How LOW Can It Go? | Steam Deck Thermal Modding

hey everyone and welcome back to the channel today we're putting the steam deck through extensive thermal testing from the stock configuration up through new thermal interface material back plates external fans and more today's episode is brought to you by the letter P for fan funding from patreon pledges with me moving back from the US and being extremely picky about my sponsors I'm incredibly thankful to all the patrons for the support this video took close to 50 hours to Benchmark burn and test and Swap all the hardware for so every bit of financial support helps a lot if you enjoy the content and want to help out you can get some cool Patron only benefits like a patron only section of Discord and ad-free viewing just go over to patreon.com cryobite33 and check it out as part of this video I'll be testing the following each product will have an affiliate link if available in the description below for back plates we have the stock backplate the older J socks backplate with a metal panel in it and the newer jsox backplate with an additional vent for a thermal interface materials I'll be trying the stock paste noctua nth1 thermal paste and a Honeywell PTM 7950 thermal pad and for additional accessories I'll be testing the jsox clip-on external fan and the jsox Magnetic external fan as a full disclosure all the jsox products in this video were sent to me free of charge but no money changed hands and they don't even know that I'm making this video those who have been around a while know that I don't pull any punches when reviewing products even if it's sponsored and there will be some of that in this video too so make sure to stay tuned lastly the thermal paste and thermal pad were purchased by me personally out of my pocket if you're only interested in a specific result from this video then skip to the appropriate timestamp on screen now for that product but I recommend watching at least the next section for my testing methodology before skipping speaking of testing methodology all results were gathered with a q1 2023 256 gigabyte steam deck upgraded to a one terabyte internal drive and with an ambient temperature of 18 celsius if you have any concerns about the implications of this SSD over the stock one then definitely check out my previous video where I go into exactly why I chose the to stick with this drive and what implications it actually has the steam deck was running Steam OS 3.4.8 and I tested three games Horizon zero Dawn cyberpunk 2077 and return them for each game I would start the game Run the built-in Benchmark go to the menu and then run the Benchmark again recording all of my results on the second run to make sure I got the worst thermally in addition to these three games I also did 30 minute burn in testing with unigen's superposition for each and every configuration we're testing today I ran the native Linux opengl version of superposition and while it performed poorly I was able to get some really good results off of it and was able to simulate an extended gaming session in the interest of brevity we'll be focusing on the returnal and superposition results throughout most of the video but at the end I'll also show The Horizon and cyberpunk results for the winner all results were spot checked with gtk stress test before moving on to the next batch and I used the updated fan curve for all testing as I figure that's what most people are going to be using for audio Comfort anyway aside from the obvious performance and internal temperatures I also tested fan noise and external chassis temperatures as well noise levels were tested by putting a decibel meter in DBA mode approximately one foot or 30 centimeters away from the steam deck in a dock and then running The returnal Benchmark recordings were taken of each test run and each video is then used to grab 10 readings one every 12 seconds of Benchmark time since the returnal Benchmark is exactly two minutes long those noise readings were then averaged and those are the numbers that you'll see in this video as a note these noise levels can vary in perception quite a bit depending on which fan your deck has installed the deck used in testing today has a Delta fan which is known to have a much more abrasive frequency and I can personally confirm this because my other deck has a Hawaiian fan which is known to be much more pleasant at the same volumes if you want to see which you have go to settings system then hardware and look at the fan manufacturer in the page external temperatures were measured using an IR thermometer the same process was used as for the noise levels but with the thermometer in view of the camera instead pointing at one of three locations the first location the rear vent is meant to measure the heat of the intake Grille and should give us a good idea of the airflow into the deck the second location the back of the device was taken from directly above the Apu and should give a good idea of how much heat is being transferred off of the chip as well as how uncomfortable the deck will be to touch off the grips the third and final location the top vent will tell us how much heat is actually leaving the deck with all three of these locations and the internal temperatures of the chip itself we should be able to tell where heat is being held and whether there's a bottleneck inadequate thermal transfer lastly for each and every Hardware configuration I also tested both stock clocks and an overclock the overclock was significant with the CPU being set to 4GB Hertz the gpus being set to 2 gigahertz and the wattage being set to 20 watts if you're interested in how I did this or seeing how to do it yourself then check out the video on screen now where I go through all of my results and the entire process alright and with all that said let's get into actual testing starting with a baseline using the stock thermal paste and the stock backplate first up I tested with superposition's 30 minute stress test as you can see CPU temperatures are exactly as you'd expect with a high overclock of 4 gigahertz GPU temperatures are also exactly as you'd expect over an extended period touching 85c once and very often hitting 84c as a point of reference here are the actual test results superposition runs very slowly never breaking out of the single digits here but it acts as a phenomenal test for thermal saturation next let's look at returnal returnal gains a bit across the board from the overclock making it much smoother although still not hitting a solid 30. CPU temperatures are high touching 92c right at the end of the test and 91 Elsewhere on the OC the GPU is much the same although it only hits 88c a few times throughout the testing last for the Baseline temperatures here we can see that the temperatures are largely the same on both stock and overclock except for the top vent the stock numbers aren't particularly worrying with the temperature above the soc on the back being warm to the touch this makes sense as the top vent is the Dex exhaust and it needs to remove much more heat that leads me to the next part of this story noise levels looking at the noise it generates gives us a fuller picture of what's going on here to my understanding this Delta of 2.7 will result in a noise that is 2.7 times louder as perceived by the human ear and it certainly sounded this way during testing this means that the overclock is outputting much hotter air and also much more noise speaking to the stock result 39.6 DBA is definitely noticeable according to the site decibelpro.app it clocks in around a quiet library but nearly 10 times louder than a whisper next let's move on to the next configuration the next round of tests were done with the newer style of jsox backplate installed this backplate has additional thermal Mass events above the fan itself and even interchangeable back paddles installation was very easy just eight screws and a shim to open it up then peel the protective plastic off of the thermal pad and close it looking at the thermal results for the CPU the new jsox backplate averages about 5 degrees cooler at stock clocks which is a huge jump the GPU also lost 5 to 6C average on the new backplate which again is huge the difference in temperatures didn't lead to any different results but let's see how it fares when overclocked surprisingly the jsox backplate dropped temperatures from 84c down to 73c which is an 11 entire degrees it's night and day again I'm astounded by how much the temperatures dropped here we went from 82c on the normal backplate all the way down to 72c with the new one all while overclocked as if that wasn't enough the CPU and GPU temperatures are both cooler while overclocked using the jsox backplate than at standard clocks when using the stock backplate yes 5 Watts more into the Apu is still resulting in lower temperatures using this new backplate lastly for superposition here we can see that while the 97th percentile decreased quite a bit the 0.1 percent lows actually increased by 124 percent over doubling the results with the stock backplate picking up the pace a bit we can see that the CPU temperatures are much lower using the back plate and even being nearly identical despite a hefty overclock the GPU temperatures aren't nearly as different but the jsox backplate still keeps a healthy lead over stock in all tests looking at how performance is affected we see something interesting despite operating at the same stock clock speeds the results with the new jsox backplate are six percent higher on averages and the lows are 19 and 17 percent higher the overclocking results are much closer trading blows across the board but at the expense of the stock backplate heating up much more as we saw before moving on to the chassis temperatures we can see that the jsox backplate runs a little hotter at both the rear vent and over the soc this is easily explained by the new vents over the fan as it doesn't Force nearly as much air over that area of the internals and will build up a bit more heat prior to moving it away the top vents however is a massive improve been on the jsox backplate with a 22 percent or 10 degree difference this is because of two reasons first the fan getting cooler air means that the exhaust will be cooler as a result second the fan is much closer to the exhaust meaning that the air won't need to travel over nearly as many hot components before it gets to the exhaust this is both negative and positive and I'm going to address some concerns of mine towards the end of the video but for now let's move on to acoustics the noise generated with the new back plate was a lot lower 6.2 times quieter with the overclock present to be noted though is that while the fan noise might be quieter the model of fan you have starts to matter much more without the fan being covered the more grading frequency of the Delta fan becomes much more apparent so this is definitely a balancing act that depends on your specific Dex fan model next up is replacing the thermal paste in these tests I use my personal favorite noctua's nth1 the actual process was what I describe as intermediate removing the back plate and heat spreader was simple but it would be easy to accidentally cause damage to the board itself if you aren't careful regardless the entire process took roughly 20 minutes including cleaning the stock paste off keep in mind that these results will be with the new paste and the stock backplate we'll get to the jsox backplate results with the new paste next the first test is CPU temperature as usual with the noctua pace coming in around two to three C cooler on both stock clocks and overclock the GPU temps are similar with the results being only one to two degrees apart as you can see it actually takes a little longer for the noctua paste to allow for heat saturation indicating that it's more effective at carrying heat away from the chip performance was identical with both pastes and superpositions so let's move on to return all where terminal is a bit different here with the noctua paste actually being a little hotter post heat saturation on the CPU at least when overclocking the GPU however has very similar results even when overclocked usually when the results are asymmetric like this it means that a bottleneck was lifted and the chip was able to work a little harder and these results confirm that to be the case the averages and 97th percentile were noticeably higher on stock clocks while the lows were much higher when overclocked with the noctua paste as well quickly looking into the chassis temps the results are all within margin of error this definitely makes sense as we haven't changed the airflow in any meaningful way last up for this section the noise levels for each the overclocked noise was virtually identical but the stock clocks came in at 1.9 DBA lower with the noctua paste which makes sense given the small temperature difference at those speeds next up let's look at the noctua paste with the new jsox backplate for these tests I'll be comparing with the jsox backplate itself using stock paste since the difference was so massive with the back plate before it's only fair to see if the paste makes any difference too in isolation starting with the CPU temperatures on superposition we see something very different than before the CPU is running much hotter with both the new paste and the new backplate this was very odd so next I needed to look into the GPU temperatures much to my surprise the GPU temperatures were higher as well the only explanation for this aside from Bad paste application that wouldn't align with our previous results is that the deck was able to work much harder with the new pace for some reason so I needed to confirm and this was definitely evidence of exactly that while Lowe's largely remained the same the averages had their first deviation of four percent on superposition and the 97th percentile results were a massive 25 percent higher with both the new paste and backplate I needed to confirm this on return all so let's go see how that went I wanted to look at these results from the other direction and see if we could guess the thermal implications of the results here the new paste and backplate together had better results than averages and lows while overclocked by a few percent each the stock clock results were interesting having similar averages but more extreme highs and lows it seems to me like the pace allowed for higher burst speeds based on this chart alone I'd guess that the CPU and GPU would both be running hotter on the overclock but cooler on average on the stock clocks let's see if I was right the CPU results are both exactly as I'd guessed with the overclock running almost 7 degrees hotter by the end of the test on the new paste what's more impressive to me are the results on the stock clocks with the new paste coming at a five to six degrees cooler than the back plate alone we see similar results for the GPU again coming in far under standard temperatures with a combo of new paste and backplate taking a look at the chassis temperatures we can see that even while overclocked the maximum deviation is a single degree over the soc this is a really impressive showing and reinforces that the paste isn't indicative of chassis temp last up for this round is Acoustics and I'm still very impressed with the paste and jsox backplate combo both of these results deviate by only half a decibel which is awesome news for the performance we gained despite generating more heat alright let's move on to the next round of testing next up I'll be testing the Honeywell PTM 7950 thermal pad which I'd never tested before I ordered this pad for mod DIY as I'd heard that the pads from AliExpress get mixed results to install the pad I followed the recommendations of vagif Abdullah in the Discord server first I cut a small rectangle out of the pad and put it in the freezer for about half an hour while it was chilling I took the deck apart cleaned the noctua thermal paste off and then covered it I grabbed the pad from the freezer cut it to size and then used the included sticker strips to remove one of the protective plastic pieces I put that side down onto the chip then peeled the other side's plastic off and pressed any bubbles out with tweezers before mounting the heatsink back on also let this be a lesson for those who do this make sure to have your morning coffee first not only did I almost break my micro SD in half disassembling the deck but I also completely forgot to unplug the battery before doing this and I'm genuinely lucky that I didn't break my deck I buttoned the deck back up and then started to burn the pad in unlike thermal pastes the pad needs a burn in Period to reach optimal efficiency otherwise it won't transfer heat as well as it can in order to burn it in I did two two hour runs on gtk stress test maxing it out and then I did two hours of 15 minutes on 15 minutes off in the superposition burn test it's very unlikely to need nearly this amount of burn-in period but I wanted to make sure that the PTM pad put its best foot forward for testing today as a quick final note these results will be compared to the noctua paste tests as those were the best we've had so far alright let's see how it did the CPU results are pretty impressive right out of the gate being three to four degrees cooler with an overclock and one to two degrees cooler at stock clock speeds the GPU also ran between one and three degrees cooler with the PTM pad than the noctua paste showing off just how well this little thing can transfer heat this is another situation where the actual results are within margin of error so let's check out eternal these CPU temperatures are very interesting to me in the previous round of testing using the noctua paste we saw that the temperatures were much higher due to pushing more frames will the same hold true here speaking to the Chart itself the overclocking temperatures are nearly 10c cooler and the stock clocks come in around two or three C cooler the GPU temperature here is equally mystifying dropping about 8C when overclocked surely the performance would suffer as a result of this much cooler chip not even close as a matter of fact these results are the best I've ever seen on returnal every metric is higher but especially noticeable are the lows with the stock and overclocked results being notably higher than the noctua paste variant and here we can finally see the Tipping Point for chassis temperature the PTM pad made every point one to two degrees hotter which is a testament to how much heat it can siphon off of the chip itself let's see if it made any acoustic differences fortunately the cooler chip means quieter operation at the cost of a single degree on the chassis I think that this is a clear victory over either the stock or not to a thermal pastes the 0.2 millimeter thickness coupled with insane thermal conductivity definitely makes the PTM 7950 pad a contender for the best Tim have used and I look forward to testing it out on other devices in the future for now let's see how well it can do when paired with the jsox backplate the CPU is performing as we'd expect based on previous results with the PCM pad being one to three degrees cooler than the noctua paste the GPU is actually a bit hotter on the PTM pad and although I ran the test back to back it seems like it started off a bit hotter looking back at footage this is because the deck actually took about 20 seconds longer to start the second test for that particular result as a result of my keyboard's batteries dying I decided to see if there was any major performance impact because of this and here we can see that the results for the PTM pad are ever so slightly higher meaning that the difference in temperature was minimal enough to not require a full retest let's check out returnal to see if the results hold true the CPU and GPU are running hotter with the PTM pad again could we have another bottleneck being lifted looks like it the overclock results are a bit higher across the board definitely a bottleneck being lifted here looking at chassis temperature we can see that most results are relatively close the vents are just over a degree warmer each with the PTM pad likely attributed to the faster thermal transfer off the chip but the chassis over the soc is almost 2 degrees cooler let's see how it's cooler by checking out the temperatures in correlation with the fan speed and noise as anticipated the higher chassis temperature was joined with higher noise therefore higher fan speed stock clocks come in around 1.1 DBA higher and overclocked comes in around 1.9 DBA higher these fan speeds are definitely why we saw the chassis over the soc drop two degrees in temperature these results go to show that the deck will generate the same amount of heat regardless it's just how much fan noise there will be to compensate alright let's move on to the older jsox backplate with the metal plate in it this backplate was a simple swap similar to the other one the only thing you need to be aware of is the larger thermal pad on the metal plate designed to increase the thermal capacity of the deck this testing will be comparing against their newer back plate to see which design is better metal panel or additional vent lastly all further tests will be using the PTM PAD as it was the best we found for heat transfer and I can't see a reason to go back the CPU temperatures are much worse with the older backplate about 5c average and the GPU follows suit not really putting up much of a fight and because of that the performance suffers quite a bit at least when overclocked it definitely seems to me like the older metal backplate is worse than the newer vented version but better than the stock backplate let's see if that's true yup it certainly looks that way the newer backplate is definitely remaining cooler but the metal plate is helping the older iteration cool the CPU quite a bit the story for the GPU is much the same and performance is exactly as we'd expect with the newer back plate pulling ahead by quite a bit particularly in the lows looking at the chassis temps we can see something very unfortunate with the older back plate aside from the exhaust being even hotter than with the stock backplate the metal plate gets hot to the touch on the older one according to ecma International a company specializing in compliance standards for electronics a surface temperature of 49.1 degrees on something metallic like this could only be held or touched for 10 seconds or fewer at a time anything more than that would be a breach of ecma287 and a burn hazard at this point I feel obligated to reiterate that my ambient temperatures are only 18 degrees Celsius in a hotter region like I don't know India or somewhere equatorial then this could be much much hotter if you want to read more about ecma287 and these compliance standards I'll leave a link to their PDF in the description below for this reason alone I cannot recommend the older backplate when overclocking or with small children around at least not without the product that we'll be testing next before we move on to that though let's look at acoustics with great thermal mass comes great heat which necessitates great cooling the older metal plate back plate is much louder than the newer one and it nearly matches the stock backplate and how loud the fans need to get to deal with the heat unfortunately this is another L for the metal backplate but let's see if we can fix that with another jsox product the magnetic fan the jsox magnetic fan is a multi-purpose fan that's magsafe compatible but comes with a perfectly sized metal plate to adhere to the plate already on the old back plate you need the additional plate to make it magnetic as the regular one is not but once you attach it you can just attach the fan connect the USB cable and turn it on while the directions specify that the USBC port on the fan is a charging port I can't get it to hold to charge and it seems too light to have batteries of any reasonable size for that reason I'll assume it has to be plugged in for the duration of use the fan has some basic RGB functionality and two speed settings for my testing I'll be using a fan speed of high and centering it over the the soc on the back of the deck now let's see if it can tame the unruly metal backplate and here you can see we're off to a great start the CPU is about 5 degrees cooler using the fan and it stays that way for the duration of the test the GPU is much of the same with a five to six degree drop we didn't really regain any performance with the fan but what really matters is the chassis temperatures and Acoustics let's start with those dangerous temperatures this was great to see the temperatures of the back plate are well within safety range now and the top vent is cooler than stock let's see how it affects Acoustics unsurprisingly slowing the internal fan down by using a quieter and larger external fan actually quiets The Deck Down by quite a bit these volumes are closer to what we experienced with the new back plate and can be made quieter using the low setting on the fan but would compromise some of the cooling as a result it's unfortunate that the metal backplate requires cooling of this kind to remain safe and quiet but it is great to see that it can be done we'll touch more on this in the conclusion of today's video but for now let's move on to the final product we're reviewing the jsox clip-on exhaust fan is a battery-powered backpack for your deck it's based on a clamp that you just need to pull and clip on the top of the fan sits inside the top exhaust and will suck air out it doesn't have any RGB or anything but it does have a variable speed knob it's a dial in exactly how much you want it to cool I tested everything on maximum speed keep in mind that this fan will conflict with the new backplate's additional vent so I'm testing it using the stock backplate and the PTM pad lastly I want to mention that because it's so chunky it won't fit into any of the docks that I have on hand so for testing I actually had to balance it on the clip at the end here on the bottom and I would definitely not recommend doing that it's very prone to falling over you'll have to come up with another way of holding the deck up if you want optimal airflow while docked with that said let's test it out jumping straight into the CPU temperatures we can see that this thing doesn't disappoint the stock clock temps dropped by about 18 entire degrees and the overclocking results by 10. the GPU was similar with about 20 degrees off on stock clocks and another 10 on overclock with internal temps like these how are the chassis temperatures the chassis over the soc matched the intake vent and the top event dropped 17.3 degrees leaving no part of the deck anything more than warm the results from the clip-on fan are downright and impressive when coupled with a PTM Pad but what's the downside the noise generated by the fan at these speeds is high and the frequency makes it incredibly noticeable while gaming at 47 DBA stock clocks with the clip-on fan results in 9.5 Times Higher volume nearly an order of magnitude the same chart that I referenced earlier indicated that this is almost the sound of moderate rainfall so it's definitely not something I can ignore unless you're playing with the deck docked or aren't sensitive to Fan noise this is entirely too loud while gaming medium speeds are better but you sacrifice some cooling performance and the steam deck can perform comparably at the same volumes in those cases [Music] alright so now that we've tested everything what are the key takeaways starting with the old jsox backplate the one with the metal plate I recommend avoiding it if you're thinking of overclocking or have larger hands the metal gets hot enough that I was able to feel it a few centimeters away I accidentally touched it a few times and definitely feel like I'd have been burned if I kept my fingers there if you really want the metal backplate for either aesthetic or performance reasons then I recommend the magnetic fan to go along with it aside from not having a battery this tamed the unruly metal backplate and made it Pleasant to hold it's a pain to have the cable plugged in but it wouldn't be too hard to have this plugged into a dock and then slap it on the back of your deck whenever you want to play on the TV it's too glossy for me the r2b leaves something to be desired and it doesn't feel super premium but it's really good at what it does if you don't want to swap a back plate but you still want some extra cooling for your deck laying around then I definitely recommend the clip-on fan it might be noisy and a bit cumbersome but the speed adjustment and excellent performance make this a great choice for someone who doesn't feel comfortable opening their deck if you feel comfortable opening your deck and you have some extra thermal paste laying around then I definitely recommend at least swapping this out it's not best in class but since a lot of techies have this laying around anyway it's a no-brainer and that brings us to the winners for today the jsox improved vented back plate and the Honeywell PTM 7950 thermal pad having one of these alone is already a boon for performance but having both together is a huge boost to all metrics especially compared to stock how much better let's see a few new comparisons CPU temperatures on Horizon zero Dawn were about 5 degrees cooler at stock speeds and 7 degrees cooler while overclocked GPU temperatures were similar losing 5 degrees on average and performance was much more stable even considering the hitching inherent in the game cyberpunk 2077 ran five to six degrees cooler on the CPU when overclocked but within margin of error on stock clocks and had similar results on the GPU but managed to pull a few more FPS at both the stock and overclocked settings this was while it stayed much cooler on the chassis itself and quieter by about five db8 a boot lastly I mentioned that I touch on this at the end of the video but I do have a few concerns about the new jsox vented backplate first the deck was made with a very specific airflow pattern in mind and it focuses on negative pressure having the fan on the inside and pull all of its air from the actual lower vents allows the airflow to go over the hottest part of the deck including the vrms and the chip itself the jsox vented backplate circumvents that by bringing in so much air from right behind the fan and exhausting it right here it doesn't get as much airflow there so there is a chance that it could harm those parts in the long run I haven't noticed any higher temperatures on the VR Rams or any of the other parts as opposed to stock yet but there's no guarantee that this won't cause arm long term second and much more minorly with the fan being completely exposed it has a much more direct route to suck dust in so you might have some dust buildup in the fan over time although it should be easier to see with the transparent backplate anyway lastly if you have a Delta fan you may want to steer clear or replace the fan as well the fans are usually quiet enough to not be too grating but on occasion they will ramp up and the wine will be very noticeable if you have a Delta I will personally be keeping the new back plate on this deck having both it and the PTM pad on here is a night and day difference for me especially in terms of daily usability and hey it might even help my battery in the long term keeping the fans lower all right and with that I want to sincerely thank everybody who watched this far into the video and I really hope you enjoyed it as always thank you all to my patrons super thanks and YouTube members for being there and financially supporting me during these longer videos and if you liked anything that you saw today please consider liking commenting subscribing you know all that um and I would actually really like a comment if you'll be buying any of the things that I mentioned today I have a few videos in the works now but one in particular will be a good follow-up to this video not the next one but soon it'll be reviewing the p3r monoblock something that promises to enhance the cooling on the steam deck by quite a bit but you'll have to stay tuned to that video to see those results if you want to watch another video like this one consider watching my previous video where I tested five steam deck drives and exactly what their performance implications were the link is on screen now and as always thank you all for watching and have a great day

Channel: CryoByte33

Views: 130,964

Keywords: thermal, cooling, jsaux, mods, steam deck

Id: aai2AtxenU0

Length: 32min 46sec (1966 seconds)

Published: Sat Jul 29 2023