Gadgetory

warning content may contain disturbing images of g/b torture that is against international conventions look at it look don't try and look away from the brutal slow dismemberment of gigabytes at GT X 1080 Ti extreme this video card was one of the lowest rated GT X 1080 eyes that we reviewed last year or well two years ago now ending up the least recommended out of every card we tested this was largely due to poor assembly quality using an odd mix of through bolts and heavy mounting hardware that added nothing structurally nor thermally to the design it's also because the plastic shroud creaks when pressed like McDonald's toy plastic it isn't really form-fitting and the card is bulky to a degree of pointlessness when we tested the 1080 TI extreme we found that its thermal performance was worse than competition that measured 30 percent smaller even when those normalized at 240 DBA and that's because the extreme just doesn't make great contact with much of anything now after a year of forced labor in our oldest rendering rig we get to see the real world side effect of a needlessly heavy card that's poorly reinforced internally and has been sagging for a year that's what we're testing today before that this video is brought to you by thermal grizzly and their high-end thermal compounds thermal grizzly makes cryo not paste for high thermal performance and conductivity without being electrically conductive so you don't have to worry about shorting components Crona is particularly good for replacing stock GPU pastes as cryo not is a non curing compound learn more at the link in the description below here's our test system for the GPU sag thermal testing it's about as real-world as it gets especially from us we typically put everything an open-air bench for performance testing and run through the third the the numbers in a very standardized way whether that's thermal or performance or whatever in this instance it's a one-off test so we don't need wide sweeping standardization and we can do it in an actual system now the fall here of course as always is that we have a sample size of one and that different GPUs different video card assemblies will have different performance numbers depending on how they sag over time but this is a pretty damn extreme example if only because the name of the card is 1080 Ti extreme but is an extreme example of sag because we have b-roll shots of this but you probably see it in a role too all I have to do is apply some upward pressure and there's like an inch of drop off or something so it's it's pretty severe and what we're doing is we have it in a PMO one we've got three front intake fans its stock PMO one they are at max speed everything's controlled within this system so all fans are controlled the intake fans are all 100% speed the GPU fan for the Kart that's under test is 55% speed locked for all tests which is intentional because it will all otherwise fluctuate and then all the data is invalid and we also are logging frequency which is important because nvidia pascal cards are really heavy with the clock gating to a point where if the thermals look the same it might be possible that the frequency is significantly higher and one test on the other depending on how it was down clocking based on temperature so that's all controlled for and accounted for we do test with the bottom card installed and removed depending on what we're looking at the bottom card here isn't actually producing any load so there's no radiative heat coming off the backside and not any meaningful amount the fans add about it's at the minimum speed of 23% for the lower card it's a 980 TI it's not doing anything the only reason it's there is to create a closer barrier to the sagging card because as this sags obviously it's going to cut off some of that intake area at the front of the two cards so that's why that's there and I think that covers most of what you need to know about the test setup if you have further questions we'll have a short article about this as well with the system components but that's more or less it so it's a real-world test what we're gonna do is see in this real system that we used for like I don't know a couple years but about a year with that card in it is there a performance decay in any meaningful fashion from the card sagging versus resolution and the resolution was I gave Patrick the job of I said your job is to figure out how to make this not sag anymore and he took some wire that you used to hang a pic and tied it around the the cables or the end of the card and supported it to one of the screws in the top of the case actually where it's brilliantly we wouldn't recommend it it looks awful but it works really well so for testing purposes that was our solution if you want to actually fix it there's plenty of brackets you can buy and shove in there for this test with dual GPUs you can just put something between the two cards but the goal is to not obstruct the fans so we don't change the intake parameters during the test so let's get through the numbers that's that's all the test setup let's go through the thermal numbers we've got a additional testing of thermocouples on some of the components as well so we can see how the MOSFETs and the vram fare with GPU sag this extreme in the testing with two GPS installed we measured the performance at when sagging at forty nine point five degrees over ambient for the Gigabyte extreme 1080 TI or forty eight point one degrees celsius over a means when sag is eliminated and fixed with our wire this difference is outside of our error margins but it is overall insignificant the same is true when looking at single GPU numbers where we've plotted an average temperature of 42 point seven degrees over ambient with sag or 42 point one without sag this is also within error margins and not significantly different and we don't have the test resolution here to claim that one is superior than the other they're functionally the same as the sort of interesting aside this chart does consequently point out how a second GPU even when idle will impact their own performance of the primary GPU and that effect is is exacerbated when the second GPU is doing work keep in mind that this is just one GPU and that we can't draw conclusions for all sagging GPU configurations from these numbers it's just the extreme two other important factors come into play as well both of which we'll look at next first Nvidia's GPU boosting algorithm is heavily thermal dependent which means that the difference in temperature that we're seeing which is not much could be complicated by how the GPU clock is changing pursuant to that temperature so there may be knock-on effects and frequency that means look at second component temperatures particularly toward the right edge of the PCB like MOSFETs could be more affected than GPU core temperature well attached thermocouples in a moment to look at MOSFET and vram thermals as for GPU frequency this overtime plot will illustrate that pattern well first we'll plot the GPU thermals again once we reach steady state we begin averaging over 2,000 cells of data for the thermal numbers presented a moment ago that data came from this distribution though note that we've chopped the chart at 2100 seconds it's really the same the Sagan GPU is consistently about one degree warmer than the non sagging GPU which makes sense when considering the distance between the first and second cards has increased thus allowing more air to the primary card under test more importantly here's a look at the frequency for the SAG in GPU overall one steady state is reached the average GPU frequency is exactly 1621 0.3 megahertz for the test again that's at steady-state so there's no further thermal movement next plotting the frequency line for the non sagging GPU after propping up the same gigabyte 1080i extreme with an unobstructed block we see an average steady state frequency of 1650 6.9 megahertz exactly this increase is about 35 megahertz but our error margin for frequency measurement in this test is approximately 15 megahertz based on a lot of data where we've been able to look at the standard deviation so there's not much significant change here keep in mind that fir mark is also a power virus so if you're questioning why that frequency looks low in general it's because frequency is instantiated differently in this workload than in gaming scenarios because of how the drivers are written so overall frequency will be scaled lower below gaming frequencies but we can still use it as an indicator of performance realistically this isn't all that important it's basically the difference between one of the board partners pre overclocked configurations versus and others so you can make the case that maybe a higher stock clocked GPU that sags is about as useful as a lower stock clock CPU that doesn't sag but beyond that there's not really a difference and even that difference is questionable as it's very close to error margin so there's a bit of a difference here in terms of raw numbers but not one in terms of really meaningful impact you're talking one two three FPS average difference depending on the game so not a big change we need to see what this looks like with a single GPU though plotted over time again we see just how little variance there is between the thermal results honestly GPU configuration our scripting allows the testing to align almost perfectly in this plot and with less than one degree Delta second a second we can call this a wash it's well within error margins adding frequency to the plot we can see that frequency variation is almost zero once that steady-state we average the frequencies to be sixteen thirty two point two megahertz for the sagging card and sixteen thirty three point four megahertz for the non sagging card we are within our rough fifteen megahertz error margin and also within margin of the higher frequency card or very close to it anyway for the dual GPU test so far there aren't huge differences the last place the check would be component temperatures so we removed the GPU and disassembled it then reassembled it with thermocouples on a hotspot MOSFET and gddr5 X memory module keep in mind that this test data is incomparable to the previous data as it required disassembly and so new paste was applied we ran two configurations single GPU with sag and single GPU without sag comparable to each other directly for this task we measured GPU core thermals as within one degree of each other yet again placing within range of our first round of test results note that these aren't incomparable again as thrown face change but the proximity to the previous results is reassuring of test methodology for gddr5 X memory module thermals we measured the sagging card at fifty one point eight degrees over ambient and the non sagging card at fifty one point three degrees over ambient both are well within spec and also well within error margins establishing no meaningful difference MOSFET temperatures were within half a degree of each other also well within error margins and we're not anywhere close to throttling if anyone's wondering about that were for about seventy degrees away from that so even when trying to find thermal differences in board components like the MOSFET toward the right edge we still could not establish a meaningful Delta the dual GPU configuration produced the largest gap or the most promising difference in thermals earlier but even that was minor and the frequency difference really kind of becomes insignificant given the natural variants and how the program operates this could also be used as a stand-in for cases where the PSU shroud is excessively close to the GPU same idea as a dual GPU test here keeping in mind that this is a a single card configuration it's a while it's a single configuration it's got one card or two cards but it's all the same fixed hardware it's the same GPU ultimately that we're testing it's not like we had multiple that are sagging here keeping that in mind we couldn't measure a meaningful thermal difference really in any way in this setup so sometimes there's a difference of about a degree and sometimes it was enough to start exiting margins of error but even then it's just it's not meaningful it doesn't really show up on the radar and except for in extremely controlled scenarios and testing scenarios and environments it's not a number that you will be able to to produce as a user and confidently no there's a difference with this setup now that's not going to be true globally so if you have an instance of GPU sag where you saw the opposite please leave a comment below let us know what card it was and did you do anything to fix it because of course this isn't a representative of all video cards ever there are a lot of cards that sag out there a lot of really heavy cards but it is representative of some of them and so from that perspective it wasn't really a thermal concern ultimately it wasn't really a performance concern we do have concerns though one of course aesthetically it just looks bad it's not a good look anyone who doesn't know computers look at it probably won't know the difference anyone who does know computers if they look in the system all they're gonna see is GPU sag so that's enough reason for some people to to be bothered by this to a point of fixing it or maybe buying a different card next time maybe something that's not as heavy or pointlessly heavy as this one is a second concern over time the card can pull itself further and further away from PCIe slot now will it ever get far enough where contact between the pins is reduced to a point of maybe creating some flicker or you lose display out occasionally that depends again on the configuration depends on the case depends on how secure everything is it's possible though we found some forum posts where that happens so potential concern but something that can be fixed by just restocking the cards though it's gonna fall out and crush anything it's just that it's it's slowly pulling away from that socket and it's it's just a contact could become poor with time how relevant that is well probably not super relevant but something to keep in mind something annoying the troubleshoot if you run into that issue third concern is that over time some cards that are particularly heavy on the right side like this one is the right side being towards the camera right now those might have less right side support you might start seeing pads pulling away just slightly from some of those farther right vrm or power componentry so like the MOSFETs would be about as far to the side as you can get before you start getting into minor voltage rails and things like that that aren't that important to measure or even that hot but the MOSFET line is definitely the the more susceptible to a thermal of had pulling away with time as the heatsink sags and this is more likely going to be a problem where the PCB is supported to a point either by back plate or or just by better structural engineering than this card as if the PCB is supported to a point where it's relatively straight but the cooler is sagging which is not the scenario we have on this card then you're gonna start seeing some separation of the cooler from the components it's contacting and that's where you'll notice the MOSFET the vram temperatures maybe they're closer to the core though start to rise so that is the bigger concern here and how relevant that is to you will depend on the card so for this one the card that we have the 1080i extreme gigabyte card the whole thing is sagging all of it so the upside of that is that the PCB remains in contact the components remain in contact with the cooler the downside is that the entire video card is sagging so I take the good with the bad I guess but yeah the bigger concern is just going to be check the card that you have if it is drooping a bit and look in there and see can you still see the thermal pad as contacting the component or is it starting to peel away with the heatsink as the PCB might remain straight and that's just going to depend on the design of the card and sort of the the structural configuration of the cooler what does it have in there to support the backplate versus support the cooler and is it lopsided to the point of just only the backplate in the PCB and not the cooler but thermally performance-wise not a huge difference so we didn't really expect that to be honest the the most noticeable difference will be if you have a case or a shroud or a video card component or something close to the card that's sagging because then all it's doing is is just sort of truncating that window where air can get in and that'll reduce the amount of air flow to the card but that's a secondary effect not a primary effect of the GPU sag so pretty interesting case study but that's all it is if you have cards that are really bad maybe you have one that you know we have already like you've seen us review the card let us know what it is and we can consider looking at it this was in here for well since we got it we take the worst components we review and we put them in production systems sometimes it works better than others and so it's had a long time to just let gravity take its course and we have plenty of other cards so if there's one you're especially interested in and every one up with that to the top and we'll consider it for future look that's it for this one as always you can subscribe for more and you should or you can get a stored on cameras excess net to pick up one of our shirts this one is no longer sold but we have other designs like the Swan mod matte or otherwise or you get a patreon.com slash gamers Nexus tops out directly thank you for watching I'll see you all next time