Gadgetory

hey everyone welcome back to another episode of ask GN now that we are back from copy text we had a lot of questions to dig through through the discord chat for patreon backers if you're not one of those you can go to patreon.com/scishow to join the discord chat a lot of fun with everyone on there and they've submitted some great questions but we also have some from YouTube from the previous sgn video before that this content is brought to you by custom backplate makers at v1 techcom the UN tech builds GP back ways to order with their online customization tool making it easy to theme upcoming PC builds back plates cost $25 and up and are installed via magnets and can be seen in some shots of the cards reviewing that lately used code gamers Nexus 5 for five dollars off your order or click the link below before getting started as always you can leave your questions below if you have them for next episode and right now the one I want to start off with is pretty timely and topical this is from I'm a Jedi bra huh this cord who asked am I allowed to ask a question about this image and he sent me an image that we can probably put on the screen and said given the 28th lane at CPU and 24 lanes on the chipset it seems to give you more lanes than x99s at 40 Lane CPU I'm curious to see if the grief Intel got was premature now regarding the the grief part of this this is in reference to a lot of the talk online the past week or so about PCIe Lane count on Intel's upcoming sky like ax and KB like xcp is mostly sky like axe really and how those CPUs the $600 one especially have a lower link count than you might have expected so the question being well if you add the lane count on those CPUs to the lane count available through the chipset why is there an outcry because that's still a lot of lanes so first thing to go through here is a primer on PCH vs. CPU lanes this is actually this is a question I've answered a lot and asked gns indirectly over the past probably six months or so but it's worth doing again in a more dedicated format PCH and CPU lanes are not the same at least on Intel platforms or and tell platforms the way Intel does it is that they have a certain amount of lanes on the CPU maybe it's 16 for some of them those lanes are pretty much reserved for graphics so those will go from the cpu straight to the PCIe device whatever that may be normally a video card in the first slot maybe you have two video cards and you can run 2 by 8 configurations or if you're running AMD and you can crossfire then you could run three but you'd be stepping down to it by four in one of the slots and you probably want some kind of pecks or plx chip on there but the way their chipset works is the PCH has a certain amount of HS i/o lanes or high-speed i/o lanes high-speed i/o lanes on the chipset are somewhat assignable by the motherboard manufacturer so let's say you have 30 HS i/o lanes as with the current G series chipsets if you have 30 HS i/o lanes some of those will be reserved but most of them you can assign out as the manufacturer not as the user and send to devices like Gigabit Ethernet maybe you use some 4 10 gigabit ethernet if you have the capability on a workstation board you might send them to other PCIe devices like MDOT - SSDs that might be on an nvme protocol or something like that but you can send out these lanes to different devices on the board you too would be another example and those devices high-speed i/o are normally going to use something like 5 4 so because of this the PCH can only assign its lanes and chunks of four so by that I mean if you wanted to pull lanes off the PCH and use them for a device in a full length 16 PCIe slot what you would actually be doing is pulling four lanes off the PCH or peeling off the lanes we say and applying those to the slot so the slot might be wired for 16 but it will only receive four from the PCH now realistically what will happen is as long as the CPU has lanes available it will assign all those lanes down to the PCIe slots for all your graphics devices pretty intelligent about that but if you exhaust those lanes and run to them and you have to pull something off PCH it's an a by for config that's not compatible with SLI SLI demand by eight at a minimum it could work with crossfire but your quality of lane goes down off the PCH the the arguments I think or the the concern or whatever it may be from all the the discussion online is that generally speaking you want as many lanes as possible out of your total lane count to come off of CPU it's a bit higher value because those lanes are going to be assignable to graphics devices and they're on the CPU whenever you're eliminating steps from the process and having to go physically else around the board it's always a good thing it's a bit faster so I think that's that's what a lot of it is because the value of the lay on the CPU is higher and for multigp you kind of need it but also I think what happens or what has happened with some of the grief as you put it that Intel got was probably to do with with perception of how these lanes are created where it looks like I'm just kind of interpreting what I've seen online here not making a comment but I think what people see is it looks to them like Intel basically throws a switch and turns off or on a certain amount of lanes for a product to potentially artificially create a product in a different price category that would otherwise be identical to a higher tier product with more lanes that cost more and there may be some truth to that I'm not 100% sure how Intel manages their lanes I don't know the validation steps that go into having more lanes maybe there is actually some cost involved there there probably is but is it the cost that we're seeing in these new products I can't answer that however I think that's where a lot of the discussion online has come from is that that understanding of the situation and the fact that the CPU lanes are higher value than the PCH lanes for things outside of HSI oh like graphics cards so hopefully that that primer on how PCH and CPU lanes are assigned and used helps out with understanding why you would want more CPU lanes even though the total CPU plus PC a Chilean count is higher but if there's more questions on this leave them below that is by all means not the most comprehensive discussion you could give on this topic but given that this is an osteon we've got other things to go through I think that's not bad to get you started next question is from Jesse Shelton who posted on YouTube and said I'm choosing a 1080i for water-cooling do the better VRMs on aiv partner cards actually make a difference for overclocking or are they just for show some say the founders Edition has ample power capacities others say it does doesn't so as specifically speaking to the founders edition 1080i we have a PCB analysis which includes a vrm analysis done by build Zoid on our Channel and as build voices in that video the 1080i founders Edition card actually has a pretty good vrm it is one of the better products that intel has deserved Intel's talking out into all the time NVIDIA has designed in the past year or so with regard to reference boards actually longer than that probably if you look back farther but it's a good board the founders Edition card is actually not bad the problem is the cooler the cooler is kind of bad but if you're going to buy it and water cool it then the board itself is pretty good already that said it does have lower power capability it will it will physically push less power through the circuits than some of the higher-end cards like you might get with a Strix or if you go really extreme like a kingpin or a lightning card but it depends on what you're doing how useful those extra phases and those higher-quality VRMs might be so with something like a Strix you get you're buying more than just the different vrm the custom PCB you're buying a better cooler than the founders Edition has some of that value is lost when your water cool in any way and with a kingpin or a lightning card there is definitely value to be had in those boards because if you're doing XOC stuff lnto or otherwise you eliminate a lot of the footwork where you would have to do hard mods to the card in order to get it to carry the same capacity as a as an out-of-box kingpin card something like that is eliminating a lot of the work on the FB cards but for just normal water cooling and for maybe some very basic overclocking by which I mean you go through EVGA precision or MSI Afterburner or one of those tools and you set some kind of offset to the core and call it a day founders Edition is really not bad it is fine for that application and is basically exactly what it's meant for with regard to what are cooling there are better options but to answer your question I don't think that you will really get a significant gain in your clock out of those alternatives that said if you're running at stock the aftermarket boards generally do have a higher stock clock rate than the FE card almost always and in those cases you would see the difference in fps it can be anywhere from six percent well it's probably about a six percent but if you're moving from EFI air to liquid-cooled other card or air-cooled other card like Astrix it's six to eleven percent so hopefully that helps with that one basically epi is fine if you're doing something that some fairly basic overclock with it and what are cool next questions from ryerson on discord Ryerson says in the deleting video there Bauer mentioned re pasting this is something I hadn't really given much thought to I know stock control placed on om computers is usually pretty poor especially after five or more years I've never really noticed any thermal issues with my own build although I don't really check once it's set up and configured is this something that an enthusiast would want to check on a long-term or regular basis or perhaps preventative maintenance item nobody really covers long-term effects with respect to thermal pace and I suspect that's why because there's not much hard data out there on the subject let's not wait true anytime you're looking at something that takes years to accumulate it's hard to really do research on it so yeah they're bauer mentioned repay Steen he was I can't recall off the top of my head if we were talking about repay Seng after you've deleted the CPU or repainting the outside of IHS but either way let's just assume that we're talking to a more normal application where you the CPU installed it's not been deleted and you just have some thermal paste on top of it and then your CPU cooler I have seen the effects of thermal paste aging to some extent with poorer products laptops come to mind we did some thermal paste changes on one of the labs that I was working on recently I think we even have a video of it and it saw a pretty massive change in thermal performance so that also involves blowing out the fans but that paste on the laptops a lot of them it kind of turned into like a crystallized really hard useless compound that all it does is inter interrupt contact between the two metal plates that want to make contact or the dye in the metal plate so yes definitely there are long-term effects where their own compound can age not all of it ages the dow corning stuff that intel uses inside the IHS i actually does really well with thermal cycling which is how you measure the life of a thermal paste by the weights in thermal cycling so that does really well the stuff i've seen is a lot of laptops is garbage and should be redone in every couple years especially because laptops are not airtight and dust gets in there if dust can get into where you're thrown paste is in your desktop or your laptop it will rapidly age of the compound you start getting little particles in there that interrupt the contact between the services and overall the aging is a lot faster and the compound loses performance so the best thing to do is test your performance using something that you're going to remember how to do in a couple years and repeat that test maybe yearly or something like that and just see how it's doing as preventative maintenance it should be pretty obvious if your performance really gets bad because you're going to start dropping clocks or hearing the fans go crazy to try and keep up with with the CPU temperature so the fan speed starts ramping a lot over time and it's louder than it used to be it might be if their own compound contact issue or a pump fail or something's or a liquid permeation in the loop or something like that about it yes it's actually real thing you should repay every now and then but it does tend to be a couple of years it's not like something you're gonna have to do more than once a year or even once a year that we're talking a couple years for most pastes the next question is from no twist on discord he said quick question if there is a temperature line for CLC's which is allowed by it well I suppose to rephrase this is there a temperature limit for CLC's which is ill advised to surpass is a hybrid GPU that is 70 degrees under load okay or will the liquid at that temperature give me headaches down the road going back to the focus on silent computing what's the max recommended temperature you would run a CLC at the max recommended temperature well the maximum temperature by spec for a subject coolers which is what you would find on a traditional hybrid bod is 60 Celsius for the liquid not for the silicon but for the liquid liquid temp is a whole lot lower than the silicon temperature but the 60 Celsius number is what ASX specifies for when there starts to be problems so those problems would include things like permeation almost exclusively where at 60 Celsius liquid temperature you start experiencing permeation into the rubber tubes at which point you've got less liquid in the loop so the pump has to work harder you're going to hear wine a lot more and your temperature will be worse over time because you're losing liquid out of the loop so 60 C is the number that you're looking for for a Sutekh products and other products like cool it or Kuwaiti although I don't have their number are not going to be dissimilar from that as for the max temperature for a hybrid GPU we've had pretty good luck running those EVGA hybrid kits specifically including their stock fan at 40% fan speeds and still staying below most air-cooled cards like less than 65 C for sure but I from top off top of my head I think most time we're in the 50 to 60 Celsius range out of 40 percent fan speed with the pomp at max so you should probably stop that down 10 percent also so you'll be fine there and it'll be really quiet quieter than everything else in the system on is you've got a fully passive like very quiet or passive power supply and very quiet or passive CPU coolers up forty percent fans beat is perfectly safe liquid temperature is well within control it's under 35 C and all of our testing but if your ambience really high it would still be under 40 C so that should hopefully get you started there the spec though 60 C is the number last question try watt toss it shun says I believe this is from YouTube says with new processors going to more than 140 watt TDP well air coolers continue to suffice while remaining same proportions dissipation is roughly thin area which is pretty much the occupied volume of the thin stack so we take a Noctua and 15 or c1 and scale by 140 over 90 5 or 190 over 95 in the case of thread Ripper that arrives at quote fucking huge hanging off your motherboard surely which is one of the most accurate specifications you could give for that system yes no hate on a iOS but my only worry about them as reliabilities because especially if one has a few systems to maintain do you think we'll be seeing a whole new generation of larger or more efficient air coolers or will those processors be limited liquid only there will definitely air coolers will still be fine they can handle 140 watts if they're designed properly and have a good fan and have enough surface area and things like that thread Ripper is different story I'm not really sure what thread Ripper will be equipped with I'm thinking that it'll ship with a stock liquid cooler kind of like what we saw with the FX 9000 series that's my guess so I'm thinking stock liquid for those I'm not really sure what the air cooler market looks like once you start getting into the higher TDP is past 150 which is what thread Ripper may end up in with at least with overclocks and things like that so I'm not 100% sure of the answer to that part of the question for 140 watts there will still be coolers they it should be possible to buy some that aren't huge I mean like the be quiet dark what is it the dark rock 3 pro dark rock pro 3 that cooler is gigantic as one of the worst mounting mechanisms I've worked with but it's huge like you're saying and I'm not convinced that it's within the Intel spec for some of the sockets that are like LGA 1150 exits so it would be fine for 2011 sockets so there's there's definitely a point where you can start entering into CPU coolers that should not be used on things with the smaller sockets like 11 5 X but keep in mind that the 2000 series sockets from Intel 20 66 now and the tr4 socket will have a stronger mechanism so that you shouldn't be as at risk of warping the board which is really what you're worried about when you have those higher weights and densities in a small area with the tr4 socket I don't have this back for it but if you look at the thing in its current state it takes three Torx screws I think make my b allen keys but three Torx screws to undo giant piece of metal with a huge really thick backplate like 1 2 1 2 3 millimeters thick on that on that backplate so it should be able to support a pretty large cooler but I do think those will be liquid for Intel the answer is yes you will still be able to get air coolers the thing you start running into is they might become a little less efficient if you were to run them out of fixed noise like if you normalized for noise between air and liquid you start running into problems with some of the air coolers on the market if you're trying to stick at something like 40 DBA but I agree with you that air coolers tend to be a little bit more preferred for an application where you really you don't want to have that one extra potential point of failure in a liquid solution like a pump or whatever it may be or just aging because those CL CS only really live usefully for about five years before you should replace them anyway so I agree with you there but I don't have a great answer for you I think we saw some of the coolers coming out of Computex are trying to solve for this problem some of them are going to be crazy with copper that seems to be one of the things this year which won't really help steady state temperature is a ton if you're looking at two identical coolers other than top in aluminum but it will help with controlling the fan curve and with ramp up and ramp down times so that we'll be testing all that stuff I have a lot of thermal testing plans for both Red River and skylake X not sure about KB like eggs but we'll be testing those so stay tuned for all that because we should have that information out pretty shortly after launch well the products so you can make your decisions before building the system that's all for this one as always you can leave questions below for next time patreon.com/scishow stuff is that directly or to join the discord chat where some of these questions come through it's just it's really easy to interact there and by and out there most today so easy to get some questions thanks for watching as always subscribe for more I'll see you all next time you