Gadgetory

how frequently should I replace liquid metal is probably one of the most common questions we get liquid metal is applied between the CPU die and the IHS or integrated heat spreader to improve thermal conductivity from the silicon to the heat spreader but there hasn't been much long-term testing on liquid metal endurance versus aid cracking and drying are some of the most common concerns from commenters leading users to wonder whether liquid metal performance will fall off a cliff at some point due to age one of our test benches has been running thermal endurance cycling tests for the last year now since September of 2017 just to see if there's been any aging that's impacted performance before that this video is brought to you by us and the gamers Nexus store at store gamers Nexus net you can pick up one of our mod mats now on backorder because they keep selling out or you can grab one of our other products like the Raglan two-tone hoodie which we have in stock or the full gamers Nexus logo t-shirt that we just restocked in cotton and tri-blend flirtin more at store gamers Nexus net so this is a case study of liquid metal and dealin and you've all heard of deleting you've probably seen our dewetting tests we tested the I 979 60 X about a year ago and deleted it at the time using liquid metal to fill the gaps one thing we didn't do is remove the silicone adhesive around it we found later that that improves temperature a bit more but we left it there as a guide and then let it run for a year or so with liquid metal on it using thermal grisly conduct or not there are multiple different types of liquid metal we didn't test all of them that's one of the big things here again it's a case study we are testing with a sample size of one so we can draw some conclusions from this but it is by no means a sweeping result that you can apply to all liquid metals on all platforms ever so keep that in mind this is an experiment that is overall difficult to conduct and it's even harder to do with lots of awesome with enough platforms to build a significant sample size I mean you'd need like 100 systems plugged in for a year doing that and we don't have the capacity for that so so we've got some long-term testing on this bench next to me here and the platform is using an Intel I 970 960 X that's been overclocked to 4.6 gigahertz it's running at one point to two volts on a gigabyte gaming nine motherboard so this is a 16 core CPU that's overclocked and locked to four point six Q Hertz one point to two volts as noted all of the over current protections and other power limitations have been disabled and the 79 60 X has been deleted and is using thermal grisly conduct and not as it's liquid metal with leftover silicone adhesive still in place on the double substrate we use the thermal grisly cryo not for the thermal paste between the IHS and our cooler but also did some tests previously with some ACE attack stock pastes we used the higher conductivity pace here just to make sure it's not a limitation or bottleneck in testing we also use the G scale Triton C black memory at thirty six hundred megahertz across four channels which does impact CPU thermals because the IMC is inside of the die and the power supply is an ax 1600 I in single rail mode which is important because we're pushing over 40 amps and some of our testing which means that multi rail mode would trigger OCP and cause a shutdown so for testing the system ran FFT workloads with automated cycling for the past year rather than thermally strain the CPU 24/7 we did cycling so this is something that we learned a long time ago if you want to simulate aged on any kind of electronics or any kind of thermal interface it's better to cycle between hot and cool down or ambient temperature than it is to just blast it 24/7 because you put the component or the in this case the Tim through expansion and contraction periods which significantly impacts how it ages as opposed to just running constantly at a high heat you're putting it through more cycles more thermal cycles and that is more abusive so that's what we did to simulate age we're rapidly over the past year and in addition to that we alternated loads with break periods so it's more similar to a real use case in that way and then for final thermal results rather than using FFT as we used blender because it's been consistently reliable for a load generation for this kind of testing so we as blender for final results both before and after the liquid metal aging and then used F of T is for the thermal cycling and between and the blender workload is an in-house test we use for CPU benchmarking so it's the same one we've used for benchmarks elsewhere and additionally initial testing was done before the aging process and letting it age so we did testing at the beginning we did testing after with no changes at all even the thermal paste remained the same nothing changed we did more tests after but changed the thermal paste between the IHS and the cold plate of the cooler and then we did tests with new liquid metal which is where the interest really is today and then all these tests were conducted multiple times for parody averaged and we dropped some of the outlier results that will be discussed at the conclusion there are a couple of limitations of testing here so we'll just put those out there to start with for one we're only using one sample it's not globally conclusive of all liquid metals we only tested one the most popular one and we also don't know how this will change with different components different dyes for example if you are using a smaller dye versus larger so it's limited in that regard but we can draw some conclusions just not hard all encompassing inclusions that you might if you had a huge sample size with lots of variation between parts tested so it gives us a great baseline for a case study and the liquid metal it's not the only thing that ages the thermal paste ages as well and that's why we have tests before after and test with new paste and old liquid metal new paste a new liquid metal and then also keep in mind that the cooler although unlikely could have had some aging itself it could have had some permeation of liquid in the tubes that's something we can't really account for for in any meaningful way and that should more or less cover it there are lots of things that can change in a year with a test platform that's undergoing thermal aging so we did our best to control but just keep that in mind it's these are not results you should be plastering all over the internet saying that liquid metal never ages or always ages it's results that you should say here's a case study where they tested one and they found these numbers so keep all that in mind let's get into the results for results we measured the original application at 62 point nine degrees Celsius delta T over ambient and that's data from one year ago today without changing anything we measured their all performance as sixty three point three degrees Celsius over ambience and this is well within our normal plus or minus one degrees Celsius error margin for testing and given that we have a whole lot more variables with a one year gap between tests then normally these numbers are close enough that we can't state any difference and again potential for changes or error are introduced just in the software testing as usual introduced in the diode monitoring on a CPU level introduced in the cooler the pasted motherboard aging and a higher chance of technician error because the tests were conducted so far apart so there was no significant change that we could measure the change we are observing is within error margins and if it's not error it's more likely the pasted aging than it is anything else just from kind of drying and cracking a bit so for that we applied new thermal paste but kept the CPU socket it and didn't touch the liquid metal application with new paste applied and the old liquid metal on there we measured performance as just under 65 degrees over ambient again this is within reasonable error we are also using a different tube of cryo not this year versus last year so that could further factor into the results if the batches were different with the new liquid metal application we measured performance at around 63 point five degrees Celsius over ambient which is within reasonable error of both our original results and the new thermal paste result so that's it it's basically the same for this platform this testing we were not able to find any meaningful difference and we did control quite a bit here as much as we could within reason there are things that you could learn potentially with a higher sample size for sure but this gives us at least a baseline to say with this specific system the specific liquid metal application that was done there was no meaningful aging that we saw within the one year time period now does that change over two years we don't know and unfortunately we won't be able to test that because it was pulled the test at one year so maybe we'll set something up we had other liquid metal application results that didn't make it into the charts but I want to highlight them because from our the one you saw the final liquid metal application that was done days ago as opposed to a year ago that was our our third of three attempts to apply liquid metal for the new test just like the original data that you saw was our third of three attempts to apply liquid metal for or the original one-year long test and the reason for that and the reason for bringing it up is that it's sometimes tricky you can take a few tries to really get liquid metal to work the way you want you can get it to work pretty damn well most of the time but if you're really being a stickler about it then the surface tension the amount of liquid metal the specific way that you applied the IHS the way the socket clamps and potentially moves the IHS all of that impacts performance sometimes more significantly than others where you can get a couple cores that are maybe ninety degrees while all the others are in the high 70s and so that's something that we didn't like and specifically applied liquid metal until they were all pretty consistent and about where we wanted them and considered to be the best liquid metal application we could do but again the point of this and the point of detailing it is that the third of three attempts for each the before and the after looking metal testing result we're taken and the other two weren't that great and that means that there's more variance more room for thermal performance change when you initially apply the liquid metal then there is over a year of using it as far as this system tells us anyway we can't draw a firm conclusion there but as far as we know from this testing there's far more chance for poor performance from just a bad application than there is from aging and that means also that it's really hard to measure differences because if the difference is significantly greater from the way the technician applies liquid metal than it is from the end of the other factors like aging it's it's pretty difficult to tell the difference in any kind of testing unless you have some kind of automation to apply the liquid metal so keep all that in mind but overall there's your one year of testing liquid metal on a bench pretty controlled as noted we couldn't see any meaningful difference it's with in all variants and error margins that we allow for so the conduct did not test we just we didn't see much of a change cryo not is another story we'll look at that under the IHS next not sure if we'll do it for it here or not there that might be a bit extreme for just the thermal paste as opposed to a liquid metal and additionally if you've had your own experience with liquid metal aging and you had a different result than us please let us know below because this something we're having a greater sample size will help and having some audience participation be great for that so leave your experience below whether you've been running your liquid metal for a year plus and have had no impact in a negative way or if you've had negative impact and then if you have other specific liquid metals you'd like to see us test there's stuff from cool labs there's plenty of other stuff out there too that's some of it's pretty old actually if you have specific things you want us to test leave a comment below with what it is and we'll look into it now that we have more space in the office it's possible we could set something up that's that's a bit more intensive even than this one was so thanks for watching as always subscribe for more go to store document access dotnet to pick up one of our merchandise products like our new mouse pads or go to patreon.com/scishow and axis to house that directly thank you for watching I'll see you all next time