Gadgetory

this video demonstrates briefly what it looks like to trip over temperature protection on a vrm it's a total offshoot unplanned video we were working on deleting bit wit Kyle's 79 80 XE for his new PC build and while testing a heavy overclock and over voltage we encountered spy keep our behavior and clock throttling when we weren't even hitting t.j.maxx this is a matter of inadequate cooling on the BRM even with the half-hearted attempt of adding fins that some of these manufacturers have reacts to 99 you still need airflow or a huge amount of surface area and it's not enough to come part way we'll be looking at that and some delayed benchmarks in this one before getting to that this coverage is brought to you by EVGA and Nvidia with the destiny to 1080 TI bundle a 1080i SC to comes with a synchronous fan control for its dual fans and nine thermal sensors and again includes destiny to learn more at the link in the description below let's start with the vrm stuff first this is a complete side shoot discussion but one that's worth exploring we were doing some thermal paste versus liquid metal and sealant versus no sealant testing with IHS and the substrate using Kyle's CPU and ended up running into issues where the multiplier was bouncing between 12 X and our set 45 X and that was with the actually with liquid metal and with thermal paste with the thermal paste testing we were hitting t.j.maxx so it looks like clock throttling because of t.j.maxx so you write it off as that with the liquid metal version we weren't really getting that close to t.j.maxx which is 104 C we were below 90 so the fact that we're dropping clocks from 45 acts to 12 says something's very wrong and looking into it further in the first test that I do in that situation is touch the heatsink of the vrm and if it feels like it's going to burn your finger within three seconds it's probably too hot it's not the most accurate test but it's one that's right nine times out of ten so doing that simple test put a fan on it blasted the vrm with air and therefore dissipated all of that heat and lo and behold the clock stopped the throttling it was now old in 45 acts stable and that is people of what Deborah was talking about earlier this year when he was doing his ex 299 vrm disaster video so basically this was on the X 299 ROG extreme whatever motherboard I don't the high-end board and it's really not got the best heatsink on there the problem is it's jammed in between memory a high TDP CPU a GPU back plate that has radiative heat coming off of it and it's got no direct airflow because it's got these big towers of RAM and whatever coolers on the CPU surrounding it and no using an air cooler on the CPU is not gonna fix it either because all that air goes over the top of the vrm heatsink not through it so it's a challenge and to fix it you basically put a fan above it and blast it with air and that solved the clock throttling this chart of the average multipliers tells the story pretty well it was all over the place but we weren't actually tripping t.j.maxx went under liquid metal we were about 83 to 89 Celsius intermittently across all cores averaging ultimately 283 point four degrees Celsius this isn't anywhere remotely close to the 104 C throttle threshold for skylake X so obviously again something's wrong and it wasn't the current limit either but in this chart you can see that the current fluctuation also reflected the frequency modulation so it would be easy to think that it might be a current limit of some kind on the board this line is ideally fairly steady and straight in a blender type workload as each thread renders a one tile at infinitum until render completion the orange and red lines are crashing against an unknown wall at this time and it's not current that's not t.j.maxx but the blue line remains steady the question of course remains for what we changed with the blue line and that change was adding the fan so that fixed the problem basically the vrm one went back and tested it later was running about 120 degrees Celsius because it's sitting there with a part that can pull 500 Watts easy and it's got 60 degrees Celsius sticks of memory on all sides of the vrm there's rear i/o next to the the left part of the vrm on the l-shape and on the top side you have the part of the CLC or the air cooler or whatever it is that's really not going to be pushing any air on to that the only way to cool that thing the top part of it is a downdraft cooler and those aren't going to be any good for this type of processor so that's the problem and we fixed it with a direct fan now something for Kyle to maybe consider with his ten thousand dollar PC build he's he's taking the CPU for would be to look into the option of a full coverage block that would cover the V RMS as well depending on which board it is I don't remember which one of these using that may or may not be possible but that's something to look into otherwise just I would say Kyle try and get some kind of fan on the top or in the side or somewhere to push air onto that thing if you're running it overclocked now you're just gonna film it for pretty b-roll run some quick tests I'd say throw the fan in there for your test take it out for your b-roll call it a day but the BRM that we have on our ACS board was not happy it was very not happy so yeah that pretty much shows you what Deborah was talking about earlier we might look into it more stick a lot of thermocouples onto things didn't do it for this just yet but it's an option for the future so for the next part of this we've got the benchmarks with Kyle's CPU doing D lid versus lidded and shields vs. no sealant this was just a quick round of test so we do fixed frequency and voltage to make sure there's no fluctuation I have some auto charts in here at the end to restore perspective as to what the CPU does without any influence from all this stuff and for the overclock applied a simple 4.5 gigahertz 1.2 4v ID which is what we found to be stable on our CPU now 1.2 4v ID is pretty much pushing the limit of what I'm comfortable with with a 79 80 XE I'd rather be in the 1.15 21.2 range but we thought what we needed for our CPU I have not tested what Kyle's max frequency and voltages I just applied what I knew would work keep that in mind but all that matters is that they're fixed so we can determine a delta between the lidded and deleted with delayed results from our 4.5 gigahertz overclock and 1.2 4v ID tiles 79 ATX e with Tim and blender had several cores bumping into t.j.maxx at 104 Celsius causing clocks to drop multiplier to 12x this isn't just the vrm thing here this is actually just overheating on the CPU itself we were not able to sustain the overclock during this test with our X 62 at max speeds we'd either need a better cooler or lower clocks and voltage after a D lid and some liquid metal our heavy silicon adhesive attempt where and we used the adhesive on both layers and on all sides of the substrate and the IHS that had us at t.j.maxx once again so even with liquid metal it wasn't any better which tells you that the adhesive was too thick because the liquid metal either wasn't doing its job or there were air pockets manually spreading the adhesive into a thin film on just the top layer of the substrate and only in the pertinent corners resulted in an 83 degrees Celsius average core temperature resolving our clock throttling issue completely at least once we had the fans going the overclock is now stable and operates at 45 X constantly as for the GM parts we were at 75 degrees Celsius with our unsealed IHS previously and this was with a different chip different liquid metal application and without any sealant so a couple of factors there to consider but either way it was running a bit cooler part of that is because of the lack of sealant and part of it is just a different CPU with prime95 fixed at 3.6 gigahertz and 1.15 V ID which by the way is aggressive but that's because we want it to be fixed the GN Tim unit operated an all core average of 89 point five degrees Celsius peaking at nine DC with the unsealed liquid metal variant at 72 point four degrees Celsius that's a 17 degree improvement Kyle's Tim unit ran out of box thermals of 86 point six degrees Celsius peaking at 89 degrees more or less what we saw for the most part with a light sealant again and only the most important areas we managed the 76 degrees Celsius average and 77 C peak for peak to peak drops we saw a 12 degree dip in terms of degrees Celsius for Kyle CPU with the light sealant version keep in mind that these are two separate CPUs once again so we can't definitively state that the sealing caused the 5c difference in entirety but what we can state is that the combination of these factors resulted in a 5 C Delta and Kyle's CPU still benefits from a 12 degree improvement from comparison against itself with liquid metal and a light reseal versus Intel's Tim out of curiosity we also tested a medium seal on Kyle CPU finding that the output temperature was 78 to 80 degrees Celsius up from 76 to 77 this basically had a line of silicone adhesive on both the top and bottom substrates but it was thinned out from the original heavier attempt we can definitively state in this case that the sealant caused the 2 to 3 degrees Celsius leap and thermals just because the rhetoric often gets out of control with d-lighting tests let's keep in mind that these chips are 100% operable with the stock Tim it's just a matter of how far you can push them with Auto settings which introduce variants and voltage and power consumptions so we can't do an a/b comparison the CPU is run at around 50 degrees with blender or 50 to 60 to with prime95 we don't use auto for comparisons between Tim and liquid metal because it's unpredictable and not really repeatable for testing however we're showing this here to again give perspective that yes these CPUs cannot operate without a deal it'd it just helps to do it so then while on that point to recap why and when you would deal it'd deal it in Al's primarily into things helping with thermals it's not just because the temperature is lower that you do it it's because by lowering the temperature you give yourself Headroom to lower the fan RPMs or buy a cheaper cooler for example in the blender test we could throw a 360 cooler on there and fix the problem of running out of thermal Headroom on Kyle's CPU with the original Tim but it'd be a 360 cooler it'd be bigger it'd be more expensive if you delete it like we did you can run a 280 on there and get away with it just fine so that's something to consider another factor of course is noise because the lower the fan speed to lower your noise problems and also power leakage reduction so for every 10 ish degrees Celsius reduction in CPU thermals you get roughly 4% of your power back it'll consume 4 percent less power for that tendency drop and that comes down to power leakage and we've demonstrated that in our past content with the 7900 XD LED video and the 79 ATX a review and so d 960 x review where we showed again power leakage reduction from deleting so those are the two primary factors the main one is noise which is a byproduct of lowering your cost to entry by giving you Headroom thermally to buy cheaper cooler as well those are the main reasons 2d lid not necessary it gets necessary though as you start overclocking towards the max potential of the silicon because the silicon can do in our testing an extra 100 to 300 megahertz depending on the chip with a d lid because you've reduced the thermal Headroom issues so that's all for this one you can go to patreon.com/scishow sexist helps out directly or subscribe for more of course as always and kyle will be uploading his own video of his $10,000 PC with the CP that we deleted I have to ship it back to them so hopefully the liquid metal doesn't move around and cause problems for him when it gets back but I talked to dare bear about how to ship it so we'll do heart best check that out when it's live subscribe for more I'll see you all next time