Gadgetory

everyone welcome to another episode of ask G and today we've got some questions on process manufacturing process CPUs and their impact on 120 Hertz 1 40 Hertz displays when you're looking at FPS talking about beat bending for CPUs and GPUs stuff like that so as always questions below in the comments if you have questions we're joined obviously by senior AMD analyst snowflake so now would be a good time to post your questions relevant to that but questions below GPU other components all welcome cases CPUs all that stuff Ram this content is brought to you by I way power and their new element gaming PC with arc LED fans underglow and a large tempered glass side window so first question this verse sparse of d I don't know how to say your name sorry ask what is the difference between a 14 nanometer planner and 40 nanometer FinFET to go into that in depth would would be a whole separate video but just to really briefly go over things the first part of that question 14 nanometers that is the size of the process so the manufacturing process basically the tools used to make the components make the all the stuff that goes into your CPU your GPU so when you're looking at 14 nanometer that's what that's referring to its referring to how physically large the manufacturer number refers to a size of the process and the next part planar versus FinFET refers to sort of the the type of design for the gates and the transistors so when you're looking at FinFET the the fin part you can think of it like a literal thin from an animal so the gates basically surrounding the gate there's raised thin on the sides like that and we have images of them so raise fin around the gate and that helps trap basically your power so there's less power leakage this way other benefits too but power leakage is one of the big ones that and the N Nvidia are harping on for this generation so that reduction of power leakage is because it is trapped by effectively a wall found the gates and that's something we talked about in the gp100 architecture deep dive if you're curious about that and that isn't a video forum as well next question I by the way that's super top level but I hope that answers the basics a planar by the way all planar means is flat so just like planner and any other instance planner means flat versus FinFET where it's kind of 3d not the same as a 3d transistor from Intel but we can use the words 3d and 2d to give a basic understanding the next question is from ikea dune he says can you talk about if it's true that 120 Hertz and 144 hurts are actually CPU intensive that's it's kind of a vague question the very basics of it by nature of drawing more frames you are going to be hitting the CPU more frequently every time the CPU gets hit or every every time I frame is drawn to the screen there's work done by the GPU and the CPU so the CPU has to do work on every single frame just like the GPU does and that's especially cool true if it's if it's something like the x11 or something opengl where draw calls are hitting the cpu more noticeably so in that instance the answer would be yes 120 Hertz and 144 hurts insofar as drawing frames at that speed would actually certainly be more intensive when you're looking at games and that's that's true for CPU and GPU but CPU more probably than people think that's the top level next question very very simple from t-mo for life says what's Bend I think we talked up bend chips in the last episode so when I asked about does asus been there strict chips for use so the answer what's bend bending is a process of selecting it's a selective process for going through the silicon that's output from a factory given to manufacturers from the supplier so bending happens at a couple levels one is at the initial creation of the chip it's kind of tested and validated canvas should perform to spec does it have anything needs to be disabled if there are things that need to be disabled you end up with stuff like the g3 258 where igp is disabled or the athlon x4 series where the APU the graphics part of the APU is disabled so those things come about as a result of failing validation or just needing to fill a gap in the market the other part of bending is when you look at a manufacturer board partner in the case of GPUs EVGA AC set aside gigabyte all of them have more or less the same process so just choosing a high end card like a cane pen or something of that nature that those kind of super high-end cards will be selectively chosen because the chip quality is not the same from product to product so if we have two 1080s I don't know what this is actually 1080 very convenient so we have two 1080s from the same manufacture the same make and model say we have two of these exact cards they will not overclocked the same and the chip quality will be different almost certainly between the two that's because there's imperfections of the manufacturing process like any other silicon is very very selective in its in its function so when you have differences in manufacturing process you have differences in overclock potential and things like that stability at different clock rates if there's failed validation that might get down steps to a lower end product which we've seen in the past when you can kind of unlock a GPU from a low NGP to a higher NGP that's normally because something in there is not performing the spec of the high end so that's why that happens so binning is the process of selecting components by someone in the line selling it to you it's not something that consumer does some of the lines selling it to you is selecting components to perform based on a speck normally its associated with high end meaning that they bend it to be a better overclocking next question is from Jordan cave jordan says hi Steve my question is does water cooling reduce CPU or GPU power consumption for the same sustained clocks when compared the standard air cooling no so reducing your thermals all your it's it's still outputting the same amount of heat still drawing the same amount of power when you put a higher end or on a component like when we put liquid on these things actually right there there's your example so when we put liquid on one of these things the GPU is still drawing other than the added power for the pump which we can ignore the GPU itself is still drawing the same amount of power so it does not reduce power consumption does not reduce the heat that the component is capable of generating but what changes is the dissipation ability of the cooler so with a better cooler on there because in this case we're using a copper cold plate very conductive it's about 400 watts per meter Kelvin at 25c copper cold plate sinks heat away from the GPU from the silicon into a chamber through which liquid runs and dissipates it babe yeah carries it up to the radiator and all that stuff that we've talked about before so the dissipation potential is far greater on one of these that means all that really all that means is that your GPU itself will perform at a lower temperature which does impact a lot can impact life it can impact the overclock potential depending on how hot it's getting before the better coolers applied and generally just changes kind of stability of the clock so if you look at clock rate versus time charts you'll see that the clock rate does dip a bit when it hits throttle points like 83 Celsius on the 1080 silicon so that's the main thing that that better coolers change as far as power consumption if you kind of build an all liquid system versus all air let's just ignore the power required for the fans the pumps ignoring that the components will draw the same amount of power so that's that's more or less how it works there may be some efficiency differences in there but it's going to be really small so we'll ignore that next question is from Chris sue who says I was a long question the short of it how long do you test your a iOS if viendo CP cooler bench when we do we haven't done it in a while for publication but I'm actually working on some of that now so how do we test the all-in-ones further says a oh hi yes so basically saying that a iOS can take a bit longer to heat up then something like an air cooler which is true because you have a coolant in here as well so the coolant does take time to heat up it's not going to be the same temperature as the cold plates or anything they're all kind of different temperatures within a range of each other some cooler manufacturers have specs for TC 60 CE whatever they've got the specification where they won't allow their coolant to exceed that temperature or well I should say not won't allow because I could certainly happen if you really tried but within normal use scenarios they won't allow it to exceed a certain temperature that's for many reasons mostly longevity permeation of the tubes things like that evaporation but answering the question how long do we test them for cooler ties that we do two types of tests so when looking at most things like GPUs CPUs whatever the first test is an endurance test that runs for two hours it runs with metro last light on loop basically infinitely until we conclude the test it's all scripted starts and stops shuts down on all its own starts off the logging on its own that two-hour period is looked at that generates the overtime charts temperature versus time temperature and frequency for is a time of the CP and the GPU it's all important metrics to know because if the cooler is not performing adequately you could be hitting thermal throttle points or your temperature could be varied enough that the CPU kind of freaks out in down clocks temporarily GPU also so that's part of it the second part is a shorter test runs 25 minutes we've found through a lot of testing over the last year's 25 minutes is basically perfect for case cooler any GPU fan all that testing basically perfect it's enough time to saturate whatever is whatever the cooler is liquid or otherwise and from there we pull data averaged from a point in the spreadsheet on word for the load and then from an earlier point to the beginning of this house for idle so that's the basics of it next question is icebergs hi Jan I have a soundblaster audigy two zs platinum on an 8320 i'm a sigh 970 motherboard is it every day and it's better than my onboard and he's saying I want to upgrade I still want to use these sound card for the audio card the sound blaster audio card what can I do to keep it upgraded so icebergs here is looking at a PCI Express x1 2 dual pci riser and asking if there's going to be degradation and audio quality generally no because the the sound card is not going to saturate the throughput potential of that interface it's not fast enough if you buy a really crappy quality pci riser then you probably might have some loss but most that loss will be from poorly insulated cabling or something like that so just by something that that's decent don't spend too little on it because you don't want to lose either lose performance through they wired it poorly and the electricals bad and there's not enough throughput which I've seen or it's just again wired poorly or there's EMI or some kind of crosstalk and so you're losing performance audibly just through introducing crosstalk and basically static those the only two things to look out for you all need to worry about cable like a performance degradation over cable length if you're not running a really long cable anything that within reason if it's if it's something you can fit in your case without winding the cable several times then you'll be fine so I wouldn't really worry about it sound maybe if you did that for a video card then you can have some performance degradation if you're doing it for and even then even we've shown like one percent difference in some cases you'd have to do it for like multiple SLI setup 4-way SLI or something so I wouldn't worry about it that's all for this video oh wait now we got one more last one was from himanshu got a guy concur Oh parenthetically the Iron Man excellent ok so the iran in parenthesis says hey gamer nexus i guess i guess that's me how can i get long hair like yours don't cut it so that's all for this time eh on like the postural video subscribe for more content all that stuff leave your questions below I opens everything for GPU CPUs k is all this stuff I'm only saying that because we've had several episodes like GPU only so feel free to post about any components not just GPUs but I'll see you all next time you mm-hmm