-card used as a spreader is adequate. I recommend a rectangular razor-blade of the traditional design, but even the trapezoidal shop-grade razor-blades will suffice. (Just exercise a bit of care. If you're epileptic or have some nervous disease with spasms or shaking of the hands, get someone else to do it while you "direct" the operation.)

About the photo. It either shows "gaps" between the aluminum and copper pipes, or it shows a "shadow" if the copper pipes with aluminum splines form a consistently uneven surface.

This is important to ME -- because I'm contemplating use of IX on the ACX cooler, and other pictures I've seen suggest no gaps or channels between the pipes and splines. IX guides discourage use with direct-touch coolers which have apparent grooves or channels between the base surfaces and the pipes, or even between the pipes themselves if they lay side by side as with the 212 EVO.

So if I've convinced you to do the work and lap the base, I'd like to see a new photo -- equally as clear and sharp. IF it doesn't come out that way or you simply don't want to do it, you can just tell me after you've inspected the lapping result.

You could just as easily spread ICD on the processor cap before dropping it in the socket and latching it. It's important to avoid "tearing" the layer; an oval of it can come loose if you move the razor too quickly to spread it. Personally, I "guesstimate" the position of the processor on the bottom of the HSF base before securing it, and I spread another square of diamond paste on the base surface. It can be a really fine layer on either IHS or HSF-base, but it must be continuous, and you cannot rely on it to spread itself unless you use copious amounts.

 

[crashtech]

The heat pipes can shift in their block, too. I literally had to take a mill file to one, I think it was a Hyper 212, to get the surface level enough to be smoothed out.

 

[BonzaiDuck]

The heat pipes can shift in their block, too. I literally had to take a mill file to one, I think it was a Hyper 212, to get the surface level enough to be smoothed out.

The H212 coolers were good for the price and performance. But if my aim is overclocking -- even with heatpipes -- I'll pick something different. [EDIT:] You can probably get good OC's with a 212 EVO, but it falls short of other coolers I know of: specifically the D14 and the ACX.

With the ACX and perhaps other coolers, the pipes won't shift in their base mounting. Last I looked, it appeared as though the aluminum splines of the base follow the contour of the pipes and widen to full coverage on the application surface.

But I wouldn't apply any heatpipe cooler without lapping the base. On the nickel-plate units, that might void a warranty, but I doubt so for the ACX. The cooler can only benefit from better contact with the processor over the widest surface area of the IHS -- uninterrupted by gaps or channels. It can only benefit if the nickel-plate of a different cooler is lapped off; no less for the processor cap. Either way, you void a warranty and possibly depress resale value (while some minority of enthusiasts would think you saved them some work.) But lapping will not affect the MTBF and longevity of either item.

 

[soccerballtux]

1st off -- lap the freakin' cooler base. You don't need to make it "shiny." I usually start with 240-grit wet-or-dry, with the grit becoming "more refined" with use. Then I use 400-grit. For some TIMs, scratches in the surface may actually improve the performance, but you likely couldn't measure it. Lapping the base on that cooler shouldn't void the warranty on it -- as if it would ever matter with heatpipe-MTBF = 1 million years.

2nd. If you're showing no difference in the ICD and NT-H1 or AS5 performance, the inconsistency of your result with published or first-hand observations is easily explained by your very sharp photo (a thank you follows later.)

I don't care what Noctua says, nor can I remember. But even their directions -- no less those of folks who've tried it and think it's too unwieldy -- erase any doubt that just putting a blob or a line of the paste on the center of the cooler won't spread it in any optimal way. Micronized diamond is an abrasive; it will form its own "locked matrix" even with additional layers of diamond piled on a hypothetical single layer, but that's not so much the problem in your case. You may not have used an appropriate amount of the paste; what I see there seems grossly inadequate for the surface area. Whatever amount is used, and howsoever you take pains to make a thin application on a nice flat surface, you want as much as possible of the IHS copper and heatpipe-base to make contact with diamond particles in between like an Oreo-filling.

A credit-card used as a spreader is adequate. I recommend a rectangular razor-blade of the traditional design, but even the trapezoidal shop-grade razor-blades will suffice. (Just exercise a bit of care. If you're epileptic or have some nervous disease with spasms or shaking of the hands, get someone else to do it while you "direct" the operation.)

About the photo. It either shows "gaps" between the aluminum and copper pipes, or it shows a "shadow" if the copper pipes with aluminum splines form a consistently uneven surface.

This is important to ME -- because I'm contemplating use of IX on the ACX cooler, and other pictures I've seen suggest no gaps or channels between the pipes and splines. IX guides discourage use with direct-touch coolers which have apparent grooves or channels between the base surfaces and the pipes, or even between the pipes themselves if they lay side by side as with the 212 EVO.

So if I've convinced you to do the work and lap the base, I'd like to see a new photo -- equally as clear and sharp. IF it doesn't come out that way or you simply don't want to do it, you can just tell me after you've inspected the lapping result.

You could just as easily spread ICD on the processor cap before dropping it in the socket and latching it. It's important to avoid "tearing" the layer; an oval of it can come loose if you move the razor too quickly to spread it. Personally, I "guesstimate" the position of the processor on the bottom of the HSF base before securing it, and I spread another square of diamond paste on the base surface. It can be a really fine layer on either IHS or HSF-base, but it must be continuous, and you cannot rely on it to spread itself unless you use copious amounts.

you can see in post #17 how the 'blob' did NOT spread nicely.

 

[BonzaiDuck]

you can see in post #17 how the 'blob' did NOT spread nicely.

You mean post #19? That was your AS5 test? Or was it NT_H1?

I just don't leave it for the pea, rice-grain or any other method to cover the IHS. I spread the stuff for full coverage -- whatever the TIM.

You're last pics for the ICD just didn't inspire my confidence for coverage. A rougher surface like the ACX cooler base (the particular unit you photographed) would make it imperative to use even more ICD, but the only way to be sure is to cover the IHS with it. Like I said, it can be very thin, but it has to cover the entire IHS and the HSF-base mating area.

For whatever TIM, any unevenness in either the base or the IHS may degrade performance. Usually, we assume the IHS is "flat enough," and we just as likely think so for the HSF base.

Nickel-plate on the IHS may inhibit performance by a C degree -- just ranking lower than copper for thermal conductivity.

But I'd lap the base of the ACX until it felt flat and smooth, with the two metals forming a consistent surface. And I'd spread the ICD for full, even coverage. Priority would be to spread it thinly and evenly without "holes" or tearing on the IHS/processor-cap.

 

[BonzaiDuck]

you can see in post #17 how the 'blob' did NOT spread nicely.

I made a mistake in my post -- which you answered. I said "Noctua" when I meant "Innovation Cooling" (ICD).

Even THEY think that heatsink clamping pressure is sufficient, and it may be. However, look at the photos in their application summary:


http://www.innovationcooling.com/application.html

There was enough paste in the 5.5mm-dia. blob for full IHS coverage.

It just doesn't look to me that your line-of-paste method is going to work on that cooler for expecting complete coverage or an even spread.

If you want to USE that cooler, I'd lap it to assure flatness and elimination of the roughness or even any discontinuity between the pipes and the base. If you only want to test it with these TIMs "for the testing" -- do as you please.

Even so, if you get the coverage shown in the Innovation Cooling photos, it should be "almost as good." Think of the processor's IHS as the cake, and the TIM as the icing. You may spread the icing thin, but you're not going to deliver the cake to the guests with the cake visible in patches on top.

You'll lose your job as pastry chef!

 

[DrMrLordX]

Here's a decent video showing proper application of TIM to a HDT HSF:

https://www.youtube.com/watch?v=IG_GfSTKQ98

Note that the HSF featured in the video has gaps between the aluminum dividers and the heatpipes, which have been prefilled with TIM before applying the main "lines" on the dividers themselves. In the case of the ACX, that should be unnecessary. All you would need there are the little lines on the dividers.

I second the recommendation of lapping if you have the patience for it (and yeah, stopping at 400 grit is probably a good idea).

 

[BonzaiDuck]

Here's a decent video showing proper application of TIM to a HDT HSF:

https://www.youtube.com/watch?v=IG_GfSTKQ98

Note that the HSF featured in the video has gaps between the aluminum dividers and the heatpipes, which have been prefilled with TIM before applying the main "lines" on the dividers themselves. In the case of the ACX, that should be unnecessary. All you would need there are the little lines on the dividers.

I second the recommendation of lapping if you have the patience for it (and yeah, stopping at 400 grit is probably a good idea).

I'm a crusty, procrastinating old fart. I try to avoid needing to repeat the simplest things more than once. I'll probably sit around for weeks rehearsing in my mind what I intend to do for installation of Indigo Xtreme and the heat-up/re-flow step. It's the old military platitude for basic training or OTS: "Do it right the first time!" I'll take the time planning the "first time."

Soccerball, if you're going to take the time to test all or at least a couple TIMs and two coolers, you should have the patience to lap the base of the ACX. I'd even venture a guess that you could stop at 320-grit wet-or-dri sandpaper, if you grind the cooler base enough to degrade the grit so it's beginning to abrade like 400-grit. And I'd say, for that cooler base, 240-grit will get down to business of leveling the base.

Use a few drops of water to lubricate the wet-or-dri, hold the cooler by the base (not the fins or pipes.) The best surface to use is a glass table-top.

Tape the sandpaper to the glass with scotch-tape loops affixed to the paper side, so that it lies flat and secure. Use a firm, circular motion; avoid any "stuttering" as the base moves across the sandpaper. Turn on your favorite TV cartoon show. The lapping might take a half-hour, and you'll want to reduce the boredom.

I say -- for the $55 investment in that cooler, it's worth the trouble. Same with getting total coverage with the TIM you use.

Except for the testing regimen you've chosen, you would only perform all these actions once. You shouldn't need to do any of the muscle-work after that for the life of the cooler or the computer.

 

[soccerballtux]

You mean post #19? That was your AS5 test? Or was it NT_H1?

I just don't leave it for the pea, rice-grain or any other method to cover the IHS. I spread the stuff for full coverage -- whatever the TIM.

You're last pics for the ICD just didn't inspire my confidence for coverage. A rougher surface like the ACX cooler base (the particular unit you photographed) would make it imperative to use even more ICD, but the only way to be sure is to cover the IHS with it. Like I said, it can be very thin, but it has to cover the entire IHS and the HSF-base mating area.

For whatever TIM, any unevenness in either the base or the IHS may degrade performance. Usually, we assume the IHS is "flat enough," and we just as likely think so for the HSF base.

Nickel-plate on the IHS may inhibit performance by a C degree -- just ranking lower than copper for thermal conductivity.

But I'd lap the base of the ACX until it felt flat and smooth, with the two metals forming a consistent surface. And I'd spread the ICD for full, even coverage. Priority would be to spread it thinly and evenly without "holes" or tearing on the IHS/processor-cap.

sorry, yes. that was AS5. Did not spread like on the lapped TR-120. That was when i switched to drawing a line.

I have some other photos but most of them spread like this:

this was after I started using the line method (because of the AS5 shot in post #19).

even then, covering the entire IHS is not imperative considering only a portion of the middle matters:

I do think lapping will be in order. Looking at the footprint of the HSF and TIM, I think the disappointing rough cut that the ACX came with probably gave room for micropockets of air.

 

[soccerballtux]

Here's a decent video showing proper application of TIM to a HDT HSF:

https://www.youtube.com/watch?v=IG_GfSTKQ98

Note that the HSF featured in the video has gaps between the aluminum dividers and the heatpipes, which have been prefilled with TIM before applying the main "lines" on the dividers themselves. In the case of the ACX, that should be unnecessary. All you would need there are the little lines on the dividers.

I second the recommendation of lapping if you have the patience for it (and yeah, stopping at 400 grit is probably a good idea).

why would I stop at 400? if I'm going to do it might as well keep going!!! no? I've got the grit for it.

I still have leftover paper from my TR-120 lapping in 2009. Considering the difference it seemed to make for NT-H1 on the TR120, I'm probably motivated enough to, just need a break for a bit.

Surprisingly, throwing the included 3000rpm 100CFM included fan [the noisy server fan one lol] only made about 2C difference, if I recall. Perhaps my 130CFM is good enough? I will re-test to make sure, but it was mostly underwhelming and definitely not worth the noise or expense or effort.

Also, I discovered the cutout in the side of my case was 140MM, not 120MM, so I upgraded that after completing testing with the old config. As I look back, I'm quite surprised at my fan placement decisions some 5 years ago! Mostly I think, I was looking to do it on the cheap.

Finally, I still have a bit of 10 year old AS5 that was performing horribly on the TR-120, I'm going to throw that on the ACX to see if the age is why the temps were 10C worse, or if it was the application.

 

[soccerballtux]

Here's a decent video showing proper application of TIM to a HDT HSF:

https://www.youtube.com/watch?v=IG_GfSTKQ98

Note that the HSF featured in the video has gaps between the aluminum dividers and the heatpipes, which have been prefilled with TIM before applying the main "lines" on the dividers themselves. In the case of the ACX, that should be unnecessary. All you would need there are the little lines on the dividers.

I second the recommendation of lapping if you have the patience for it (and yeah, stopping at 400 grit is probably a good idea).

I guarantee if I spread like that I'm not going to see any difference.

 

[crashtech]

sorry, yes. that was AS5. Did not spread like on the lapped TR-120. That was when i switched to drawing a line.

I have some other photos but most of them spread like this:

From the pattern the base looks rather convex.

 

[DrMrLordX]

why would I stop at 400? if I'm going to do it might as well keep going!!! no? I've got the grit for it.

I read an article years ago, before I first attempting lapping anything (x2-3600+ Brisbane yeehaw), I did some research. Turned out that a guy who used to do in-lab lapping of quartz surfaces (I think) for the purpose of keeping them cool with a heatsink found that lapping to too fine a grit actually impeded performance of TIM by significantly reducing the effective surface area of the mated surfaces. One-dimensional heat transfer from a hot to cold surface through an interface is based on thermal conductivity of the interface, temperature delta between the two surfaces, surface area of the two surfaces, and thickness of the interface. Obviously having too rough of a surface is going to increase thickness of the interface to an unacceptable point.

The article claimed that grits beyond 800 impeded performance. I used to lap to 800, but now I just stop at 400 for numerous reasons, the main ones being that 400 is as high as I need to go to assure flatness of the surface (not counting microfissures and other irregularities that you actually want when using TIM) and that I've gotten a bit lazy about lapping. But that's besides the point.

The main benefits of lapping are getting rid of nickel coatings and to make sure that the exposed copper surface is uniform, at least in a macroscopic sense.

 

[BonzaiDuck]

I was slightly unsettled by the Innovative Cooling "Application" page of their web-site.

It seemed that the common wisdom when ICD was released advised applying a thin but consistent layer with credit-card or razor-blade. IC says they "discourage" that method.

However, I can see the advantage of dropping the 6mm-dia. blob dead center on the IHS, setting and securing the heatsink after that. There's not likely to be either air-pockets or "tearing" of the layer -- which causes the same air-pocket problem.

On the matter of the IHS, since copper conducts heat in all directions, I'd think there's only an advantage to covering the entire IHS, even if the size of the silicon die is smaller.

 

[soccerballtux]

I read an article years ago, before I first attempting lapping anything (x2-3600+ Brisbane yeehaw), I did some research. Turned out that a guy who used to do in-lab lapping of quartz surfaces (I think) for the purpose of keeping them cool with a heatsink found that lapping to too fine a grit actually impeded performance of TIM by significantly reducing the effective surface area of the mated surfaces. One-dimensional heat transfer from a hot to cold surface through an interface is based on thermal conductivity of the interface, temperature delta between the two surfaces, surface area of the two surfaces, and thickness of the interface. Obviously having too rough of a surface is going to increase thickness of the interface to an unacceptable point.

The article claimed that grits beyond 800 impeded performance. I used to lap to 800, but now I just stop at 400 for numerous reasons, the main ones being that 400 is as high as I need to go to assure flatness of the surface (not counting microfissures and other irregularities that you actually want when using TIM) and that I've gotten a bit lazy about lapping. But that's besides the point.

The main benefits of lapping are getting rid of nickel coatings and to make sure that the exposed copper surface is uniform, at least in a macroscopic sense.

that makes complete sense, I remember in 2008 I was asking 'why would you lap to a perfect mirror finish, you're losing 50% surface area by doing that!' no one had an answer....

at any rate nickel's thermal conductivity is still pretty high, I don't see a reason to remove it

 

[soccerballtux]

I was slightly unsettled by the Innovative Cooling "Application" page of their web-site.

It seemed that the common wisdom when ICD was released advised applying a thin but consistent layer with credit-card or razor-blade. IC says they "discourage" that method.

However, I can see the advantage of dropping the 6mm-dia. blob dead center on the IHS, setting and securing the heatsink after that. There's not likely to be either air-pockets or "tearing" of the layer -- which causes the same air-pocket problem.

On the matter of the IHS, since copper conducts heat in all directions, I'd think there's only an advantage to covering the entire IHS, even if the size of the silicon die is smaller.

the thermal conduction through that thin layer on the IHS past the edge of the die is going to be so miniscule it's not worth it IMO

 

[BonzaiDuck]

the thermal conduction through that thin layer on the IHS past the edge of the die is going to be so miniscule it's not worth it IMO

Look at it another way. If you're testing TIMs and heatsinks, you want to ASSURE a consistent test-bed. If the TIM's are spread inconsistently, you can speculate about the IHS area not in direct contact with the silicon die, but you can't be sure for comparing apples to apples.

The silicon-to-copper interface is a bottleneck or limit based on the area of the die itself. But the entire IHS will absorb heat, and the better surface contact with the heatsink can only improve conduction.