Cutting AMD Coolers Open & Flat vs. Vertical CPU Cooler Benchmarks
video description
Date: 2020-05-06
Related videos
Comments and reviews: 10
Dave
Just to confirm what GN is verifying on, Many years ago when I was on OCN discussing about heatpipes and stuff and convection, it was one of the key things that people got wrong but alas, gravity doesn't affect heatpipes in forced convection situations (nor does a top to bottom airflow design help a force convection design, that is simply because of silverstone's large fans that don't suck and plenty of breathing space) Heatpipes aren't purely pipes, they have sintered copper powder in them for the liquid to seep through the side walls creating capillary action. Some people still think they're literally pipes and don't do jack shit Or that phone manufacturers advertise heatpipes as liquid cooling and people believing their BS. however vapor chamber is different as they're pumped with pressure into them, lowering their vapor temperature and being able to reach steam phase faster. Basically pricier heat pipes where heat pipes usually cannot go like say, very high heat flux (modern CPUs really, although coolermaster is making a mess of that) or where the pipe is becoming insanely flat. On top of that, you get much larger contact surface. So that's why the earlier old copper wraith stealth coolers perform pretty good compared to what you were expecting out of them because they have a really big dollop of a vapor chamber in them. In fact for a real world test, put a non heatpiped heatsink on a hair dryer vs one with a heatpipe, you'll quickly find that the one with the heatpipe conducts insanely well compared to one without a heatsink i. e wraith stealth and intel stock heatsinks (and the new aluminium wraith spires ugh) When your amount of aluminium starts to get further and further away from the source in terms of height of width or whatever, the thermal conduction of the material heavily comes into play so you lose efficacy really fast
reply
Just to confirm what GN is verifying on, Many years ago when I was on OCN discussing about heatpipes and stuff and convection, it was one of the key things that people got wrong but alas, gravity doesn't affect heatpipes in forced convection situations (nor does a top to bottom airflow design help a force convection design, that is simply because of silverstone's large fans that don't suck and plenty of breathing space) Heatpipes aren't purely pipes, they have sintered copper powder in them for the liquid to seep through the side walls creating capillary action. Some people still think they're literally pipes and don't do jack shit Or that phone manufacturers advertise heatpipes as liquid cooling and people believing their BS. however vapor chamber is different as they're pumped with pressure into them, lowering their vapor temperature and being able to reach steam phase faster. Basically pricier heat pipes where heat pipes usually cannot go like say, very high heat flux (modern CPUs really, although coolermaster is making a mess of that) or where the pipe is becoming insanely flat. On top of that, you get much larger contact surface. So that's why the earlier old copper wraith stealth coolers perform pretty good compared to what you were expecting out of them because they have a really big dollop of a vapor chamber in them. In fact for a real world test, put a non heatpiped heatsink on a hair dryer vs one with a heatpipe, you'll quickly find that the one with the heatpipe conducts insanely well compared to one without a heatsink i. e wraith stealth and intel stock heatsinks (and the new aluminium wraith spires ugh) When your amount of aluminium starts to get further and further away from the source in terms of height of width or whatever, the thermal conduction of the material heavily comes into play so you lose efficacy really fast
reply
Louis
Hey guys, I noticed that on some of your results, you didn't report error bars on the upside-down position. Not sure if that was a mistake or intentional. I'm a little concerned about the error bars on these tests in general, and where they're coming from. They don't really seem to fluctuate with the underlying values, which is a little odd. If you're just reporting error based on the precision of your measuring device, then some statistical inference would give you more confidence by averaging the error. If you're reporting error bars using a population average and standard deviation, that's probably not the best way to handle the data for a limited number of tests. That potentially discards useful information, and it makes the results very sensitive to population sampling bias. From a UQ perspective, the testing error is an unknown quantity that you would expect to be well-correlated between similar tests, so variance in one of these orientations and locations would inform variance in the other orientations. Using a Bayesian approach and having some prior normal distribution, you can then update it with each new test you run. The more tests of one type you perform, the more certain you can be about what your expected variation is, and the less information you need to collect from your population to validate that.
reply
Hey guys, I noticed that on some of your results, you didn't report error bars on the upside-down position. Not sure if that was a mistake or intentional. I'm a little concerned about the error bars on these tests in general, and where they're coming from. They don't really seem to fluctuate with the underlying values, which is a little odd. If you're just reporting error based on the precision of your measuring device, then some statistical inference would give you more confidence by averaging the error. If you're reporting error bars using a population average and standard deviation, that's probably not the best way to handle the data for a limited number of tests. That potentially discards useful information, and it makes the results very sensitive to population sampling bias. From a UQ perspective, the testing error is an unknown quantity that you would expect to be well-correlated between similar tests, so variance in one of these orientations and locations would inform variance in the other orientations. Using a Bayesian approach and having some prior normal distribution, you can then update it with each new test you run. The more tests of one type you perform, the more certain you can be about what your expected variation is, and the less information you need to collect from your population to validate that.
reply
Mike
I think the key factor that makes a difference in performance is the height distance between the evaporator and the condenser area, and that's where the relative positions between them would make a difference. The reason is that capillary action does have limits under gravity, which means that a long distance between the heat source and the heat sink would then show up in the results, and with a long-enough distance meaning that the heat source being above the heat sink would eventually render it useless. In the applications we're involved in here, where the heat sink is very close to the heat source, and therefore where at the worst, the difference in height is relatively small, with very fine capillary action it's possible that the difference is negligible, as they found.
reply
I think the key factor that makes a difference in performance is the height distance between the evaporator and the condenser area, and that's where the relative positions between them would make a difference. The reason is that capillary action does have limits under gravity, which means that a long distance between the heat source and the heat sink would then show up in the results, and with a long-enough distance meaning that the heat source being above the heat sink would eventually render it useless. In the applications we're involved in here, where the heat sink is very close to the heat source, and therefore where at the worst, the difference in height is relatively small, with very fine capillary action it's possible that the difference is negligible, as they found.
reply
Greg
Science! Just an OT-ish thought. I sometimes wonder if upside-down mounting might interact negatively with board-flex due to mounting pressure. i. e, conceivably, upside-down orientation could ameliorate board-flex-related stress on PCB soldier joints and, consequently, improve the effectiveness of the board as a last-resort heat-sink and board longevity. Probably, H0 ought to be no relationship; but, I wonder if some board/cooler combinations might have stronger correlations, i. e. sockets with bigger packages. Hard thing to test rigorously, I suspect. Regardless, it's nice to learn/confirm that, in the short run, at least, for air CPU cooling, there's little point worrying about orientation! Great work.
reply
Science! Just an OT-ish thought. I sometimes wonder if upside-down mounting might interact negatively with board-flex due to mounting pressure. i. e, conceivably, upside-down orientation could ameliorate board-flex-related stress on PCB soldier joints and, consequently, improve the effectiveness of the board as a last-resort heat-sink and board longevity. Probably, H0 ought to be no relationship; but, I wonder if some board/cooler combinations might have stronger correlations, i. e. sockets with bigger packages. Hard thing to test rigorously, I suspect. Regardless, it's nice to learn/confirm that, in the short run, at least, for air CPU cooling, there's little point worrying about orientation! Great work.
reply
DEEPCOOL
Hi Everyone, This is the thing: We initially used the two processes of sintered and grooved in our lab, but we found that the performance dispersion is relatively large after massive performance testings. So in order to ensure the stock that are all with stable high performance, we adjusted the sintered process with some internal packing powder, vacuum and other parameters, to achieve the same high performance that we targeted for the product, and maintain the stability of each product that we ship to end-users. So every piece of the ASSASSIN III in the market is just with sintered heatpipes. We are terribly sorry for the confusion that we made to our users, thank you! - DeepCool
reply
Hi Everyone, This is the thing: We initially used the two processes of sintered and grooved in our lab, but we found that the performance dispersion is relatively large after massive performance testings. So in order to ensure the stock that are all with stable high performance, we adjusted the sintered process with some internal packing powder, vacuum and other parameters, to achieve the same high performance that we targeted for the product, and maintain the stability of each product that we ship to end-users. So every piece of the ASSASSIN III in the market is just with sintered heatpipes. We are terribly sorry for the confusion that we made to our users, thank you! - DeepCool
reply
salut
ahh teledyne technologies. used to work briefly for one of their subsidiairies here, to help assist with upgrading from win 7 to 10. they knew NOTHING about automation, and I did the most basic of setup things (removing start menu items, pinning shit, setting up the BIOS when they use dell's imaging tool which is more than capable of this) and I got done when the guy told me to replace a redundant power cable on a switch, dude, the cable's already there, redundancy or just leave it because you think the power cable is going to melt because it's not 14 AWG when this thing draws nothing. and they're running windows enterprise, it's not like there's limitations in automation
reply
ahh teledyne technologies. used to work briefly for one of their subsidiairies here, to help assist with upgrading from win 7 to 10. they knew NOTHING about automation, and I did the most basic of setup things (removing start menu items, pinning shit, setting up the BIOS when they use dell's imaging tool which is more than capable of this) and I got done when the guy told me to replace a redundant power cable on a switch, dude, the cable's already there, redundancy or just leave it because you think the power cable is going to melt because it's not 14 AWG when this thing draws nothing. and they're running windows enterprise, it's not like there's limitations in automation
reply
Daharen
Without the raw data, I can't run the numbers myself, but while the variation is within error-margins for each individual set of tests, the upside-down average is so consistently lower across all tests compared to the vertical and flat tests, that it seems possible there would be a statistically significant difference if a collective analysis was done across all of the heatsinks and all of the tests. This is obviously interesting because it's counter-intuitive if true, but I understand not wanting to run a multi-variable analysis. Still I'm left wondering why that would be the case if it is true.
reply
Without the raw data, I can't run the numbers myself, but while the variation is within error-margins for each individual set of tests, the upside-down average is so consistently lower across all tests compared to the vertical and flat tests, that it seems possible there would be a statistically significant difference if a collective analysis was done across all of the heatsinks and all of the tests. This is obviously interesting because it's counter-intuitive if true, but I understand not wanting to run a multi-variable analysis. Still I'm left wondering why that would be the case if it is true.
reply
freddan6fly
At 3: 15 you said testing with the cooler like this and held it the wrong way. When you have gravity (I will sponsor you to space soon, You will get better convection if you have the fins oriented so hot air can rise. It is very important if you are gonna use passive cooling. I had a computer in my bedroom with a 35W intel I5-4590T and passive cooler and fanless PSU. The passive cooler did not work unless the fins let air pass vertically. And the case had to have big openings at top and bottom. It was the only way to get a dead silent PC in the bedroom.
reply
At 3: 15 you said testing with the cooler like this and held it the wrong way. When you have gravity (I will sponsor you to space soon, You will get better convection if you have the fins oriented so hot air can rise. It is very important if you are gonna use passive cooling. I had a computer in my bedroom with a 35W intel I5-4590T and passive cooler and fanless PSU. The passive cooler did not work unless the fins let air pass vertically. And the case had to have big openings at top and bottom. It was the only way to get a dead silent PC in the bedroom.
reply
Secto
The liquid at the block vaporises and the pressure forces the rest of the liquid to disperse evenly across the entire surface area. This means there is a continual 'stream' where the gas is condensing, and the liquid is forced along the surface area to the block, where it vaporises again. As such the orientation doesn't matter, as the momentum is driven by the pressure, not by gravity. There would have to be a huge pressure differential for gravity to matter. What Steve should do is run a cooler with a progressive number of heat pipes cut open!
reply
The liquid at the block vaporises and the pressure forces the rest of the liquid to disperse evenly across the entire surface area. This means there is a continual 'stream' where the gas is condensing, and the liquid is forced along the surface area to the block, where it vaporises again. As such the orientation doesn't matter, as the momentum is driven by the pressure, not by gravity. There would have to be a huge pressure differential for gravity to matter. What Steve should do is run a cooler with a progressive number of heat pipes cut open!
reply
P8qzxnxfP85xZ2H3wDRV
The reason there was no difference is that even in upside-down orientation, you didn't reach the heat transfer limit of the heatpipes. If you further increase the wattage, you'll see a sudden breakpoint, which depends on orientation. This does not matter for computer-end-users though, because they will never reach those wattages and even if we're talking about overclocking, it doesn't make sense, because the condenser on the cooler is too small anyway.
reply
The reason there was no difference is that even in upside-down orientation, you didn't reach the heat transfer limit of the heatpipes. If you further increase the wattage, you'll see a sudden breakpoint, which depends on orientation. This does not matter for computer-end-users though, because they will never reach those wattages and even if we're talking about overclocking, it doesn't make sense, because the condenser on the cooler is too small anyway.
reply
Add a review, comment
Other channel videos
