Gadgetory

everyone welcome to another ask GN it's been a little while for these so this is a community questions than answers segments you can leave your questions in the comment section below we also have some from our patreon backers on discord and there will also be a second ask GN going to the public Channel so we're doing two of these from public release over the next day or so and then there's a patreon ask patreon patrons the looking for patrons ask G n that will be available via at this court or the patreon group for anyone who wants to watch that separate episode with a with just some extra fun questions from discord so getting into this one oh I should also note the second public asked young one we're doing will hopefully feature snowflake I think so check back for that I want to try and film the intro at the home studio for that before that this video is brought to you by the gigabyte zi 390 ARS master motherboard which comes equipped with one of the more powerful zi 390 VRMs for heavier overclocks on the new 9th Gen Intel CPUs the ARS master is also one of the few motherboards with a really heatsink this generation featuring a mix of high surface area fins and looks oriented hover blocks oh and it's also got updated our DV illumination learn more at the link below alright so as always the format for these were going into I don't know a hundred and fifth that so format for these is leave your questions below and I'll do my best to get to them and I might have to pitch them to people who know more than I do depend on how hard they are so also I'll just note these are still sort of in stock so if you want one they're on store like humans axis done that but all right first question is from ultimate fate who said why are the GT 10:30 gtx 950 and 1050 TI manufactured using a 14 nanometer processor is the 16 animated FinFET process used for the rest of the pascal cards how did anybody have it lower end cards manufactured using the smaller process and why were the other cards that not manufactured using the smaller dode so the process size doesn't small process i their process no it doesn't just mean it's better as a first note Nvidia it does not give a great answer on this question if you ask officially their responses something to the tune of we believe all of our manufacturing partners are great and all of these process nodes are excellent choices for our GPS and our gamers so not particularly helpful my understanding is that the switch between different different solutions providers for silicon manufacturing by Nvidia TSMC or otherwise is really just to spread the load so that's that's kind of how I understand it is it's a a means of ensuring that there's a constant supply of GPUs and that they're not putting too much too much strain on one source and it also allows to in theory provide some additional negotiation leverage when negotiating prices for the silicon parts that come out of the fabulous Nvidia is fabolous so that's why my understanding as for technical differences I mean it's just I don't think there are any meaningful differences in a way that you would you would see the manifest in performance or in the capabilities of the cards at least that's that's that's how that's how I was how I was told when I asked this question to a few people in the industry so yeah my understanding is really just it's to spread the manufacturing load and to get multiple sources just just like they do for memory they get hynek's micron Samsung memory whatever they can get to put on the cards and sometimes some is a bit better than others but it all fits the spec so but back to the the question of node sizes I'll just note that things like like density do matter as well the definition that matters so AMD for it an easy example and these seven nanometer is similar to intel's ten nanometer in actuality and well assuming intel ever you know makes it but it just it comes down to again density definition who's making it what is their technology so not necessarily as simple as like you get 40 nanometer over here in 16 over here 14 sweater than 16 next question we're gonna have a couple of hard cuts on this one because I'm still kind of safe so next was from dar he says new motherboards often come with one or more thermocouples Oh thermocouples which motherboard component or installed device would you choose to monitor great question so yes some of the high-end ACS boards do come with thermocouples I don't think too many boards come with them but certainly the high end overclocking ones do the problem is that typically those thermocouples are like the fatter kind so the you can get kind of two types of thermocouples you get the flat film kind which is really good it's they're expensive as heck to buy but but they're very good you can sandwich them between just about anything doesn't really impact thermal performance your measurements will be accurate down to fractions of a degree and the resolution is often like 0.1 degrees Celsius or something like that so those are really good and that's what we use for for example if you want to sandwich them between small SMDs and a pad or something but the problem is the ones that come with the motherboards have their I can't remember exactly they have like a bulb and so it's a bit thicker and round so you can use them for sure to some degree of accuracy but the problem is you can't really like sandwich it between a MOSFET and a pad because the the end of the thermocouple where the measurement is actually taken is not touching the MOSFET directly in that scenario it's probably sort of floating up in the pad or in the air with the bottom side of it touching the MOSFET so you're not gonna get the most accurate reading that way which is unfortunate but also it is now creating a height gap between the device and its heatsink which could pose issues with mounting you could crush the thermocouple when you tighten the heatsink or you might not tighten it enough and there's not good thermal contact the results are meaningless stuff like that so if you get a board with a flat thermocouple properly flat like what are they like 1/100 of an inch thick typically for the good ones then I would put it on a hotspot MOSFET which you can determine by kind of I mean ideally you measure them but for all intents and purposes for consumer use it's about the same to just kind of look at it and then you know count the MOSFETs and pick the one in the middle assuming they're all vcore and if it's not be corn there's one that's like GT over on the end that ignore that one so yeah that's how I would do it I would pick a hot spot like central MOSFET that's sandwiched between some inductors because those are really hot there's some capacitors maybe is in poor areas through couples which they often are then I guess my choice would be if I'm doing an open loop maybe I'd stick one of the reservoir just to get a liquid temperature that's really useful kind of interesting to look at so that's a good choice for you you could you kinda you could shove one between maybe if it's tight enough in between a MOSFET and an inductor as opposed to on top of one of them because then you're not going to cause a height gap issue and you can just kind of get an ad-hoc measurement of between two really hot things and make sure they're not crazy that doesn't tell you anything useful sort of but I mean it's you'll know if they're too hot and up to until that point it's maybe not the most useful measurement but could be fun to look at other places you could stick one I mean if you think there's like if you have concerns of I don't know heat with your GPS now you have multiple GPUs or sandals together maybe you have an m2 SSD under there or something like that for the most part you can look at software measurements for all of these devices but if not for some reason then I would consider a thermocouple down between the GPUs or maybe under where the SSD is because that's gonna be a really hot area that's sandwiched between a bunch of things and one more option here that's actually pretty practical is just a case ambient thermocouple you just kind of it won't be great for Kayla man ahead but I just kind of tie it somewhere like on a hard drive cage or somewhere that's between the the CPU intake fan in the front of the case or if there's a water block in the water block in the front of the case between your intake and where they're supposed to go basically consider just mounting one kind of in the case floating an air and that'll give you an error an ambient temperature in the case so if you know your external ambience 21 degrees and you see your case is 40 well there's plenty of room to improve your your case airflow so that's another really good and practical place to put one hopefully those give you some ideas they're not all that helpful for everyday use but they're definitely fun to use thermocouples it gives you some more measurements you don't normally get otherwise these are typically for high end overclocking scenarios they can be used to insert into the the through-hole in the socket sometimes there's one of those and high on board so you can get a socket temperature I would say you can use them for like a dry eye spot or Allen to pot or something like that but you probably want to use something better for that better insulated better better build hopefully those could be ideas though great question if you have any more like that please ask them lots of fun to work with those types of questions next one human person says yes I'm sure you are that you are not a robot or an alien at all human person what PSU wattage is reasonable do I just add the TDP of all my components and then take double or something like that really good question this is a pretty sort of basic one but also one that's important to answer and and is I'm not particularly well documented so power supply wattage what's reasonable I wouldn't just add the TDP so I wouldn't take like like my GP is 250 my CPUs 90 I wouldn't then call that 340 44 for the power supply requirement at that point I would I would here's what I would do I would go to reviews and look up like the the actual power consumption tests these are kind of tricky I'll tell you why in a moment but look at power consumption tasks between hopefully us we might have some an attack normally does a good job of them PC / in the past it's done a good job with them and there's plenty others out there too so I would check different reviews for the power consumption of the components you're using now here's the the trap you need to avoid a lot of the time we reviewers will use total system power consumption for our benchmarks and what you're really looking at is kind of a delta there so if if device a consumes air quotes around that 500 watts on this chart imaginary chart and device B consume 600 watts ultimately the difference is 100 watts and assuming the test platform itself is controlled properly and I know we control eros properly and I'm sure other reviewers do theirs as well then that's a reliable means to determine the difference relative of the two devices so you really look at a relative difference not necessarily an absolute difference because unless you know exactly the this power consumption of the rest of the system then you've got kind of a variable there and the reason that measure is taken is because for things like GPUs it's strictly speaking it's easier and it is I mean that's that's really mostly it because you can current clamp the power cables go into the GPU that's not difficult at all but to then also get the power going through the socket that's kind of hard because ideally you'd do something like a riser cable and then you splice off the the 12 volt lines maybe make a bit longer and clamp those but that is also kind of difficult to do properly I've spoken with build Zoid about this in the past wave some solutions but there are ways to do it absolutely Allan mal Ventana when he was at PC / he would do this type of thing - to just get only the GP power draw in a GP power consumption test by clamping the right things by door through a riser measuring that with I don't know if it was an external controller or a clamp but there are ways to do that it's just that typically you see total system power consumption so if you see 500 Watts check to see what the the chart is for if it says total system power consumption that's not the power consumption of the video card or the CPU or whatever you're looking up it's everything that it's under some kind of load remember that power fluctuates based on load so I would account for my expected maximum usage scenario which is probably going to be at some point 100 percent usage and then I would look up the power consumption of different devices and if I can find an individual test for it for example our CPU tests are not total system power consumption they are CPU only power consumption before VRM efficiency losses are taken into account but close enough so for our test for CPUs you'll see an accurate number to what's going to be drawn and then once you figure out your power supply I needs your wattage needs add a bit to that so maybe 20% to 50% overhead just to pay on how much efficiency you want at 50% you're gonna be a peak efficiency of the power supply but you don't need that much you give me at 20% or something if you want to save some money so figure out the wattage just by adding things together from reviews and then add some overhead on top of that so I mean that's the way to do it honestly is to look at reviews I'm of course biased in this we do reviews but that's what I would do so that's what I would suggest you would do because TDP is a it's a tricky number everyone's got a different definition of it for your CPU is it's more measure of how much cooling potential is needed to keep the CPU at spec or are under t.j.maxx now typically TDP at least well in general is pretty close to the actual power consumption but not always as we've Illustrated depends on things like MCE and then GPUs you also have board power versus GPU power so depending on how the manufacturer rates their TDP that may or may not account for memory power consumption which is somewhat significant as well so anyway that's what I would do hopefully that helps a bit next one Erik Ernst a month ago I noticed when you overclock processors that temperatures of each core vary sometimes by a lot I also noticed that on the 79 80 X II you raise the multiplier by one time by one axe more on the first two cores than the rest of the course could you optimize your overclock by raising the multiplier on cooler cores and lowering it on the hotter cores to equalize the temperatures across the entire package would that stabilize the package for more total frequency or is a straight overclock a better option so let me try to jump to the the answer on this yes so technically different cores will overclock better than others it's typically a pretty small difference so for competitive overclocking when we're fighting with while anyone on the leader boards then it is worth it because it's it's a lot of the times a great way to gain extra maybe even couple hundred points on times by extreme which is significant and when you're at the edge anyway to try and regain a lead so it's worth it there I don't know that I would say it's worth it for for my home system excuse me because it's just it's a lot of work and ultimately unless you're competitively overclocking you're not really gaining that much for real performance a couple FPS maybe and probably not even that so I would not say it's worth it but to answer your question yes core to core overclocking is technically the way to get the most out of the CPU because I don't want to use really the word binning but each core will perform differently and that can be for a few different reasons can be for silicon quality it can be for where is the core on the package if it's in a cooler area then it might be able to run a bit a higher frequency because it's cooler so it's not fighting the same thermal challenges that a hotter core may be in the center of all the other hot cores would be fighting giving you some extra thermal Headroom to push the frequency a bit higher so anyway could you optimize across the entire package the answer is yes and in an ideal world if you have infinite time and you sit there and you tune each of the in our case 18 cores for that scenario and you find which ones can do 5.1 which one can maybe do 5.2 and you set the others to 49 and 50 something like that so yeah hopefully that answers your question it's the answer is is yes you could optimize your overclock but the answer to is it worth it is no almost certainly not except for competitive overclocking so next one this one is from Lamont who said hi Steve thanks for the informative content you produce been enjoying it since I stumbled into it researching my new builds what are the most basic benchmarking tools or the more basic and validation tests you'd recommend for someone who wants to validate system performance with an emphasis on gaming and maybe simple pre and post overclock performance to be primarily I don't want to become a benchmark junkie around hours of tests but I'd like to know if I've got any problems where my hardware is underperforming for example I'd like to determine if cpu GPU are being thermally throttled and see what impact it has if I open the front case door for more air intake when gaming or modify fan curves or locations I'm looking for the high school yearbook greatest results least effort winners here excellent way to phrase that question and I can help out with that so I think the the first thing I would point out let's see your thermal question first that's really easy is it throttling so the answer is you can use hardware info 64 it's free it's really good we like it I've never I would love to speak with the developer one day but we don't know anything about them other than they make really good software so hardware in foot 64 is the one I use GPZ is another one you can get the data you need just from hardware info 64 it's just it's it's a bit easier to use that for CPU and then use GPZ for GPU because um this is too much it's too much data with harder and for 64 to scroll back and forth so I'd recommend that and use that for CPU is it throttling well the cool thing is even if you don't know t.j.maxx you get two really obvious three four really obvious indicators if it is throttling one of them is a straight list of thermal target or thermal limit it says so it's a whole section that says like thermal limit yes no and if it says no all the way down then you're good and if it says yes anywhere and the thermal limits for any individual core then that core has hit the thermal limit may be 105 C on x2 99 1 on 109 on one of those CPUs a hundred ish on the average Intel stop CPU and then depends for AMD so that would be the the first list to look at this this talking after a cold thing much harder than I remembered the second thing to look for is also it'll highlight red if the answer is yes it's throttling second thing to look for is going to be the actual temperature area where you'll get core temperatures and if those are hitting t.j.maxx or too close to see it with thermal junction max then that is a problem and you're going to be thermal throttling probably or very close to it so TJ max you can either look up for the CPU using or you just kind of I think it tells you in hardware info in one of the columns you look at a distance from t.j.maxx that'll tell you how many degrees Celsius you are away from the maximum junction temperature that's the one you care about an ideal you see something like ten or more and then you're probably pretty good at that point as for other well there's other places you can see throttling in there too like just looking at the frequency of a CPU and seeing does it drop down if it does then it's throttling it's just a question why is it throttling because you forgot to remove the current or power limitations that's pretty common one as a throttling because the thermals is it throttling because of a turbo duration limit which would be stock in toss back for example AMD has it too so and those answers you can get through the other column where it says like throttle reason yes no and it's in words not in numbers in Hardware in 464 with Intel you could also use XT you which will give you some additional yes/no readouts on throttles I don't particularly like XT you but sometimes if we're really puzzled it can be the best way to get an answer just because Intel makes it so it does detect accurately in almost all cases but hardware for 64 on average better than x2 you at this point so short version of the first part the answer hardware from 64 second part gpu-z for a GPU it's harder to tell if it's throttling because it will completely modulate the clock all the time it changes constantly and what you're really looking for is a rough average just do a mental average and depending on the card on average under 84 degrees Celsius you're going to be running at boost specifications but if you get a sounds like 60 then it'll run maybe 25 50 megahertz high or something like that get it down below 60 down to 50 something like that 40 it'll boost even higher so is it throttling well sort of in that yes it is is clock gating based on thermals but it's not truly thermal throttling until you're beyond like 84 degrees on on the average NVIDIA GPU or 87 on the new ones and I am these target barriers I I think most of theirs are right around 80 something 90 somewhere in there so you'll start seeing thermal throttling pretty obviously in the frequency you want to cross validate with performance you can use 3d mark it's free as well at least the version you would need is free you can use that run the test a few times or run it on loop or pause the demo you can do that by clicking and and then just look at the frequency over time in GPZ and see if it kind of does that and levels out and you'll see that to some extent but you've seen it more than 35 megahertz then you might be throttling and you can fix that with fan placement as for validate and performance recommendations 3d marks good it's easy you can compare results online pretty easily so that that makes it nice you could use for stability I hesitate to recommend Prime at this point prime95 is is very easy it's a free application it's small you could use prime95 26.6 specifically that version for non AVX testing it'll be less thermally straining we have custom settings we use but for something easy I'd recommend just clicking the small FFTs button and clicking go and that'll tell you pretty quickly if you're stable under non AVX workloads that's really useful that's gonna be your average game as non AVX and it just generates a ton of heat takes a ton of power and if it survives that the CPU survives that for an hour plus that it's probably going to survive indefinitely maybe run it overnight if you're feeling adventurous or really want to be sure and then the next option would be probably 95 29 point whatever is current - I think is what we've often used there's newer ones and that's an AV X workload that's straining and if you don't have a negative AV X offset then it might not survive whether or not that matters to you depends on how much you use AV X workloads if you use blender blenders and AVX software AV x accelerated them you would want it to be stable in in AV X workloads but that's the most stressing you might not have a CP that survives that not not like physically kill it unless your settings are really bad it won't physically kill the CPU it just won't be stable and you'll have to down clock or something so those are the good options that I could get through without just making a whole video about this so hopefully that gives you a starting point short version hardware info 64 gpu-z prime95 26.6 and 29 point something small f of T is for that and 3d mark and if you want fur mark for power virus for the GPU be careful with that one and it's also not it won't enumerate the clock in the same way you're used to seeing because it has some protections and it so it doesn't kill itself so you won't see the same hi clocks you used to seen but it is a good stressor for like BRM thermals next one Ethan L when overclocking memory on a GPU does it follow the same rules as standard system memory as in your system memory always reads half of its actual clock speed due to reading at per channel also based on ddr3 DDR does GPU memory works the same way the spec page for my EVGA 20 atti FTW 3 shows 1400 megahertz but when in precision the clock shows at 7000 megahertz so by that math when increasing the memory clock by say at 350 is it functionally increasing it by 700 is that in fact the case great question so let me let me break this down for you if what you're talking about you're talking about effective clock and actual clock so what you're talking about for GP is so an example actual clock you might see 17 50 megahertz effective clock you'll see it advertised as 7,000 or 7,000 10 megahertz so if you have a gddr5 GPU that's rated as seven gigabits per seconds it's kind of the same as saying seven 7,000 megahertz it's interchangeable the way GPUs are rated then then that is an actual block of 1750 so the reason it's easy math seven thousand divided by x equals 1750 so it's four and this is because for every every clock cycle I can I can do it for every clock cycle gddr5 can achieve four data transfers and with DDR you you double with DDR eww just quadrupling so you can think of it as a formula of bass frequency your actual frequency 1750 multiplied by the number of sort of of pumps or of data transfers per clock it gives you your effective clock so 1750 would be 7000 1750 actual doesn't really matter which number you use just be consistent use the same one for everything and it's fine you'll also as I said often see gigabits per second and gigahertz used interchangeably in GPU memory video memory no rhyme or reason to it as people switch between them once one's kind of a rate and one's a speed so hopefully that helps a bit pretty simple I think did I answer every point of that question I think I got most of what you asked so yeah I mean that's that's that's what it is quadruple basically next one scooby-dooby asks so Unreal Engine is owned by Epic Games yes we're aware they're there just that way they're out the street Epic Games just launched a steam competitor 2 on PC that not only gives developers a bigger cut of the pie but also incentivizes them to use Unreal Engine by waiving the fee for its use in games released question exclusively on Epic Games store Unreal Engine heavily favors Nvidia architectures do you think this could be the beginning of a profound shift toward Nvidia we're pretty familiar with the the new store and with its incentive so it's it's cool to see it's good to see competitors team that's that's got a good chance now know is my answer though I I don't think this will change anything meaningfully unreal engines already used everywhere that it's not that's not a triple-a game so I think Triple A games you'll still see them using their own engines you're still gonna see far grind dunia still gonna see sniper on like a sewer or something you're still gonna maybe not triple-a but they're not gonna use someone else's engine theirs is like the best you'll still see Ubisoft on whatever engine it is they use I forget you'll still see modified crystal for a Tomb Raider so I don't think it will meaningfully really change that much Unreal Engine is very appealing to mid-size studios getting used by some of the larger studios of course we see Kingdom Hearts I think is is getting or was remade with it it's certainly big and we use Unreal Engine for some stuff we do here we greatly respect its abilities but I don't think it's going to really change the spread it is or introduce concerns to the GPU market just because Epic Games has a store now I don't think so I I'm also not sure that it's always true or well to what degree it is true that Unreal Engine favors and video architectures I know we've seen that but I don't know we're like what is that what does that mean you have to get two cards that are directly comparable so maybe an RX 5/8 and a 1060 you see the 1060 outperform the 580 or a 590 may be a better example because it probably shouldn't on average then yeah I would say there's an engine benefit there but I mean I don't know I don't think it's I don't know I'm not worried about this at all so my answer is no I don't think it changes anything next one Axios am I correct in assuming that I will be able to put a rise in 3000 series CPU in the am 4 slot on an X 370 motherboard obviously there's no chance of a BIOS update for PCIe gen four on that but I'm curious as to whether X 370 would for example handle the horizon 7 or 9 3000 chip when it's eventually released my understanding is yes you'll be able to use it on X 370 be mindful of vrm requirements and because those may go up with a higher core count CPU they will go up and mindful of BIOS updates but my understanding is yes you would be able to do so last couple here or last last to Andrew or says when doing live streams many of the astute overclockers among us give you helpful tips on which motherboard settings to tweak like v mesh can you go through some of the more obscure but still important CP overclocking settings what they do this was a hard question for me to think about there are really obscure overclocking settings and I scroll past them a lot but that should tell you something a lot of time we don't use them and I've spoken with builds worried about this and with they're Bauer and with Stephan Z and with Vince kingpin and there are absolutely settings they use that I don't use but there are also settings that really no one uses so I don't know how to address those I can't remember what they're called but they're obscured to a point of being meaningless in 99% of use cases so it just as an example build Zoid front of the front of the channel and does some work for us occasionally builds right in the past when he was helping with some memory overclocking I asked him ok so what did these ones do and he could answer that almost every time and he's he is extremely knowledgeable really knows his stuff but I read every now and then I there's a million tertiary memory timings and every now and then I'd hit one where he'd say honestly I'm not sure I just I make it as low as possible and that's completely valid that's all I needed to know is should this never be big or small I didn't need to know that's really how it worked now I would love to but the point of me saying this is that builds oi whom we consider to be one of the best overclockers out there also does not know every single setting there's a lot of them and they aren't all important so anyway to give you a couple examples of some that I think are not obscure but are probably reasonably obscured to a mainstream audience or maybe some of you so one of them memory times in general very important for overclocking very competitive overclocking not often tapped into by by most of the audience I think they most the audience thinking back to when I was kind of starting it's kind of like set the core ratio and then look up what's a safe v core and then set that I think that's kind of how people mostly do it so memory is really important T faulty f8w is a good one so this is for active window and it's the the length or it's a period of time where in the row activations happen four times in one rank that's an important one the more you can drop it and maintain stability your performance will improve a bit tre fi tre FC I don't consider those obscure I talk we talk about them a lot but if you're not familiar with or the average overclocking probably isn't then maybe obscure safe refresh cycle and refresh interval you want those to be heavily modified as well so if you're working with overclocking or competitive overclocking the important thing to remember is that you want the the memory to retain its information for as long a period of time as possible and to refresh it when it has to as quickly as possible so that's what our efi interval and re FC RFC excuse me will do so you want for example a lower TR FC and a really high TR efi like 30 2006 2000 whatever the board will allow for competitive overclocking not necessarily the best settings for everyday use but that's a good one too - no DRAM current limits maybe not that important but you can actually you can current throttle on your memory so you need to kind of max those out ia current limits the mesh is a good one you named so mesh voltage it sits for uncor on the X to 99 CP is really important on core be basically mesh on something like this Saturday on ATX II and that handles internship communications it handles the memory controller speed it handles the the cache so stuff like that io between the cores that's really important and that's that's mesh and probably not too commonly messed with I wouldn't consider it obscure but I'm hoping I'm hitting a few here that some of you that don't know about IO and SAE voltages those are important you really need to be careful about SOC 4 Andy you need to be careful about how high you set those voltages those the ones I can quickly degrade a chip and you won't necessarily know that it's happening spread-spectrum basically disable it it's my approach to this one spread-spectrum is supposed to reduce EMI but it does so I potentially the cost of a bit higher frequency cost of stability potentially and then I don't that's that's pretty good list that's about all I got right now there are way more obscure ones I apologize I did not pull out a boards and go digging through them but I guess my goal for this was obscure ones that we still use and then they're not that obscure but anyway last one steamed or as t-33 med said are we going to see more of hypebeast Steve I sure hope not I really hope not and that uh that's well the fact that I can barely talk right now let's just let's just run with that and end the show thank you for watching you go to store it I Cameron taxes net to help us out directly by picking up one of our mule mugs or mod maps by the way finally shipping to us the mob Matt so we'll have them at about ten days from now I think so they'll be going out soon or subscribe for more adrianna episode will be up shortly and a second one to the public channel if not already I can't I can't talk anymore so I'll see you all next time