Gadgetory

everyone welcome back to another ask GN episode as always you can leave your questions in the comment section below for the next episode and this time we have a lot of good ones it's been a while but we're back and I have several questions that have been excuse me we're Kim here that have been in the waiting so thank you for your patience let's get into it before that this video is brought to you by be quiet and it's straight power 11 series power supplies the straight power 11 PS use ship from 450 Watts up to a thousand watts accommodating most of the gaming PC build requirements you'd encounter and focuses on delivering a higher quality power supply that doesn't sacrifice on efficiency or stability noise is also a heavy point for the straight power 11 using a 135 millimeters silent winds three fan that can spin as low as 200 rpm for quieter low load operation learn more at the link in the description below so as always this is our Q&A section I'm currently filming at the old studio at the house right now with the og wood wall over there and the snowflake joining me and we might teleport to the normal studio in a moment first question is regarding the Kington card someone asked if you were mr. moneybags gamer how high would you run the power limit for daily use assuming temps were okay not in an LM 2 application end because my co-host has laughs we might as well move back to the other studio now so one second and we're back so the question about basically it comes down to is it safe to overclock GPUs 24/7 especially something like a kingpin video card where it's built for really high power overclocking and we had some thoughts on this but when I hadn't reached out to kingpin and ten anyway and we also had Jacob from EVGA and copy as well who's our main point of contact over there we got answers from all three so here's kingpins answer is pretty straightforward it was quote I would for sure max it out 144 percent or whatever allows the card to get the highest stable overclock without throttling simple enough the next comment from Jacob went a little bit more detailed on some of it and said if the viewer means power target just max it the card is only going to draw what it needs just raises the ceiling and on a normal cooling setup he would never reach a point that would be detrimental to the GPU regarding GPU lifespan under overclocking without mods there are enough safeguards built into the GPU to prevent any long-term issues and went on to say but if modded of course it's a bit unknown I don't think there is really good data on the long-term effects if you mod the card and then finally 10 Ilya responded as well and 10 does the the power engineering type work on these cards and also an hour lab tour you can see a lot of his equipment it's really cool setup Ilya said this is really two questions one is how power limits affects GPU operation increasing power limit allows a higher currents or power consumption however even at 400 watts this is 35% higher than rated design spec typically there is quite significant amount of headroom left as you know there is a process variance and some GPUs leak more some less maxing out power limits on stock biases 144 percent on KPE without touching voltage will not cause problems playing with voltage or load line however is a different story I'm still working on more details about voltage tweaking for RTX KPE which will be published on EVGA 0 C forum section with the cards launch and just intermission here it has launched since this answer was provided to us so that should be out there now Ilya went on to say the second part of the question is what are safe let's assume we want to keep the card in operation for two to three years at least power levels before the GPU starts having issues this is a bit of a gray area as nvidia and any other ASIC vendor Intel AMD etc will never publish this sort of data as it directly represents margins and process parameters which can often change over time from our experience if you keep voltage modest eg 10% increase or less then it's hard to kill or degrade the chip all that being said temperature is much more important as a factor than the power limiter itself as thermal stress affects not only the silicon die but the packaged PCB assembly and components life like caps fans pumps you would want to keep the GPU and 30 to 60 degrees Celsius region for maximum lifetime temperature is one of the factors why modern smaller nodes cannot operate at high voltages like the old-school hardware as nodes get thinner operating voltage levels will continue to go lower and power delivery will continue to be even more important in the future there are already chips out there that can operate from as low as 0.4 volts core voltage for comparison old cards like the 9800 GTX could work with voltage droop between vrm and GPU as big as 0.4 volts when doing zombie mods even with Allen teapots directly attached to the silicon die transistor density is so high that with high-voltage little transistors already overcooked by the time the heat reaches the ellentube hot service you may know that all transistors electronic circuitry on the modern FPGA chips like any CPU or GPU of the last 10 years it's located on the bottom side of the silicon die one facing the PCB answers the question very well so the answer is you're fine you can max the slider the more detailed part of the answer I got all of it but just to emphasize a few of his points heats of concern so Ilya mentioned that temperature is a point of failure for things like capacitors fans and pumps so to expand on that with a pump like an ace attack pump the spec for ASA tax coolers is no higher than 60 degree Celsius liquid temperature before there's some permanent loss of liquids through permeation and EVGA does use a stack as its primary partner for a lot of its cooling products like the EVGA CLC series that's all ace attack so the spec follows there but 60 degrees Celsius liquid temperature is really high you would need an extremely hot GP or CPU to reach that level of liquid temperature so I don't want people to get scared here and see that their CPU or their GPU is at sixty degrees Celsius which is actually really good if it's under load and then think that their liquid will be 60 C and freak out and think about it's killing the hardware it's not so 60 C GPU or CV temperature is way different than the liquid temperature will ever be liquid leave much lower than that so anyway keep an eye on liquid temperature some of the coolers out there like the modern ASA tech stuff the some of the kulit solutions have liquid temperature monitoring built-in they have little thermocouples in there in the liquid and you can often monitor it through hardware info 64 as long as you have that USB cable attached the same one that controls the led colors and you can monitor it liquid temperature that way if you're really concerned about it or you can stick a probe in yourself you have an open-loop setup it's much easier to do but then you're not dealing with the same type of issue because it's an open-loop setup there are a lot of different components in that so there's different specs for each one and obviously it shouldn't be too hot but permeations not really that much of a concern because you can easily maintain it so anyway that's one aspect the other is capacitors typically you have caps that are rated at something like 105 degrees Celsius for 5000 hours or 10,000 hours or you might have something low-end one that's 85 degrees Celsius for 2,000 hours caps not always gonna be that hot unless you keep it under a hundred percent load in a really thermally abusive environment so you don't have to worry about this all too much but on low-end components certainly the capacitor life is one of the most common points of failure so if you're in the habit of trying to keep things alive and the your you have a motherboard that's been perfectly good but fail there's something like that get good at replacing capacitors because that's probably the most effective way to bring a component back to life it is often the caps that fail depending on obviously what you were doing with it before then but anyway there's their answers I it doesn't seem necessary to expand on this further but yeah thank you Ilia 10 and Jacob for helping out with that I was straight from EVGA so hopefully that helps you out this would apply by the way to most other products as well so if you buy most video cards off the shelf you max up the power limit unless you're going to do what they're saying here like voltage mods you'll be fine so the obvious issue is that it draws more power just make sure you have a good enough power supply and you're fine how do factory tours influence your conclusions this was a great question so Addison Martin asked asked this I think this was a youtube comment I believe and Addison Martin asked asked GN how will all these factory to all these awesome factories where's thank you very much a vector judgment when reviewing new products going forward do you think you'll be able to give us consumers are more fair and in depth review absolutely so this is something we've talked about before but the really cool thing the behind-the-scenes stuff that hasn't made it into those factory tour videos is that we we get to sit down with these p.m. as the product managers and talk at length before we go through the factory about like okay what is all this stuff really cost and you know how long does it take what's the product development lifecycle what's the manpower cost what's the biggest bottleneck in the pipeline what are your material limitations do you have shortages or issues acquiring materials things like that and that really informs review of content because you know that probably the most common complaint in any review is well this is a pretty good product but to be competitive it really needs to be like 20 bucks cheaper or 50 bucks cheaper and it's easy for all of us to say that of course because we're not footing the cost to make the product and you look at the competitive marketplace and often it is a true statement to be truly competitive with existing solutions on the market products do generally need to be cheaper but to recover some of those initial costs they often launch at a higher price then we think is fair for the consumer but doing these types of tours and having that information behind the scenes obviously informs us to know that well okay there's a lot of moving parts here the price isn't unfair in that it's not asking a significant amount of margin over the cost to make the good even if it seems like it's unfair as compared to competition so there's kind of a few points a few angles of fairness there and one of them is is it fair to the consumer from the manufacturers perspective and then is it a reasonable price for the consumer to pay from their perspective and fair from the manufacturers perspective might be something like with maybe say a case or a keyboard maybe you're looking at like thirty percent margin might go up to fifty percent margin on some of these products and if you're in that range with that type of product you're feeling pretty good about but obviously everyone wants to profit more so it just becomes a question of how much can you push that margin before you restrict your market to a point that you have diminishing returns so you're moving enough less volume that you're making less money versus you know switching it a bit but that's not something that we really know a whole lot about typically but these tours do help us inform become informed on that and be a bit smarter when we say what we think was a shortcoming or we're not now another thing that these help with is we do learn what are the manufacturing limitations so if there is well I wish it had X great example USB type-c a lot of people comment on our case reviews when a case doesn't have USB type-c and say I wish this case had USB type-c in the front and we've certainly counted on it in the past now Patrick and I you know we've talked in the last episode about bias in the industry and biases Patrick and I don't really care about USB type-c all that much so when we do a case review it's reflective that we don't care like we barely mention it and some of you like it a lot but either way one of the things we've learned from these factory tours USB type-c not so easy and not so cheap so we did our factory tour of the cable factory we learned that the operators are able to make per operator a maximum of about 6 cables per hour and that's a skilled technician and it's because of the amount of manual wire routing involved so that's the bottleneck is the individual routing the sort of splaying the wires now other machines the process can make tens of thousands of these types of cables per day it's just at some point you have a bottleneck and six per hour per person force playing the wires is obviously a massive bottleneck we've also learned that the pricing for example unfortunately I can't give specifics on this but I can say that USB type-c is somewhere in the range of seven to ten times more expensive than just a normal USB three header on the front of the case and depending on how much it's been validated some of you thought you were making whereas comments in that video saying well but 10 times 20 cents isn't that much or 10 times 2 cents it's not it's a it's a lot more than that like I can't give you a number I can tell you it's in the dollars range not in the pennies range so this is a big deal like if you want to add USB type-c to a case it's not gonna happen on a $40 case it will barely happen on a $50 case if you make a lot of sacrifices like if you start cutting corners on quality if you eliminate fans then you can make it happen so this is not a cheap thing and that has informed us as well other things we've learned just like I guess understanding this supply chain so this was kind of a funny thing to go on a tangent here some people were we're sort of upset about the like how cardboard is made in house screws are made videos even though the screw and really well so the the kind of funny thing to me is these are the same people often who praise our content for going like really hardcore in-depth on detail and then when we do it for factories they're like why are you talking about how screws are made well because it's in it's in literally every single product we review so it's important to know about and that's part of like the GM like let's let's just go really ham with the depth on this specific Thane until there's nothing left to talk about ever again so what we learned though supply chains massive obviously and interestingly all these companies the the factories are pretty close to each other so for Coolermaster coolermaster works with a cable factory and i think it's i want to say it's in shenzhen they work with a screw factory in either like dong-gwon or shenzhen as well they work with a their main factories in hey Joe and if these words mean nothing to you they're different parts of of China and they're all within roughly about maybe an hour to an hour and a half of each other driving so this is important and it's it's interesting to learn because now it's like oh okay so Coolermaster makes the product they assemble it in this factory maybe who a Joe and once it's done away Joey it can it needs a box and so they go down the street 25 minutes to the box factory and there's your box and it's a factory that makes just millions of dollars worth of boxes annually for all kinds of companies and so there's all these different moving parts to make the product you ultimately get where by the time you get it there's been dozens of hands on the product from the box to the PCB that you may never see inside of the product and there's also been potentially half a dozen companies that have had hands on some part of the product and all of that is logistical II challenging and expensive obviously but you know it makes more sense still to do this than to for coolermaster as another example to make its own screws because that's just why why would they bother doing that when they go to a factory that makes millions of them a month and get them for far cheaper than using their own expense of floor space to put like a screw shop like a screw machining shop let's rephrase that so anyway yeah it's really taught us a lot and I don't know I think going forward what new perspective will we have for consumers I think that perspective we will have is I'll be able to to give you all more detail on you why does this product cost what it does and then okay let's go through some of the materials now that I know more about material costs time costs and let's see if it's fair let's see if it's actually a reasonable price for the manufacturer to charge whether or not it's reasonable for you to pay it given the competitive landscape we'd always look at it in a vacuum from that perspective so that'll help and then stuff like knowing timelines will help significantly because now I understand better like how long it takes to make different types of products and we kind of knew this before but now we better understand which part takes so long so tooling can take several months to get tooling made and finalized and can cost upwards of a million dollars to make all of it and and so that's valuable to know because now we know when we get a case in here as an easy example we know that the tooling for that case if it's not going to reuse has been finalized for at least six months at this point likely maybe three depend on how fast the company is at making their product so it's all valuable to know but I don't want to go too crazy on this but yeah I guess one other thing to the MSI radiation lab tour we did it was very educational because we learned you know about all these different certifications that companies have to get to sell in different regions of the world and that's tremendously involved I mean MSI as an example they have their own semi anechoic chamber it's not like the most advanced one I've ever been in but it's more advanced than most people have and they have their own radiation lab and these things aren't to slap a C label on their products but it's to validate that their product needs to spec they have a pretty tight tolerance for it and as long as it meets the spec what they then do is send that product out to be validated by a third party really expensive lab and get the government or the organisation stamps on it that they need so that was really interesting to learn as well as you know like all these small steps you don't even think about like a video card is tremendously complex obviously not even counting the GPU and you kind of take some stuff for granted like the i/o DisplayPort HDMI not something we really think too hard about and certainly not something we think needs validated at this point for the most part because it's so known but every one of these cards and motherboards has to go through validation and that's a lot of extra steps so what so what's the next let's do the next question that was a great one though I do want to talk about it some more but we'll stop there next one where does Gao actually quick note please for those of you watching this please leave more questions about the factory tours below I saw a lot of questions about the factory tours about like manufacturing and China being in China working with the companies touring the different facilities how automation works how different products are made and stuff like that please leave your questions below I saw a lot of them on the factory tours but it's much easier for me to collect them if they are down there in the comments section of ask to you you can also press on the patreon discord if you would like to so where we have the bonus episode of ask GM anyway yeah I would love to answer more fashion questions please leave them in the comments next one where does GM draw the line for depth atom McQueen said hi Steve I'm always wondering which level of knowledge a good tech journalist needs to make great content one of the advantages of G n is sometimes GN has much deeper dig into the implementation details other than just benchmarking but on the other hand I never see GN talking about source code level things like compiler ir which is commonly used in graphics drivers and would impact performance I totally understand thinking too deep into source code if there is would make G n more like open source mail list but too shallow also makes G oh not so GN so how does GN balance between a hardcore source code fpga spec level details against plain benchmarking although the GN method of benchmarking is already more scientific than a lot of other tech supers so good question driver level source code isn't something we're doing so I and I suspect you obviously understand that given your question where do we draw the line is how I'm going to interpret this question I think where we draw the line is largely based upon our level of competence and knowledge on the topic so over the years one thing longtime viewers will have noticed is that we've gotten more detailed and more confident in the details as we've gone so one thing that we've sort of phased out is in video card reviews CB reviews we basically never show a spec sheet anymore of what the product has respects like oh I'll talk through it for a minute or so before we get into the charts but we don't just show the spec sheet what we used to do is spend maybe five minutes talking about just basically reading the spec sheet and then might deviate a little bit and explain what some of these things mean but years and years ago my knowledge was an hour knowledge as a team wasn't there it's really expand on it so it was reading a brochure and I didn't really like that content approach we've now eliminated it we understand these specs at this point we've explained them in great detail and other content pieces but we've mostly eliminated going through the spec sheet because it's like we know enough now about other aspects of the product that if it's something you can just look up on our website or another website and look at a spec sheet there's really no reason for us to spend minutes reading on its it's so inefficient to read it on camera compared to just looking at it so that's something we've eliminated we've drawn a line sort of at the low end of like this is not interesting enough to get into the video it's it's not something we can provide enough value on because it's literally reading a spec sheet so we drew a line to the bottom end cut that out and made an extra couple of minutes of up to five minutes that we can spend on other stuff which is obviously very important we draw lines that we try to keep it at like 20 to 25 minutes max for most reviews for really important content I'll allow up to about 30 occasionally we go over 30 but 30 is like a really hard limit of it's a psychological limit it's kind of like when you see something that's $9 99 times 999 instead of 10 so 29 minutes versus 31 minutes there's a big difference in how many people click on the video so that's where we draw the line in terms of content length now that dictates a bit where we draw the line for content depth because at some point we have to cut it off for time reasons to make sure the video scene is to cut it off for financial reasons because if I want to sustain the channel then sometimes it makes more sense to split content into multiple pieces and have two 20 minute episodes instead of one 40 minute episode and this isn't just a financial thing either it's a we put a lot of work into this and want to make sure people actually see it because it's disappointing if they don't so sometimes that makes more sense in terms of depth answering your question if it's a topic I don't feel great about in my depth of knowledge and none of my team members know much about it then we'll do some research we'll try to talk to some experts if it's an important topic and maybe include some quotes from them but we're not going to expand on it from there because it's obvious when you're out of your depth and it doesn't provide any value to speak in vagaries and ambiguity and sometimes we do speak in ambiguity and vagaries normally it's for reasons that are related to NDA's or embargoes to try and get like a little bit of information out there without breaking any rules but yeah it's just if it's not something we know a lot about that's that's the line so obviously our knowledge base as a team is expanding as we grow and he's been working on things like RT ax and DX are in Unreal Engine so that has informed us greatly in that aspect of things Patrick's been learning more about the games he's testing the software he's testing we've done research on all these different software solutions for CPU benchmarks well any more about how case thermals behave and so all of this allows us to get more detailed in our reviews but it's a it's a really gradual growing process and just we're not to use your example again we're not going to jump into something like source code level discussion because we don't understand it so I could find someone who could put some words in my mouth but that doesn't really seem genuine so the most will do for that kind of stuff is if it's a really interesting topic it's likely I don't know what's interesting because I don't know enough about the topic to know that it's interesting so if it's an important topic always feel free to let us know Twitter is an easy way at Cambridge Nexus patreon discord is a great one you got a patreon.com slash gamers nexus tag me in the discord I see them I don't always respond but I do see them and read them and and then what we can do is reach out to some experts and hopefully get some answers and minimally mention their answers but might not going to death beyond that if we're not knowledgeable in the area hopefully that answers it though it's I mean it's it's pretty straightforward and genuine it's it's do we understand it yes or no now obviously stuff we understand doesn't always get in the video either just because of content length or maybe it doesn't really fit the topic or whatever but you can always post a simple comment or tag us on discord or whatever Twitter and let us know like hey I noticed that you guys didn't mention this this aspect of the product could you expand on that and maybe we'll find time to do so in a news topic or something when is replacing GPA it's worth it gam says how old does a GPU have to be before disassembling to replace stock paste becomes a worthwhile endeavor so dust mostly is the answer it is worth replacing it as soon as like the coolers just discussed in once you notice that your thermals start you're on a happy with them and it might be immediately then it's worth replacing it now some card is worth replacing right away depends on the quality of the card and in general the pace from factory is not always that great so you can typically get improvements by throwing like even just some cheap like Arctic on there RQ comics for or something will often be an improvement over the factory level application thermal grizzly of course is a longtime sponsor of ours I don't know if they're I don't think they're sponsoring we I think we've served their ads for the month but so cryo not if you want something really high-end and expensive and good that's another option but the the answer is is it hotter than you're comfortable with and also if you're overclocking it and you feel like you can get a couple more megahertz improvement with a lower temperature which every five degrees matters then it's time to replace the paste yeah I would probably go based on when you do your first dust cleansing of the whole system it might be a good time to replace the paste and just you know don't expect wonders but you should expect some improvement especially if dust is like getting in there it's you know it's small it can kind of get around the edges of the cold plate and really make things gross let's do just two more apparently this Sony a7 3r turns off after like 30 minutes so or doesn't it stops recording so we're redoing the last two questions here swifty bastard from from the GN discord asks Steve what do the lines at the end of your bar charts mean why are they only for 1% letters point one percent low bar is not the average for FPS comparisons so we answered this already in the GN discord which you can join by going to patreon.com/scishow cameras nexus put your questions I answered it previously but there's also a bar on the average one it's just it's so small because the deviation on that chart is basically zero for average that you couldn't really see it but so the bars at the end of the chart there's this little eye that was rotated 90 degrees and that eye indicates the sort of a bracket of our standard deviation or our margin of error depending on the test on the chart to illustrate when a benchmark is is meaningfully different and measurably different than another product that was benchmarked so the point of adding those was bring some reality back to the benchmark in space where people are so often attached to one company exceeding their succeeding over the other one that they forget that a difference of two fps might be completely margin of error and totally irrelevant now even if it's not margin of error and it's measurable it's still imperceptible in that examples are still irrelevant but for a different reason so not every test is the same tests are all imperfect and the variance for different types of tests can be quite large or it could be basically nothing like in the chart you noticed 0.1% low as 1% lows have a pretty wide variance because the nature of them is that it's a more restricted set of data it's excursions from the mean and so that ends up being where you'll have that that margin of error that standard deviation bar on the chart be a bit wider so if we see something like a result that's 80 FPS 0.1% low for product a and 68 FPS point one percent low for product B if the two numbers there are within that deviation of each other there with an error of each other you cannot fairly state that the one with 80 is better than the one at 68 because if they're within especially if they're well with an error maybe there's like a 12 FPS variance on that chart then it is unfair to come to that conclusion because you can't we wouldn't have the test resolution to declare that one is significantly meaningfully better than the other even if the data reflects that these two numbers have ones bigger than the other so the point is just you know when you're looking at the charts we do this and script around it so hopefully you don't have to do too much work but if you're looking at the charts and trying to determine which products better than the other take a look at those bars we try to get them on every chart sometimes we miss it but take a look at the bars and if you see two results are within range of each other they're really not meaningfully different like we can't we can't see the difference in in our testing procedure our testing method for that game so yeah what we do is we take all the test results we get standard deviation or we calculate margin of error depending and and then with that we can determine how much variance there is Rondon for each game for each product so if you have like a game that's scoring in the range of 200 300 FPS you could easily have depending on the game the engine a variance of maybe like 10 and so if we have a product that's at 300 FPS another product that's at 292 FPS you could say that the one at 300 better or you could be realistic and look at the error bars and see that actually they are functionally the same and that's normally the phrasing that we use when we describe results that are within error so that's what it just you know keep keep your expectations within the realm of reality because testing is all imperfect and so everything is going to have some error baked and not all results are significant in terms of their superiority over the next result down on the chart and you could have things that are fairly close to each other that are actually basically the same at the end of the day so not getting too caught up in really small differences and then I guess other than that we we build these data sets based on when you see a chart with like 20 CPUs or GPS or whatever and you see a review go up that might have multiple different games tested with a lot of different products there's thousands and thousands of cells of data there that we go through for all these tests for all these products and all these games and so on you end up with really complicated reviews there's really there's a lot of data to look at and start piecing together what kind of range you have for each game and should we eliminate this product or this game from the test because the variance is too high and it can't be trusted or whatever I'm sorting right yeah the answer is those bars illustrate the expected or the known test range based on looking at all the data we have and coming to a margin of error or a standard deviation for that title so that's the answer that one next one last one was smoke father two months ago said Steve looks like a retired wrestler I think you're thinking of hype B's Steve who last made an appearance at CES my long-lost cousin that's over this one as always if you want to leave questions in the comment section below I'll try to get them for next time if you come up with a question in two or three weeks from now and I want to make sure it's seen posted on this video not on a future video come back here and post it here so I can see it or you can go to patreon.com/scishow and nexus support us there and join our discord through the integration and post your comment in the ask GN questions section where i will look through for the next episode we have a patrons ask TN episode going up separately on that channel so if you want to check it out and you're a patreon backer you can get access to that or you can sign up and get access to it and yeah I stored our cameras access dotnet to pick up a shirt like a teal version of this one this was limited but there's still teal one out there or the mod mats thank you for watching subscribe for more I'll see you all next time