Starfield GPU Benchmarks & Comparison: NVIDIA vs. AMD Performance

[Music] today we're benchmarking Starfield bethesda's Skyrim in space fun fact and one of the retakes uh I called it skyfield so we've really made it Skyrim space whether or not they want to be branded that way we have some cool research tests here we've done a lot of performance benchmarking the game launched in Early Access about 16 hours ago at the time of filming so I've been here all night running those tests and really excited to talk about some of our findings for it this is cool because almost certainly there will be driver and game updates that will change the performance so getting something done early allows us to collect some interesting historical data as those patches roll out over time so we can plot the trend and see how the cards are improving now for Intel because you can't divide by zero their Improvement will be difficult to calculate it doesn't work on Intel Arc right now and that's a known issue Intel has acknowledged it they said they're going to have drivers ready on the public launch day but that's kind of what you sign up for with Arc at the moment and this was an AMD exclusive title so it launches with FSR Intel's behind the ball on the drivers because maybe it seems like they got to access the same time we did which was the same time everyone else was able to if they paid the 100 bucks for it and Nvidia they've got a driver but it looks like that might have some updates anyway so given that a lot of people watching this are probably planning new pc builds around star field and other games ton of them coming out right now we have our own exciting launch with this which is the 15-year anniversary mod mat the new gn15 pc building mod mat is launching as part of this Benchmark video instead of putting an outside sponsor in this spot and it's launching on store.gamersnexus.net right now this special new anti-static conductive mod mat is our best one yet and it's part of our GN 15 round of products it's our newest large mod mat at four feet by two feet or 122 by 61 centimeters for PC Building and parts modding we just received this at the warehouse alongside our gn15 shirts and both will be shipping out soon the limited run mod mat has the gn15 logo centrally plus a host of new diagrams to make PC Building and maintenance easier new helpful images include the most common motherboard front panel wiring layouts we've seen for LEDs and buttons 8p8crj45 wiring diagrams for those of you building your own ethernet cables some helpful angular and length references and more we also just included stuff that's cool like a Zen 4 CPU dilocation diagram for its unique dye layout and locations the simplified compositional cross-section of the same type of CPU and some quick references for typical fan orientation to help new pc Builders the screw sorters on the left feature two GPU Silhouettes for screw tracking for your water block or rebasing projects plus a row of alternating color blocks to help keep mental track of where different parts came from the DDR modules are also to scale in the top right these are also high heat resistance mats making them great for Tube Bending and working with heat guns if you liked our recent CPU sample size testing and you find today's video helpful for your build please consider buying one of the large mod mats to help support our work and get something useful for your projects these gn15 mats are rugged durable surfaces that can take a beating and protect your table and your components from Dentons and scratches grab one on store.gamersaccess.net today and help us celebrate 15 years of running benchmarks like this okay so some context whenever we're benchmarking a brand new game that rolls out like this uh especially when it comes in as hot as star field did we immediately set forth trying to collect some data on how does it just perform in general in different environments of the games we'll be talking about that a lot today came in early access at 8 pm Patrick and Jeremy set up the benches for me during the day I shifted my schedule basically came in right when it launched worked through the night and then Jeremy and Mike came in early to help me finish up the testing and the editing so to my team who enabled all of this and my overnight Obsession benchmarking thank you we've got some cool content all right so these tests used the latest Nvidia and AMD day one drivers we have fresh bench to table benchmarks and again Intel's not on here it doesn't work yet there are a lot of different ways to test games and so there's going to be a ton of room for differing results as for our approach here the choices we made included disabling FSR disabling vsync and also disabling variable rate shading as well as Dynamic resolution these will impact results compared to tests that might leave them on so you should look at the numbers largely against themselves and avoid cross comparing if there are unknown variables in the settings our reasoning for turning those things off was mostly just to control the test environment so FSR for the most part are it's fairly predictable scaling so we're testing that native and then from there what we care about is the relative distance one card to the next we might revisit this with additional like specific FSR tests in in the future but since things are kind of in flux right now with the game and drivers and everything uh we kept it very simple with just native resolution and then we can extrapolate findings from there as for the rest of it uh standard to turn vsync off and variable rate shading does introduce variables as well so for the rest we stuck with fairly simple high and ultra settings we have some medium and low in there as well and we ensured that Dynamic resolution features were all completely off tested at 1080 1440 and 4K in the old school method which is just the resolutions as they are our goals for finding a test location primarily revolve around a couple of things one is repeatability of course so we can't have too many random elements like surprise combat for example happening into the test scene because we need that reliability the second point is that the test location should be close to a worst case but ideally it's generally representative of gameplay so it's not the absolute worst case possible it's more of like a heavy load like a city environment for example that's what we end up choosing it must still be a common gameplay scenario in other words but still a heavy workload for this process we captured the frame times for the first couple of hours of play and logged them this included the tutorial cave the initial Outdoors location on the moon or asteroid or whatever it was the ship take off space flight Combat on foot and combat in space Crete and its indoors research lab and New Atlanta City games are highly variable dependent on location and choosing a location has all the impact on the results that sets us up for a long-term Benchmark It This research was captured on a 4060 TI at 1440 PM High we want to emphasize here that the settings on the card are basically Irrelevant for the research stage other than one important thing prior to getting all of the gpus involved we start on one card and that card's job is to be GPU bound we want it to be around 50 or 60 or so FPS generally speaking assuming that there's a sufficient CPU Headroom and there is here and that's so that we can find locations that show scaling so ideally you want to be able to see a wide range assuming the game has a wide range of types of play like non-combat combat dense cities and open areas and space in this case so for this we choose settings that typically put us closer to that sort of middle 50 60 FPS range and then we can scale from there here's the chart we have FSR upscaling Dynamic resolution and variable rate shading all disabled here so there's no FSR or upscaling at all the 4060 TI at 1440p high held in the 70s in a few locations notably inside one of the areas of the early cave and inside the lab on Crete and again the importance here isn't the absolute frame rate of the 4060 TI because if you don't have that card or want it then that's not a helpful number but what is helpful is the range so we observed in the 40s FPS average for initial conversations where High poly detail is loaded in for characters faces Outdoors roam in range but we saw mostly in the 40s there as well space combat and space flight had us typically around 50 FPS plus or minus a bit takeoff had a severe frame time spike a couple of times actually but as this moves to a hands-off animation that just doesn't seem as important as say an actual gameplay we also captured two sets of 10 minutes of gameplay shown here in New Atlantic City these numbers are useful to give us a fuller picture of averaged frames across large parts of the game with loading screens removed from the numbers because those are pointless these should also be considered fairly representative of play we're going to spare you all the details of going over each of these numbers individually and talking about it and if you want to learn more about sort of our selection process for benchmarks you can check out our sample size test for CPUs recently we show a lot of the behind the scenes for that but there are two important entries at the bottom one is called all gameplay average that's the average FPS and one percent of each number above it in other words we're producing a common frame rate of 53 across all scenarios tested including averaging for the tests in the 70s and that's across the first few hours of play by measuring in different areas once we have this number we chose an area of New Atlantic City that was fairly close but leaned slightly heavy that way we have a representative sample of a higher load City and the end result was 48.6 FPS average here remember that it's possible there are areas later in the game that are heavier but given that we captured again a couple hours of play and count with these numbers we think the test area selected is a safe sort of lower end frame rate representation for high density areas without being ridiculous we tested in the residential district of the city and had a fixed walk-in path we did a number of one-off tests in this area but before we get into the comparative charts let's look at some quick settings research as the next part of our research process we started using our actual bench pass and for this we tested each preset with the one research card with again FSR disabled all types of dynamic resolution turned off and VSR or vrs turned off as well and this was in the same location using the same test using four passes minimum each and then we did this at 1080P and again we'll get to all the cards but here's this chart here we found that the 4060 TI at 1080p held an average frame rate of 53 FPS at Ultra improving by 17 by dropping to high now at 61 and remember the previous chart was at 1440 so that's the change here and cutting that to medium gave us an additional 16 frame rate cutting that to low boosted us to 82.6 FPS or another is 16 Bethesda has remarkably consistent scaling percentage-wise from one preset to the next year with our sort of dynamic scaling changes that makes it seem very intentional how they assigned these settings doesn't necessarily mean it's well optimized or anything but it does seem like there was some intent here it's about 10 FPS Improvement on this particular card for each graphic settings decrease one final research chart this isn't a frame time chart but rather just an FPS over time chart we're plotting the frame rate as we played through various parts of the game this is 34 000 or so frames of data it's about maybe 10 minutes of gameplay we truncated the chart to 100 FPS the spikes above that go to over 200 FPS in like menus inventory loading screens and that's just it's not important for the rest you can see the range was typically 40 to 70. with most of the numbers clustering around 50 to 60 FPS average on this device and with these settings the game has some plunges but otherwise it's relatively consistent we'll look at the run run consistency with standard deviation of just a few of the gpus that we've tested for this one this is not an indicator of the gpu's quality but rather the standard deviation for the card's test passes now each pass is an average of thousands of frames of data and we have a minimum of four pass says here we didn't need to discard very many we did run additional for some of these cards though upwards of six or seven for a few of them generally speaking we were within about one FPS standard deviation run to run lows are always wider here since there's less data to work with by their nature but the averages are extremely consistent from one run to the next making this overall relatively easy to parse and meaning that we didn't need to discard or as stated we run many tests so that's good up first our 1080p High chart this is the densest one and remember we did all of these cards in like 12 hours or so so this all got us started pretty well and then we'll keep adding to these as we approach the 7800 and 7700 XT reviews in this one the RTX 4090 the RX 7900 XTX and the RTX 4080 can all be ignored they're all bottlenecked these just didn't have enough scaling room you're not seeing a difference between them because there's not any room to see one we'll have to wait till we get to 4K to start seeing that and we'd need either a much higher end CPU with a higher clock or an area of the game with less CPU load and this has some crowd density to it shifting attention to Unbound comparisons the 6800 XT remains a strong modern option and it sits Atop The 4070 here with the 6700 XT technically leading the 4060 TI but they are functionally equivalent the RTX 3080 remains completely viable here as well not much of a surprise it's not really not old and it was a good card the 3060 TI and the 4060 TI have more of a gap here than they have at higher resolutions but that's Behavior we already saw in our 4060 TI review they tend to converge and the last gen option can even surpass the 4060 TI in some games at 4K resolution the RX 7600 and the RTX 4060 are roughly tied in average frame rate with the 7600 having a tactical lead and a price Advantage at least right now as for older Hardware like the 1070 and the 2060 we'd have to turn to lower settings for these we added these at the last minute just because we know a lot of people are still on them and it's nice to see the older Hardware in modern games now we didn't test many cards at 1080p medium but here are the ones we did the movement and absolute FPS isn't much the 1070 and the 2060 of course benefit from the drop in settings but not enough to move the needle without other settings reductions as well the RX 7600 overall does well relative to the rest of this group especially with its newer sub 250 dollar price that we've been seeing around we only ran two cards at 1080p low here are the results for them the 4060 gained about 10 FPS against its test at medium settings previously with the 2060 gaining a little less than that moving to 1440p and high settings the top remains bottlenecked although it's reducing a little bit the equivalent FPS between two resolutions for the top three cards indicates that bottleneck pretty clearly to us the more relevant comparisons happen below the 95 plus range the 4070 and the 3080 are about equal here with the 4060 TI and the 6700x see also close in rank none of the lows in this chart jump out to us as particularly bad in general they're not keeping Pace with the average as well as we'd like normally those that red and blue bar would be a little closer together and we've seen that in other tests in the past but they're consistently sort of disparate across all devices so that Gap remains true it's not like it's any one particular card that is the result of it it's just this is the game at this point we're on to 4K High next we didn't do as many tests on this one as the prior settings due to the intensity of the configuration the RTX 4090 predictably leads the chart but it's biotechnicality by reality it's tied with the 7900 XTX they're within run to run variants of each other for average one percent and 0.1 percent lows which is actually really impressively close the RX 7900 XT manages 87 of the performance of the 7900 XTX at 64 FPS average the RTX 4080 is just behind that at 58 FPS average which means that the gap between the 4080 and the 4090 is 30 here that's about the same as we've seen in other games dating back to the 4080 review the 3080 is also near its usual distance from the 40 90 although a little higher than we've typically seen it allows the 49 to be a little over an 80 lead it's not unheard of but it's a little on the high side you can check our old 40 90 review for more of that as for all these cards under the RTX 4080 remember that we're one at 4K and two establishing scaling in a relative sense matters more than the absolute FPS for the way we run benchmarks you can reduce settings or add up scaling and increase the frame rate to good playability on something like a 1600 XD here or a 3080 and with FSR even lower than that but for testing purposes today we're showing the ranking of the cards relative to each other the 1080 TI once was among the few early cards that could really handle 4K games well but for this one it's just not able to do that at least not without external assistance from modern upscaling techniques and to be clear it would probably be kind of rough even then we haven't tested that on the 1080 TI yet the game's still too new 4K Ultra is pretty simple we don't have many cards that can run this well so it's a limited list the RTX 490 ran 4K Ultra without upscaling and in the city at 68 FPS average with lows paced roughly proportionally behind that has the 700 XTX and the 4090 once again about tide with no meaningful difference between them and any of the three reported metrics the summary 100 XT allows the 700 XTX a lead of 14 while the XT non-second X leads the 4080 by about 14 percent again the 4080 it leads the 6800 XT here by 26 percent now even at the 6800 XT level you could still run at 4K Ultra with some use of once again FSR and upscaling it's an option here it's not a particularly great experience at 40 FPS but it's not far off from something that would be fine and drop in settings or adding apps SAR would get it the rest of the way there if that's really important to you so recapping some interesting findings first of all the reminder once again repeated it three times sorry but this is important this is day one and so we we push really hard to get as much good data out as we can early to help people make those early decisions it is going to change that means if you look at this video in six months from post and you should probably check our channel for a follow-up or for the most recent GPU or CPU review for that period because this this situation with drivers and game software it's gonna change but the data's still really interesting and the data itself May remain relatively scaled the same as it is today we'll see right now AMD has very competitive performance in some cases so as compared to the 40 90 it's fairly com it's actually particularly competitive as compared to the the 7600 the 4060 they have a strong foothold there compared to some of the other games we've tested in the past and that's interesting maybe the case that this is a jump that AMD had with thanks to their exclusive partnership it could also be the case that this particular game engine or the way it's built in general is benefited by the hardware the architecture for amd's ardene A3 uh more so or disproportionately so versus Nvidia and we see that sometimes we're one game versus another you you might have a 15 swing or something and the number is based on how it behaves on that particular architecture it could be a memory thing could be a bandwidth it could be rops there's lots of options there Intel will revisit as soon as they have a driver so we'll have news for you soon on that probably about a week and as for the game itself it's pretty taxing so just strictly speaking performance not talking about optimization or subjectively what we think of the graphics or anything strictly on the hardware it is a taxing game the top cut of cards are bottlenecks on other components that say 1080P and 1440ps but older cards like a 1080 TI a 2060. they struggle a little bit more than we would expect at 1080P and you get into a situation where at least right now with the current game version as of uh Friday the 1st at you know middle of the day Eastern which becomes relevant because who knows when they'll badge it right now those cards need some help it could be upscaling it could be a reduction of Graphics but they they need some help to really run this well now there's no fsr3 unfortunately uh there's fsr2 in there there's no dlss yet I think everyone knows that at this point and the Intel Arc drivers are kind of an interesting Trail to follow as well because we're not sure if that's an Intel like the the 100 plus billion dollar Corporation then or if that's a Bethesda Thane as in Bethesda slash AMD Nvidia has it working on their cards uh it doesn't appear that they have optimizations or many of them yet we did use the game ready driver quick cursory glance it looks like maybe the same as the previous driver we haven't 100 confirmed that uh it's been a long night of testing but the fact that Intel doesn't have Day One support it's disappointing we just don't know which party to be disappointed at right now it's also not totally unexpected you kind of know what you're signing up for with Arc everyone's talked about it at this point all right a couple things uh so the settings we wish they had an fov slider is one of the things I noticed right away it had kind of limited controls you don't have many Graphics options um and in terms of expanding that it's probably going to come from Mod support the findings we had throughout the game were pretty interesting so in that sort of case study with the 4060 TI seeing the 20 FPS range in conversational dialogue versus say indoors combat and lower density areas that was pretty useful for understanding the the plus or minus versus The Benchmark area we chose and we'll keep looking around for more variable performance in the game as well for now though that's it uh I'm gonna go to bed and get this video live not in that order and we'll be back shortly with more star field content really interesting games of test it's an exciting time right now there's a lot of cool games coming out or that have come out and we're excited to run the numbers on them so check back for more of that subscribe for more store.amers access.net to grab our brand new gn15 mod mat it is for the 15-year anniversary we have a lot of quilt diagrams and graphics on it it's very helpful screw tracking uh GPU Silhouettes for disassembly water block and stuff like that so grab one if you want to support us thanks for watching we'll see you all next time

Channel: Gamers Nexus

Views: 776,183

Keywords: gamersnexus, gamers nexus, computer hardware, starfield, starfield benchmarks, starfield best gpu, starfield gpu benchmarks, starfield pc, starfield max graphics

Id: 7JDbrWmlqMw

Length: 22min 34sec (1354 seconds)

Published: Fri Sep 01 2023