Gadgetory

hey everyone we're back for another ask GN this is episode 24 I think that we were just debating the idea of pre recording every number possible and then inserting it in the event that I ever miss speak and say the wrong iPod number I think we're on 24 and this as always comments below if you have questions for the next episode there are plenty of them we'll do our best to get to to whatever's out there but it can't them all so first question is from Jonathan Poe who says hashtag ask GN that is an artifact from the first one I'm noticing that with the 310 ATS I have owned it so far there that's a lot of 1080s there seems to be some scaling issues when overclocking I would call it diminishing returns basically I can manually overclock get some extra performance but when I push further you know if stable performance drops off and the card runs worse in benchmarking than if I had an overclock at all why is this so some of this is described in part when I talk about under Bolton with the rx 480 and probably the previous episode or the one before that and it's sort of the same concept where the diminishing returns would be the correct thing to call this but we talked about this in our 1080 review as well in our 1080 hybrid part 3 or 4 but basically you're beginning to hit a limit somewhere so depending on what card you have that could be a thermal limit it could be a power limit or it could be a voltage V core limit imposed by V bios and for the 1080 we found that with reference there was a slight thermal limit if you're running a IV partner cards is probably not the case so then you start looking at power so what happens is let's say you overclock by just throwing numbers out there so you hit a 250 megahertz offset we did this with the 1060 Hybrid 250 megahertz offset and you can pretty much sustain it but if you dip down to 225 it's gonna be a little bit higher performance or equal performance minimally the reason that happens is the frequency plot over time which we make charts for this if you haven't seen them basically if you look at the you measure the frequency versus temperature dureena maybe wanted to our burnin you'll see that the frequency of spike when there's less power available or not enough voltage because V cord is limited and by down stepping your overclock you can produce a more flat frequency line which is what you want and so that means that instead of boosting up and down kind of variably based on power availability or whatever thermals other issues you have a flat line so now your FPS output will be more consistent so that's where that comes from we kind of plot all these in the hybrid videos especially but also in the reviews so that's that's most of it if you want to test this yourself I would suggest a 264 I think they have a free version and you can use that to check the frequency if there's also a gpu-z if that's good too that is free use gpu-z turn on the logger and plot the log file after and if you see a flat frequency curve then you're in good shape and that will hopefully resolve any of the issues that you're seeing but yes sometimes it is better to actually down clock slightly from your peak and sustain a more stable output even if it looks stable because you're not crashing it could be not stable because of the jumpiness of the frequency as it tries to accommodate whatever limitations are going on under the hood depending on your card next question is from - penguin if you said what would you consider to be the maximal safe target temperature for the RX 480 the goal is to have a car that operates as quietly as possible without a fire hazard don't know that it's a fire hazard to begin with but we did have the the fire extinguisher on the table that one video so the the question here it's it's it tries to sit around 80 Celsius and that's pretty reasonable as is that's where the reference card drives to sit anyway I don't think that's that's a bad temperature but if you start pushing past 80 Celsius if you enter the range of 90 then I would definitely either increase the fan speed or get a different cooling setup in your case or different card if you're just trying to operate quietly while kind of maintain as you said the maximum safe temperature I would set the temperature target to about 80 Celsius to a nice few cells he's in that range somewhere and let the fan take care of itself you can manually configure a fan curve the Whatman software has some very basic function of this you can do something more advanced with afterburner once it's fully supported for whatever card you're using but that's the way I do it I'd set around 80 to 82 Celsius for the maximum temperature in the let it take care of itself from there next question is from Reza and L who says I have a question planning on building a new rig as a GPU I'm thinking about ZOTAC 1080 should I go with the a MP or amp or amp extreme does it worth the is it worth the extra money so the Valastro tack card we looked at was an extreme card and I was not kind to it I have not worked on any since then because I don't think so Tech has worked up the courage to try again so I can't really comment on the 1080 amp or amp extreme personally but as a general statement with the 1080 and the 1070 line unless you're really serious well even if you are really serious about overclocking we hit enough sort of limits with V bios and the voltage the vcore setting for the GPU that were not able to really advance the frequency noticeably from one AI v partner card to the next Japan on which one it is so what you want to look for first is just a good cooler and for the most part Twin Frozr AC X Strix wind force they're all fine coolers on the 1080 you don't need anything insane you don't need a three fan cooler or anything like that so that's why I look at first a dual fan pushed cooler will probably do the job just fine unless you're building in an ITX box where you want to blow her fan instead that's generally the idea for thermals as dual push will work out well for the 1080 and then in terms of overclocking like I said I think we worked on the 1080 gaming X and that performed about the same or slightly lower than the found addition card and if you look at any other 1080 they pretty much all sit around the same overclock max so I wouldn't buy based on claims to overclock higher for example if someone says they've got eight or ten or twelve power phases or phase power design it's not gonna change the world that much for you so I would buy based on the cooler and then the price if I don't know what the extreme costs off the top of my head but if it's more than a few bucks extra over just the apps and I would say forget it but again I haven't looked as though tax and serve 980 extreme review which was titled something like do not buy the 980 extreme the next question is from I'm not sure I can say that next question is from that shitty chat Hall who says what does that spec PCI lanes in CPUs mean Wendy told me it means it more bandwidth for SLI configuration that's why you need a high-end CPU for SLI it's not really exactly how it works it's not bandwidth it's so bandwidth there's they're sort of two different things we talked about this a bit in the x8 versus x16 video for SLI basically your CPU and your chips that both have PCIe lanes and that means that they have a fixed number of lanes that they can assign to PCIe devices whether they are in the PCIe slot or whether they're in interfaces that are enabled by PCIe the bus so an example of a device that uses PCIe lanes might be an m dot 2 as d or a u2 SSD or something like that nvme drives and depending on how they're set up basically an m2 SSD will pull its lanes from the chipset but I think this is a common misconception so with skylake onward there are HS i/o lanes which are high-speed i/o lanes I think the z170 has 26 of them I've thought my head and I think H once the H 170 chip has something like 24 26 H 110 is less than that 22 maybe or last 16 but either way there's a stack of PCIe lanes per chipset we have a video on it I don't have if you have a Z 171 it has 26 h SiO lanes in segments of for those can be diverted to PCIe so that means if you had an AM 2 device which uses 4 lanes or if you had some other PCIe enabled device or not a GPU generally unless it's a crossfire setup which can support X 4 but you can't run an Nvidia card off of the chipset lanes with the new Intel chipset because they are limited to segments of four an Nvidia requires a buy eight minimum set up for PCIe lanes the CPU has something like sixteen lanes if it's a lower on CPU or it goes up to 40 on the high-end extremely serious stuff so if you have 16 lanes the CPU will basically say these 16 lanes talk to GPU number one and if you have two GPUs it'll say eight of them talk to one eight of them talk to two based on which PCIe slots you plug it into on your motherboard and depending on how that demands you set it up if you have a 40 Lane CPU it'll assign me it might assign two sets of 16 millions to two GPUs if you have them in SLI certainly it gives you more bandwidth in that regard because you're getting something like 16 gigabytes per second versus 8 with Gen 3 but in our testing that we did recently with I think 10 70s are 10 80s the PCIe lanes really didn't matter that much because the bandwidth of it by 8 setup already supports these new cards pretty well so we were seeing about a 1% difference so hopefully that answers that question kind of topo we've talked about this a lot before if you search a channel for PCIe or lanes I'm sure you'll find it next question is oh right this one hey Steve what PC do you have can you say the specs and show a photo of it if you do have then does your job leave enough time for you to play games my compute the worst computer possible out of the parts that have been retired so all the good parts may be contrary to common belief with tech channels I don't use those all of our parts her review stay and read you systems they stay on the bench if they're not used in the bench then they're on a shelf waiting to be used because it's far more efficient to do that than to pull them out of different personal rigs and use them in whatever so when a part gets fully retired like a 960 then I'll move it into our render bent or render machine or a production machine for content creation but if it's still an active product and we're not doing regression testing yet then it'll stay in the bench if we're keeping it for regression testing it'll stay on the shelf my system has an asrock 990 FX board of some kind so basically a couple years ago I said I'm gonna build an AMD FX 9000 series system because I want to play around with overclocking and I want the experience of building and these top-of-the-line system because we were working with Intel already at the time pretty regularly so I wanted to know what is it like to build an AMD top-of-the-line system as opposed to the ones I already had experience with Phi sevens and things like that cuz I did know halen before this and frankly Nehalem worked better for me for what I do then this 9000 series chip that's for a lot of production reasons the asrock board I'm using whatever 990 FX extreme 9 or something that it's they're more expensive aim through plus board it has serious vrm issues and if you use the liquid cooler that came with the FX 9000 chip like I did I forget what ship it is but then I think it's 93 70 maybe but if you use the CLC you lose that potential for the fins the lower fins and the heatsink to kind of siphon away some of the hot air from the vrm that's a problem so the VRM well occasionally hit ninety to a hundred plus celsius i've seen a hit 105 celsius terrible machine it's in an old thor or no not through authorities in a throne a white throne from Roseville that we reviewed ages ago I think so that's the the Box it's in which I also don't really like that much but and then I had to down clock it to save the vrm because it was overheating and playing games it would start screaming the fan actually just watching a YouTube video the decode works so hard right now that it screams with the fans so not a good system the video card I think I have a 780 in there it's enough for when I occasionally do get a chance to play games which has not been a few months it's been a few months since I've actually properly played a game other than testing because we basically work constantly with benchmarking and content creation so there's not enough time to really get into stuff other than work so 780 works fine for when I need it at some point I'll upgrade actually as I keep tying Andrew our video guy he to rebuild my system extra month now I've been meeting to build a new system with I think I bought an i7 6700 K and I bought a good amount of RAM and some other stuff and then we made content with those products instead and they haven't gone into a system so I am using something that you shouldn't build and you should instead use our build guides for what to build because I need the parts for content creation so that the parts can generate money so we can keep my heart content so that's how this works maybe at some point I'll retire something like a 780ti into the system and build something properly which is still an old card so hopefully that answers that question last question was from Super Mario who said Steve I'm going bald any advice Super Mario I would suggest maybe doing one of those implants where he takes on the mustache hairs and put them up there and see if that works out for you otherwise there was a good count reply to this that suggested using thermal paste on your head to sync the heat and allow more hair to grow which I think would probably work in my professional advice so that's all for this video page on commercial video thank you for watching comments below for the questions for the next video I'll see you all next time you trying to show me off