Gadgetory

it was 2014 when we last benchmarked DB softs watchdogs IP and now with the second incarnation of the game here we're starting our benchmarks with GPU performance analysis across the entire ten series from Nvidia and the AMD rx series along with some older cards from the last generation of hardware from both vendors there are a lot of cards on the bench and we spent a good amount of time studying the graphics options to better understand what it is that watchdogs to finds the most stressful for GPUs and hardware before getting to that this coverage is brought to you by AMD freesync displays like the LG 34 um 57 - P which is a $350 budget ultra wide display at 2560 by 1080 learn more in the link below let's start with the watchdog's - benchmark and we spent the first hour of the game just looking around and studying how the graphics settings and the different elements within the game impact performance this is looking for things like meshes the complexity of them and their impact on processing ability for whatever card we happen to be testing look at texture resolution and the vram consumption of that texture at least so far as we can tell and other things like shadows and how they impact gameplay which is actually pretty heavily in this particular game there's really complex foliage trees things like that and they do interact with light in a very intricate way with watchdogs - this creates a pretty good representation of how a world would actually interact with the Sun but obviously you've got a pretty big performance trade-off depending on the graphics setting so that's part of what we looked at and the full methodology for this benchmarking course is of course defined in the article linked in the description below so if you're curious about how we tested it why we tested a certain way what drivers we use things like that it's all down there I will make one quick note the drivers used just because every time it's a question and people don't seem to click the article the drivers used are brand new we have access to them normally early the Nvidia cards we're test with the latest Nvidia drivers the numbers in the article and those do officially support watchdogs - and the same is true for AMD tested with the latest and these drivers so that's all done as it should be and the rest is defined below now getting in Thane's loosely starting graphics settings include geometric detail or the mesh complexity and number of primitives drawn extra details which adds to mesh detail and uses a percentage offset to do so terrain and vegetation are both next which also both impact LOD and view distance on their respective elements so for some specificity vegetation detail directly correlates to draw distance of those more complex trees that I was talking about including individual leaves and branches and how far away you can see them shadows are a bit interesting as well although we primarily tested with core shadow settings that would be low to ultra you could also go beyond the presets and configure them to use things like percentage closed or soft shadows or high fidelity frustum traced shadows both of which are end video technologies that we previously covered in depth and you can learn more in our coverage of those technologies the rest is pretty self explanatory and has very an impact on weather and fog effects both plentiful in the game's recreation of San Francisco and generally just a couple of other small tweaks things like anti aliasing which is a large tweak and then your normal filtration settings as always we conduct multiple test passes per device so for some devices depending on how many resolutions we test in quality settings that could be dozens of tests for one video card and that's to ensure parity we're looking for deviation we're looking for consistency in the frame rate measurements not just from an analytical perspective but from a gameplay perspective can you actually feel the difference one test run or one day to the next when you're playing the game and that's averaged as necessary and we of course use our 1% end point one percent low methodology which is basically taking the slowest 1% and point one percent of frames is all defined in a video elsewhere and we look at frame times when doing that but we look at the slowest percentile metrics and then use that to extrapolate the low end performance rather than a minimum or maximum which are both outliers and should be ignored neither hfts nor p CSS were used in the larger benchmark charts though we did take a quick look at them and that's just really for some fairness between brands and because they're kind of unnecessary technologies and do have an impact on performance well looking at the game with 1080p ultra very high and medium for the 1920 by 1080 settings and then we'll be looking at 1440p ultra and 4k with very high settings the game was tested in near Mission Park in San Francisco where we mix in cars grass and complex trees we carefully tested at exactly the same time of day with the same conditions which is a sunny day that cast more shade from the foliage and that more heavily impact performance so yes it does actually matter quite a lot what time of day you benchmark if it's raining or not and what direction the sun is shining depending on where you're walking when interacting with those shadows let's get to the first chart this one's simple with this chart we're showing it rough performance scaling across settings by aggregating a couple of our mid-range cards note that this isn't a perfect representation of how all cards will scale but it gives a rough idea devices with less vram for example will struggle disproportionately toward the high end where these averaged here might not a note also that max settings the quote unquote max is tuned with ultra and then further tuned with extra details Max and the shadows set to hfts let's move on to some hard data here's our 1080p ultra results EVGA is GT X 1080 FTW unsurprisingly leads the pack operating this configuration at about 76 FPS average with lows consistently around 60 fps the gtx 1070 a card around roughly $400 is capable of juggling 1080p at Ultra settings and that's with a 64 FPS throughput and closely following lows no trouble at all for this one and then AMD's r9 Theory X pushes a frame rate comparable to the 970 SSC not benefitting noticeably from its HBM that's not the first time we've seen this performance out of the fury X either or its stack up against cards like the 970 SSC moving down to the mid-range cards we're seeing MSI GTX 1060 gaming X operating at 45 FPS average about 39 1% lows and about 37 0.1% lows that's fairly tightly times across the board and remains reasonably playable we'll look at how a settings reduction supports the GTX 1060 in a moment though for folks on the GTX 970 as shown here you'll be at roughly performance to a GTX 960 at least the ones we've tested and moving to MSI's gaming X version of the rx 488 gigabyte card the card is producing an average FPS about 41 with low is at 32 and 28 respectively our gaps between averages and lows are a bit larger than on the GTX 1060 here though the visible difference is not necessarily that substantial at this point next the RX 474 gigabyte which is 35 that's 7 FPS average with 0.1% low is also somewhat distantly timed this card begins to exhibit some stuttering during play and that's a result of frame time disparity and can be resolved largely by a settings reduction speaking of these are our very high charts and they're a little simplified the high end devices remain left to the ultra tests and at the top of this particular chart we see the GTX 1060 gaming X and 970 SSC both at roughly 60 FPS average the our X 480 gaming X follows closely at 56 FPS average effectively equal to the 970 SSE and performance and largely insignificant in its differences from the gtx 1060 the new cards on this chart include the gtx 1050 2 gigabyte and our x4 62 gigabyte the 1050 non ti card is pushing a 32 FPS average with sort of rough lows really the RX 460 is in similar shape they're worse off at 26 FPS average and 13 FB a 0.1% lows this manifests itself noticeably in jarring stutters just for quick reference we ran the 1050 2 gigabyte in 462 gigabyte at medium settings for a more realistic look at performance and the result was an output of about 50 FPS average for the 1050 and about 4045 FPS average for the rx 460 so a lot more playable if you drop down to medium on those cards 1440p ultra it's a bit taxing but we gave it a shot anyway the GT X 1080 FTW pushes about 55 FPS average with the lows consistently north of 40 next the 1070 runs a 45 FPS average and remains fairly playable with how watchdogs 2 is built anyway the gtx 1060 fury X and 970 SSE are all in the stuttered range of gameplay and they would be better matched to a 1440p resolution if using settings closer to high watchdogs intensive settings we only tested three cards at 4k resolution because we were pushing for a very high quality the GTX 1080 FTW isn't quite 4k ready at very high settings but could do so with a slight settings reduction 1070 SC would also need to be dropped though closer to a mix of high settings maybe a bit of medium for FPS sustained in the 40s the fury X isn't much of a contender here at all and stutters like mad with those low frame rates but if you really really wanted it I suppose you could drop down the quality quite a bit I would generally prefer a 1440p output with a higher quality in this scenario so watchdogs - it's a very intensive game it's not really that easy on this hardware even stuff like the 1080 FTW is struggling at resolutions with quality settings that in other games wouldn't really make it struggle quite so much and that's not necessarily an optimization issue so much as it is just what's being drawn in the game I mean think about it if we walk by a park it's drawing tons of grass and just from looking at it it seems like most if not all of those blades are individually interacting with light with shadows you walk under a tree with the Sun behind you shining down over the character shoulder and that's casting a ton of really finely detailed shadows from an already complex tree onto the character and from a lighting standpoint it's pretty cool graphics but it's not easy for a GPU to handle so this is a game where the FPS does not look quite as impressive as some of the other games we've tested recently however watchdogs - just sort of subjectively speaking here is one of the games that is fairly agreeable to play in the 40s for FPS I was really not complaining that much when in the 40 45 FPS range obviously a bit higher a bit better but it wasn't as big a deal as say battlefield one where you really want that high FPS for competitive play that's not to say watchdogs - can't be competitive but certainly with the way the game is designed the lower FPS is a little bit more agreeable than in some of the higher speed games so overall if you're seeing FPS metrics sort of in the mid to upper 40s that is generally acceptable for the card that you're seeing the numbers for however you could always do a slight settings reduction and in this game a good place to look for you to start with settings reductions would be things like shadows because shadows have a huge impact on performance and then next would probably be things like foliage and vegetation which again a lot of things to draw and a lot of light interactions if your shadow settings are set so that they are interacting so that's where I'd look as far as cards that can play watchdogs to obviously like always most things can play technically it's just a matter of how low you're willing to go on the settings so for my notes here I've got that the 1080p very high settings and 1080p Ultra settings that look more or less the same when you're actually in the game playing it not actually sitting there staring looking for differences between them but when you're actually playing and engaged they pretty much look the same it's there's a couple differences but it's not bad with that said very highs a lot more playable than ultra in terms of FPS the jaunt is pretty massive here if very high is acceptable for you and it probably should be for most people then the arcs for 70 and up are playable on the AMD side the GTX 970 dot are playable on the Nvidia side the GTX 1050 Ti is pretty close a couple settings weeks would make it possible to play with fluid frame rates and have good gameplay that's acceptable and that's with a reduction to the settings obviously the GTX 1060 and our X 480 really have no problems at all with 1080p and very high settings Ultra is a lot harder and that's really where you're gonna need something like an Rx 480 or a GTX 1060 and even those we're getting by and about the 40s but like I said that's okay in this game for the most part now as for the golden standard 60fps metric that's where you've got a push for something like a 1070 SC or there abouts and the currently current generation doesn't have an equivalent to that but with the advent of Vega that will obviously hopefully eventually come out but that's not here today so we can't really tell so for today the 10 70s he looks about right for that settings output higher resolutions are pretty rough anything short of a 1080 will generally need to operate at a settings reduction and even the 1080 FTW we've got here we have to drop the settings quite a bit for 4k to be happy that was generally in the range of high rather than very high or ultra so that pretty much sums it up now I know there's always questions of why didn't you test X card or Y card we've got limited time and we've got limited hardware I know we've got a lot of video cards but it's not necessarily the ones that you always want us to test and then there's obviously again limited time so 980 I unfortunately did not make it in today I would like to test it and we will probably test it in the future the same goes to the 960 and the same goes for some of the r9 series cards like the 380 X and the 290 X which I'd really like to test but we have content pieces for those coming up shortly so as always to see those subscribe for more patreon link in the post roll video you can click on that to help us directly if you like this type of reporting the article is linked in the description below for more depth I'll see you all next time you