Gadgetory

there are a lot of feel-good stories about the Windows 10 fall creators update providing huge performance gains on various system configurations we're here to ruin that for you we isolated variables to target just one component the CPU and focused solely on the r7 1,700 to see if the reported uplift on rising platforms was due to GPU or CPU bound testing as always you can't just take this data and scream that the FCU caused huge gains or no gains or the opposite you've got to take the whole picture into account and with many GPU bound tests out there we figured we'd run a CPU bound test to try and pinpoint performance uplift and look at the other side of the spectrum before getting to that this video is brought to you by synergy the software that lets you share a keyboard and mouse between multiple systems if you have limited desk space and multiple computers to command synergy removes the need for a separate peripherals or a KVM and works as over the network software use our link below to get 50% off the home or provision with SSL we're aware that some users and other outlets are reporting performance uplift with FCU or fall creators update but we have to highlight that a lot of these benchmarks were conducted with Vega 56 and that has two factors playing towards the performance uplift one of them is that it seems like Vega may have had some specific optimizations and the other one is that just GPUs in general seem to have improved with FCU we're focusing on the CPU side to see if there's any uplift there but on the GPU side we've received multiple reports from our trusted user base in the patreon discord that for example g-sync and free sync stability in terms of frame times have both improved and both of those are GPU specific items we've also heard reports that Vega specifically has seen improvement haven't looked into its competing 1070 yet but it definitely it's become a bit clearer that there is GPU uplift with up with FCU to some extent and some of that is on frame time stability and some of it's just overall in general uplift but what isn't fully clear is how much of that uplift comes down to the CPU so we're all a ting for the CPU here and we're using a 1080i FTW three to do that as usual our testing can only represent our testing of course so rather than jump to the charts and then scream about the results being end-all be-all keep in mind that we're only testing the 1700 and the 1080 TI FTW 3 we can't yet speak for results of Vega cards or Intel CPUs Vega uplift again could be either from specific optimizations or just GPU performance uplift in general in the drivers testing is conducted in an a/b fashion so we have some older data it's not that old it's from the same proceeding creators update build that we're using the test here versus FCU when I say older I mean like tested a week ago as opposed to yesterday we have some of that we have some a B tests where we ran CU vs f CU in the last 24 hours from filming this and that provides a pretty good look at the performance uplift if there is any we're focusing only on 1080p today eliminated the usual 1440p testing because it just wasn't necessary as a side note we have been working with shadow of war for the past few days trying to get GP benchmarks up for it and we looked at it for CPUs shadow of war has some of the most variable performance we've seen to date haven't fully pinpointed why that is just yet but basically the you could run a set of four tests and see an average of maybe for example 20 FPS at an unspecified setting and with an unspecified GPU but you might see 30 FPS and one test and then 20 in the next set of 4 tests and then you're back up to 33 and the next set of 4 tests so it's kind of all over the place in performance with the built-in benchmark and we don't really trust it yet so we've eliminated it from this CPU comparison before today and we're focusing on the normal CPU suite instead starting with ashes of the singularity at 1080p we measured the CU 1703 dot 608 build the previous one at 42 FPS average with 1% lows @ 31.6 and 0.1% lows at 30 fps the FCU if all creators update tests planted the r7 1700 stock cpu at 40 point 3 FPS average with lows of 32 point three one percent twenty nine point eight zero point one percent this is within standard deviation which we measured at 0.5 FPS for average and also 1% or 1.5 FPS for zero 1% lows we were not able to discern an appreciable performance change between Cu and FC u and ashes of the singularity with data plotting the difference as within error and test variance here's a frame time plot between the two just to show the differences in ashes of the singularity at 1080p high as you can see they are fairly consistent what you're really looking for is more consistency and ideally a number lower on the scale rather than higher for frame times total war Warhammer at 1080p is next for this test our original r7 1,700 CPU plotted an average FPS of 128 with all those at sixty at five point seven and sixty the FCU test had us at 127 point 6 FPS average with a sixty five point three one percent and fifty nine point five zero point one percent low enabling game mode in FCU plotted one twenty eight point four FPS average sixty seven and sixty one point six for the lows and these results are all within standard deviation of each other which we've plotted as one point nine FPS average in this particular game one point three for the one percent values and two point three FPS for the 0.1% values where we start losing some test resolution again we can't determine statistically significant changes in performance from these numbers total war Warhammer was around more than ten times for each of these that was the title we spent the most effort on as it was the first one we tried after the update one outlier result was tossed from each test batch but overall the results basically point that there's no real change here watchdogs to plotted to see you are 717 hundred tests at eighty seven point seven FPS average when retested yesterday so this is a different batch of tests than our CPU review batch with a one percent low at sixty seven point three and zero point one percent lows at fifty four the FCU test had us at eighty eight FPS average and game mode also had us at eighty eight FPS average standard deviation and this one is about zero point nine FPS average with lows deviating by about one point two and one point five FPS respectively project cars to at 1080p is next and thoughts our seventeen hundred stock CPU from the creators update previous version of Windows at seventy eight point five FPS average with lows at sixty two point five and fifty eight point-five FCU i had us at 79.5 with loaves at sixty two point eight and fifty six point eight standard deviation is about zero point six FPS average 0.71% lows and two point six FPS 0.1% lows civilization six turn time completion is measured in seconds not FPS average FPS actually increases in civilization with worse CPUs for the AI benchmark because more time is spent with the screen unmoving and Static whereas faster CPUs process turns faster and so the screen experiences more dynamism and therefore a lower frame rate the turn time completion for the r7 1700 with creators update was twenty point nine five seconds with FCU at twenty-one point zero two seconds that's within error and variance and we're within 0.7 seconds of each other GTA 5 is last at 1080p we're at 95 FPS average for the original test with creators update sixty six point three four one percent and sixty three four 0.1% FCU plots us at ninety six point three FPS sixty six point three FPS one percent lows and 61.7 0.1 percent lows again this is with an error the frame time plot doesn't show much difference between the two runs once again and demonstrates that frame time consistency is largely unchanged in our our seven seventeen hundred cpu-bound testing now again before people start just jumping to the charts and posting well I guess they probably already done that but hopefully when you encounter people who just skip to the charts and ignore the entire thing at the beginning talking about to what extent our test is applicable what this tells you is not that you can't or won't see performance uploads with fall creators update certainly people have it's that with our test configuration and our seven seventeen hundred stock on an ACS crosshair hero six motherboard with thirty two hundred megahertz CL sixteen memory and most importantly with a 1080 TI FTW three we're not really seeing a change in these games there might be other games where we could see a change but more likely we would see a change with Vega or just with a lower and GPU like a 1070 relative to the 1080 TI it would be lower and so that would be the the next thing to look at would be basically GPU testing not just CPU testing but on the CPU side when left unconstrained by the GPU and with an NVIDIA GPU which is another important factor to consider there is really no performance change from one to the other it is possible once again just to reiterate that you would see a performance change if we stuck a Vega GPU in there for one of two reasons one may be because of an AMD driver thing that plays better with FCU and - just because Vega is probably going to be more GPU bound in some scenarios than others than a 1080 TI for example so yeah with this configuration we're not seeing a change maybe with others we'll look into it we do know however with relative certainty that there is improvement in FCU for frame time consistency for users with g-sync and freesync from what we've seen and heard it looks like those two technologies have tightened up their consistency with frame times with FCU so in the very least that's changed but if you're strictly looking at a cpu scenario for our test it doesn't look like there's a change if you find a game where there is one let us know but that's all for now as always patreon.com slash gamers Nexus 2 helps out directly you can go to store that gamers nexus net to pick up a shirt like this one the anniversary edition tear down shirt and subscribe for more I'll see you all next time