Gadgetory

this benchmark took a while to complete we first started bench working CPUs in battlefield 1 just after the GPU content was completed for that game and following that I had to follow up with AMD to look in some of their test methodology so we can validate it on our end with CPUs but the CPU benchmark is finally complete we tested the i7 down to the i3 and the x-48 4583 7080 through 20 or more before getting to the results this content is brought to you by an tech and the new cube mini ITX case which positions the GPU faceplate so that it's visible through the top of the case learn more in the link below we've posted two content pieces for battlefield one already the first was a benchmark of video card performance in the game using an i7 as we normally do because we're isolating the video card as a variable and the second was a look at HB AO vs. SSAO technologies in the game and their impact on performance which was for the most part negligible today we are looking at the cpu performance explicitly and cpu benchmarking is a little bit unique it's not quite as straightforward as GPU benchmarking because isolation of the CPU as a bottleneck is sometimes more difficult to do and extrapolating the scaling between one CPU in the next is also not necessarily easy so we've defined all of that as always in the article linked in the description below but I'll go through some of it here just for a quick recap tests were conducted using ultra settings at 1080p with two GPUs one was a GTX 1080 FTW hybrid used to place emphasis on the CPU performance interestingly the 1080 is powerful enough that just running a lower resolution is enough to demonstrate CPU scaling and we can leave the graphics options more realistically high this isn't new with this generation of GPUs the next test was an Rx 488 gigabyte gaming X which gives more of a middle-of-the-road look at fans and we'll get to that later the point of running a lower resolution is to show scaling performance and as you increase the resolution load will be placed more heavily on the pixel pipeline and the GPS ability to draw and sample all of those pixels which of course shifts the load DirectX 12 performance was measured using the on present variable from presentment we extract 1% and 0.1% low metrics using a Python scripts that game razaaq's created game graphics are configured to ultra 96 degree horizontal fov gb memory restriction off and be sync of course off and the test methodology in the article probably answer all of your other questions including drivers used why we tests a particular way we did and what the realistic hardware impact is we've also got our benchmark courses there our article also has some raw data from test runs and combat scenarios versus non combat which gives a better idea as the scaling and this is important as multi players effectively impossible to reasonably benchmark between this many devices so our scaling test in the article should answer those at multi player performance questions we're starting with DirectX 11 DirectX 12 will come next the ISO 159 30 K at stock frequencies hits a frame rate of about 158 FPS average with lows well above 100 FPS the 6700 K with hyper-threading enabled that's the HP one in the benchmark chart is next in line at 155 FPS average with one percent lows of 111 and 0.1% lows of 96 there's not a huge gap here and a good amount of that is because of the 6700 KS increased clock frequency over the 59 30 K which you'll see it in a moment is actually a pretty big impact on performance between CPUs interestingly disabling hyper-threading results in a reduction in FPS average of about 4% down to nearly 149 but our lows take the biggest hit and now 95 and 80 FPS this isn't significant in terms of visible framerate you'll still experience the game more or less the same way but hyper-threading is clearly beneficial to low values in this test scenario moving down the line the more mainstream I 566 hundred K at stock speeds is achieving a 140 FPS average landing it about 15 FPS behind the hyper-threaded 6700 K this is where we see CPUs of beginning to limit the GTX 1080 FTW hybrid we selected and that's intentionally used to demonstrate exactly this type of performance scaling these 6600 KS performance trails of 6700 K with hyper-threading by about 10% moving further down the list the I 560 409 K CPU at 2.7 gigahertz a good deal slower is starting to show a frequency advantage to the 6600 K with a performance disparity of nearly 20 FPS between the 66 and 6400 i-5 CPUs finally we're met with the i3 6300 at 111 FPS average now a full 44 fps behind the flagship skylake CPU and then the FX 83 70 at 90 FPS average and average FPS is fairly tightly timed on most these devices but that has a lot to do with one the API in battlefield one DirectX 11 tends to do well here and to the GPU itself which is responsible for a lot of the frame times but still the FX 83-70 does well to keep it's low frame rates within a reasonable range of its average despite a lower average overall the FX 83-70 is about 65 FPS behind the 6700 K trailed next by the power saving 8320 II that's what the e means and it's suffix it's a lower powered chip and that's at 77 FPS average the a-10 7870 K follows this but note that these 78 70 K is effectively identical to the athlon x4 880 K so that CPU interests you or is the one you have you could compare it pretty directly to the Sony at 70 K and the 880 K may even perform marginally better because it's got a 100 megahertz increase in clock rate next the athlon x4 845 runs at 63 FPS average just a few frames behind where the 80k would be although there's a clear choke point on these CPUs when running on a high end GPU one useful thing we can learn is that the x4 845 would perform close enough to the 80 K that the price difference could be enough to encourage 8:45 pickups over the 80 K in this case though there's some overclocking differences as well let's look at the scaling of all these devices in percentage form for a moment this chart shows devices with the i-5 6600 K is a baseline of 100% performance and the important part to note as stated or using a 1080 FTW explicitly to draw out the limits of the cpus and see where each chokes as we go further down the stack you'll want to make sure your cpu choice does not limit the GPU choice at least too much it's always possible but note that also the higher-end device is used with high resolutions as expected would put more load on the GPU so if we this 1080 at 1440 the story might be a little bit different many of the differences between the i5 and i7 devices vanish when moving to 1440p or 4k on the GTX 1080 and that's because we're now shoving more pixels into the pipe for the GPU let's look at DirectX 12 D X 12 destroys frame time performance at the 1% low and 0.1% load spectrum and post- scaling when moving between api's with Intel we see a negative hit of approximately 25 FPS on the I 767 hundred K and on the i5 6600 k FX 83-70 increases its average by 1 to 2 FPS but like all the other devices GPS included post worst low frame performance the FX 8320 ii also posts an improvement of about 2 to 3 FPS average on the FX architecture and the 7870 k nearly an 880 k equivalent with the x4 845 both see slight negative scaling of about 1 to 3 FPS but these chips are different architecture and they run at fewer cores so we're confident in these results but we're curious about them after looking through some of AMD's own public blog posts about their performance we decided to reach out to Andy and ask how they tested battlefield 1 and after a few days of back-and-forth emailing with one another we talked to several people from their test team and engineering team in the test lab and validating each other's tests and methodology and things like that GN we decided to run more tests this time with a lower end GPU and AMD said they were seen upwards of 28.5 percent gains with an FX 6350 and an RX 480 so they said that they used a three minute long test which is a bit longer than what we do and they played on a different map completely they played on through mud and blood which is one of the British levels and other than this their settings were pretty similar to ours in terms of quality we decided to adopt their methodology and use that for further validation and given the information that we had I think we have a pretty good example of the way AMD ran their tests but I still don't know everything about their setup I don't know exactly what Rama is what speed it was I don't know if they're using on present and there's a couple other things we don't as well but we've got a good amount of it so it would be odd but it is possible that running the game for a longer period like their 3-minute test run would be enough to begin hitting some sort of draw call or other threshold that you might encounter with a poorly optimized game and might not encounter with a shorter test given battlefield ones already questionable optimization this did seem worth testing a few more things to note here MD ran four test passes per setting per device and then they averaged all four of those results in gns methodology we run six to eight test passes and we always throw the first two sets of data away this is because the game outputs horrific ly varied performance when first launched and we see drastically reduced performance while still loading everything into memory some examples of a single device going through multiple test passes around the screen now you can actually see how it scales and sort of levels and equalizes as you run through a couple of the test passes and we believe our test is representative of what happens in a multiplayer map and I've tested that too after you've spawned and run around for minute or two performance improves greatly with the x11 across the board and in dx12 0.1 percent low frame time performance that first run though will look like garbage compared to the rest and so we do what you do with garbage we get rid of it this is even visible when the level isn't loading you can see prefab elements at various LOD is attempting to pop in especially environment elements like foliage and grass now another final important point to note Andy did not provide us with frame time metrics and only offered us average frame rates and we've already shown as a few other outlets have that battlefield ones DirectX 12 performance can actually be worse in terms of frame times than DX 11 and as a result it can stutter through gameplay so the our X 480 these tests were conducted on the through mud and blood campaign level and were tested using a much different methodology we ran three tests for three minutes each with highly consistent frame rate output because we'd practice the test run and could accurately repeat it each time so it's pretty accurate despite being three minutes long these tests were run after receiving input from AMD on their own internal testing of the game with an Rx 480 and an FX 83-70 we're still seeing no meaningful scaling technically in this testing the DX 12 results about four percent better than the dx11 results but that's still four percent and the negative hit the frame times is definitely not worth it even with these altered test methods we're not seeing the performance winds that am the reported and told us and we're definitely not seeing favorable frame time performance in dx12 though and ii did not make any comments about frame times now this doesn't mean that AMD is wrong it also doesn't mean that we're on and really it just comes down to a few things there's differences in testing that we may not know about different variables potentially measured for performance or different means of measuring those variables and and he also did use an FX 6350 which although it should not scale a difference of four percent to 30 percent when looking at something like in 83 seventy it could certainly scale a little bit differently so we don't have one of those and obviously didn't test it so GN spent about a full week validating these results it's more time than we spend on any of these benchmarks really in terms of validating existing results that were confident in so now we're definitely fully confident in the testing having retested it a thousand different ways on many different devices and we're confident that the performance difference between DX 11 and the X 12 remains largely the same for multiplayer use especially we're seeing no scaling performance that is noteworthy on DX 12 regardless of whose GPU or CPU we use and this fits our previous testing which validated extensively the GPU performance with both DirectX 12 and DirectX 11 and saw similar scaling that is to say effectively zero and negative scaling when looking at frame x battlefield ones DirectX 12 performance remains spotty and suboptimal regardless of whose CPU is used or even GPU for some extent anyway but definitely on the CPUs and overall we're seeing poor 1% and 0.1% low frame rate metrics and that's across the board even when conducting extended benchmarking 3 minutes 3 passes of that which is a lot longer versus the regular 30-second buy eight pass test runs and we've defined all of that and why each one works and works well in the methodology section below but the point is that with both sets of methodologies ours and Andy's we're still seeing poor framerate scaling at the and the frame times kind of stretch out and have a larger latency between them on occasion and that is reflected in those pretty spotty 1% 0.1% lo metrics again regardless of device the stack remains mostly the same in terms of hierarchy the high end i7 and i5 k skew devices do perform significantly better with drags x11 the i7 6700 K posts a definitive performance gain over the i-5 6600 K in dx11 at about 10 to 11 percent averages and the 6600 K runs about 15% faster than the i-5 6400 9ks CPU largely a result of the faster frequency on the 6600 k FX series CPUs and the i3 6300 CPU are struggling in terms of performance and scaling overall and looking at the other devices and the disparity between the FX CPUs and the i3 6300 could be related to the increased memory speed support of the i3 which is what we're researching next so stay tuned and subscribe for that then the x4 8:45 and x4 8 ATK surrogate post performance that is pretty limited and dragging the GTX 1080 down with them to around 60 to 65 FPS average but would still be perfectly playable on bf one if reducing settings and running an equally low end GPU like an rx 460 GTX and 50 tons of DTI or similar of course depending on your GPU CPU selection could be less relevant if you're running something like GTX 1060 or an R X 480 the 10% difference between an i5 6600 KN is 760 700 K could become largely irrelevant in the face of that lower and GPU that's what you'd expect the GPU starts becoming more of a choke point than the CPU depending on your graphics quality settings and next we're looking at memory so that's my next point of interest for Battlefield 1 I do suspect that memory has a bit more of an impact in VF 1 than what we might normally expect in other titles so we'll be testing that not 100% sure yet that's why it has to be tested but do subscribe to that content as always a comment if you have thoughts on this testing and the patreon link is in the post roll video if you want to helps out directly thank you for watching I'll see you all next time you