Gadgetory

everyone i'm steve from gamers nexus tonight and this is our star citizen alpha 2 point 0 GPU benchmark where we look at the public test universe or PTU and do some tests for a few different game modes including the new Ark Corp and Crusader maps we just tested art Corp for now we also look at the murray cup racing free flights and the hangar module for performance and we'll look at all those in terms of Delta scaling and an absolute FPS metrics as we roll through the content here so previously we've looked at the original arena commander dogfighting module we've looked at the hangar module and now we're back as the game has developed quite a bit more since our last look at the game we've expanded our test methodology to be much more comprehensive as well and try and really look at all different parts of the game because there is a very wide range of one gameplay type and 2 FPS performance depending on which game mode you select so we look at a lot of those it's unrealistic to look at all of them on all of our cards right now that's just it's not going to happen because it's a lot of reasons that I'll explain in a moment but we got some Delta scaling stuff and some GPU verse GPU testing in the standard racing part of the simulator alpha 2 point 0 for those who do not follow as closely is the first major really launch for star citizen in terms of looking at what the grander open world will feel like as the game approaches it's more complete stages so part of this includes the ark Corp and Crusader maps which are two maps within the PT you and those allow for some basic multiplayer interaction including seen other players on the map chatting with them and collision that's that's about it in terms of the on foot gameplay right now and wandering around there's some pretty cool stuff I had one of our testers Mike gaglione he he works on the benchmarking with us he was wandering around and collecting some b-roll of the art Corp city or station I guess you could call it and it's pretty neat some of the stuff that you see is there's a lot of potential for immersion there and it's definitely an intensive game which makes it fun to benchmark and that's what we're doing here today so let's dive into this first of all big disclaimer this game is not done this is alpha two point oh you did not miss the one point o launch of star citizen proper this is how the second original game it's still an alpha and because of that that means there is effectively a guarantee that everything's changing this is not just some pre-release week before launch test it this is if you're watching this video in the future when the game has come out just ignore it move on to our more recent benchmark because I can guarantee these numbers will change but the major things here are drivers the GPU manufacturers obviously haven't made drivers to support star citizen yet because it's not a complete game so there's no manufacturer side driver support to a large scale and then on the star citizen side there's very little optimization at this time including some limited LOD scaling and that is visible in the pop in effect in game so pop and just briefly very briefly explain it is when you see objects and the distance begin to pop into the view like they just materialized almost and in star citizen that's really it's not great right now and that is for I would imagine performance reasons because you see a lot of stuff like fences and sign posts and things like that popping in maybe 30 40 meters away where in most games it can be a bit farther than that so that is definitely a concern in terms of the benchmarking with performance you will see numbers chains as the developers change the pop and distance change the LOD scaling create more assets finally it's art assets so all of that disclaimer stuff out of the way if you are trying to play alpha 2 point 0 this is what you will be looking at for your performance metrics first of all we tested a few things here we tested the Murray Cup race for the actual card vs card performance we tested the PTU for delta scaling on a 980 TI and then we also tested free flight for delta scaling on an ayiti I when I say delta scaling this is a new chart that we introduced with the overwatch benchmarks and delta scaling is basically a percentage difference look at the frame rate increase or decrease between settings and between so for star citizen we say we're not going to test all the cards on the PTO and the race in free flight that's it's far too much work we it's just not feasible to test all four different existing game modes right now on 10 12 configurations that's a lot of testing to do for a game that's not done yet so we do the big test on the race and then the delta scale takes a 980 TI as a sort of cross section and says what is the percentage performance gain or loss in game mode X versus game mode why those would be things like free flight and Hayner and then we also look at what is the percentage gain versus loss at very high high medium and low settings those are the four presets available right now there is no fine tuning available without going into i NIS or something like that and we're not going to talk about ionize at all here because it exits the scope of the article which is the benchmark the game as it exists right now so in free flight spaces largely empty and if you're exploring that empty space you are FPS will be higher than what we're showing here because we aim to create a worst-case scenario in these benchmarks that's for a few reasons I've discussed many times before not the least of which is reproducibility it's very important to be able to reproduce these tests and their results with great consistency and eliminating things like variants introduced by just being way too high of an FPS because empty space or eliminating variants from other players I EPT you testing on all cards is it's very important for us to do that so we do the free flight test by flying towards some objects in space including sort of a space station capsule type thing and an asteroid and then for the racing tests we wait until exiting the sort of hangar area because that is that tanks your FPS hardcore when here in that area so we exit that we do a 30 second run of the race take one corner and then for the PTO testing I had Mike figure out a course to test for that that presents sort of a mid-range FPS we're on the higher end cards we've seen 50s to 60s in terms of frame rate and in some parts the PTU you'll see up to 70 even 100 p.s some parts you'll see as low as a dip of maybe sixty percent so it one depends on what card you have and two depends on where you're testing all of our test courses are shown in this video and an hour separate video on the channel that's two minutes long if you want to see those and replicate our results so let's dive into this first of all we're going to talk about the Delta performance tests if you want to see the full methodology the other components in our bench rig hit the link in the description below for a lot more depth and you can find that there we're starting with the Delta performance charts then moving into the charts that compare multiple GPUs against one another these first charts are used to assist in extrapolating the theoretical FPS throughput in each area of the game the first chart shows performance scaling in the PT or persistent universe module and that's looking specifically at the arc Corp map on a 980 TI at 1080p and the bars represent percentages a few quick notes here the PT right now is tough to get into in terms of the instance will fill up there will be network drops there's players which can impact FPS load time is a massive issue so that is why we excluded it from our all GPU testing and instead was something that is tried true and creates reproducible and easily validated results which would be the race course later on so that stated we've established very high as our baseline for FPS assuming that this is sort of the target setting and that's why you see that at one hundred percent each reduction in settings would be high medium low shows an improvement in performance and this offsets the percentage positively showing a scaling of around twenty percent going from very high to high effectively zero scaling in the PTU specifically between high and medium and about a forty three percent performance win from very high to low which translates to roughly a 20-percent gap in medium and low now this is on tonight ati so on some cards lower end cards in particular you will see slightly different numbers here going from the higher settings the lower settings because of the just the level of load that they're put under but this is a basic representation of what you can expect and helps with extrapolating results from the main test in testing our core proved to have the least noticeable difference between high and medium and falls within a margin of error the art Corp benchmarking doesn't scale the same way as other game modes do right now in the PTU and in the game in general I don't know how much of that is the team the star citizen teams graphics optimization or settings changes or network optimization with reduction of elements or things like that for passing on the network I'm not sure what's going on there but medium and high and our testing look to be effectively identical on multiple cards actually we did validate that result and the rest shows a swing between 100-200 forty-three percent performance gain at low moving to free flights the maximum performance gap is thirty-one percent between a very high and low with high seen a six percent gain over very high and medium seen almost a sixteen percent gain over very high or ten percent over high there's one more chart to look at before we move the GPU verse GPU tests so we're still on the 980 I stand alone for now this one shows a game mode preset scaling at 1080p and high settings again high not very high specifically as it pertains to our benchmark courses the asteroid Hayner which is what's in our press account right now was the least intensive area of the game running at 90 FPS average and with relatively consistent low performance metrics free flight and just somewhat on the complexity of the cockpit obviously so in our all of our tests it is key to note that we flew an M 50 interceptor I think that's an origin ship and our free flight testing involved a linear course in which we approached a couple of satellite objects in space and they're a bit of a worst-case scenario for the testing so if you fly around an empty space understandably it's much less intensive and you're going to see a higher FPS than we saw in free flight testing when flying at more geometrically complex objects the race course provides a more intensive look at ship gameplay by planting us in a city and offering significantly more object density and geometric complexity than in free flight so is a lot more polygons being drawn there's a lot more draw calls there's a lot more stuff hence the density of objects and that makes for a slightly more intensive test it's not until we get to the Ark Corp testing that FPS falls heavily presenting a swing of forty-two percent between the ark Corp map and the hangar with low performance metrics proving a lot looser than previously and that's a big performance gap depending on which part of the game you're in and is a major contributor to y star citizen is so difficult to benchmark so that sort of shows what you can expect in terms of the performance scaling between each device let's look at card vs card for the 1080p 1440 and 4k resolutions on all presets except for low in this case at 1080p very high we were only able to exceed 60 FPS on the 980 I reference card one percent and 0.1% low frame rates are extremely consistent and tightly timed with the average something that we want to see for improved fluidity of gameplay the GTX 980 4 gigabyte card runs 10 frames slower in the race map than the 980 I planting it around 17.2 percent slower than that card and play becomes noticeably choppier as we approach the 40 and lower FPS range occupied by the r9 290x and slightly exceeded by the r9 390 X and these new 380 x can't cut it at tennety and very high it's just it's not really that playable right now without doing any additional tweaking drop into high settings still at 1080p sees mixed FPS increases with most cards landing again between three and seven percent the 390 X now approaches 50 FPS and the 290x still shows its poor one percent and 0.1% low performance the performance across the board slots the same card hierarchy as the previous chart so they are all listed basically the same order but we introduced the GTX 950 to this chart falling short of any playable ranch at 26 FPS average medium is significantly more forgiving than very high but we again see fairly linear performance across the board dropping to medium allows the 290x and 380 x to enter general playability worst case scenarios but the 380 x does struggle a bit the gtx 960 also isn't particularly well suited for play right now but can get by until a time at which optimization improves or cars are upgraded just don't expect precision now again these are sort of worst-case scenarios for testing except for the Han or bit right before our test begins where you'll see even worse performance than this but generally in free flight or in Hannah it is going to be much more playable than what we're seeing here on average the 970 sits above the 390 X in all of these charts and that's something we've seen increasingly commonly as we've tested the newest games out including Assassin's Creed and some other games that we've tested in the game benchmark section on the website or on the channel so the nine sanity just slightly outranks tonight the 390 X right now and that's consistent across all the charts 1440p is understandably abusive on the system introducing it 1.6 million more pixels than 1080p nothing is perfectly fluid at 1440 and high and it's really not much better at medium with 14 44 k is effectively out of question right now with the 980 TI falling to a laughable 23 FPS at 4k and medium so really don't expect to play 4k right now with even though 980 TI so that is our benchmark for star citizen alpha 2 point 0 including some of the public test universe testing that we did on our corp and other maps if you want more detail on these tests you want to see what our cpu eyes and all that information then hit the link the description below for full information and a more in-depth article look at this obviously right now we're not going to really recommend GPUs to purchase because the games not done so I'm not going to do the usual what's the best GPU at different price ranges thing that I do for these videos but as it stands right now generally the game is playable in the 40 ish FPS range as long as you're okay with a little bit of chopped in the worst case areas of each map because it's not like it's a competitive game right now it's not csgo it's there's not intense dog fighting with actual high consequence as their theoretically will be when the game launches so right now you can get away with that lower FPS more so than perhaps you normally would be able to and it's it's playable on some of the lower end cards as long as you're willing to live through some choppiness and deal with a bit of issues on art corp and the other more intensive maps but again not recommended a GPU this time and that will look at the game as it continues to iterate going forward so thank you to all of you who have watched our content and continued following the channel we're producing a lot of it and we'll continue doing that in the immediate future if you like this kind of stuff hit that patreon link in our postural video huge thanks to those of you particularly in this subset of the gaming community for backing us and helping us out because it does help us upgrade or upgrade our equipment like the lights and microphones and all that stuff so thank you all for watching a lot of fun producing the stuff and we'll be back very shortly I'll see you all next time