Gadgetory

we just published a video talking about 1% losing 0.1% Louis for frame rates and frame time testing and methodology accuracy things like that and part of presentation of data when we do these benchmarks on GPUs or cases or CPUs is talking about delta T over ambient temperatures so that's a pretty critical point that I wanted to make a separate video explaining because it doesn't seem like it's necessarily universally understood so here's the thing with case testing test something like this versus an H 440 or whatever whatever the fantex one is that I hated on when you're doing case testing and you're looking at temperature values just pulling an actual measurement from straight from the CPU or straight from the GPU without accounting for the ambient temperature or the environment temperature of the house or test area there could be variants in that ambient temperature that makes the one to two degree differences and a lot of instances basically gone or you can even have cases trade spots when they shouldn't just because you haven't accounted for ambient so let's talk about that the ambient temperature of this room I'm in right now is probably somewhere in the range of 68 to 72 Fahrenheit depending on whatever the thermostat set to but it also depends on things like how hot that wall is from the Sun if we're doing something during the day or where the vents are located in the ceiling and so when you have these different variables doing one case test one day to the next there can actually be a very large change so as an example many years ago when I first did case reviews him before taking about a year break between them I had a case like this I ran a test on it one day it was X degrees and then the next day it was X degrees plus two ended today after that it was X degrees minus 2 or 1 or whatever and so there's this big range of a couple degrees so to get rid of that we came up with this solution that I talked about in our thermal test chamber video so we validated our methodology using a thermal chamber and proved that it was basically a hundred percent reliable and you can find that video in the end slate I suppose or the description below but we use one of these this is a thermocouple reader this is a k-type thermocouple it can withstand a serious amount of heat I think 800 degrees Celsius or something and we run it actively and I just I keep one of these measuring ambient and if we want extra validation I'll put another one on and measure somewhere else in the case like the intake or the exhaust or whatever in front of the CPU fan but the one measuring ambient right now says it's 23 degrees Celsius twenty two point eight degrees Celsius twenty three point one so a bit of a fluctuation there I logged that value actively every second and we logged the CPU diode every second the GP diode every second PCH whatever and then take this data put it in a spreadsheet in one column take this data for say the CPU put it in another column and then you subtract this from the CPU temperature that gives you what we call a delta T value or the temperature difference when subtracting ambient and so now let's say I take this case and I test it against what something not NZXT that we liked the 400 C or the 600 C so we take this case have a 600 C here run them both through the bench and I subtract the data from test one from this out of this case and test two from this out of that case now we have basically perfectly comparable results accounting for and again this is still fluctuating it's at 21.5 now accounting for that fluctuation in ambient and that fluctuation can be big just as an example of some instances where you will see a change in ambient positioning this case here versus where the camera is there would be a difference in performance so this eliminates the issue of needing to be a hundred percent of the time in the same exact spot it also eliminates the concern of is AC on is heating on or is neither on and then there's further concern there if your heating and air if your HVAC system is toggle II not an author of the task is trying to sustain one temperature that will change the results too so there's a ton of variables another one opening the door and walking into the room and walking out of the room if it's enough of a gust it will change the temperature by a couple tenths of a degree maximally if you have more people in the room it gets hotter if you're running multiple systems doing multiple tests it gets hotter if it's winter or summer hotter colder type of thing so it's very important to do this ambient subtraction and create a delta value and if you're looking at results that don't do this or don't account for this somehow there could be a couple degree variants and data which often times and the instance of a case or a GPU cooler between vendors EVGA versus MSI or a CPU cooler often times they're fighting over one degree max because thermal is hard and we've kind of gone a long way with it and there's not a lot of room to go further without increasing the cost substantially for these cooling units so when you're fighting over a single degree it's important to make sure we're illustrating who wins that one point because even though it's mostly irrelevant for everyday use it's still differentiating sort of a superiority of design for a specific use case now there are instances where several degrees will differentiate cases so for example by the point of this thing going live probably our origin Kronos system build review will go live and in that mini ITX system we saw the GPU with a 55% GPU vrm fan speed at fluorine 55% of the speed we saw that hitting something like 90 Celsius for the actual total temperature or that would be in the 70s for the Delta and then you might see another Mini ITX like the NZXT Manta bit bigger but hitting something like in the 80s that's a big different that's a 10 degree difference and that impacts things in the form of frequency over time so if you look at a frequency over time chart as the case gets hotter and sort of equalizes at whatever temperature it's going to equalize that the frequency might start at 2 gigahertz but fall to 1.6 gigahertz and that impacts your frame rate pretty massively so that's an instance where there are real temperature differences 10 degrees plus in some cases but most the time it is about one degree we do this delta-t subtraction so when I talk about a GPU running at 0.45 Celsius idle which I think one of our liquid-cooled GP is ran away with arts for a tea I was around one degree Celsius idle when I say that I don't mean that it's near freezing that's not what we're saying because it is not possible for your GPU your CP it's not possible for any of these components to be less than ambient because the air you're pulling into the system is gonna be 70 degrees Fahrenheit or 20 21 degrees Celsius that air just by going through a fan doesn't get colder than it was when it went in it might accelerate it and that certainly helps things dissipates the heat faster through the fins but it's not gonna be colder than ambien so when I say a temperature idle is 1 degree Celsius what we actually mean in real terms without the Delta values is it's 1 degree plus the ambient temperature so it's effectively an offset so Bamian was 20 then that means we'd read the GPU as 21 Celsius so that's kind of that's what that means now the next thing is our ambient temperature is way different from yours potentially there's a big range so folks who are located in dry vs. humid areas or in a desert that's hot versus in somewhere colder that's going to impact your ambient as well so what you should do with these numbers is when you see us reporting something like a 7 degree 70 degrees Celsius GPU load temperatures so 7 0 Celsius you go to your own thermostat you look at the temperature that it is in the house and convert that to Celsius and then add that to the value so let's say that your house is 30 degrees Celsius because you're in a hot climate 30 plus 70 that's 100 degrees that's a big deal so that means that your GP is probably past or hitting teeth the t.j.maxx and it's throttling down to control the the frequencies that it can lower its temperatures that's where something like that would really matter maybe you're in a 20 degree ambient environment 70 plus 20 is 90 you have much less of an issue than person at 100 degrees still an issue but a lot less a lot lower magnitude we'll say so that's why these things matter because if we just tell you the definitive temperature I'll call it to avoid using the word absolute and offending somebody if we use the definitive temperature pre Delta subtraction one the case will change in temperature or the GP whatever the products will change in temperature from test to test because fluctuations in ambient and - it will be different than your own ambient temperature when you add that back in so that's why we do that and that's really I think that's all I have here for notes as well so if there's a lot of reasons to do this there's a lot of variables things like sound testing which is why I have this here sounds house and you can't do the same subtraction we'll talk about that in the future but if I record something at 60 decibels which is about what I'm talking about right now 60 70 decibels they record - 70 decibels for product a and then 70 decibels for product B we still have to subtract the noise floor which we measure before every test and then that subtraction is not just for example say let's say the noise floor is 25 dB you can't do 75 minus 25 equals 50 decibels that's not how how noise work it's it's logarithmic so we'll explain that in the future but still the same kind of idea of presenting a Delta value where we subtract out whatever the ambient environment noise or temperature is to produce a result that you can use so there's a bit of work on your end but it's really not that much if you want to make it easy at 20 Celsius to every temperature value and you'll basically see these sort of definitive value with a 70 ish fahrenheit house but otherwise it's all comparative so you can see comparatively which product is better than the other one if there's a big difference then you know which one's superior if there's a little difference you know that because of our testing methodology the difference is accurate and what we would call measurable but imperceptible which has a very specific meaning it means that we can accurately measure and we feel confident that is outside of any margin of error but it is not something you would perceive as a user to be impactful to your experience so that's what that means when I say that but hopefully that explains delta T over a means if you want to learn more about our testing methodology in the post roll video there will be two that are interesting to you one the thermal chamber video where I put a case in a thermal chamber and validate this and show that it works and two the 1% lows and 0.1% low FPS video which explains what those two things mean so as always pay traveling the postural video if you want to helps out directly by funding future efforts but as always just subscribe to get more content I'll see you all next time