Gadgetory

everyone welcome to another episode of ask GN as always if you have questions leave them in the comment section below I'll try to get to them for the next episode we're trying to do one of these per week now pretty consistent lately so before getting to this episode which has a couple of pretty interesting CPU and GPU questions this content is brought to you by Roseville and there : encase which is right here I just revisited this as an aside to the 570 accent to Sony are the two new coarser cases that we reviewed on Monday which is the day of filming for this but we talked about those and revisited this in that coverage this is a tempered glass case tempered glass on the side and on the front and also on the right side from the perspective of the camera it's got a tint to it which some of the other cases don't and then as far as price it's on sale this week for $120 so that's pretty significant drop from the original $150 asking price if you're curious the original vendor of this case actually also makes the Anna DS AI crystal so you've seen the AI crystal which for a brief while was one of really the only new tempered glass cases on the market right after Computex after this was shown they came out with theirs and they actually come from the same source they're both from Jones bow which is a sort of OM designer company and they put together a lot of cases like this so really comes down to the logo that's in it and then things like warranty and price which right now this one's cheaper so that's the sponsor first question that this time is shoot I don't have the oh this is from an older video but I finally got an answer to it so I don't have the contours name written down but they asked hey Steve I just wanted to know if there's any particular reason for Andy to stick with the PGA style of stock sockets for a m4 instead of LGA so this question was asked a while ago I thought it was a pretty interesting question I sent it over to AMD and had them paint a few different people internally and figure it out so the feedback the answer from the tactical marketing team which is the only group that really provided an answer is as follows direct quote PGA packages generally provide a thicker processor package that is more resistant to bending under the forces of a mounted heatsink PGA's sockets are also less expensive than LGA sockets which can help reduce the cost of a motherboard and for those who don't know basically the options LGA is what Intel pretty much uses for every socketed CPL does use for every socket at CPU the alternative being something that's soldered to a board which it's not really applicable here they use LGA that's where the pins are in the board in a socket and the CPU just has the contacts on it with am these processors as many of you likely know the processor itself has the pins like AMD saying it's a thicker package of the processor as a result of that but the motherboard is now less prone to potential damage like bent pins so the motherboard socket basically the pins feed into a cluster of the holes for the pins and then there's the contacts underneath hidden under a plastic protective and mounting layer so that's the socketing setup for both devices and and the uses pga and their answer to me it was that it's just because the resistant to heatsink bending and that the cost is lower both valid answers so I didn't get more than that but that's that's what we got for now next question is from smoky dots 500 spoken says I have a typical AMD or Nvidia question regarding releases of different GPU architectures with Nvidia's 10 series of GPUs we are seeing releases of GP 107 106 104 and 102 do these represent different architectures or unique dies or do they represent the common adage of flagship architecture with components on die that failed Quality Assurance question goes on for a bit but that was the the basics of it so this is kind of referring to for those who don't know instances where you end up with a product that is a or was originally meant to be a higher end product and has been sort of locked or restricted because it failed in some ways validation it's not a hundred percent how it works with GPUs that is how it works in some places an example would be some of the older AMD GPUs you could actually unlock and get some more power out of them and that was really more of a thing where there was demand for a certain type of product more than another and so Andy would release more of those to fill the demand that's it's kind of how it works in general this also happens else from the industry but for GPU specifically this is another one where I reached out to the vendor so ask them Vidya you know the obvious thing to kind of point out here is that when you look at the cards like this is the EVGA FTW that we've been working on for vrm temps that's got a couple thermal probes still attached with adhesive when you look at that dye the dye package is a pretty visually different size from what you would see that's a 1080 for example on a 1050 Ti GP 107 is much smaller so in that regard if this concept were true what would happen is you'd see something like this a GPU 104 or whatever 102 package in its full large size being placed onto something like a 1050 Ti and then it would be more limited in its performance but the die would look the same size as that that's obviously not how it works you've all seen the tear downs at least the ones that we've done if not elsewhere and they're pretty obvious smaller size as you go down the listing the tearing of cards so then what happens well when I reached out to Nvidia they basically said for their cards their GPU specifically if they are underperforming or if they for some reason don't function just from normal yield yield is not 100% perfect for really any electronic device especially silicon so there's always something that fails and what happens is they recycle them somehow so not in a product as far as I know but recycle them as in normal recycling where you try to reuse stuff and I actually go off-camera somewhat have one I just remembered that I got this years ago I think it's got the GPU identifier rubbed off of it but this is a failed NVIDIA GPU and I got this at their silicon failure analysis lab we toured it probably 2013 or something and the guy there is pretty pretty damn smart was talking about how all their equipments used like scanning electron microscopes and things like that and magnifying silicon to 1.5 million X so they kit like mil a whole into the service middle hole down into it and look at almost basically a transistor level what's going on when something fails pretty cool stuff but I got this from him and it's basically from Howard Marks his name so it's a failed chip in this case they just have a bucket of them and give them out to visitors so that would be one place that we used as opposed to again in a lower to your card so I hope that answers that question there are places where parts are I won't use the word recycle because confusing but are reused in a lower tier product but the current ski series of GPS is not one of them with the 10 series I did ask and I'm waiting on a response what about the GTX 1060 s because there's two SKUs of 1060 s so in that specific case it's a little unique because a 1060 you've got the base model and then the 1063 gigabyte model which has ten percent fewer SM so the question is 10 percent fewer sm's is it a higher skew is it a 1066 gigabyte that's been restricted or is this a separate line and manufacturing that's producing these 10% lower SM count chips specifically for the three gigabyte cards have received an answer die yet but I only asked a couple hours before the video so next question Wesley Runyon says Steve why is it the different benchmarks ultimately give different answers for instance passmark versus fire strike hash mark rates my score with a 1070 from MSI gaming Z better than a 1080 or even a Titan XP then very similar pcs with less speed ram clock so you see outperform mine it's impossible to cheat the system or software yeah yeah I know turn off all the background apps etc etc not buying it times blah blah not cutting it that's kind of rolling over a lot of what we do I monitor my hardware closely and that excuse is nonsensical I am curious to see what your opinion is of synthetic benchmarks so synthetic benchmarks are tough even just 3d mark on the same system ignore all the other systems you're talking on our cards you run it a few times on the same system there is variance and that can be the difference between a card with one clock trading places with card with another clock and that's a problem so there's reason we don't use 3d mark scores and our reviews that may change at some point but right now I just don't like the very incident and I don't like the idea that I have to run it several times to average an answer that seems reasonable because it's just it's as extra work and it makes more sense for me anyway to spend that time benchmarking another game instead so we can spend that time doing that but that's three mark so past mark things once you start mixing and matching tool you can really can't look at these things outside of their vacuum that they're in so a pass mark 3d mark fir mark they really are not meant to be compared one against the other when you're looking at scores it's sort of look at it in a vacuum everything passed mark everything 3d mark whatever as far as why your card might outperform others I don't I don't know specifically what you're looking at but like I said because the tools are so different it could actually be the case that there's a difference in memory or in the memory clock on the GPU or something like that well that might actually be significant to the particular tool that you're using an example of this would be fir mark fir mark and combustor the MSI benchmarking tool combustor contains variants of fir mark that are branded by MSI I find them to be a bit more abusive fir mark is behaves like a power virus it burns the vrm so basically that's what I've been using on this thing it sort of sits there and incinerates vrm to a point that is not realistic for what you would get in a game and the testing as you'll see this week from this card will see a difference of sometimes 20 Celsius and vrm times probing the VRM the MOSFETs directly when using a game like Doral a Metro last light any number of modern games battlefield as opposed to using fur mark for mark always being the hotter of the two so that's an instance where when you're comparing two different cards temperatures do actually matter quite a lot because if you've got as a better example maybe an AMD card where there are different throttles in place than on that video cards running into those temperature limits with fur mark would produce results that may look worse for a car that's cooled worse at like the Sapphire Platinum Series rx 470 which is a reference cooler that would look worse than something like maybe a gaming X even if both have the same everything else and that's just because of the cooling and because of boost limitations as well so on the 10 series cards because of the way Boost works if it doesn't have thermal Headroom you're going to get a lower clock and that will impact your score so I don't know that temps blahblah uh cutting it really understand where you're coming from but also temperatures are pretty important depending on what you're testing now once you're looking at the good cards like a bunch of me a CX twin frozer whatever else is its tricks those types of coolers that are all pretty much always below 80 Celsius then the difference comes into play from things like power target and power availability so again an example might be if you're testing over clocks or if you're even without over clocks just because the way GPU boost 3.0 where it's on Nvidia you've got more power Headroom because the vrm is designed to supply that power like this card we can toggle to 130% offset and that gives us something like i'll top my head I'm not sure I think it's like 245 watts total available power something like that anyway that's a pretty good impact also again on the clock rate and on the clock stability specifically so the clock rate may look about the same in terms of the maximum number but the stability of that number will be greater as you reduce your thermal Headroom as you increase your power availability and as you increase your voltage and those three things to some extent are done automatically by the card so this isn't just you going to the MSI after burn or whatever in changing it in general synthetic benchmarks are not my preferred way of testing things to really put your question to the test you'd have to kind of take one of each device put it in the exact same rig and test it because you don't know how there's those tools or programs what they care about in terms of system resources like again is another final example 3dmark if you're on the full 3d mark test it will run a physics test in there and that's not graphics processing for the physics that's system physics processing that goes on the cpu and so you've got to graphics tests and one physics tests three total test passes in 3d mark you're running a lower M CPU or one of the different clock or maybe the temperature was different this time or maybe the voltage was different this time because of all the C states and all the ist and all this crap that's in BIOS they have to disable to really get an accurate run to run test something changes your physics result could be lower and that drags down how the card looks and the same is true with these other tools so that would probably be your answer I think after going through the rest it but I think that gives everyone a bunch of different contingencies to consider next question is by the way that was a good question because I think that is one needs to be discussed but next question Peyton Pringle says how comparable is performance sustained at SSA a or super sampling anti-aliasing at 1080 to 4k assuming all other values are the same a game settings computer hardware etc and then Peyton Rangel goes on to elaborate that super sampling and anti-aliasing is basically rendering the game at a higher resolution than sampling it down to the display resolution for 1080p monitors have the same resolution as a 4k monitor so yeah 4k monitor if you got 3840 by 2160 something to the tune of a bit over 8 million pixels I think it's the same number if you do 4 times 1920 by 1080 in terms of raw pixel count so to answer the question which is how do they compare native 4k vs. super sampled I did a test of that when D sar & VSR were first posted the AMD and NVIDIA techno and videos DSR and AMD as VSR so those were their super Sandler they're scaling examples where the game would be rendered at the higher resolution that you specify which could be over the native monitor resolution and then fit to the screen so you end up with what appears to be AI or pixel density and greater quality of image this reflects itself mostly in things like textures so if you look at a ground texture or a cloth on an arm and something like The Witcher 3 it's a really good example where with DSR or VSR you'll see greater detail in that texture normally to the tune of things like a more clear line in the fabric rather than sort of a jagged line or maybe you see lines and cross hatching that you couldn't really perceive before so that's where you see it performance or I should say visible quality when we took screenshots side by side years ago they looked pretty much the same I'm not sure if that's changed I don't think it has but back last time I tested it they did look in terms of like one screenshot 1080p sampled rendered at 4k down sampled to 1080p screens whatever versus a native 4k screen they looked about the same so I I haven't actually used as a user dsrvs are I always just use the native resolution but I'm sure there are good reasons to use either one it's just not something that I've done but it also depends on what kind of hardware you have so if you end up with this is a good use case if you buy say you want a 10 7 8 10 80 whatever now or even an RX 480 and you have a 1080p display with a 480 and up all three of those devices you could easily do 1440p so no harm in rendering the game at 1440p and fitting it to your 1080p display it will look like there's greater pixel density and quality will improve detail improve mostly so hopefully that answers that basically the answer is they look pretty much the same to me last time we tested it which years ago also the the scaling performance was about the same as well.there within and percentage point or support less being point one percent one test last question is this yes let's get techy asks Steve have you ever considered growing a beard maybe not a full-blown hillbilly beard but a beard at nonetheless I think it would match your hair like a boss hashtag of beard life I'm going to stand very still so that Andrew can mask in a beard on my face and we'll see how it looks and hopefully it is a full-grown hillbilly period so thank you for watching I will see you all next time on we remain right here so that the mask doesn't break and a subscribe for more patreon like the postal video see you next time I'm the bad man