Gadgetory

Mass Effect Andromeda posed an interesting research opportunity for game graphics and so we set forth benchmarking the game with our updated GPU test bench and about 14 video card will be defining some of the game's graphics settings in this video and talking about performance issues that influence test results and how to avoid them and also we'll be going over GT performance at 4k 1440 and 1080p resolution before getting to that this video is brought to you by EVGA GTX 1080 and 1070 hybrid cooling kit which we recently used to hack together our 1080p ife hybrid mod the kit offers an efficient way to adapt compatible GPUs to better cooling performance and allow some more Headroom out of the cords link below for more information nice Effect Andromeda had a couple of interesting options in the graphics settings menu one of them in particular stands out that we wanted to go over just to clarify for everyone and that is frame buffer format under display settings we going over that and then afterwards we'll be walking it through some of the research that was conducted on Andromeda prior to running the full benchmark suite this is stuff that we wanted to start publishing to get a better idea frame times of specific performance issues or anomalies and this is kind of new how the game performs in different areas of the game that's pretty cool because we normally don't have that information and thanks to our patreon backers in discord we were able to get multiple save game files to put it all together frame buffer format offers two different settings half 16 and compressed speaking of with technical artists that the company we learned to that have 16 that doubles the size of all colored buffers in the rendering pipeline the more bandwidth is needed during rendering it also doubles the precision of the intermediate color buffers and is something that won't be visibly noticed on a standard displays that is what most of us have considering the performance hit though half 16 is generally not worth it for users and so is something with disabled in our testing as for compressed mode that one keeps five to six bits of precision per color channel but doesn't have artifact of 16 bit color because the rest of the 32 bits are used on range it's possible that you'll see some color banding issues in places with s er but should be uncommon bright skies for example can cause from exposure issues sort of like Ian be mods but it's otherwise it better than the performance hit of half 16 HDR however stretches range and precision and that's where you'd probably want have 16 instead because you'll start noticing color banding otherwise as always with our game benchmarking we started with analysis of how the game performs in different options configurations or in this case different places in the games that would include planetside testing testing on Tempest testing on the Nexus things like that and looking at the data for all of them to figure out where makes the most sense for our benchmark course so that we can most accurately represent the game as you will most likely be experiencing it whether that's a worst-case scenario or a best-case scenario or somewhere in between so we have some of that information for you a big shout-out to our patreon discord numbers for sending over save game files to accelerate that process and most immediately important before all that we notice that just like battlefields one the first test pass of Mass Effect Andromeda has some interesting variants it happens on both AMD and NVIDIA but the performance hit is the biggest with ambient so we have some data on that and how we've elected to resolve those issues for our benchmarking this first set of several test passes with an r 9 fury x at 4k and ultra settings that shows that some of the variants that we're talking about the important thing here is the disparity in frame times this is repeatable and seems to happen consistently on first load of a level but goes away after exploring the immediate area you can see it in this frame time chart that we're seeing massive noticeable stutters in the first pass of testing but significant performance improvement in subsequent passes like past two in this case the most exaggerated spike in this instance is a two hundred 80 milliseconds frame time at one point which means that you're standing around for about a quarter of a second waiting for the next frame that is noticeable to further illustrate here is a set of data just a raw data table from our RX 488 gigabyte gaming x at 1080p ultra and this one our first half past shows 72 FPS average with 56 FPSO lo and 6fps 0.1 percent lows this shows itself in stutters during the first pass but it smooths out in subsequent testing within about a minute normally and after walking through the area and loading everything into memory or vram as necessary we improved from roughly six FPS 0.1% lowest performance to 53 in the second pass massive difference here and the contrast this vs nvidia the GTX 1066 gigabyte gaming ex at 1080p ultra posts a 60 FPS 0.1% low output for the first pass then 65 for later passes and we also see a difference in averages the difference here isn't visible and isn't anywhere near what we were seen with andy but it is consistent and measurable so this is obviously some sort of issue with Mass Effect Andromeda though it is exaggerated on AMD just be very clear here we're not saying that Andy is worse overall nothing like that what we are saying though is that this is something to keep in mind because you might see stuttering as you immediately load a level but it should mostly go away as you start playing that level this is particularly noticeable in the planets where you're probably going to be on the planet for a long period of time so it's not a huge issue after you immediately load the area surrounding your character that said it does mean that we have some considerations for testing namely do we include the first pass in the data and average it which would really bring down the 1 / 7.1 for something low as mostly point 1 percent on Andy Hardware in a way that would make it look a lot worse than nvidia or do we get rid of the data so we did what we did with battlefield one testing which is run a minimum of four test passes per card / resolution and then we're eliminating the first test pass that means that the AMD data will be more consistent because we're eliminating that sort of outlier although it is repeatable in the first has passed so what that means is looking through the data what you are seeing is the latter the later three test passes and we've eliminated the first one for all cards that also means that you're seeing numbers that are more immediately comparable so it should be fairer in that regard but keep in mind that there is a potential stutter there that we're not showing it but you've seen it now in our research piece so that should give everyone even ground understand where the different cards are anywhere we're assuming here that there's some sort of driver level or game level optimization issue and given that this was an Nvidia partnered game it's possible that there's still some work to be done on and the devices by either Andy or by Bioware as for other research we ran several test passes that lasted a few minutes for the first planet for The Tempest and for the Nexus zone overall performance numbers suggest that planetside driving as the highest performance no surprise there as it is the least geometrically complex while planetside battles tend to mostly equal the average frame rate of the tempest or of the Nexus zone the tempest ship has consistently the lowest frame throughput but overall all test environments are roughly equal when you average the frame rate output over a long period of time a couple of minutes which benchmarking is a pretty long time and so because we can see that the averages of mostly align it doesn't matter a whole lot where we benchmark as long as it is not in the open driving areas of a planet because that's when we see frame rates that are at least a couple percent if not ten percent or higher more than what we see on campus and Nexus or in battles planetside because of this we decided to run our tests on Nexus the Nexus seems to be the most representative of overall average performance when considering the lower frame rate battles and higher frame rate desert driving and mixed frame rate ship or station traversal all sort of land at around the same place for average and frame time performance having to find the interesting discoveries let's get into the benchmark charts for 4k 1440 and 1080p note that as always the full testing methodology is linked in the description below along with the rest of the analysis from this piece if you are curious how we tested something what drivers we used what specific versions of cards we use it's all defined in that article check that first just to quickly note on drivers they were the latest at time of benchmarking which was official supported drivers from nvidia and from AMD so those drivers were the ones released by both video cards vendors specifically for Mass Effect Andromeda that's what we use you can find the specific numbers in that article at 4k Ultra and the Nexus benchmark our 1080 TI hybrid we built is performing around 55 FPS average with one percent lows at 54 and 0.1 percent at 50 moving down the list a 1080 FTW from EVGA runs about twenty percent slower than the 1080 ti at 52 FPS average at whitlows about 10 FPS below the 1080p I is one percentage point one percent values this is still plenty playable even at Ultra as we generally found that Mass Effect Andromeda feels fluid enough for play at around the 50 FPS range your mileage may vary depending on how picky you are about frame rates but that's what we thought based on how this game plays and feels at 50 FPS or higher the GTX 980 TI hybrid from the maxwell generation runs about thirty-nine percent slower than the pascal pen a DTI at placing at 40 FPS average with low is just above a 30 FPS we can find and these are 9p reacts at 32 average with lows in the 20s at this point alongside the 1070 at 38 FPS we're beginning to experience more frequent and observable stutters or tears the experiences poor enough on the fury x the 1070 and potentially the 980 TI that we'd recommend dropping to medium settings instead of ultra one note here there's really not much of a performance benefit by going to high as it turns out in our test course so it's really ultra medium or bust on 4k for what it's worth the msi RX for any gaming x 8 gigabyte card hands on pretty closely to the fury x at 25 FPS average tying the 390 x and outpacing at both the former flagships from both andy and nvidia the 290x and the 780 ti moving on to 1440p ultra was the same necklace workloads at 1080p I gianna hybrid operates about two times its frame rate as at 4k and now pushing 126 FPS average as well as lows at 90 FPS one percent and 79 FPS 0.1 percent the 1080 FTW two trails at 101 FPS average making the 1080i roughly twenty-five percent faster at this resolution that's not quite Nvidia thirty-five percent faster claim there's a gap between a 980 TI and the 1080 TI is a noteworthy sixty percent improvement from the pascal TI the gtx 1060 gaming x6 gigabyte card now appears on chart as well placing that around 58 FPS average with lows at 47 and 46 and these are 9px has nearly identical performance across the board at 58 average of 47 1% low and 44 0.1% lows both the Andy fury X and NVIDIA 1060 output frames it was fairly even frame pacing which is somewhat remarkable for a newly launched title we don't always see that performance regardless of vendor the EVGA GTX 970 SSC runs about ten percent lower them 1060 but is still playable with ultra settings this is followed closely by the RX 480 which in this current state of the game seems to perform about where the GTX 970 and r9 390 X perform interestingly we see both former flagships continue to exhibit their weaknesses and higher resolutions with the 780 ti and 290x falling below the modern art 480 and the GTX 1050 and frame rate at 1440p these two cards perform more closely to an RX for 74 gigabyte card dropping any of the three tailing devices to medium settings would push you to 50 FPS if you prefer to 1440p to the higher setting moving now it's the most widespread resolution 1080p posts the 1080 TI nearing 200 FPS average the 980 TI at around 120 average and the 1070 tied with a 980 TI everything here even the worst-performing devices have tightly x frame pacing which makes for a generally smooth experience is above roughly 50 FPS average the gtx 780ti and r9 290x do a lot better here thanks to the lowered pixel throughput which is something that these cards were more built for and the 290x lands a few frames ahead of the gtx 780ti from a cowboy generation that said nvidia holds the top half of the chart at least until Andy pushes its highly-anticipated Vega GPU of the same name architecture as the currently best performing AMD card is the fury X at 82 FPS average which is at trailed about eight percent by the RX 40 Gaming Act and then that is followed closely by the r9 390 X as for low-end cards the 1050 does reasonably here though dropping to medium settings would be ideal the GTX 1050 TI would do fine but we didn't test it due to the usual time and origin activation restrictions the RX 462 gigabyte card struggles and definitely needs a settings reduction for better playability for 1080p and ultra settings it seems that Mass Effect Andromeda runs play ibly on something like a GTX 950 TI and up just extrapolating or an RX 470 end up that's the 4 gigabyte model with before 70 offering about a 60 FPS average throughput at the bottom end and then if you look to things like the gtx 1060 or our X 480 we begin to exceed that 60 FPS mark with a 1060 pushing nearly 100 FPS in the nineties or somewhere around there and that is also capable of 1440p play ability to have some extent though you may want the slight settings reduction depending on how much of a stickler you are for 60 FPS gaming the same is true for the fury ax hanger on that card tends to perform around where the 1060 is in some of these charts depending on the workload and resolution again as for 4k it's looking like 1080 class hardware is required for this game if you want to play on ultra at forte so you could take the settings hit and get some other cards up to play ability if you determine 50 FPS as playability for this title your mileage may vary on that and then finally we notice that the difference again between high and ultra is negligible so between AMD and NVIDIA was never more than a few percentage points which tended to be a couple of frames difference at best and that was with the Nexus and the tempest workloads so it's mostly an option of old traverses medium for this game with potentially low if you're running something like an RX for 60 or a GTX entity as an option as well so as always the article is another link in the description below patreon.com/scishow styleside directly subscribe for more thanks for watching I'll see you all next time you