Gadgetory


All Cool Mind-blowing Gadgets You Love in One Place

Ask GN 57: Chipped GPU, Pascal Temperature Response

2017-08-26
everyone welcome back to another episode of ask GN as always you can leave your questions in the comment section below or on the patreon discord in the ask GN channel on that discord where I comb through all the questions for each episode and try and fit as many as possible in the next one today we've got some interesting questions on one one a discussion opportunity on tech battle scars and then some questions about use cases for thread Ripper questions about overclocking and multipliers questions about cleaning systems things like that overclocking monitors as well before getting to that this coverage is brought to you by the EVGA 1080 TI sc2 and NVIDIA Destiny to bundle running up through September 4th the 1080i sc2 comes with a synchronous fan control for its dual fans and nine thermal sensors and again includes destiny to learn more at the link in the description below we had a couple good questions from the last episode that were more like myths kind of mythbusters-style things that won't be going into here but they're interesting ideas one was does holding the power button down rather than clicking shut down still hurt the OS like the the myth or the legend once was and those moving the PC while it's in use hurt the hard drive because you've got a spinning platter those relocating it hurt that or misaligned the header which I guess and in fact it is possible to cause some kind of header misalignment and mess up the drive if you're moving it around while it's in work but don't really know I just thought they were interested if anyone knows about the power button one I'd be curious though because I haven't looked into that so the first real question is this was brought up by I think whiskey in the discord chat the patreon chat we're just talking earlier about overclocking and how he had encountered an issue where 10th his 1080i was not overclocking as high as previously and I mentioned that it might have something to do with temperatures and I think we got into a good discussion there where the result was a higher overclock with a lower temperature and so it's time to kind of bring this up once again but in the past we've briefly mentioned at least once that pascal is very sensitive to the GPU core temperature i think we talked about this with Kane and previously but every five degrees can really help you out and what we were seeing the over clocks in discord chat were improvements of on the magnitude of maybe 10 12 megahertz at a time by dropping clock and I think whiskey's quick testing was more or less validating what we've seen previously which is if you really wanted the highest overclock you blast the fans you'll get a couple extra megahertz maybe 20 extra megahertz with the higher fan speeds as the GPU approaches 40 Celsius it seems to really do a lot better in the 30 to 40 C range then above that and now how appreciable is that in gaming not a whole lot but if you're just going for Maddox clocks Pascal is really sensitive to temperature so just wanted to point that out because I don't think we ever really talked about it a whole lot in our reviews never spent a day dedicated to proving that point but it is something we've seen routinely it's something we know to be fact basically if you're wondering why your overclock today might be lower than they were previously it's probably more likely because something happened where your temperatures are higher maybe you decided that you didn't need the aggressive fan curve because temperatures are fine so you lower the fan speeds and that would potentially be a cause of the higher temperatures or the higher clocks rather even if the temperature is well within operating range for a 1080i kind of under bring that up next question that was also GP related from it's a YouTube submitted question from bents who says sgn I chipped the corner my 1050 TI is GP 107 core it still works but how well it just randomly died on me I swear I saw a transistor in the in the removed corner fYI it was the top right corner so I don't you might have seen something else probably not a transistor but maybe a cap or something up there I'd have to look at that that died and substrate I haven't looked at it in a while but there are often tiny caps surrounding the outside of a CPU or GPU and you don't need all of them technically speaking that we were talking about there Bauer about this at Computex and how he as on several occasions accidentally ripped out like a capacitor or something like that when trying to deal it processors and they still work fine you might not know what functionality lost or maybe you do and it's just not a big deal but they can still work without all those components so it's it's not instantly dead but if you chipped the corner I'd have to see a photo and I would love to see a photo if you get a chance but it's possible that you only chip the cover of a die the actual die is underneath that metal looking sheath on top the reflective surface it's under there and it's possible that you just didn't dig into anything important which in that case probably don't do what you did again and count your good luck and move on but yeah I would say it's you can remove small parts it's not always going to kill stuff you don't want to do that but a lot of the times these parts are a lot more they will tolerate a lot more abuse I think than people realize you can miss handle things quite a lot and they'll be fine I we work with this with all these components on a day to day basis and when it's in a work environment it gets abused so that's part of the job that's part of owning the equipment and that's why we have redundancy because it's not a one-off build for us like it is for a actual consumer personal machine and I can tell you a firsthand they can take a hell of a lot of abuse in fact this is where the tech pals carts discussion comes in so we were talking with other media at one of the recent events including Gordon from PC world and a couple other folks and talking about different battle scars of things where like you might remove a PG a cooler and rip the CPU out with it or something like that or whatever the case may be and one of the ones that I brought up was there was a time when I forget what I was working on but one of the CPUs were working with actually bent like like the CPU was like that and hot so you're like oh well that's dead oops but now so I took the CPU the subs you can physically see it like crooked and not not bent from like if you bend a PCB bend the PCB on a video card by over torquing something not bent like that dent like legitimately if he pushed it any further it would snap in half and the IHS was more or less holding it together so that part was no longer used in testing after that but I wanted to see if I could reclaim it for other machines that were not using and testing so I took it to a vise you wrapped a towel around it took it to a vise and it kind of slowly bent it back to shape and it worked as a CB worked fine I think we didn't even lose any memory channels all testing validated all testing past temperatures were fine memory speeds were fine memory bandwidth was the same Layton sees were all the same and the benchmark performance was the same that CPU is still in use today and one of the systems I've got and it performs just like if you ran a 3d mark on it it would perform the same way it did years ago so 4000 series CPU so I would just go it to show if they can take a lot of abuse as long as it you don't hit a mission-critical area I wouldn't recommend that you go bend to CPUs because there's a very good chance that it does not unbend or that you do in fact damage something there's all those layers of PCB there's wiring in each of those layers Jay's got an old video or he drills through a motherboard which we've done too by the way and depending on how careful you are and what you know about the trace layout on the board which is often nothing because that's not something that's really published unless you talk to the manufacturer and try and learn like hey is this a safe area to drill generally the answer is no but you can definitely kill stuff but they can always take a lot of abuse the board that I drilled through was because it was a z97 I think board something like that and it got it's just it got some mounting hardware stuck in it it just wasn't gonna be possible to remove it so I'd rolled it out and I mean I was drilling through a saga hole anyway so there's no risk of damaging any and it works fine because you just don't clip the edges don't clip the PCB and it's fine you're just hopefully just drilling straight through a hole anyway other than the screw that's in there that you want to get out so I kind of closed but it came out relatively unscathed it still worked and worked fine so yeah they think some abuse now as far as your GPU I would say just be happy with the luck and probably don't mess with it again but it sounds like it just happened to miss something important next question from scooby-doo beyond discord he said your latest video this is referencing the thread Ripper one the use cases for thread Ripper one the way this video talks about use case scenario you found four thread reproduction compression in it you mentioned how Andy's PR department couldn't give you straight answers about rise ins ideal use case scenarios due to a lack of communication with their technical people because well anyway you point out that in blender and Adobe Premiere that CUDA acceleration is faster much faster in fact making a powerful CPU rather redundant even including the CUDA render times with a 7600 K and a graphs prove a point now another tech Tube channel I watched keep saying that his videos don't include a lot of quote effect and so he doesn't benefit from CUDA acceleration as much making a strong case for using a powerful CPU your videos don't seem to use a lot of effects either though so I'm at a loss to know what he's referring to since I don't render myself so the question is is there a use case scenario where rendering with a strong ZB alone outperforms a weaker one with futa acceleration is a stronger CPU it in such an instance redundant or superfluous so that's a really complicated question and there are a few different answers to it Linus has run into this as well so - did a video years ago I think either the thumbnail or the title was something like our $4,000 mistake or something like that he talks about how they buy or bought hi i'm cpus for editing machines and found that cuda was doing the job for them the same as if they paired it with a $300 CPU that's changed a little bit since then not a whole lot premiere has been updated several times since then CeCe's kind of a thing and so there are differences but in our testing very directly to answer first of all when I say in our testing I mean internally for use just kind of gauging it I ball in it look at performance that way it's not like we're benchmarking this stuff for publication so in that testing what we've seen is that things like warps so like warp stabilized seems to be pretty CPU intensive when it's happening so if you look at the thread utilization while a warp is being applied at least in the footage we've worked with that is a CPU intensive process and a lower-end CPU will hinder the speed at which that task completes so it seems to depend on what you're doing warping with what we've done I haven't seen that warping as CUDA accelerated I'm not an authority on that topic but that's what my observation has been so it seems to be a CPU accelerated task and so you would benefit from a better CPU I'm not sure if you can do to accelerate it maybe you can not something I've looked into CUDA acceleration definitely speeds things up I mean if you look at our charge for the 1080p AVC HD render we do for though the CPU benchmarks which is an old test at this point but if you look at those charts an r5 and i-5 perform the same as an i-9 and thread Ripper if they're all cuda accelerated it just it doesn't matter I mean they're within one to two minutes of each other on a twenty to sixty minute render depending on what you're looking at and that's basically variance at that point so CUDA seems to pretty much be the great equalizer in that test but that's a 1080p AVC HD video whereas what we do now is 4k and it's some other codec and that seems to interact differently in rendering than what we used to do with the 1080p60 AVC HD stuff so I know that for sure upgrading CPUs can't help so we moved from a 49 60 X 4.4 gigahertz overclocked I think that's what is that six cores 12 threads or something we moved from that to the Xeon something from the same generation x 79 CPU and that's a 12 core 24 thread part from years lowered clock speed and the render performance improvement was something like 10% so clearly there's still a benefit there that's I mean it's not a linear gain because you've doubled your thread count or something like that you've increased your thread count a lot and either it's eight it's either eight and sixteen or six and twelve up to 12 and 24 is a big jump more than 10% but we're only getting 10% improvement out of it so there's a game to be had but I don't know it just depends on what you're doing like the effects we don't do a lot of effects as in like green screens and stuff like that but we do use all kinds of color correction we apply warps we work with other various small things like there are ultra keys in there for the intro on the outro and lighting changes in post but I couldn't tell you if it's more CPU or GPU limited what I can tell you CB does matter it's just there's a limit like the GPU in everything we've done going from something like a Titan XM to a 1080i or from a 10 but what's 980ti to a 1080i those have shown bigger jumps in our type of rendering than a CPU upgrade but we still see a CPU upgrade jump so the point I would make here is I think the benefit is more than just rendering because we see the bigger gains and things like scrubbing playback while a warp is being applied so if you're worried about more than just rendering then you need to consider the editing scenario to where during editing you have scrub playback and how quickly the timeline responds to movement and so if you were doing a warp that's consuming a bunch of threads while also trying to still edit the rest of the video that's where you want the more powerful CPU to try and keep up with both of those things I want because that's starting to get intensive especially if you're scrubbing stuff that has other CPU intensive effects on it I couldn't tell you what those are I just haven't figured it out yet but if you're scrubbing stuff that seems to be more CPU enzymes of while doing something else let's see the intense stuff like a warp then you have trouble so I don't know it's a lot of a lot of words - more or less say it depends on what you're doing but I think that gives you some of the some of the angles we look at it next question is from yummy Frank who says Steve I have a question when you overclock a cpu what happens with their boost speed for example if I overclocked a 1702 3.1 will a boost the 3.8 or only 3.7 what happens to boost speed if you go over the boost speed with OC when you start changing the multiplier basically turns off boosting that's all there is to it and that's why with Rison or threader for CPUs if your overclock is lower than the XFR numbers and so it might be 4.1 or something and you're working with an application that only engages the amount of cores that will be hit with xfr so 4 on thread Ripper then you'll actually where's performance but as long as you exceed xfr and your work or you're working with applications that use a lot more cores than xfr can work on then you'll be advantaged but yeah basically it's an override once you start tuning it overrides all that stuff next question Zeta discord also says question for Sdn when cleaning your PC is it bad for the motherboard or the fast to blow through blow air through the fans causing them to spin at high speed well I haven't had this one because of all the counts that we see there like be careful with the fans don't touch those this fragile like there's some truth to that but also get what calling back the beginning of this video they can really take a lot of abuse these components but that's worth taking an air compressor or something I don't know I really like spray and air compressor at a fan for very long the bearings probably won't like it but I don't know I have no problem with taking a can of compressed air and just kind of blast it at the corners to get all the dust out and once the dust has gone just stop I don't know I don't think that's really gonna cause any serious damage if you're doing that everything honestly like seriously anyone who's gonna pick a fight about that causing damage just kind of you know a couple like quick blasts at the corners to get the dust out the point to make is that the dust that would have accumulated if you're looking at a scenario we're like either I spray this out with compressed air or I'm not gonna ever touch it if that's the scenario then basically you're looking at your the dust accumulation will be far worse for the fan than whatever meager damage that might cause to the bearing lubrication or something like that from three seconds of spinning in a circle which they're designed it to anyway and yes they're not not supposed to be spun when they're not powered and the electromagnets off but I don't think it's gonna be a problem I'm sure there are people all the people who are the great defenders of fan rights will probably dispute that but I don't care it's like like spraying air at a fan doesn't bother me out you're not just don't sit there with an air compressor and like watch it spin for five minutes that's probably suboptimal for the fan but yeah I think it's I don't I think it's fine next question dr. guns for hands says Steve I heard some talk of people overclocking monitoring fresh rates how does this work and is it something that would be compatible with free sync do you think or would it break them I'm actually not sure about free sync or do you think but I'm sure some of the comments can tell you I have not looked into overclocking monitors for years and the last time we did was in 2014 we ran an article on it which is probably show on the screen or something and Michael Kern's to that article for us and he did the monitor overclocking at that time so I can tell you how it works outside of free sync and gsync most the monitors have on the market have some overclocking range like 60 Hertz displays you can often get them to 80 if you're really trying you can get them to maybe 96 it kind of depends on the scaler so if there's no scaler and they're like the old UNIX displays that were popular in 2014 those QX displays didn't have a scalar in them and you could achieve 94 or 96 rather Hertz on them without much issue in most cases otherwise 80 is somewhat common so it just depends on that I'm not sure what the current climate is on scalar inclusion and qyx monitors monitor coverage is not where I specialized Michael Kern's kind of covered that of stuff a lot and Eric Hamilton covers stuff a lot in the news segments but the next item item of note is unstable monitor overclocks are pretty obvious so if you try to push to 80 Hertz and the monitor doesn't like it you'll see artifacting or there'll just be no display output and at that point you just try and roll it back it's it's really like most other overclock and it's trial and error until something stops working too many revert your changes and next one is this so overclocking this these are notes from Michael's article if anyone's curious overclocking display is 10 darken the screen it can lower the gamma and of course there's warranty concerns we don't really care about those on this channel though and then also with DVID this is probably less relevant now than it used to be but with DB ID you had to raise the pixel clock of the cables so you could do that through an video control panel or catalyst whatever it's called now crimson Radeon settings that one Radeon settings you can install a patch that's available online from modders or I think Nvidia might support a natively now and you change the pixel clock of the cables to help with the overclock and then in precision and afterburner there are other options as well for tuning and I think I wrote a few of them down here yes sir precision gives you options for LCD timing modes so Kearns and his article suggested leaving those alone for starting and mentioned that just overclock the display to start with and then you look into tiny modes later to try and get more out of it if you want to it's kind of like anything else with timings where you're just kind of tweaking at that point for the sake of tweaking but yeah the article on that will answer more if you're curious but those are the very basics of it it's not too hard to get started with it's just you depending on the display you're kind of fighting for not a lot of gain a lot of the time the qyx displays we're a great example of something where you get some serious value out of it a lot of the sort of Korean be panel displays are like that where you can buy them for cheaper because they didn't meet the requirements for whatever was and then trying to overclock it yourself to hit the high refresh rates but I think that's it for this one so as always leave questions in the comment section below for next time subscribe for more you can go to patreon.com/scishow and access top side directly and I'll see you all next time you
We are a participant in the Amazon Services LLC Associates Program, an affiliate advertising program designed to provide a means for us to earn fees by linking to Amazon.com and affiliated sites.