Gadgetory

everyone welcome back to another ask Jian episode as always you can leave your questions for next time in the comment section below and we also are doing a part two of this video so there's a two-parter this week this is the main one we'll have sort of an after-hours or extras ask GN segment that is going live for patreon backers only and we'll be publishing that on the patreon page so keep an eye out for that one we're gonna film it right after this before that this video is brought to you by Thermaltake and the view 71 enclosure the view 71 is a full tower case that's capable of fitting three video cards and most configurations it's also one of the better cooling cases in our recent case testing bench lineup the view 71 has hinged a tempered glass doors on either side that make it easy to open and show off and it comes with at least one rain fan though you can get the RGB version if you prefer learn more at the link in the description below I wanted to start this one off with just a quick news item on GDD our six because it'll probably be a day or two before we do our next news video and this is kind of noteworthy so before getting to questions g-d dr6 has had some news lately the newest bit of info is from SK Hynix who are noting that their GDD are six chips are available for purchase now from the partners who buy them to make stuff so they're selling the memory modules at 10 12 and 14 gigabits per second this is a bit slower than the modules that Samsung has talked about however Samsung's aren't available yet and they were still pre-production when they last spoke of them I think Samsung was indicating maybe as high as 16 to 18 gigabits per second for its speeds whereas hynek's is capping out at 14 here and other than that the speeds risk or the bandwidth respectively would be 40 gigabytes 48 gigabytes and 56 gigabytes per second for the hynek's modules at 10 12 and 14 gigabits per second so that's kind of what we got for now voltages are 1.25 or 1.35 depending what it is and we'll talk about this more in a news item sometime soon but I want to put that out there because it's kind of interesting news and it has implications for the GPU market we already know that new GPUs are coming sometime this year so that sort of I don't know we'll see if they actually end up on the cards or if that's next-generation but either way they will be ready for purchase by manufacturers so first question for this one was pretty interesting this is from YouTube commenter Ben Grogan who said open van GPU heat sinks run their fin arrays transverse on the card is there a reason why this is the case beyond it longer heat pipe runs would there be any benefit to longitudinal fins which would allow the case front fans to encourage front to back airflow so as I understand it the question is basically is convenient questions basically why are these fins running up down versus left right and I talked to a few people about this including VSG from thermal bench and I think our answer is all pretty much aligned which is it's the same idea as for radiators for liquid cooling where you're in this case you're trying to expose as much of the primary heat transfer heat dissipation element as possible and in the case of an air-cooled heat sink as opposed to liquid really what's doing most of your work is those heat pipes so these heat pipes in this instance are all running left to right and that layout allows the heat pipes to be as long as they can be going top to bottom looking me shorter left to right if you did the horizontal fin layout you wouldn't get as much contact to the heat pipes and that's kind of from what I understand a bit less of a concern than just the heat pipe run length in general and to catch our one-up I'm sure most of you know this at this point but heat pipes do a phase change so that's why they're so effective at cooling there's basically a liquid in there we've talked about its composition in the past and content pieces there's a liquid in the heat pipe when it gets over the GPU in this instance heats up turns into a gas that's a phase change goes to the other end of the pipe cools down condenses uses capillary action to come back down the heat pipe as a liquid where it repeats the process turns the gas again and so in that phase change process going from liquid to gas and back you're losing a lot of your energy in the form of heat and that's what really does the cooling so to get the most benefit out of that it's got to be a longer pipe and running across the card this way is obviously going to give you a bigger heat pipe than up and down so if you go across the card can't really have horizontal fins as well it would expose to the front case fans better but those are doing such little work compared to the fans on the card anyway perhaps with a passive card maybe like a really low end passive card with no heat pipes or something maybe that makes sense I could definitely see that making sense in a small form-factor case or something that's just getting really putting a lot of requirements at that point getting really niche but other than that let's see bite any notes from VSG so VSG told me for his own take on it saying as he says it's the same as for liquid radiators the heat pipe and coolant tube has to be on the longer side of things to allow for as big of a delta temperature as possible so fins are stacks perpendicular and sometimes in columns and rows if there's room for more than one set of heat pipe or coolant tubes you want to give the best option possible for the coolant medium so that it will do more work in this case it would be the phase changer in the heat pipe if you have longer fins that's also more solder points and logistics of making more fins versus more heat pipes also comes into play the solder points are up on the weakest point of transfer in an air cooler like this it's typically 10 there are other types of solder though and I haven't really thought up without any manufacturers on when they use 10 versus something else so that's something we dig into more people are interested but yeah that's your weakest point of contact and then I think everything else that has pretty much been inside at this point so that's most of it very good question though I think we can count that as answered for now but if anyone has add-ons to that question let me know or if you're just interested in air cooling parts in general we also ask the EVGA about the heat pipes and they had a pretty good point here they noted that for the heat sinks they want the shortest distance for the hot air to be exhausted so that's accounting for the fans on the top making sure that when they blow air down into the card it needs to have a clear path to get out of the card and having longitudinal fins would stretch that out and make take longer for the air to get out of the way so I was a good point as well next question is it's from ten teraflop APU barrier who said ask Tien sorry if this has been asked before when benching the performance of a baseline GPU boost 3.0 enabled card is it fair or should these cards have their boost clocks locked to the manufacturer specified boost maybe I'm misunderstanding something but it would seem that you could see exaggerated performance gains as now a stock card is effectively being thrown against an overclocked one so to be fair the stock card is also technically overclocking / boost spec the boost number that Nvidia gives in this instance it because we're talking about 3.0 that's their thin the boost number they give is not normally what you're actually running at it tends to be higher if there's thermal and power Headroom I believe it'll be lower that's the reason there's a base and a boost it can be lower than boost under the bad power or thermal conditions that may be encountered on the card it should not ever be lower than base that's the the actual clock specification unless you lower your power target but to answer the question should you basically normalize cards you can definitely lock the clocks it's not very reliable it tends to be a little bit it's unstable - like with K boost or something so I don't really like doing that but also you have to realize that when you're reviewing a card anyway the silicon variance GP - GPU is the primary driver and differences in how are they overclocked it's good to have a strong vrm it doesn't necessarily matter that much because aside from a select few cards like kingpin cards there's really not a lot of bending that goes on and binning is kind of a misnomer - because even in that case what they're really doing is just pre-selecting so they're testing the chips they find the good ones and they put it on whatever the spec is for the card so for example I'm just using EVGA here because I know what they do if you have something like an FTW three high-end card and then you might have a lower end DT card in the same cooler design and same PCB what they do is they'll run the cards all of them through testing when it's on the PCB and pass or fail if it passes it becomes maybe the high end up tw3 if it fails it becomes the DT Series with a lower specified boost clock so the point of all that is to say that silicon quality varies a lot more than really anything else that the AIB partner can do for you aside from the cooler quality so when we're testing a card what we're really testing in terms of its clock performance is one what's the quality of the silicon they used which is not really part of the review because that's basically random and it's controlled by Nvidia more than the board partners the next part is how good is the cooling solution and that's what really determines how the frequency will fluctuate from boost parameters that we're talking about here so to normalize these at a fixed frequency or something like that would really be taking away the advantage of a better cooling solution or power delivery or whatever it may be that establishes the real differences in the cards and ultimately again silicon variants per Nvidia's manufacturing run will vary more than really anything else anyway but yeah it's a good question from a pure if you're trying to collect really pure data it could be interesting but you'd have to have a big sample size to have any confidence in it and and ultimately we are testing the entire package the whole card so to do that properly does require not really messing with those things for a stock baseline test because you are trying to see how much Headroom does that cooler give you for its clocks anyway but yeah that's I think that's that's kind of most there's a lot more to say on this when I've said it a lot in the past too and GP reviews but boost makes things difficult and silicon quality tolerances make things difficult so that's why when we're reviewing for example a bunch of 1080i is a 10-7 tt is after the first one or two all the FPS testing stops because it's it's just it's to be the same it's gonna be the same plus or minus silicon quality differences if I line up 10 1070 T is here asu's EVGA gigabyte everybody the difference is between them in terms of gaming performance unless one has a really bad cooler those differences will amount to silicon quality if you put an air cooler like some Aero card on there that would be worth testing because then your cooling performance is gonna be so bad that you could actually have a clock deficit so that's kind of my parameters for testing for reviews we mostly focus on you know thermals and things like that anyway next question is from David Watson who said I think it's from YouTube Steve this has been driving me insane why the hell are manufacturers still using toothpaste compound for their CPUs when we all know that by using liquid metal you are dropping the temperatures from anywhere between 8 and 20 degrees why haven't they cottoned up why I think that's supposed to hey why haven't they caught on to this obvious and glaring fact by now surely they realize the temperature drop only significantly enhance the lifespan of CPUs and make it superior product it's insanity I've talked about this in the past too I think it was with maybe the 79 AD XE review or one of those but first of all liquid metal is not really a great compound to use what you would use instead is solder that's what I am to use is on the rise in CPUs and intel has used it in the past on some of their high-end desktop even before that I want to say Sandy Bridge I'll top my head I think some Sandy Bridge CPUs used to solder as well that would be your alternative to liquid metal it'll be a bit better it's better logistically because you don't have to worry about like cracking or we're just kind of moving around and things like that aging solder is gonna be a better solution now why don't they do it well liquid metal is too expensive and it's impractical solder I don't have a great answer for you I've tried to get one for years and it's come down to a few things I have a few theories on this now so for the reason why doesn't Intel use solder on their CPUs you could say cost cost a bit more than doing their own paste you could say there intel has argued that the smaller dyes have potential problems with solder where they could crack over time or develop micro fractures there's not a great way to validate that however you could point to any number of millions of Sandy Bridge CPUs and other CPUs that did use solder and were small dyes and did not have micro fractures the larger dies there's no that argument does not apply so the high core count CPU is there's no real argument there and potentially you could start arguing things like well maybe there's a conflict-free side of things and tell does that conflict free marketing I don't know if there are any government grants or anything like that but using thrown paste from a conflict-free factory or whatever it may be or source or supplier I don't know if that gives them any kind of incentive in terms of money to do that whereas sourcing solder without being hyper expensive may interfere with that I don't know if going like the going green side of things is part of it as well I've heard that kind of rumor does also with this so I don't have a good answer why Intel doesn't use solder I have asked to them I have asked every PR rep that's worked there in the last several years I've spoken with it with about it with their Bower a lot and we've kind of hypothesized things on this channel actually about why that may be the case but anyway so liquid metals not a great solution solder would be good there are ups and downs I would like to point out here that despite some of the rhetoric Intel CPUs lately have actually improved the high core count their own taste is different than what the 7700 K used as far as I'm aware and its thermal performance is better and also at that same base as used on the 8700 K and overclocking them within reason is really not that bad now if you want maximum overclocked performance you do have to basically delete it because it's just you're going to run out of thermal Headroom but they do operate pretty reasonably with the stock compounds the biggest problem with these high core count compound CP is not the thermal paste it's the motherboard vendors blasting really high Auto voltages to them to the point where as we've reported in the past you get seventy seven hundred KS operating basically stock at ninety degrees Celsius then you go in and manually drop the voltage to something that's perfectly stable and much lower and you could be at sixty seventy degrees Celsius with a two hundred eighty millimeter cooler and max fan speeds so a lot of its the voltages I will certainly not argue that that solder would be better but at the same time liquid metal is not the ideal solution I think that addresses pretty much everything I know about the issue from what intel's told me or not told me and what i've spoken with other people about but next question is from salad who said in your recent d lidding and liquid metal video your recommendation for preventing liquid metal overflow or spillage from getting on the cpu contacts was DS nail polish an optimum techs recent deleting liquid metal video he recommends using silicon sealant the same that you use on the IHS in this video so any reason why you would not recommend simply using the silicone sealant to protect the contacts why didn't you just use your silicone sealant that you had on hand instead of nail polish a couple reasons for that one so the main one here is removability so what kind of sealant you can you have to basically pick it off to get rid of it and there's a really good chance that you're gonna rip off a surface mount device when you do that those capacitors are exceedingly fragile and you get away with ripping one of them huff but no I'm doing much more than that the resistors and other things on there it's not worth it because if it's ever coming apart again you should just use nail polish because you can remove it with acetone if you need to it's it's got a low profile it lays perfectly around the surface of the device you can still see the device it doesn't create any kind of Z height between the PCB and the IHS and again it's easy to remove whereas silicone adhesive if you have to remove it because maybe your liquid metal application has aged or you didn't apply it very well the first time which happens you have to remove it there's more risk and damaging the parts if for some only goes wrong so that's that's my primary reason and also it's again it's just super easy to remove nail polish and it does basically the same thing so that's really all there is to that also I hate where you can with silicone adhesive it kind of sucks to work with but next next question is from I don't write down who asked this one it was in the SDN questions challenge discord for patreon members GM staff and Vidya ampere and have your computing unfortunate naming coincidence I think that the name that people think the next generation architecture is going to be called might be incorrect but I'll leave that there it might be a code name or something we'll see if it actually ends up being that the public name but I that's I don't think it will be next question is from webhead sm webhead sweb head thank you for the difficult name a stupid question I finally received it new 1080 FTW if installed in the top 16 PCIe slot is a bit close to the CPU cooler I watched the video that you did regarding performance and 16 versus 8 lanes would it be stupid to use the eight lanes to gain distance from the GPU and the air cooler this would be on my old and moldy but still rock-solid gigabyte z77 motherboard appreciate the input I think it's fine it's pretty pretty straightforward PCIe gen3 I think that might be done I don't know I think that's still gen 3 right it was either way it doesn't really matter Gen 2 or gen 3 16 verses 8 if there's difference it's gonna be like 1% so yeah if that's what you have to do to make sure it's got some distance and the CPU cooler is not going to short the card then just go with that I didn't look up your motherboard some boards do have by 16 slots lower down on the board like a lot of the time you'll have you know by 16 up here and by 8 and then maybe another by 16 right there so check and see if that's the case it doesn't hurt to use 16 if you can but I also don't think it's gonna hurt to use 8 next question vagin 64 says we're vegan 64 I guess in this case says with all your research how much data did you produce in 2017 and gigabytes I don't have an actual number on that for the data collection from testing I have a number on it for media production things like that and gets complicated because we we compress our video files sort of once they've reached a certain age point but we keep them around it's fairly lossless and we also delete our a roll files once they're rendered into a video so it would be a lot more but as of today I think we've used about 10 terabytes in the last year or so and then just reduced that by about a terabyte today by running one of our scripts so we can control it pretty well using the compression scripts and and getting rid of stuff that we definitely don't need anymore has allowed us to keep the same set of hard drives for years at this point rather than just adding on infinitely more and more hard drives it's a lot of cost involved in that a lot of maintenance things like that so just easier to organize stuff get rid of what we definitely don't need and and then compress everything else test data definitely is in the hundreds of gigabytes if you start counting the captured videos for things like VR and console benchmarking it's even more hundreds of kilobytes getting two terabytes range for the last year but we do try and compress those as well got a couple more here that a quick Street guru I finally remembered what happened to the Radeon SSG could the effect of one terabyte of VRAM be good for miners did they realize that the ten people with red wasn't going to be enough of a customer base so the Radeon SSD is available you can buy it on B&H photo and a couple other places it's like seven thousand dollars it targets a professional market I don't know what kind of sales volume they have but it's out there and you can buy it the bigger part of the question though and I think we can all agree the more important one is how does it do for mining well given the deterrent GPU storage and the mining demand for cards I would like to think that if I say the cards very good at mining they'll all go by the $7000 card and leave the other ones alone so we'll just pretend that's the case in reality I did actually talk to some people have this question I thought it was kind of a funny one to bring out so despite having 1/2 hair by the vram and please if you know a lot about crypto mining let me know below what you think but I talked to some people and from what I understand crypto mining is more bandwidth intensive than it is capacity intensive I don't know if there are any algorithms where that's not the case while they also remain profitable on a $7,000 card but if they are not capacity intensive then in fact SSD storage will be lower bandwidth than the 16 gigabytes of HP m2 that's on the card and at that point you might as well buy Vega frontier or 64 or whatever if you're really doing mining and looking for bandwidth because I think it'd probably be about the same or better especially for the $7,000 price although the gtx 1080i will soon be priced at that same level so anyway if you know where the capacity might come into play with my name let us know in the comments I'm just kind of curious at this point last one not twist said an EVGA tech suggested I should remove the cooler off my 1070 TI replace remove all the thermal pads and replace them with an actual thermal compound is this a bad idea yes it is most certainly have had idea and I'm sure that's just a one-off instance but don't do that it's a home don't pads are there because they're thicker they're not runny they don't really age the same way it will sit there always as a gap to or to bridge the gap between the components and thrown paste is not there specifically because it probably isn't a good application so I would say if you have to replace them and buy more thermal pads and replace it with those I don't know what the the technician was thinking but last one we had some fanart from someone without a screen for a second and that's it for this time they were watching you could subscribe for more as always go to patreon.com/scishow so you can get the second part of this ask g and it'll just be a couple more questions if I do that or Reese more frequently in the near future so as always thanks for watching I'll see you all next time