Gadgetory


All Cool Mind-blowing Gadgets You Love in One Place

Ask GN 36: GameWorks & GPUOpen, Handling ESD

2016-12-01
hey everyone welcome back to another episode of ask GN as always before we get started if you've got questions post them in the comments section below had a couple repeat questions from previous episodes finally we'll be getting to those today so we've got some interesting stuff on SSDs on GPUs open jeep you open and gameworks things like that before we get into that this content this week is brought to you by the a.m the RX 470 and 480 cards which can be had for under 200 depending on where you're looking right now because cyber monday and black friday and of course regardless of when those this video is posted those deals i'm sure will continue basically forever because that's what retailers do now it's it's not black friday it's black november as you well know if you've been to newegg they actually call it that so without further ado the first question is from ghulam fifth alam five who says see why they're noticeable real-world benefits to using m2 SSD is like the intel 600 p which by the way we just included in one of our buyers guides or sounds on 860 evo over conventional say 23 SSD is like even shorter boot or load times or are there insane insane numbers from m 2 drives being bottlenecked by other stuff like software CP GPU first part of the question obviously and i'm sure this is clear for the most part but m dot 2 drives do have the benefit of they can be used in places where satan might not be like a smaller system but ignoring that and looking just at the speed part of the question it depends so for boot times i am reventon a bunch of SSDs including i'm about to drive so i'll be doing boot time measurements i don't have an answer for you right now if it's actually noticeable if anything it's a couple seconds because boot times are already so fast that yeah you will be bottlenecked at some level by things like windows ability to sort itself out and then just in bios alone whatever your settings are there you can enable things like fastboot get rid of the usual posts kind of delay that you have and things like that get rid of the splash screens but those will be limiters almost more some extent than the SSD depend on what SSD you're using but as far as the other tests games will be limited by a couple of things mostly the game and its engine that is programming so those can only really load so fast most the time I don't really even with a hard drive like I can't remember what games we tested last time we did the an SSD review but for the most part the loading times between levels was pretty minimal now there are a couple of exceptions to this like Battlefield one which and I think battlefront make sense same engine I think those two games actually load noticeably quicker with an SSD so we used like a hyperx predator SSD which is an m2 a I see so it's an m dot to as is DM on an add-in card for PCI Express I use that back then for that testing then also used traditional hard drive for another test system and it was actually noticeably quicker with the load times so there are instances where an SSD at all makes a difference whether or not you see it with m dot tubers aceite i would say those differences are almost entirely relegated to things like production workloads where if you're whatever you're doing is capable of writing say upwards of two gigabytes per second because there's no bottlenecks anywhere in the system like maybe for example if you're executing a script to do mass file transfers where there's no encode decode or transcoding involved and it's it's just basically moving stuff then yeah you'd see a difference there because the sequential times especially the reads are so much faster on some of those m dot two drives but I you'll be limited for games by plenty of other things in the system where I don't think you'll see a huge difference in loading times that said they come back in a week or two because i'll be visiting that topic more thoroughly and actually have some test data for you that i can provide to give more actual numbers about the differences and not just say isn't noticeable or is it not noticeable but in general the use case the main use case if you set out to buy and i'm about to drive or an nbme enabled drive you are primarily looking at things like am i limited by the physical layout of my system can I fit a SATA SSD in there if it's 2.5 inch and there are SE 2dr it'sit's again I'll talk about this in the future but there's SATA enabled m dot two PCIe lanes are split off depending on nvme versus Zeta interface and stuff like that all dictated kind of by the HSI Oh set up on your chipset or by dictated by the drive itself talking about that soon basically the answer is not really huge difference unless you're doing production things with mass file movement or a lot of really fast right times from whatever your production environment is in those cases you'll see a difference for things like capture server for example we've set up a capture server for things like fcat we're live captures the gameplay and records it that's an instance where you would want rate SSDs if you're running off of SATA or something like an Intel 750 series nvme drive which is what i'm using for our capture server because anything short of that is not necessarily fast enough and you start to lose data and that's bad but that's a really specific use case that's a kind of reviewer use case that gives you an idea though next question tech Pascal asks I was wondering if you can explain in greater detail overclocking processes and what they mean as far as power target memory offset voltage etc also what's the process you go through reach overclock so for specific architectures we've got videos and articles about how to overclock Maxwell Pascal and Polaris so you can check those content pieces for that to very briefly go over this the question about a power target so power target setting is basically you have a GPU it says maybe the instance of higher on GP it says something like I want 180 watts and that's at a hundred percent of the power target and that if you offset that normally you get twenty to thirty percent offset depend on what it is is AMD GPUs some of them do fifty percent offset within the same GPU from different vendors you might get twenty or thirty percent offset the reason that is is because the percentage as always is kind of a little bit deceptive because it's with gigabyte for instance the extreme water card that reviewed the 1080 I think technically has a higher percentage offset than ebj is hybrid but the base wattage is lower as well so it comes out to be about the same if once you account for that but that percentage offset basically says if we're drawing a hundred watts now we offset it an additional 20 watts now you can draw 120 watts so you've got more power Headroom for the clock so with NVIDIA and with AMD basically you're looking at a my thermally limited no okay so we'll push the clock okay so now let's check again am i limited by power no not really as then next check and we limited by voltage yes okay so then the user can do a voltage offset you've got really limited playroom with that for current GPUs that are out there you only get so much in terms of millivolts offset and GPUs three point I was a little bit strange with it it's kind of hard to work with but basically the the question is power target gives you more power Headroom which allows you to increase the clock rate of the card which gives you of course the higher FPS to some extent and then does increase thermals though it also increases clock rate stability because if you've got let's say an offset power target but you haven't touched your GPU clock rate now there's some room where if the GP runs into a scenario where it becomes power limited normally under one hundred percent offset or what sorry one hundred percent power target it would throttle the clock back temporarily until it regains its footing so to speak whereas if you offset the power target and you have extra Headroom but you've not overclocked now you can sustain that boost clock basically at a flat level if you were to plot it over time and that's beneficial for reducing things like the occasional stutter in your gameplay and things like that I think that answers the basics though next question is from a poor of galata says hey Steve I want to ask if desktop replacements like the acer predator 21x really makes sense i mean battery life will be laughable so what exactly makes them viable yeah so the battery life is not great on any of cyber powers units we've looked at things like that you might get kind of best-case scenario if it's a single GPU like an hour gaming which is not great but then it's also not great battery just for normal use either because the nature of having higher end a hardware bigger screen that demands more power things like that but or higher resolution but i guess the tiny use cases are first of all these are all for the most part either lost leaders or halo products so non sa meant to be how a company makes its money it's more of like let's get eyes on our company and then just like with the corsair obsidian series cases like the 900 d not a ton of people buy that that's really meant to be more of a halo product and the idea is that people who see the 900d are like this is a really high quality case so then they recommend whatever's below it that's the same idea with the laptops now as far as usability the use cases for these desktop replacements are basically are you a person who commutes between work and home and you need your own system in both locations if so and if it needs to be a high-end system then you get one of these and you plug it into a dock on the other side and you use it as a desktop the question is well why would you use that why would you need that much power if you're doing something like rendering like we do we need effectively desktop replacements for when we're traveling to conventions to render the videos out fast enough and with whatever filters we want to apply it's really just not possible on smaller laptops for the most part what's the speed that we need but that systems basically plugged into a wall one hundred percent of the time so it's okay it's just not super mobile so the market is is again basically production or strange use cases may you something like is someone who's moving around frequently between two or more locations where they live or work that's basically it but they aren't really meant to be kind of the main way for a company to make money next question n 3 is any 5 saying how to ground yourself when working on a build I heard that you can touch the power supply grill to do that but does it have to be plugged in turned on also how do you use an anti-static wrist bracelet do you attached any metal surface or to a PC plugged in I have one of those so the setup we have is a little more elaborate than you necessarily need but the way I do it we've got an antistatic mat not necessary but this plugs into the mat and that will handle any ESD basically from whoever's wearing it dissipated through the surface rather than to into devices and then for it to actually work properly it's got to go somewhere obviously so this is just a metal hook it's just a circle that goes around to ground pin in our case on anything that plugs into a wall so the way I do it for anti-static bracelets is you've got two main modes normally you'll buy them like the roseville ones for a couple bucks you buy them they'll often have an alligator clip at the end of it and you can actually pull those off it's not going to hurt anything and then this is under it which is a banana plug it's called you can plug those into special banana plug outlets in a wall so you can buy and basically an adapter for your normal wall outlet and plug this straight into it and you connect this to a bracelet on the other end that's one option another option is you get something like we've got set up where you have the loop at then the metal loop this hooks around the ground pin in the US at least that's going to be the third pin the large circular one hooks onto that and then you plug that into the wall and anything that would be coming off of me through the wrist bracelet would go basically down the cable into the wall into Earth and be gone so that's how we deal with it and you you can do it kind of you can kind of hack an ESD solution one option I did previously did a video on was taking a power cable like a standard power spy power cable and you just basically cut the head off one end clip the hot wires and leave the ground wire that strip the housing off of it leave it exposed copper then clamp into that that would be a way to hack it and be successful you can kind of touch a metal surface um it becomes a bit questionable how effective that is so one it's if you're touching something like a power supplier case it's got to have a path to ground so ideally you're plugged in with the power supply not on because you don't want it on your building but then there's issues with like for instance paint paint is really not a great interface to have between you and your means of grounding it'll probably work ok but some of that might just be because you've not encountered a scenario where ESD was an issue so hard to say how effective that is but that would be my backup yeah clip into the power supply grill plug it into a wall and then you should probably be ok but these solutions are a bit more effective and some of them aren't too expensive this next question I'm not going to answer but I wanted to point it out so it's from ill to xbox who said basically asked about are there any games that use GPU acceleration for character AI if not is it something Deb's can do in the future if I understand correctly ai and machine learning algorithms can generally benefit a lot from using a GPU basically parallels amount of GPU I agree it's really interesting question so I emailed that out too and the in 10 video and we'll see if I get an answer and if so i'll address it next week but i do not know that's that's out of my area of expertise next question is from mike smith who says how long do heat pipe last I've had one for three years or more three different rigs now still working well so basically forever heatpipe coolers don't really die the not like liquid coolers they do have liquid basically liquid in the heat pipes but that's a sealed system so unless it's encountered some damage that liquid is not going to escape anytime soon because it's designed to sort of evaporate and then use capillary action come back down so it's not getting out of there but yeah they basically last forever the only thing you're going to replace is the fan that's a cheap replacement next question orvis 25 he says hey Steve second time asking can you explain GPU open AMD and gameworks Nvidia specifically what the differences are and what they do and why game works is still used far more than GPU open and games specifically when GPU open is free and open source game works is also free and video doesn't charge for that so I'll briefly do this we talk if you want GPU open stuff I do have a pretty long interview with Raja Kaduri on the channel you can watch that he explains it pretty well just search GPU open one word and that interview will come up he does of course a better job explaining it than I can in a short amount of time because that's what he does but to get the basics here question was why isn't GPU open used as much even though it's free and open source again both are free the GP opens pretty new relatively gameworks game works in its current form has been around a little while but also was preceded by individual effectively plugins it wasn't always the game works sdk it was individual stuff it's all kind of unpackaged into one now so let's do a stick game works first game works is effectively a collection of libraries that developers can pick and choose and plug into their engine or game so an instance of use might be if I'm a developer and I say I want some fire effects that are pre-calculated pre-written maybe it's dynamic fire that spread to something like that and also I don't want to write it myself because that's too time-consuming too expensive too hard in that case you'd go pull the nvidia gameworks sdk sdk and the library for fire and plug it in and it's mostly done you do some fine tuning things like that and Nvidia and AMD both have engineers available to developers to help with that sort of thing and optimization another option would be same idea but for fur or hair like what the witcher did and obviously these plugins these libraries are going to be optimized for nvidia hardware and video made it they're certainly not going to optimize on behalf of their competitor and they're probably going to build in a way that would work well on their hardware and maybe that doesn't work so well on their competitors hardware again tessellation with hair works in the witcher 3 at least natively at launch anyway so that's kind of the basics now the next question what about GPU open GPU open does it's it's a bit more things it does kind of the same thing at a top level which is a collection of libraries that can be used to better effect in a game but also GPU open is trying to sort of collate and these knowledge at least into more accessible package for developers to get to an implement in the games and what i mean by that is AMD and nvidia have very specific knowledge and how to optimize drivers and games to interact with their hardware and this is done constantly this is why you can see improvements twenty or thirty percent sometimes with drivers because the developer doesn't necessarily know i'm working with this particular particle effect or fire effect or whatever it should be drawn this way with this video card they don't necessarily know that or optimized for it and that's what the two companies will do the vendors now with the new api is that's sort of going away so this driver layer is becoming very thin as the GPUs interface more directly with the software so you've got several abstraction layers removed still a couple on there like the compilers making sure the code is all legal there's not as much driver interaction as there used to be so GP open is trying to go out and inform developers how to optimize their games properly without relying necessarily on a third party package like a driver so that's kind of a newer effort and again the interview with Raja kind of answers the rest of that but I think that gives you a starting point at least as to what each does and so I have to say either is better or worse than the other but that's what they do most we have here very quickly i'll answer this because i've answered it so d to ricky said why aren't we seen huge gains in dx12 out of the top end of dx12 GPUs so much a draw call api test is this due to the majority testing on super high-end hardware or more testing is being done at higher resolutions now we've tested both it's basically that DX I've said this several times I think most recently in the battlefield one CPU or GPU benchmark if you're curious the articles might give you more on that but to very quickly recap directx 12 and Vulcan are not free performance gimmies for the developers it's not let's plug into DX 12 and we get free performance with the CPU pulling draw calls or draw calls being pulled from the cpu to the GPU that's not really quite how it works they have to do a lot of low-level optimization they have to build specifically for their effects their processing their job management within the engine so it's not a free gain and a lot of the time at least earlier with the x12 what you end up with is what's called a rapper so it's a Direct X 12 rapper it's not actually doing anything at a low level it just calls the dx12 API so it looked like dx12 to the user but performance might actually be slightly worse than the x11 or opengl and that's why with early testing you saw negative scaling because it wasn't actually built for the API not ground up anyway and it's just being wrapped and it's basically executing it another API or at last not quite as low level on the hardware as you would hope for one of those low-level api's so that pretty much that answer is really mostly questions one from last week and I a message to Michael Kern's one of our writers back because he's an expert in keyboards a bit more than I am question was from Bassam BOTS hereof he said basically was asking got my hands on g610 orion brown keyboard i haven't had any mechanical keyboards my possession before so i really had high hopes for this one to put it mildly I was disappointed the keys felt mushy almost an attack how about modifiers were loud felt more like a membrane and keyboard to me I trade tried razer blackwidow before and those were satisfying the press and I asked Michael Kern's what he thought he said quote okay so it sounds like you don't like cherry MX Browns ie you don't like light tactile switches a decent amount of people myself included don't on the other hand you like like clicky switches like the cherry MX blue or razor green clears and ergo clears now yeah there's decent of people who like uh what am I reading here there's a decent amount of people like you and what a lot of people like cherry emma is cherry MX clears and ergo clears thank you Michael for spell checking this for me that says MX clears are so much stiff in that they take more force than cherry MX Browns or razor greens but ergo clears Ahmad and he's got a link to that desk Authority net will have the mod for her go clears are like cherry MX Browns what you used but with a more pronounced bump so you may like them more i would suggest trying amex clears as they are somewhat easy to find in keyboards and while ago clears maybe a bit better for you they aren't easy to find and require hand customization generally so there's your answer to that as always if you've got more questions anyone post them in the comments below thanks for watching page on like a post your videos subscribe for more 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.