Gadgetory

everyone welcome back to another episode of ask GN the first one for 2018 this is a series where we answer questions from viewers and readers you can leave yours in the comment section below or you can go to our discord on patreon.com slash gamers nexus to join the discord and then leave questions in the ask en channel we've got a few good ones this week about sort of predictions which we try to stay away from but predictions about how mining can damage or if it damages PC components talk about if it is possible to quote fry a system from user error with regard to thermals and a couple of other pretty good ones in here as well PCIe lanes all that kind of stuff so before we get into that this content is brought to you by the Thermaltake flow RGB closed-loop liquid cooler which is a three hundred sixty millimeter radiator plus three 120 fans that are RGB illuminated if then we'll take it ringg fans at that this is a 4.5 done a stack pump which is one of the faster pumps you can learn more at the link in the description below quick announcement before the first question we are going to be at CES in about one week's time so we will be there the the event sort of starts pre-show on Monday ish there'll be a couple of meetings and and pre-show announcements and embargo lifts stuff like that and then goes through about Saturday so we'll be there for the event we mean with pretty much every company that you can think of that visit CES from the computer hardware side and a couple of other media outlets as well folks that you watch on YouTube so we'll have lots content make sure you're subscribed and keep an eye out for all that it's a big show for us we'll probably be doing two to three videos a day on news announcements while at the event and if you see anything that you think is two things here I'd like to request from you during the event if you see anyone cover something or you see tweets from companies or whatever Facebook post by companies if you see something really cool that we haven't already covered tweet it at gamers next as that is where I am most likely to see it during the show and we'll do our best to get over there if you have I guess that's the main thing really is if you have requests let us know especially if it's before the show because now I can plan it better we've got most of schedule planned already first question comes from cluster recluse for 1:3 who says what's your honest guess for the gaming versions of volt release frame would late Quarter 1 be likely so disclaimer here I don't like to try and predict things in the market for releases we really don't have that much more information than you so all I'm doing here with Volta and quotes is I'm going to give you some general ideas of how Nvidia launches products and generally when but that doesn't mean they always abide by those timeframes but that might be the one thing I can offer that you may not already know so first of all Volta will almost certainly have large parts of it grafted into the gaming architecture which we we think we know the name of it but you can't talk about that yet but the gaming architecture is not going to be exactly Volta as we know it today I don't know what's going to be removed and what's going to be added but it would be reasonable to assume that Nvidia is following the same the Lawtons design philosophies for the gaming architecture that would probably include for example a focus on asynchronous compute performance we saw a big uplift in Volta I would suspect that'll carry through to the gaming cards so that stated being that this is not a direct v 100 or the Titan B will not turn into an 1180 TI or whatever it's going to be modified in some form before then but the thing I can tell you is that generally speaking and video likes to launch in two primary windows they tend to launch things either at GT c or there abouts that would be their trade show the graphics or GPU technology conference that tends to happen in May each year just before Computex if they don't launch things there they sometimes launch them near GDC the Game Developers Conference but a lot less often way less off I think it's happened twice in the last couple years so May would be the first window to look at for something like that at least an announcement normally they announce it's changed its it's kind of been alternating the last few years they'll do a gaming architecture or a gaming card and then they'll do a deep learning card or architecture GTC has been trending more towards the business and enterprise side of things so I don't know if we'll see the trend of gaming cards being announced their continued but they do like to announce things either at their own show or at an event near a main event but tends to be their own show so that stated a direct answer to your question I would say almost certainly not quarter one but this is me going off of just experience not off of any special privileged information so I don't know that much more than you quarter one seems unlikely to me I think a sort of mid-year launch is more realistic typically and video looks at either just before Computex or just after four product launches the GTX 1080 was something like May 23rd or something it was right before Computex and and that tends to be what they do so sort of that yeah there's really only two launch Windows for flagships that companies do it's it's basically May or it's like August ish if it's not there then they're doing something pretty different from normally so yeah that's that's what I got for you March was a bit anomalous friend Vidya that was primarily because AMD launched rising in March so they want to put out a 1080i because they had no competition for GPUs at that time so they wanted to grab all that market share from the rising system's next question Sergio Gonzalez says the mining of cryptocurrency can damage the PC parts if you work in the same PC and mining question work so let's look at this from perspective of why do components get damaged regardless of how they are used does it doesn't really matter that much what kind of work under they get damaged from heat heat causes a few things he causes rapid degradation of parts it is a catalyst and he also causes things like fires so in extreme scenarios you could have not a fire like a house fire but a fire like a component like a power component just popping vrm pops it's kind of what you would say so a good example would be a capacitor capacitors a lot of them can take about 105 degrees Celsius some of the better ones might be rated for 10,000 hours at 105 C some of the cheaper ones might be rated at 5000 hours for 105 C these are real numbers for actual capacitors you find on boards some of the B 3 actually a lot of the be 350 motherboards use 5 K capacitors at that's at 105 C so if you're running higher temperatures doesn't matter if you're mining or not for prolonged periods of time you could be eating away at the lifespan of those components you are far more likely to chew away at the lifespan of capacitors than a MOSFET unless it's a bad batch of MOSFETs or bad batch of power components like EVGA encountered with their a CX card just a little over a year ago at this point and that's unlikely though generally you're looking at other component failures VRMs vrm components like chokes and MOSFETs can take 125 150 C really before they start D rating hard so yeah caps would be the thing to look at and this comes down to how the coolers design so each card is going to be different cards that don't have a good heatsink contact to those power components are going to be under a lot more duress so if you have cards that you can adjust adjust the fan speed to keep the GPU core temperatures low they're not cooling the power components you can still have a catastrophic failure and when they fail primarily will hinge on how hot those components get that are rated for lower temperatures like capacitors 105 C is relatively low compared to other power components so it's it's not going to be an instantaneous fail another source of failure we've seen so I was looking at some photos of mining machines that had burst a flame via Twitter and if you look at a lot of those they tend to fail from overdraw on the power cables so that's a scenario where you are either exceeding the current rating of the power supply for that rail or you are exceeding the current rating for what the gauge of the wire can withstand if you're using and this is true outside of mining as well so if you're doing high-end overclocking with power mods or hard mods on the card like we did with Vega 56 if you were to then take a smaller gauge wire that's not rated for as much current and plug it in there's a chance you could burn out the wires and hopefully you have a good power supply that can help deal with that sort of thing but that's generally a type of failure you see is a overdraw or power component failures it is possible for the GPU or the memory to burn out but unless you've modified the card in a way that has interfered with the contact of those components like HBM to the cooler it's pretty unlikely the card has protections to throttle itself once it gets warmer unless you enter a runaway thermal scenario where you've got racks and racks with these things and they're in a high ambient operating environment 4050 C unless you're in that scenario ice power components are still where you look for a failure and then after that probably fans how many hours the fans are waiting for you're running fans at high speeds for long periods of time you're gonna chew through they're rated lifespan a lot faster and when fans fail other components will fail unless they have a mechanism built in to trigger a shutdown which I don't I can't name a card current gen that does that so yeah there's the main things that fail this is it's a really interesting topic and it's probably worth a completely separate video with some testing because I've done some testing on this kind of stuff so I will leave that there for sake of getting to other questions and there's a lot I've left out here I'm sure there's stuff that I'm not thinking about right now that we could include or talk about in the future but we got a roll through some more questions so if you are interested in that topic more not just cryptocurrency mining but hardware failure from load in general let us know and we'll look into it I'll note one more thing here with regard to any amount of sustained load whether that's 3d rendering or mining you can stay a lot of the keep a lot of the the negative effects at bay by modifying the card with just a simple power offset if you tell us to draw something like 70% of its power it is going to draw well below what it's rated to handle those power components should be if you've left it unmodified and it's not a terrible terrible cooler design they should be fine so if you are really concerned about your rendering machine or your mining machine or whatever with high thermal environments if you're concerned about it burning up consider lowering the power target increase the fan speeds make sure there's direct cooling on the back of the card or somewhere to get rid of radiative heat off the back side of it and then you can of course tune clocks if you had to from there but under volting of course also an option although that doesn't really work that easily a lot of the time especially with Nvidia cards but easier to do through power play tables so yeah that's that's some basics for you but the thing to take away here is that heat is the catalyst for everything bad in computers it's or unclean power I suppose you can pretty much solve all of the things you'd be concerned about with good cooling solutions and with good power supplies and that I once you've done that there's not much more you can do if you have fire concerns a good power supply will solve a lot of those and cooling will keep the heat away because heat causes fire and it's not like it's not as literal as how high of a temperature do you need for metal to catch on fire because it's really freakin high it's more a question of what is catalyzing from that heat what what's going on what chemicals are breaking down what power components are starting to fail and then what is there catastrophic failure mechanism is it popping is it like bursting into a puff of smoke and brief flame pretty uncommon things you don't really need to worry about them but that would be kind of the the catastrophic failure I think you're talking about you don't really have vrm components when they fail it's not like you start getting an artifact ting you don't just start seeing memory artifacts or texture artifacts in your graphics it just dies it goes black dies shuts down whatever there's no gradual failure of the RM component it's just gone so that's kind of I think that that should give us a good starting place to look into more later I really like that topic a lot of what we do is based around it like thermal testing power testing so I would love to talk about it more next question is from honey force who says Steve is it possible to irreparably fry a system through heat by gross user error like not using thermal paste forgetting to plug in a pump or fan and then gaming I'm pretty sure there are some built-in protections but can't recall hearing a trustworthy answer not true at differences between short and long term exposure pretty sure there are average consumers out there who throttle for months before has a need to clean the heatsink etc so a lot of the previous answer feeds into this one so that's helpful but briefly I've done a couple of these things in the past a really common one is forgetting to plug in these SATA coming off of like NZXT crack and coolers because then your pump doesn't turn on you notice fast because it'll throttle hard the tubes get physically hot to the touch things like that there are built-in protections not every BIOS and CPU behaves the same you can often configure them at least in the higher-end board common failure modes would be when t.j.maxx is hit on the cpu which would be your junction temperature maximum that's what the J is this Junction temperature as opposed to case temperature so this is more or less than inside the CPU temperature as opposed to an external probe or sensor so Junction temperature when you hit max 105 C on X 299 stuffs HTC PI corrigan CPUs when you hit max on that your motherboard will dictate do I shut down or do i throttle oftentimes it's it's the latter its throttle and so have it throttles you're just gonna lose performance drop clocks hard it'll go down to 12 and then 45 and 12 45 over and over and over again for the multiplier and you'll notice that another option for failure is a thermal shutdown that's just cuts off goes black that's it so those are a little annoying for even if you're an experienced user I've had thermal shutdowns recently on production systems that I've used and it's not always easy to troubleshoot it because it just goes black so you don't necessarily know what caused it you might think it's something else first as for the question is it possible to damage stuff through gross user error certainly it is if you run a GP a GP is a great example because it will not cause the entire system to shut down unless there's something really wrong so a GP it might be a scenario where your maybe you reassembled your heat sink and properly there's poor contact something like that if that is overheating and it's still trying to operate at full capacity okay better example the heat sink contact to the die is perfectly fine and it's keeping it cool but something's wrong with the contact of the VRMs there's no mechanism in there short of erm throttling if they have something that will keep them cooler as opposed to the core if the core is running too hot it will throttle itself to a point maybe down to even two D clocks if it gets bad enough but the the short of it is a GPU or component like that is more likely to have some kind of frying using your word than something like a CPU the motherboard VRMs could also be damaged the damage tends to come in the form of loss of lifespan of things like capacitors refer to the previous answer but yeah it's always possible to cause damage from heat you try to avoid those scenarios but also if you've already had a scenario where you forgot to plug in the pump for a while or something like that I wouldn't stress about it I mean if you if you realize the error and fixed it and it seems like everything's okay just just move on there's nothing you can do at this point if some damage has been caused you might not know for years if it was just a capacitor overheating or something like that if serious damage was caused immediately to the component you would know you'd get artifact in or you might not be able to sustain the overclock that you previously sustained that's a pretty common one for people who overdo their asset their system agent voltage sa voltage or also known as VC CSA or if you are running a be 350 board with a high capacitor temperature you'll eventually start getting degradation or I should say hi mom spent temperatures get degradation of the overclock so those are things to keep in mind as well but yes it's possible it's just yeah you have to really do it pretty bad have that happen quick note here Ivan I wrote down a note for myself to address this one so for our content we were diagnosing the Titan XM where I took apart two hybrid coolers to liquid coolers during the diagnosis a lot of people were like there are two comments I wanted to address one is why did you destroy two perfectly good CLC's for this and the other one was why didn't you just weigh them so let's talk about why did you destroy first I did not destroy them you can refill them they're fine also separately that's kind of our job there are a few reasons I would disassemble to liquid coolers so the contacts here is we took about two liquid coolers and measured the amount of liquid in them for a permit a very quick ad-hoc permeation test it wasn't that scientific and the reason you do that one is so you have a baseline you understand how much liquid are we expecting from a new unit to let's say we thought there was a problem with the first one and I just say I think that it permeated there should be more liquid well if I'm wrong odds are that is a specific enough test that someone else probably won't check it for a while and you don't want to be wrong and how people believe something false or inaccurate about a company's product because that's not fair so it makes to take apart a second one and sacrifice it or in this case just empty it and refill later to make sure that we were correct at least sample size of two is small but it gives you some redundancy just to say you did due diligence to make sure you're not hurting a product's reputation when it shouldn't be hurt so that's part of it the other part is again my job is to review components not to amass them and use them personally so it's it's not like it's not wasteful to take them apart it's part of the job and it's you know it's it's not like that component would have gone to you for free if we didn't do it so it's really not any loss other than making sure the contents somewhat accurate next question why did you not weigh it weighing the coolers assumes a lot of things primarily it assumes that the the manufacturing tolerance is very tight no error because if we weigh them both empty you're probably going to find disparity just in the radiator just in the tank the the liquid is a mixture of glycol and distilled water it's possible that those mixtures changed between runs it's possible that the well really a lot of things a cold plate comes from another supplier that's not a static or evj in that case so that could have changed the mass of that could have changed there are a lot of really small things and there's going to be tolerance for all these whether or not they actually change the run yeah look take something really minor look at Noctua they had that fan controversy last year with the China versus Taiwan made fans that was a color change and it's because the factory ran out of pigment or whatever color for the first one and they made another batch and it was slightly off that's just color so that's why we don't I didn't weigh it cuz didn't really need to and also side note I need to see if the cold plate was gunked-up or if there was a gunk in the tubes or anything like that next one here mega scrapper what will happen if you use a modern desktop CPU without any heatsink will that thermal throttling esily or will it simply refuse turn on because there's no cooler detected this was aggravated by mark wolf in the comments so they will generally turn on we've done it actually some motherboards do have pressure detection where they'll refuse to boot or shut down if the inadequate pressure is detected but we did this with thread Ripper we actually turned when we first got threader before we turned it on without a heatsink so we take a thermographic image of with like we had Sharpie dit out so there was no emissivity issue but we took a thermal graphic image of the IHS which allowed us to see which spots were active which then allowed us to see which dyes were active as opposed to the other two which are silicon substrate interposers and and determine back then the question was are all four active or are two of them what people called dummy dyes also known as silicon substrate interposers and that allowed us to do that so the answer the question is you can still turn it on but it will eventually shut down or get dangerously hot and depending on where you are in the boot cycle it might not protect you from that next question Judas says a Judas Iscariot says it's kind of ridiculous that they're launching new products called risin plus or whatever that's effectively the same product but fixing the platform issues memory compatibility better clocks etc of the original rise in line I wanted to address this one because I can understand product segmentation to a degree it gets a little ridiculous sometimes see end video with their recent data center decision but this kind of product segmentation i wanted to talk about even from our perspective taking an existing products that you have and then trying to find a way to reuse it and build upon it makes a lot of sense fab plants can cost upwards of five billion dollars andy is technically fabulous now so they don't have that cost but they still spend a whole lot of money doing R&D nvidia brags about spending a billion dollars on R&D you spend that kind of money you can't just launch one product and you're not gonna profit on it and take a micro look at it from our perspective we do a lot of testing there are times when actually a lot of the time you don't want to throw out all your data so what do you do okay we we test all these GPUs for a review so we have a review coverage piece we can do and then maybe after that we can do a separate overclocking piece maybe after that we can do an SLI piece and you can do stuff like explicit multi-gpu with two different cards and you have all this comparative data that you kind of a mass over time so you're able to reuse large portions of that to get a profit on a huge upfront expense which would be the initial review because otherwise it's simply not profitable and from a much larger zoomed out perspective your company like AMD or Nvidia or Intel you have billions of dollars of expenses yeah you probably want to launch an iterative step afterward not just go full new architecture and also keep in mind that fab process the the shrinks the process shrinks are not as common as they used to be and they're getting a lot more difficult so if they want to stay in business like how to keep launching product yes it is it can get to a point where it is anti-consumer certainly but that's not always the case it's just you have to judge it product by product and make sure it's not something that screws you over the the real question is does this version of the product add something meaningful over the previous version and could they have added that to the previous version just as easily depending on the answer there you can kind of decide an example of segmentation that kind of makes sense is professional GPUs versus gamer GPUs where you're dealing with things like certified drivers professional support and product segmentation specifically with high-end enterprise products to try and make back your money at licensing fees do that to a degree as well or licensing rather not licensing fees I don't like the idea of imposing a EULA to force users to use a product certain way companies don't generally tell you how to use the product but it's part of the game ideally that particular the game doesn't exist but there are reasons for for this type of product launch to happen and that's just you know not taking andis side on it or anything but just realistically all the companies will need to do that just like look at cases ninety percent of the cases that you've probably seen in the last year share tooling it can cost a company six hundred thousand dollars a tool case and they got to make money on it so they're gonna do that by reusing that tooling alot and changing the shell and the injection molding on the outside or backing or whatever they use let's do one more last one there's a lot of good questions I'll do two more quick one David says this might be slightly off topic how does your usual upload look at the impression that a lot of the smaller youtubers spend some time in the comment section like and responding to comments after the upload is that true yeah I think objectively you can look at it and it's true we do it for a few reasons one of them is because I like to try and encourage interaction in the community that is hopefully more meaningful than just memes and jokes things like that I try to keep the signal-to-noise ratio as clean as possible and as you grow it's it's more and more difficult to do that and I'm okay with chasing away people who don't really belong here if that means we can sustain a community that is more mature and more focused on product analysis and you know development and engineering and all that stuff I'm okay with making sure our audience is focused there and being in the comments helps a bit also give us ideas it helps me know if we need to make a correction or anything like that but yeah we spend some time there the next one there are so many options okay so let's do this one quickly Ryan Frick says our number of PCIe lanes available a limitation of the CPU or the motherboard that went and Paul made a comment in passing about the rise in 400 series chipsets possibly supporting more PCIe lanes I thought rising CPUs were limited to have less available compared to Intel processors very very quickly because I don't want to spend a lot of time on it at this point but we'll talk about this more in the future very quickly the CP on the chipset both have PCIe lanes for 400 series chipsets what probably what Paul and Kyle were talking about I didn't watch it but I would assume they were talking about the PC is IG changes to the 400 series chipset those would include moving the general-purpose PCI Express Lanes these would be lanes you can assign to anything as a motherboard maker it is your choice to say you know what let's pull these four lanes off and give them to a Gigabit Ethernet controller let's pull a might not take four lanes but just an example let's pull these lanes give them to MDOT - let's pull these give it to you got to pull a whole bunch of them give them the SATA Express or whatever so it's their choice 400 series is going to jet generation 3 for the general purpose lanes I think there's 8 of them as opposed to generation 2 so they are going to be roughly 2 times the bandwidth per Lane and therefore you could pull half the lanes to achieve the same thing as 2 times the lanes previously so if previously you need it as a make motherboard maker four lanes to achieve a certain speed on your MDOT 2 device now you built two gen 3 general purpose lanes and achieve the same speed so that's the change that's happening my account is primarily dictated by the CPU that is what communicates directly with the PCI Express interface and it's slots that are hosting your graphics devices and that's dictated by the CPU it also has some generation two lanes depend on which one you're looking at the PCIe lanes on the chipset go to all the other high-speed IO devices HS IO as Intel calls it and those are divvied up in a lot of ways but for Intel the PCIe lanes on the chips that can never be assigned in greater than groups of four to a particular device so you can't assign chipset lanes to a graphics device and have it be by eight it would have to be by four which means that if you're powering crossfire you're fine but if you're powering SLI off of as e200 series CPU which has remember right now but 24 lanes I want to say something like that it's off top of my head but it's not 32 lanes basically so you would have to do an honest a lie you'd have to be limited at two cards for SLI you can't do by four for SLI but that's the very basics of it there's a lot more here as well I'll save the rest of these questions and we'll do another one we will do two this week or something CES is coming up after all thanks for watching as always you can subscribe for more if you'd like to help us out directly you got a patreon are complex gamers Nexus or you gonna store it on game designs it's not net slash mod matte let's think about mod matte like this one I'll see you all next time