Gadgetory

it works and intended twits that we tore down and the joy con which was really the nightmare to put back together and the doc are all functional so with that we were able to go forth and do some thermal benchmarking and noise level benchmarking on the switch which uses about 25 millimetres fan 24:25 so we've got some thermal and noise analysis on this it's a bit of a fun side project no frame rate stuff as of yet maybe in the future at some point but before getting to the switch this content is brought to you by EVGA and they're 1080 TI FTW three icx cards that's coming out soon and EVGA is 1080 TI FTW three is this thermistors placed all over the board to provide a reading on BRM and vram performance useful for lowering overall noise output of the GP fan learn more in the link of the description below the nintendo shield uses a modified at maxwell SOC from nvidia it has some modifications in theory from nintendo we don't know exactly what those are but is a maxwell SOC because its maxwell SOC and because we're running on a super lockdown OS thank you nintendo we have no real way of monitoring anything that's going on in the system that means for any measurement of temperature of frequency framerate anything like that we need external tools software hardware otherwise so for the thermal measurements on this reusing thermocouples i've explained these several times before in videos on the channel but you can check the link in the description below for the article version of this video where we explain test methodology in addition to those thermocouples for the temperature readings we're using a noise meter just a DB meter to measure the noise output of the fan as it ramps into a heavier workload so we're able to plot temperature versus time and versus noise which is interesting because we can see how the fan handles is heavier workloads considering we have no other way to read the fan pack we've no reads way to read the GPU clock rate or anything like that for today is testing we decided to look at thermal performance versus fan speed and match that against heuristics for framerate and by that basically mean observation because again there are no framing and monitoring tools that are public without making our own for thermals that we're using those thermocouples attached to the top center of the SOC die and top center of the memory module closest to the USB port in the switch the USB type-c port those thermocouples only give us the case temperature of the die or the memory module casing so we don't have a proper junction temperature as of now which is really the interesting one but we can look at the case temperature and still figure out if there's a throttle point based on the numbers of outputs and the sort of observation of frame rates within the game in this case that would be Legend of Zelda we do know that the Tegra x1 this is the x1 not lean in Tendo modified one will operate at higher clock speeds went under a juncture temperature of 70 C and it drops about 6% went under or between 70 and 90 C and then reaches t.j.maxx some around 105 C depending on which specific implementation you have we're not sure how comparable the switch at so C is in regard to these temperatures but it's probably not all that different still we're working with what is effectively a T case measurement not a t-junction measurement so do keep that in mind now again for full testing methodology check the link in the description below that will talk about how the thermocouples are set up and the position of the noise meter and things like that the frequency spectrum analysis is not something we do a whole lot of so it's not as scientifically defined as the other processes in this TAS seemed like the thermal measurements but hopefully it will give a baseline as to what type of frequency is emitted from the fan it is kind of shrill so that's the best we've got for you if you have suggestions on improving frequency spectrum analysis in the future post them below because this is something that I've been looking into getting more involved with in the future for PC components let's start with just the temperatures then look at noise and then plot them against one another this chart shows the switch from initial boot through gameplay in breath of the wild and the big dip around the 1,200 second mark is when we were forced into the inventory menu by a prompt which immediately drives temperature down about nine Salty's in some cases the menu and tutorial interrupts help the switch recover its temperature to some extent and this can be seen on observing fan noise versus on-screen action as well for the temperature chart red represents the memory temperature and that is again the module located closest to the type C port and orange represents the SOC temperature there is one other memory package on the switch but we're only measuring that one close to the USB type-c slot at boot temperatures are expected ly close to our ambient temperature of roughly 28 Celsius throughout this test as we launch and load the game temperatures rise until they hit about 57 Celsius on both measured devices and then drop with each menu or prompt or interrupt it's not until around the 1600 second mark in this chart that the switch starts dithering around where we assume its maximum temperature spec is which is triggered most consistently by engaging in combat with multiple enemies concurrently the SOC drives up to around 59 Celsius and each time we hit that 59 to 60 C effective case temperature we were observing a frame rate reduction or stutter in gameplay and combat it seems that the switch is oscillating it's clock between maximum boost and some lower value when attempting to impose thermal constraints so we don't know precisely what temperature in terms of Junction this constraint would start existing because again we don't have a Junction measurement and we have no access to official white papers given that this is just a package reading 59 C is actually pretty high junction temperature would be a good bit higher than this if not significantly higher and let's look now at DBA output stand alone over a shorter period of time for a baseline of the fan noise here we're seeing the decibel output increase as load ramps with dips corresponding again with loading screens or menu pop-ups and just for the records the device was muted for this towards seconds of 400 to 500 were near multiple other actors in game who are driving up the fan rpm DBA output peaks at around 33 to 35 DBA when under heavier loads including those combat workloads and this is about the noise output of our case test bench PC one in the frackle define see that we're currently testing an tune reviewing so in terms of comparing it to your PC it's really not that loud in terms of raw DBA output this is completely bearable especially assuming you're setting up the console at a TV where you're probably 5 to 10 feet away and you've likely got game audio plane these things that will completely obfuscate the noise and the handheld mode though it always becomes a bit more noticeable but that's mostly because that's spitting it into your face out of the back of the unit this is also true for temperature though depending on where your head is you might not actually feel that the type of noise produced by the switch is of a high frequency nature generally producing an annoying Y and that's particularly heavy in the 1,400 Hertz some times 400 600 Hertz range depending on just how fast the fans going this isn't a perfect test environment as stated we're not using an anechoic chamber or anything like that but we're still able to log the tone output by the switch fan with relative accuracy here's a sample of what sets and sounds like let's look at these temperatures versus the noise plotted against one another this chart shows how the temperature and noise ramped up alongside the other fan noise and therefore fan rpm climbs steadily as SOC and memory temperature increase this chart is cropped in to an earlier part of the test so we're only maxing out at around 57 C instead of the 59 C max temperature reading that we had and were able to see that the DBA output remains pretty much in a bearable range of 33 to 35 DBA and again considering you'll have volume and probably distance that's really nothing to get mad about what is annoying is the higher pitch frequency if you're close enough to hear it or if you're particularly sensitive to those frequencies what this mostly shows is that there seems to be some sort of clock management going on based on thermals and that's again without a hard measurement for framerate but we can observe a gameplay and see very clear stutters or frames dips as many other people will tell you in plane press of the wild on the switch particularly we can observe those that frame anomalies or poor performance periods and plot that versus the temperature and generally with our thermocouple hitting 59 C which really who knows what the junction temperature is at that effective case temperature but with our thermocouple hitting and maintaining at that temperature output it would seem that's about when the clock starts to either throttle or there's something going on within the frequency and boosting functionality to prohibit a runaway thermal scenario or start triggering other thermal trip points in the case of the switch seeing those lower dips in frame rates correspond with the 59 C temperature would lead us to believe that in theory there's some sort of clock reduction that's not necessarily the only reason you would have those poor performance periods but it could just be poor optimization it could be too many Polly's and too much geometry on the scene at once but generally it is during those heavy combat periods one ways to validate this theory would be to improve cooling so that's an option but for now in the immediate future we have a few new KH reviews coming out this week there will be a 1700 X review on the website only no video for that so you have to go to the website today for that review will be posted around the same time as this video give or take a few hours overall though the switch for noise output not so bad DBA is ok temperatures a bit warm definitely approaching either some kind of throttle temperature or if not that just the game is way too much for that so see to handle in those heavy load combat scenarios particularly with the more geometrically complex mobs but overall that's what we've got for you for the switch for thermal and noise analysis I wish we could do more but based on the limited tools available that is to say strictly hardware solutions no software solutions that's the best that we can do today thank you for watching help us out on patreon.com slash gamers Nexus if you like this type of reporting it is definitely a bit unique for this one the fun project gamers Max and spot Annette for the full article or for the r7 1700 x review if you prefer that subscribe for more I'll see you all next time