Gadgetory

let me ask you something why do we need to render this or this or this or this or this tunnel or this piece of road or this large hairy animal this benchmark is rendering creatures that use hair works even when they're miles away from the character and the camera we discovered this while running benchmarks in a zone with no hair whatsoever zero none at which point we realized by accessing the game settings files via a hack that disabling hair works would still improve performance even when no hair works objects were on the screen before that this video is brought to you by thermal Grizzly makers of the conductor not liquid metal that we recently used to drop 20 degrees off of our temperatures thermal grizzly also makes traditional thermal compounds we use on top of the IHS like cryo not and hydro not pastes learn more at the link below so that was a cowl we showed you or something and it wasn't an accidental cow either that doesn't mean it's a malicious cow but it's not accidental we even found an LOD line drawn across the map where you can see the transition of details so Square Enix has marked this space as being clearly not visible to the player and yet still way beyond the boundaries of this line miles away we see the cows in full detail complete with fur there are also examples of : present in the game there are examples of selective removal of unseen objects there are examples of LOD scaling and yet here we are with our friends the hairy cows rendered in full detail this investigation derived from just an innocent game works benchmark when we found a guide online to disable various settings manually because the benchmark doesn't allow you to do that out of the box and what happened was we found that in our 90 second benchmark that we had been running for game works there was no instance of hair works in that 90 seconds it was cut intentionally that way because the next 30 seconds involved the hairy mammals in the game that used hair works so the idea was we'll benchmark all these other game work settings in the first 90 seconds and then we'll get the hair work stuff later what we found though was that when disabling hair works without ever encounter something that uses hair works we were still seeing a performance deficit on both nvidia and AMD cards it's actually quite significant even for nvidia they see frame time and consistency pop up from hair works being enabled when there's nothing with hair works being drawn i just really want to drive that point home so this is where it gets kind of interesting because it makes the benchmark disingenuous it's it's not really something we can base anything off of because it's not realistic to what the user is probably going to experience when the game launches and it's fun 100 gigabyte form what's more interesting is that the nvidia cards as expected have a hit of about 15% from hair works the AMD cards as expected have a hit of about 37% from hair works now it's expected because hair works is ultimately an nvidia thing they optimized for it they've had access to the game probably for longer at least closer access to it so not necessarily crazy what's crazy is that this is a technology not being used to enhance how the game looks in its present state in our current camera position and yet we're still seeing massive hits the extent of which is basically two acts on AMD hardware and the improper use of hair works and the improper culling of things that shouldn't be there isn't necessarily Nvidia's fault they're not the developers after all but they do have their name all over the box and the game so they have some level of responsibility here to ensure that things work we've been in contact with nvidia over this issue and know that the company is looking into it for now it seems like square-enix didn't add the buffalo to the list of things that should be occluded or removed and it's something which has detrimental performance impact we even validated that the grass is specifically using turf works just to make sure that hair works wasn't being used to make grass or something like that and the only thing that might use hair works would be the characters and we know that the characters aren't using hair works presently for two main reasons one there's no dynamic movement of the hair as the characters move and two by using Nvidia Ansel yes intentionally we actually intentionally used Ansel we can fly into the character's head and see that the hair is being drawn by using alpha textures the hair is attached to a plane the characters head similar to how leaves are attached to trees meanwhile looking inside the Buffalo we see the individual strands of hair protruding from the body clearly indicating hair works integration and indicating that it's even working we can toggle it on and off and see that there is a distinct difference in how the Buffalo looks with it toggled on and off via that settings file let's look at some charts for some basic examples before even getting into the details of the comparisons this chart of relative performance scaling with game works shows how AMD GPUs and NVIDIA GPUs are affected unequally by game work settings the Vega 56 card maintains 66.3% of baseline 100% of the gtx 1070 performance at 1080p high and when we completely disable game works something that typically is impossible while still maintaining equal graphics settings we see that the vega 56 card manages to maintain now 90% of the GTX 10 70s performance hikes from a 66% baseline that's a massive gain and one that illustrates uneven performance scaling for perspective and video cards scale with one another almost one-to-one on this same chart as for what specifically causing the biggest performance hits that award goes to hair works the single disablement of which grants Vega a massive performance spike turf is the next most responsible and is what Square Enix is using for its grass technology as for a terrain tessellation our initial hypothesis for it was actually completely wrong because we had hypothesized that the tessellation of the ground was impacting performance the most noticeably potentially causing the biggest performance disparity for AMD who have historically had issues with a lot of tessellation although they've improved recently the reason that's wrong is because by disabling that we see very little difference in performance on a and the end Nvidia it's 1 to 3 FPS different and it scales equally and linearly on both so when you disable it you can see the terrain kind of pushed down as the tessellated elements are pulled up so it's there and you can turn it on and off very clearly but it just doesn't impact performs like we thought it might now we would have never known that without the ability to change the graphics settings and that's entirely thanks to the users on reddit who made a utility that makes it pretty easy to change their settings so we thank you for that contribution it's very upsetting that Square Enix didn't provide something like that because the entire point of a benchmark is to allow people to rather transparently look at the game and figure out how it performs with different graphics hardware under different conditions not to produce a poison set of data that has users aggregate from all over the place with all different hardware configurations all different CPUs into one big pool of charts on their own Square Enix website which is completely invalid because there's zero control whatsoever and further not having the ability to change settings means that until we found this sort of exploit posted by others there's no way to know what's causing the performance disparities and that's really important because it's important to understand if it's something that can even be fixed when the game launches so anyway let's get to the FPS charts now that are building the relative charge this is the settings configuration we're using for the benchmarks we're toggling off game works under high and we're leaving the default options still enabled so basically when looking at the charts what you need to understand is that they are cumulative from a high baseline the high preset sets the game with all game works options enabled except for VX AO which is presently disabled them less manually enabled and it does work and the shadow library is also disabled so all the things you're going to see on the chart are the default all game works settings on except VX AO and shadow libraries and then disabling whatever the the test setting is for that line item on the chart even with game works completely disabled the Vega 56 card is offering at a 10% deficit to Nvidia that's not completely unreasonable just note that AMD is not free from anything here either they have some optimization work to do it's not all just Square Enix but AMD has drivers planned for closer to the game's launch so we'll revisit it then ultimately though the frame time performance is still undesirable under all settings for AMD even with game works completely disabled and AMD again should be addressing that closer to launch more importantly however the performance Delta between the various settings using just the native 1080p high settings with our 90 second benchmark we recorded 38 FPS average across multiple runs disabling turf and hair works nearly doubled our framerate giving some perspective as to the perceived performance capabilities and we don't encounter flow very much on the benchmark till the very end so we can ignore that for now terrain tessellation also has much less impact than we expected and versus baseline Andy is gaining an insane 37 percent performance by disabling hair works but leaving all the other native gameworks technologies enabled and video meanwhile gains about 15% from the same conditions and enabling vxa Oh shadow libraries things like that under the all game works on category shows that the devices take a further tumble but these settings appear to be disabled by default so as for the 1070 then and videos handling of the same settings first note that we complained about frame times on both AMD and NVIDIA cards in our GPU benchmark showing that the company had trouble processing its own game works features without tripping over wide frame to frame intervals the result was occasional stutter and more disparate frame creation time we can finally illustrate this at baseline our 0.1% lows we're at 21 fps although higher than AMD's it's really not better in terms of average to low relative scaling disabling hair works specifically grants us a major uplift in frame time performance but just a moderate one in average FPS the average increases by 15% with a lows increasing by more than 2x game works seems to be hurting performance for no apparent gain here aside from some grass detail and in one location some hair detail and that's the big question here it's completely reasonable that gamers would worsen performance on even on video Hardware it's not like it's advertised as a free performance booster its advertises something to increase the apparent fidelity of the game and that costs resources so that's not the unreasonable part as a consumer however you should care about what you're getting for the technology and just looking at the benchmark which is not a full representation of the final game mind you but is being promoted as a futility and people are buying devices based on the results so that's what we're criticizing now so as a benchmark if you can enable these different game works technologies and let's just ignore the hair works magical cow that you can't see feign if you're enabling them and you can't really see a difference but there's still performance impact what are you doing why it's it's a completely invalid benchmark because you can't look at the numbers and this isn't sublime any particular party but you can't look at the numbers for something that gives you no settings control you can't disable those things unless you figure out how and no one ever expected that to happen and if you have no settings control and you can't disable things and you're forced to run on high settings or medium settings it's not knowing necessarily what those even do then you really don't have a good idea of what hardware's required to play the game and there's no actual visible advantage in a lot of the cases of this technology in the benchmark that might change it very well could ultimately this is a 3.7 gigabyte download against a game whose specifications call for a 100 gigabyte plus harddrive it is very likely that none of these numbers will apply to the final game and if they do then that is indeed a very sad thing so let's just talk about a few places where we tried to find differences here hair works obviously is functional it's a little too functional because it's applying globally and you get the performance reduction everywhere but it is functional you can see it on the cow buffalo things tessellation however is seen at the beginning of the game for terrain tessellation you can see that Square is pushing down the road and pulling the rocks out as tessellated elements there's no big impact on performance at less than 2 FPS per card even on AMD so it's something that you can kind of see although it does go away later in the benchmark but we'll see how that acts in the final game either way though this is a setting that has some visual impact even if it's not really all that necessary but there's no real performance cost right now either we looked at grass as well this one's obvious they're taller blades and strands but they don't seem to conform to player movement to like in the promised tech demo maybe that'll come later either way it's denser strands which we think might be particle meshes and that's a versus textures for grass on the ground the result is that disabling turf works gives a patch your Balder look to the ground whether or not you think that's a good look or a bad look or worth the performance is entirely up to you although Nvidia users don't have as much to worry about we still plotted a six point five percent deficit on the 1074 turf whereas AMD plotted a 12 percent performance deficit flow is visible at the very end of the benchmark where the campfire is lit up so that one does actually appear but it's not too common and other than that VX AO is presently disabled you can turn it on it's not really that visible of a difference unless you start zooming into things and the shadow libraries also take a bit of looking to really find the differences so at the end of all this there's rendering of unnecessary detail and that's problem for a benchmark because maybe the game ships like that maybe they ship a game that renders everything on the map when you're nowhere near it ok it's representative in that case and I apologize but I do implore you to revisit your development strategy that's probably not how it's going to work though so it's it's a useless benchmark in that regard it's also a disingenuous benchmark and misleading in the very least because what you end up with is charts that frankly make one vendor look a whole lot worse than the other where there is literally zero difference in what you see on the screen that is fact you can disable hair works you can enable hair works and if you're standing anywhere except for where the cows are maybe maybe one other place in the game but whatever it's a six-minute long benchmark stand anywhere else except for there you don't see the impact of hair works but you're taking a massive performance hit on the AMD cards and even a performance it down to Nvidia is just less so so just to kind of temper that a little bit that's not to say I'm videos being malicious but there is a problem where either square-enix or Nvidia or someone in between has a benchmark that is not representative of reality you're getting performance that doesn't match the visuals that are coming out of the screen so proud to say we found this bug ourselves you can find more information in the article below as always we've been working with Nvidia on this issue for a few days now and the company is working diligently with Square Enix in attempt to resolve the issue Nvidia didn't code the game but once again they've got game works branded all over it and considering how game works can be controversial it's in their best interest to work with Square to resolve this and to our understanding that's what's happening for now though this benchmark is not valid in our opinion and we will update you if and when that changes more likely that will be with the final release of the game not with the current benchmark so for a game that promises that it's a game based on reality maybe not so much but yeah that's that's how our discovery for this one so just to kind of reiterate here all the benchmarks we've done and everyone else has done on Final Fantasy 15 are completely pointless because ultimately what you're dealing with is testing a benchmark utility that has things like random pieces of road weight way off in the map background being rendered any way that are impacting performance and yes they're being rendered no it's not ansel interfering with this or anything like that I mean you could you can fly on the map and see that Square Enix has LOD scaling lines drawn on the ground like lines in this head they have intentionally and selectively absent objects it's not like Ansel is reaching out into the game libraries and finding the exact object that was removed and pulling it in and putting it right where it belongs that's probably not that's almost certainly not happening so we're fairly confident that this is in fact an instance of overdraw or of elements being drawn when they really shouldn't be and that is at least that is well probably largely on Square Enix that said it's a benchmark utility that they clearly rushed and pushed out before it was ready and the game should hopefully be different so we'll revisit this topic then but for now just know that the benchmarks really they're valid insofar as the performance of the benchmark utility if you want to know how well you can run the benchmark utility we've got you covered not a good example though for how well you can run the final game particularly cuz you should probably be turning off gamer at settings that you don't actually see the effect of especially if you're on AMD hardware but even on NVIDIA hardware because again it's not just AMD here NVIDIA has a frame time hit for these settings too so you know turn those things off but you can't in the benchmark making it somewhat invalid anyway it's parked up benchmark utility we don't recommend it thank you for watching you can subscribe for more as always go to patreon.com/scishow and access to us that directly and to fund further swearing because that always brings up audience retention or we go to store that guarantees access dotnet slash mod matt to pick up a mat like this one they're on backorder now i'll see you all next time