Gadgetory

our initial benchmarking of the AMD rx 5700 XT GPU had one primary complaint and that was predictably the blower cooler we typically do hybrid mods where we stick a liquid cooler on a GPU to improve it but this time we decided to do something simpler and replace the Hitachi HMO 3 thermal pad with four cents worth of washers and some thermal paste then we made an additional change after that and added some therole pads to the backplate for some a be testing this data shows how much of an improvement you can get with a better mounting pressure on the GPU and with some new thermal paste rather than the thicker and more thermally inefficient they're all pad with our quick changes we're able to get the card into a 40 DBA operating range without thermal throttling which was not possible previously on our test bench before that this video is brought to you by Corsair as an iron claw RGB wireless gaming mouse the corsair iron claw focuses first on comfort with its palm grip approach and also uses a sub 1 millisecond wireless connection for the pc you can toggle between wireless and Bluetooth connectivity making it easy to control multiple pcs with one Mouse like a stream PC and a gaming PC the mouse has 10 programmable buttons the pmw 30 391 18,000 DPI sensor 3 zone RGB LEDs and 50-million click lifespan for left and right mouse buttons learn more at the link in the description below we've done this before with AMD cards we did it with Vega and more depth with 56 and 64 and that was primarily because with HBM they had different packaging methods where some of the packaging would have an epoxy resin on it and some of it wouldn't and so that changed the mounting contact to the HBM and the GPU from the vapor chamber with Radeon seven we did some of this and this is just pressure testing with pressure paper we would secure it remove it secure it again and did that three passes for these and the end result was that the entire middle never really made contact with a vapor chamber until you put that hitachi throwing pad on there which is what andy's using for these cards instead of throwing pace so there's some ups and downs to this the positive side is that thermal pads like the hitachi pad or like some of the other more permanent throne pads in the market like carbon odd or something like that the advantage to them is that they aged well and that they don't really age like throwing paste a lot of paste will cure not all of it does but a lot of it does cure over time and become worse over time so that's the upside the downside is that all these solutions and including carbo not they are not as thermally effective as a good thermal paste and so much of this comes down to things like one of it is that the properties of the material but another one is the thickness of the material what their own pace you can smash it down tighter and that Hitachi HMO 3 pad is pretty damn thick so that does impact the cooling capabilities in a negative way now the upside is that Andy can try to solve some of the mounting pressure issue which we've tested again the first thing we did after getting all the thermal testing done was taking the thing apart and put some pressure paper on it lo and behold the top one here which will show a mural the top one is the original test then the bottom one is the test where we applied some washers to it so this is the simplest thing we did to fix our rating on seven-card 2 is we put four of these on here these just come with radiator screws you can get them for about a penny each at your local hardware store for reference these are one millimeter thick we might recommend something a little bit thinner maybe more like if you can find on point 5 or 0.7 millimeter thick washers we went with metal because they were around plastic would be a little safer because then you'll crack the washer before you cause damage to the hardware in theory anyway so metal and a little bit thicker is what we use but we'd probably recommend plastic and a bit thinner just so you have some more play there to to damage the washer before you damage anything else because then it would be the weakest link it's just a safety measure but anyway the point is we put some of those on with Radeon 7 and we were able to get paste working where previously when we because of this when we installed Radeon 7 with our all paste and without any extra mounting pressure from washers under the four screws on the retention plate it made worse contact on the HTML three and so the HTML 3 thermal pad was better in those scenarios but with the extra mounting pressure the thermal paste is better so we applied that here too and we've improved it quite a bit for just the blower cooler alone to the point where we can now run at a lower noise level while operating at slightly lower temperatures as well and no longer throttle previously we were thermal throttling on the frequency at 40 DBA and we've now fixed that so while this shot of the pressure papers on the screen couple things if it's a brighter or more saturated red that means there's better contact in that area if it's not red or there's it's not quite a saturated that means there's worse contact and we see a gap exists in the center for both tests even with the washers which would indicate one of two things either a lack of mounting pressure in the center of the cooler which kind of makes sense because there's no screw there since the GP is there or it could also mean that there is a bump in the surface of the vapor chamber like a non levelness which would also make sense for a vapor chamber and in our original radion 17 you can see how there's a similar pattern that emerges and we think that is probably from where the the vapor chamber might have some cylinders or different support we're not quite sure but it's a level a level this issue in the cold plate before I get into the testing data here it's it would probably be good to talk about thermal conductivity as a rating so there's different ways to measure thermal conductivity a lot of the manufacturers do it in different ways you often one pace to the next you can't just straight compare the thermal conductivity number sort of like how with fancy FM if it's not tested the same way from one manufacturer to the desk there's no industry standard that these companies adhere to if you're not comparing to like for like then the numbers aren't necessarily directly comparable and that's true for thermal conductivity especially so when you see a really high thermal conductivity rating on these Hitachi thermal pads just because that number is higher than what you would get on what we use which we use the hydro not a thermal grizzly height or not paste and I'll link them in the description below if they're not already an ad for this video they might be but we as the hydro not paste and that's fairly high thermal conductivity but the Hitachi pad is higher but it's not Hitachi pads higher therefore is better because there's also the thickness of the material and then things like orientation on those pads matter to vertically they they transfer the most efficiently which these are oriented that way but yeah so that's wanted to point that out you can't just look at thermal conductivity between two property to two product and determine if one is better than the other without any further information well it's gonna test data for this we are looking at GPU Junction and GPU edge temperature the edge temperature is the standard one that's what Nvidia it measures presently or presents anyway the junction temperature is the hottest of all the sensors on the die from memory Vega or Radeon seven had 64 of those I think and we're not sure how many 700 XD has but it takes the hottest one and then boost according to that temperature 110 is t.j.maxx or at least the maximum temperature at which it will begin to throttle and then with Radeon 7 2004 clock I'm not sure if it's supposed to with this one though so one times the max for Junction we're also measuring vrm and the memory temperatures for this too so and then finally we have a backplate test where we put some thermal pads on the backplate to see if it improves the vrm or memory thermals it would probably not improve the GPU thermals as you'll see here and then that was done without even needing to read paste the card so we can do those tests back-to-back because all that comes off is the backplate all the other screws remain in so that makes it really easy to compare without even having to repay it which would throw another variable at it and then finally we did four so at well we'll talk about this more later but we did an average of three different test passes for each configuration for the final chart that you'll see that way we can eliminate some runs run variance and ambient temperature was controlled and monitored every second so let's get into it for this benchmark we retested the thermals of the rx T 700 XT at 40 DBA the rx 5700 XT is below or cooler proves problematic for our initial testing where a 40 DB a fixed noise that will result it in a GPU Junction temperature about 110 to 112 degrees Celsius which bounces off the throttle point the GPU edge temperature was about 94 degrees compared to an Nvidia 2070 super F e-cards edge temperature of about 67 degrees you can see that chart on the screen already hopefully and that'll be the previous chart we showed in our review so no new data here yet the bigger problem was that the GPU junction temperature as that's where a throttling will start and we also noted that we ran these tests in something of an ideal environment in that initial review which is a low ambient of 21 to 22 degrees Celsius whereas most modern cases will sit at an ambient of at least 30 degrees especially if the CV is under load if you have a case with a solid front panel but fortunately those don't exist you never see those and also if you don't run AC or something in our new testing we found gbu edge temperature to plot at about 86 degrees in this chart that's a drop of 8 degrees Celsius in our controlled testing environment this is where the ambient unchanged and logged again second a second during the run with the fan speeds controlled the 40 DBA an eighth degree reduction in edge temperature from just adding paste and washers is a massive change for minimal effort again it's about four cents of washers and if you buy some good paste it's maybe another ten bucks or so as a plug we did use our GN teardown toolkit and had the cooler off in about five minutes in the second time we did this replace the paste and had it all back together and working in under half an hour so this is a pretty easy model the first time you do it it might take 40 minutes or so and you can grab our toolkit if you want it on stored at kyndra's nexus net it has every tool you need for this video card as for the GPU junction temperature we improved that from 110 degrees to 112 degrees previously throttling down to 103 degrees Celsius it's also a major improvement and means we're no longer clock throttling at 40 DBA technically this comparison isn't even one to want because the throttling clocks previously meant that we didn't know what the true uninhibited temperature would have been if it hadn't throttled it would have been higher though as for the memory and the MOSFETs in the new chart here our memory temperatures were originally at about 98 degrees Celsius at 40 DBA which is at a point of spec limitations for GT dr6 it's it's fast approaching that that limit where GD d r 6 is operating outside of the recommended temperature the change improved memory thermals for two reasons one we've increased mounting pressure of the cooling unit we can put a shot up of our teardown where we show how the cold plate comes into contact with the memory indirectly via base plate and the second reason is that we're increasing the cooling efficiency and getting heat away from the GPU faster which will reduce heat in the vicinity of the memory back to the chart vrm MOSFETs temperatures have remained mostly the same which makes sense these components are farther from the GPU socket and our heatsink tin to the base plate directly the improvement is matte similarly two to three degrees Celsius which is outside of our error margins but not particularly noteworthy as for whether adding pads to the back plate helps we have an easier tard to look out for that the answer is that it's more or less as expected we saw no reduction in GPU temperature which makes sense since all the heat is for the GPU as far from the backplate would have to go through multiple layers of substrate die copper and gold and then the the PCB eventually so that makes sense to not see the difference there we saw a reduction of about two degrees on the vrm MOSFETs and this is after accounting for the ambient again and these results are as a reminder an average of three tasks passes on each configuration which is why you've seen this chart instead of overtime that's a total of six test passes between these two thermal paste was able to be left for each pass because only the backplate is removed which means that the cooler itself can stay mounted with the same paste after all that testing we end up with an average improvement of about two degrees in the memory and the vrm thermals which in the sake of the memory is actually somewhat important because it's already run into a hot vrm it doesn't matter that much but it is an improvement so as for whether this is worth it well it really depends first of all we would recommend waiting for partner models if you're buying this card because those will in most cases solve the problem without needing extra work and will be better anyway though this is still it's still loud and it still runs hot it's just it's a lot better so a partner model will still be better than this if you are buying one of these though and you are leaving the reference cooler on it then it might be worth doing the washer mod with their old pace the or even without thermal paste you could still improve it a bit the downside is there a couple downsides one if you remove this cooler and you're not just adding washers you'd add washers without removing the cooler in which case you don't damage the thermal pad if you remove the cooler there's a good chance you damage the thermal pad at which point you're going to have to use paste and so then you lose the advantage of a thrown pad which is not really subject to the same aging as a lot of the pastes are but you can counter that by using a non curing paste and again we used hydro not here but we'd probably recommend a cryo not that one lasts more or less it'll last the life of the card at least from our experience so if you use a non curing pace like that you'll be in good shape but still you're modifying the device and if you're not experienced opening it then that could be a downside it's not difficult of course you can watch our video to see how it's done it's very easy couple screws and some cables increasing the pressure is something we're a little uncertain about if you're careful with it like we were and you don't over torque things then it should be fine if you over torque it or you use a washer that is a little bit too thick and you have to really drive that screw into the into the threads there's a chance you could cause damage to the PCB or the GPU so be careful if you do this mod it's not difficult there is an improvement we would say overall it's worth it it's just because of how many people watch these videos and the skill level of all of the people watching varies greatly I do want to be very clear that if you have no experience working with stuff like this it is not hard I don't want to make it sound too difficult it's just be careful because if you over torque things what so an example of what can happen is if you have like because this happened to me if you have the cooler not line up perfectly when you get this fan back in there it's got a triangular plate that sits in the base plate and if that triangle is not inserted to the other triangle and it's out of place what will happen is to use a kingpin quote the PCB will Bend like a banana when you try to put it back together and you'll get a really nasty bend in it and if you screw all the screws back in and that's that state it will probably damage the card if you catch it you'll be fine but anyway yeah point is this is not difficult and it's certainly easy for anyone to do we just want to make really clear for people who are inexperienced to watch out for that kind of thing because this is something that I would expect more people will do as a mod because it's so trivial it's a couple washers so again the final recommendations we use the one millimeter thick steel washer and that works fine just kind of watch the torque on it use the cryo or hydro not paste would recommend a cryo not paste or something similar and for your uses I'd probably recommend a plastic washer that's a little bit thinner 0.7 would be ideal not sure if you can really find those but and then that that would be a pretty good safety check too so anyway that's it for this one there's an improvement that's worthwhile and the reason it's worthwhile is because we went from 40 DBA and throttling which is not a really a valid use because now you have to increase the fan temperature say you're not throttling run from that to 40 DBA and not throttling so you can actually use the card at 40 DBA as tested 20 inches away because that's important in a room with a noise floor of about 26 DB so yeah worthwhile if you feel like opening it up for the thermal improvement and I think that'll wrap it so that fixed a lot of the problems with the cooler in terms of not throttling while being a reasonable temperature we'd like to see AMD do this instead because then you can I mean then it comes from the factory that way my assumptions would be that they think the mounting pressure issue is best solved by a thermal pad rather than by finding a new retention system because maybe when you're mass producing thousands of these it may be it becomes an issue of you could I don't know over tighten or something like that if it's all done by machine and there's some very and it's not really sure what the exact reasoning is but that'd be my assumption thank you for watching you can go to store documents access net if you'd like to pick up a toolkit like this one which is what we use for this modding again it has every tool you'll need for this what's your experience with it takes about five minutes to take it apart and put it back together - the pasting time and you hear the patreon.com slash gamers Nexus - helps out there as well I'll see you all next time