Gadgetory

everyone we had a clerical error in our GPU benchmark for Gears of War 4 which has now been fully rectified with this content today and that error was really a result primarily of having three different people working on the benchmarks this is a growing operation and also a mix of the game having 40 graphics settings and Python script to interpret data and then present Mon which is also sort of new to benchmarking because it is a DX 12 title so there were a lot of variables going on and it's been corrected basically the error boils down to a simple thin its chart generation all of the tests thankfully we're actually valid 2/3 results were valid tests were good the data was accurate but they weren't put on the right charts so basically we had AMD and NVIDIA on the same chart using different settings for 1440p and for 1080p 4k was fine so the chart generation error created an unintentional misrepresentation of data for that you have my apologies I have corrected that we've also done some extra tests now so now we have a throw-in card of the GTX 1063 gigabyte that was not on there originally thought that'd be fun to look at the difference of three verses six gigabyte performance we've added that we also have proper ultra tests this time for 1080p and 1440 so now we have 1440 1080 P at high and at Ultra and we have 4k at high as previously let's get to it first off Gears of War 4 is a uwp title that's DirectX 12 enabled which means that usual testing via fps overlays isn't supported for that reason we're using present model with an in-house built Python script to analyze the data and create 1% in point one percent low outputs that we use in our normal fraps and fraps testing and the game also has a few dozen graphic settings as I said about 40 that really complicate the testing but we've gone through them and defined the majority of them in the article and we've also made sure that they are applying correctly to the different settings used for testing what you're looking at now is the first 30 seconds of the Gears of War 4 benchmark this is the course that we use primarily for graphics testing with our present montt commands set to expire right around with the camera angle changes you can also see our Gears of War for maximum graphics video which shows some of this benchmark info on the channel we used high settings for 4k and properly used ultra settings at 14 piano 1080p but we've also gone back and added high settings for those two resolutions drivers used were recommended by Microsoft AMD and NVIDIA alongside the required Windows 10 anniversary update and that includes the nvidia 3:7 3.2 drivers and an d 16 10.1 which are the same as 16 9.2 for Gears of War 4 the optimizations are the same in each an AMD has done some optimizations we spoke with their team about that starting with 1920 by 1080 properly configured to high settings across the board we're still hitting a bit of a CPU bottleneck with the high end devices as limited at approximately 155 FPS average with our third 59 30 K at 3.8 gigahertz that is the limiter here the GT X 1080 the old man 980ti and the gtx 1070 are all peaking at around 150 FPS average showing no issues whatsoever with Gears of War 4 at 1080p with high settings the low framerate performance is reasonably tightly timed with each card dipping at its lowest to 105 FPS at the 0.1% low metric or the slowest 0.1% of frames output as represented by frame times this plants Nvidia close to 120 Hertz displays and actually beyond it depend on if you're looking at averages or not and gears won't work for isn't exactly a game where you need that you barely even need 60 honestly but it's at least doing well with the lows moving down the hierarchy the AMD r9 fury X is now pushing significantly better performance than what we saw with ultra hitting 124 FPS average and lows of 90 2.8 fps and 63 FPS respectively that number is deceptive though and it's for the same reasons that we discussed originally that has not changed we are still seeing critical variance between test passes with AMD devices at least some of them and here's a quick look at some of the raw data from the fury X running on high settings where we're seeing as we were seen before with ultra a dip to 25 FPS for the absolute lowest output in the 0.1% lo metrics not the minimum but a point 1 percent low and an output of 96.2 FPS for the point 1 percent low in another one of the test runs that's a range of 70 FPS for the absolute lowest performance in point 1% metrics which manifests itself occasionally at least once every minute or two as a stutter or hitch and framerate fluidity and if you look at the averages you'll see that and these are nine px remains fairly consistent in its average output it's just these low values it suffers now this is also an important note to make about testing and why our efforts to make this Python script are really worth it this time actually well it's always worth it but especially now because if you look at the minimums output by presentment if you use present modern launcher or if you're just looking through the data manually the minimum is really useless it's almost always an outlier and when we run through the numbers and look at a 1% 0.1% metric we get these outputs that are much more consistent and accurate with what you're seeing on the screen the minimum because it's one frame out of thousands is not really a good representation of the low-side performance of cards that's why we're doing this just to quickly show gears over for at 1080p with the ultra preset settings here's a look at the performance on both AMD and NVIDIA with that preset applied we're seeing 143 FPS average output on the GT x 1080 FTW hybrid 98 FPS average output on the r9 fury X the difference again is in the low performance when you're seeing some of these stuttering and gameplay and some of this can be seen in our review of gears over 4 already published on the channel those stutters do appear during gameplay but the averages are good for and the overall and really this can be problematic at low end excuse for NVIDIA - like the 1068 1440 it's just a matter of AMD and the coalition working together hopefully to stabilize the low-end performance for those cards at 14:40 with high settings the GTX 1080 FTW hybrid from EVGA is now sitting at 130 1.7 FPS average with tightly time to lows including exceptionally consistent low frame time performance as shown in this raw data on the screen with a 980ti trailing there's a noteworthy gap between Pascal and Maxwell performance when it comes to the low values and that's something which has been improved upon by Nvidia for this generation of GPUs the GTX 1070 and 1062 game well at 1440p with the AMD r9 fury X operating at nearly 89 FPS average just above the GTX 970 SSC from EVGA and the GTX 1060 gaming acts from msi at 75 FPS average effectively the 917 1060 in this tests are identical in performance and that's really only differentiated by Pascal pushing higher low values next the MSI R X 480 gaming acts at 13 16 megahertz is operating competitively to the 970 and the 1060 with a 70 3.8 FPS average though it's low performance is just as a variable as we've seen in the other test that we previously published and we'll be publishing again in a moment we saw at least one visible stutter in framerate output in the test runs but AMD has a reasonable foundation with its average FPS to improve upon performance in the lows with future updates as they did with doom with a 390 series when that card saw 25 percent increase just from a driver update about a week or two later ultra is sustainable at 1080p on the r9 380 xrx 470 GTX 970 and up the AMD cards at the low end again needs some extra adjustment to deal with those drops as for 1440p at Ultra we're seeing performance output of about a hundred FPS on the GTX 1080 with reasonable lows that are just below 60fps see 706 FPS on the r9 fury X + 62 FPS on the msi r9 390x remember the low values for AMD are a little deceptive here again for the same reasons that we just stated previously since the variance is so great that some test passes will make those values appear higher than they are during certain runs or an extended gameplay the GTX 1060 has a little bit of trouble with stutters at 1440p but it could be made playable with the drops down to high from ultra on a couple of settings like SSR and the rx 480 is in the same boat and could also be made more fluid with a few fine-tuned setting tweaks the final chart hasn't changed nor has the one paragraph of the script talking about it the tests are all using high settings at 4k and were since the start and 4k high gears or 4 is handled at approximately 70 FPS by the EVGA 1080 FTW hybrid with a 1070 SC following 34% behind at 50 3.3 FPS average last generation 980ti still have classes pascals 1070 at 57 FPS average of though it does exhibit a wider FPS range overall which plays to the one percent in point one percent metrics we're showing as for AMD the company is headed up by its fury X at 50 FPS average followed by the msi r9 390x at 43 average and the our X 480 gaming X at 38 point 1 FPS average 4k is relatively playable across most of the hardware that we've tested here particularly if you're willing to dip some settings down to medium for the RX 480 because gears isn't a particularly fast-paced game and a hard 60fps isn't mandated for a good experience though it may be desired you could always get away with a slightly lower fps depending on how you feel about your gameplay and this next chart shows setting scallion which also hasn't changed from it previously here's a look at these settings scale that using some average test passes between devices this gives an overall look at performance and how it scales between various presets so if we didn't test the setting that you're interested in you can use this chart to extrapolate performance relative to the tests we've conducted there's not a huge visual difference between ultra and high for the most part but the vram impact is actually somewhat substantial the game tends to sit at around four point eight gigabytes maybe five the RAM utilization by the end of the benchmark run alone not talking about the actual gameplay here and that's as tested using eight gigabyte GPUs the metric is output by the games I own a built in performance monitor that we're not entirely sure how it's measuring vram but it is a newer API it's low level and it should in theory be more accurate than the vram request metric you get from cheap easy or similar software and while I've been talking you should also be seeing visual comparisons using some b-roll we captured of identical test scenarios that will help give an idea as to the performance output with all of the charts updated to properly contain the data that they are meant to reflects we can now reach a conclusion and it's really somewhat similar to the original conclusion which is that for low-end performance you really need to pay attention to those low values 1% point 1% below is because those manifests themselves invisible stutters in gameplay that we have b-roll of this is seen on some Nvidia cards though less frequently the GTX 1060 for example at 1440p ultra is not something I would recommend you'd want to drop down to about high or somewhere maybe in that area to reduce the chance of a stutter being encountered Andy it's a bit more of a pervasive issue and we see this across all the graphics settings we tested for the most part and it hopefully is an issue that can be resolved with drivers or a game patch and in talking to Andy we know that they are working towards a conclusion this technically 69 2 in 1610 one contain the same driver optimizations for years over for this is officially stated by Andy but the the future can still hold improvement in that regard with the frame time disparity we'll call where there's less consistency in the frame time one frame to the next on some of the AMD devices overall the game is fairly playable even with the higher-end ultra settings with high settings 1080p you can play on pretty much every card we showed on that benchmark chart and the same is true for 1440p at high 4k is a little more different but it's obviously 4k so you're at a 200% scaling from 1080p it's an issue just because a resolution but overall the cards that are out this generation pretty much handle Gears of War 4 without too much issue at 1080p RX 4 sixties pushed it a little bit you can drop the settings maybe slightly to medium for some fine-tuning and then 1440p the RX 480 is mostly ok other than those low dips and you can do some tuning to hopefully resolve some of that to an acceptable level 1068 1440 high is okay as well for general performance so thank you again for watching as always patreon like the post well video helps out directly links in description below for more information I'll see you all next time and that's what we're doing today we're benchmarking Gears of War for did you see that