Gadgetory

we started out wanting this to be a companion to our GPU benchmark for destiny - hoping that we'd have a full suite of CPU benchmarks to definitively and confidently state which cpus are better in the current beta state of destiny - what it turned into was more of a research piece we looked into performance across different settings we attempted to optimize those settings on the CPU side and looked into multiplayer its campaign performance a lot of the stuff we do is the GPU benchmark except applied to CPUs as well and there's a good reason for that which we'll talk about through this video before that we partnered with LastPass to bring you this video LastPass is a password manager that helps generate random passwords for each account and sharing unique passwords for every log in as fast strengthens account security so that you're no longer using the same password for multiple sites learn more at the link below so let's go through a quick recap of the GPU benchmark first of all in the GPU testing we found a few important things one of them is that depth of field destroys performance particularly frame times with AMD Hardware when it's at its highest settings as in using the highest preset so specifically to address that what we would suggest is that if you're looking at playing this game consider using if you can support the highest preset and then go manually drop down depth of field to something lower ambient occlusion also had somewhat of an impact all of that and the GP benchmark won't rehash it all here but if you want to learn more about the GPU side of things that check that content it's also got all of the research that went into the initial part of this one now because this is a beta that has a limited time before it expires we were basically researching stuff encountering issues trying to figure them out while we were going which means there's no room to really adjust the core benchmarks on-the-fly so what we have for you is a more detailed look at how the CPUs perform under different parameters within the game and we do have comparative benchmarks between CPUs but they're not the core focus of this content because thanks to a few decisions that the destiny's team has made our confidence in some of the numbers are a bit lower than we'd like one of which is a limitation on SMT and these SMT is effectively unwell not even effectively it's just plain unutilized in destiny - right now from what we've seen and we can ultimately only speak to what we've seen so there's room for things to be different it is a beta keep that in mind stuff can change drivers are game ready for the Nvidia card that we use for the CPU benchmarks they should be fine but can still change so there's room for differences in other benchmarks and what other people do and test and things like that but we did not see any functional gain from SMT being enabled which means that a lot of the rise in CP is just lose that advantage that they have on the thread side so some some decisions here that we didn't like fraps being broken is another one so not just fraps but on screen displays fraps I believe afterburner and rivatuner and a couple of other solutions broke today the day of filming this and they worked the previous day during the private data so basically Bungie won and they're disabled them from what we understand and it sounds like that's for some kind of anti cheat or anti hack purpose hopefully by the time the game ships they have a better solution than just breaking everything that people use for every other game in the PC world especially since Bungie is trying to come out of the swing and saying hey PC players look at what we're doing for you relatively uncapped frame rate etc also don't use any software or else so I'd like to see that change but I will see if it does with the October launch either way all of our tools worked the first day of the beta and then we had to switch to other tools to the second day that still worked but those might get phased out as well once Bungie discovers them so let's get into this starting with the CPU side of things it's all CPU side is supposed today what we're looking at is the thread limitations and then going into multiplayer and campaign differences so straight away we noticed that thread utilization was a mixed bag destiny punishes rise in CPUs by ignoring SMT we didn't have time to test it but if you disabled SMT Mike and uplift like we saw in other games at launch of Rison and so this was ignored to the point I've seen a zero percent utilization more or less on the SMT threat horizon 7 said 900 with all physical cores engaged did not report any of its additional threads engaged so you're looking at eight cores utilization and eight untouched this remained a theme for the 1600 x as well where the logical threads did not engage with the game the good news is that we've seen this before total war Warhammer in battlefield 1 both pushed updates post Rison that improves performance on Intel and AMD alike this was primarily by better utilizing the SMT threads and again it benefited both companies so we're unclear on if this will happen for destiny 2 at some point but for now we're off to a rough start and that's more on bungees end than anything so hopefully Andy can work with bundi to try and fix this going forward but for now the game is reasonably optimized on GPUs to the point where it actually performs decently once you eliminate depth of field issues and it's lacking heavily on the CPU front so kind of disappointed see that let's look at performance scaling and competitive multiplayer versus the intro campaign effectively single-player as a reminder this was tested for our GPU benchmark as well and we found that scaling was highly consistent on our 7700 K 1080i 1050 Ti and rx 580 will cycle through some of those tests on the screen now but the full recap is in the bench video for the GPUs ultimately standard deviation was far greater in multiplayer and the average performance tended to be within a couple percentage points of campaign generally about a 6 percent difference performance between multiplayer in the campaign but with far greater consistency and tighter standard deviation in the campaign tests which ultimately made it more desirable as you can see with the Saudi 700k numbers what was the campaign tests were within bounds of the more variable competitive testing on Midtown this makes campaign a good stand-in that's representative of multiplayer during our GPU benchmarks but again more on that in the GP video so let's look at the table of our three 1200 runs on the Midtown multiplayer map using 1080p highest settings across multiple competitive matches we're averaging 85 FPS with lows at 59 4 1% and 44 0.1% the range is 79 fps 290 FPS for this test sequence with standard deviation at about 5 FPS for the averages bringing the campaign results on the screen now the share space the r3 1200 still at 1080p highest is standard deviation tighten to 0.8 FPS average the final bench score is 94 FPS average 60fps 1% lows and 50 to 0.1% lows this makes the campaign bench seen score approximately 9 FPS higher in averages and markedly higher in 0.1% low frame time consistency we've got a bigger spread here than with GPU testing unfortunately and it seems to deal with how the lower end CPU is are engaged during multiplayer the gap is about 9% between campaign and multiplayer with an r3 and that gives us a starting point for understanding performance throughout the game moving to 1080p hi our competitive matches at now range from 80 to 295 FPS averaging out to 87 FPS average 55 FPS 1% lows and 40 fps 0.1% lows compared to the previous multiplayer logs with 1080p and highest settings we've now got a delta of plus or minus 2 FPS average which is effectively equal this is a trend we've run into with the low end CPUs and comparative testing as parts bump into a wall that isn't remedied by going from highest to high bringing the campaign table up we see the final bench scene score of 93 FPS average is 1% low is 60 10.1% low is at 52 compared to multiplayers final bench averages that we're about 6 FPS higher an average framerate and 13 fps higher and zo behind present low frame time converted fps values although we saw a much smaller difference on the 7700 K so the lower and our three CPU is more susceptible to frame time variance and multiplayer than in the campaign at least as far as our testing goes today here's a frame time plot showing the r3 1200 performance at 1080p high when an a multiplayer match matched against the same CPU and settings in campaign mode frame time variance is reflected visibly if we make another frame time plot for the 7700 K in campaign vs. multiplayer using the settings detailed on the chart header it's clear that the performance impact is last significant there's a lot more to know but we won't have enough time prior to the betas end to figure it all out most of that will wait until October anyway for now we're seeing bigger swings and performance on the r3 versus the i7 CPU will have to reevaluate testing options for CPUs and destiny too once the game properly launches but right now the variance is greater than what we see with GPU testing which makes test core and planning more difficult to figure out the best we can do for now is demonstrate those differences help viewers understand where the game's performance may change depending on which part of the game you're playing and then commit to some type of tasking that's consistent so we had already committed to testing for the campaign for this one because we started that with the GPUs found it was extremely reliable and very comparable in terms of performance with the multiplayer gameplay and so we stuck with that just keep in mind going through this that the lower end hardware like the r3 is the G 45 60 things like that will experience bigger frame time variants in multiplayer than what we're going to be showing with the campaign results once we get to the competitive or comparative data so there will be a dip there if you go into multiplayer it's still relatively comparable we're within nine ish percent average frame rates so we're in the ballpark for now this is the best we can do once the game ships will reevaluate invest more time into figuring out how we can do things better to fit with Bungie and destiny to is a very strange at times requirements for the game in terms of how it behaves with different hardware so we'll look at that then but there's clearly only a few days right now so this is what we've got it's pretty good it's not the best but the point is you have all this data to look at the multiplayer the campaign performance and the performance of different graphics settings that we're going to get into so normally this isn't stuff that's really shared as part of a benchmark benchmarks tend to be CPU a for CPU B which one's better what we're trying to do is say here are all of the parameters and situations that you need to look into because the game obviously has different a bunch of different game modes and then consider all of those things once we go into the core benchmarks that are going to be representative primarily of campaign play our next topic is the impact of graphics settings specifically on the r3 1200 hopefully enabling us to better understand performance differences and where there's Headroom for improvement our first attempt at improving framerate went something like what's on the screen now there's almost no change and the campaign areas we tested these minimal impact most the impact seems to go to either the GPU or just isn't visible at all in current accessible areas of the game so there's room for this to change later once more zones are released either way going from highest settings to lowest settings produce no measurable difference in performance and a huge difference in visual quality we next attempted this in multiplayer despite greater variance tested test and found that the highest settings and multiplayer match has produced the numbers on the left and that the minimum settings produced the output on the right of the screen we're showing each test pass rather than just a chart with them averaged because there's more variance in each execution so we'd like for viewers to understand what the variance looks like when you test in multiplayer rather than just averaging them all together and muddy in the waters our difference ended up being 85 FPS average and 40 fps 0.1% lowest for highest with minimum settings being 95 FPS average and about 40 fps 0.1% lows again that's minimum that's disabling stuff down below even the lowest preset just to try and figure out what kind of changes we'd see on the CPU side the game looks awful at this point comparatively and we can show some that in b-roll but all these changes not only about a nine to ten percent performance uplift now of course this is just one CPU if you count the 7700 K from our GPU test that's kind of two CPUs so we can't definitively speak for all the CPUs but as far as these go there's more room for GPU scaling than CPU scallion in our testing for destiny to beta so all that in mind let's get into the comparative benchmarks again a big reminder that this stuff is subject to change the game is beta the performance is sort of inadequate in some ways in terms of thread utilization and tapping into the CPUs so there's a lot that can change here don't read too far into it for now but this will give you a baseline for where the beta sits starting comparatively with 1080p and highest settings the intel i7 7700 k-chart tops at 171 FPS average which is bumping against the 200 FPS frame cap that's enabled by default this means that frame time latency is faster than 5 milliseconds will be capped to 5 milliseconds and so will drag down the average below what could be possible the next CPU in line is the i-5 7600 case the Ox CPU performing 157 FPS average 104 1% lows and 8701 % lows this is trailed by the i3 1750 K at 131 FPS average with lows at 88 1% and 79 0.1% our overclocked r7 1700 effectively a stand-in for the 1700 X and 1800 X as well and so they can all achieve similar clocks performs at 126 FPS average at 3.9 gigahertz this plants the 3.9 gigahertz r7 CPU if you have PS behind the i3 and part of this performance deficit is due to destiny T's failure to acknowledge all the threads so we've only got physical cores engaging properly that's obviously a huge part of horizons argument is having all of those threads you know that this isn't the performance people want to see out of an r7 remember that destiny 2 is one in beta and to having trouble with SMT so issues arise we may see a performance gain that later on just like we did with Total War but it's hard to say right now so for now this is what we're seeing in our testing but because of destiny T's present state our confidence isn't strong enough right now to claim with authority that these results are representative of all configurations and scenarios as the game is still developing and clearly has issues down the line the r5 1600 X performs about 2.7 percent behind the overclocked r7 1700 keep in mind that as far as Bungie is concerned the 1600 X is a 6 thread CPU and the 1700 is an 8 threads GPU so the actual advantages of the 1700 are not reflected here beyond an additional 2 threads as far as the game is concerned we're looking at a 200 megahertz clock difference that manifests in a 2.7 percent boost over the other CPU both the 1600 X and stock 1700 boost to 3.7 making them effectively equal and planting the two with invariance of one another Intel's G 45 60 meanwhile performers at 110 FPS average with low at 73 and 65 frame times so far are reasonable across the board for all CPUs the next closest Intel apart is the 70 through 50k presently $70 more than the 45 60 in the US and it's about 19% faster than the 45 60 in the present state of the game with our benchmarks the r5 1400 and r3 1200 round out the low-end of the pack near 94 FPS average here's where it gets interesting switching to 1080p with high settings you likely barely noticed that the charts changed a lot of the numbers remain the same or similar with minimal scaling when reducing the high settings many of the options that change between high and highest are GPU limiting like depth of field and ambient occlusion and so won't appear in CPU constrained tests we've got fxaa enabled so that's out as a variable but if you had MSAA you'd also see a huge GPU side hit the performance stack is largely the same here we've gained a few FPS for the r7 and r5 but otherwise the results are effectively within test and test variants we'd have to manually adjust CPU limiting settings to get a better idea here like items governing geometry LED and view distance but as we already saw it doesn't seem to do a lot we also added 1440p to test and to get an idea for how or if the CPU Delta's close as more load is placed elsewhere in the system the answer is yes with highest settings the i7 7700 K falls to 109 FPS average from its 170 perch previously with 1% lows at 80 a 1001 % at 79 the 7600 ki 5 is now just behind the 7700 K indicating a GPU bottleneck at this resolution this is why higher resolutions and graphics settings serve as an equalizer for CPU performance in many instances even the i3 is nearby and averages that technically lower in frame times behind the i3 is the 1600 X now getting closer to the chart-topping CPUs and performance and this is thanks to resource limitations else we're almost assuredly the GPU we're at a deficit of 6.9% to the 77 hard K now rather than the bigger gaps earlier the 1,700 at 3.9 gigahertz performs effectively identically with invariance and the 1700 is nearby it's not until we hit the r5 1400 and r3 1200 that frame rate drops reasonably below the 100 FPS range finally switching over to the 1440p high settings we see general uplifts for parts across the board with FPS approaching original values there are no big changes here when compared to the 1080p high values if you look between them the charts all the numbers are plus or minus a couple FPS of each other so some might be lower some might be higher about the end of the day there with invariance that's it for this one the biggest thing here to take away is the front the video or the article where we talk about all of the exploration of the games options the different parts of the game the performance anomalies things like that the comparative data is interesting and hopefully put some information out there but keep in mind that one doesn t twos and beta-2 we'll see how it changes on the other maps game zones and modes that are not are not released yet or not fully released yet so there's a lot of room for changes here the takeaway being and what I'm really trying to instill into people's minds is rather than taking charts copying and pasting them places when you see people do this because they will rather than copy and pasting charts places and saying X CPU is better than y cp you just remind those people that there's some variance here we're not fully confident in the comparative results right now because of Bungie and destiny to is kind of weird interactions with well there's a lot of things a lot of stuff we didn't go over but switching between the highest low settings things like that there's so little impact on the CPU that you really start to question what's going on so there might be more room in there for the CPUs to improve we'll just have to find out and part of that will be waiting for the game to launch officially part of it will be having enough time to dig through every part of the game graphic settings things like that that's not something we can do alone so we'll look for other reviewers and users who are able to explore the game and report their experiences so please contribute it to that discussion if you get the game once it comes out let us know what your experiences in different areas of the game so we can try and look into it thank you for watching you can helps out directly by going to patreon.com slash gamers Nexus and then of course shirts like this one are at gamers Nexus squarespace.com this one technically is an older model though you've got better ones on that store than I have here so thank you for watching subscribe for more I'll see you all next time you