Gadgetory

we're coming to a close on our battlefield one story line it seems there might be one more optimization guideposts in the words but for now we've got to get back to regular component reviews today's topic though looks at memory frequency and its impact on bf1 performance particularly frame rates and we briefly look at capacity and utilization but will primarily be focusing on speed before getting to that this content is brought to you by Rose Wells : encase which is a tempered glass side panel case also tempered glass in the front available for $150 on new agate now they occasionally do bundle it with a 650 watt PSU as well so be sure to check the link below for more info on that we've been through battlefield one a few times now first was the GPU benchmark looks at the video card performance and then after that HBO versus ssao graphics settings performance in game and then most recently we did the CPU comparison for benchmarks and dx11 dx12 today we're looking at Ram methodology remains mostly the same but there are a couple of critical changes we define all of it in the article in from the description below these results are not necessarily directly comparable to the previous test because one we're changing memory here and that's obviously the focus of the tests we've overclocked the CPU running a 60 700k overclocked and the game has been updated which also had some at least marginal performance impact mostly positive and then we're also using different video drivers so those things stated this is a new test looking strictly at Ram but if you're interested in the others they will be linked in the description below we're using two primary platforms for this test the i7 6700 K with an MSI m7z 170 motherboard will be our main platform with the 6700 K overclock does 4.4 gigahertz memory used is at 32 gigabytes of course there's Dominator Platinum ddr4 easily capable of speeds up to 30 200 megahertz with its first XMP and we manually tune the speeds and BIOS before each of the tests then run six to eight passes per device DirectX 11 will be the focus for the test since we've already shown that dx12 has some optimization issues with battlefield one that will sort of override any lacking memory performance and we did do a few dx12 tests but they're pretty brief the GPU used was our EVGA 1080 FTW hybrids so we don't have much in the way of bottlenecks when operating at 1080p and with ultra settings the most likely components the choke will be the CPU and the RAM the point here is to show scaling of memory so keep in mind that as resolution increases we begin to see more taxing the performance from the GPU and that is more directly related to FPS than memory at that point so frequency becomes a bit less relevant as the resolution increases let's get to it this first chart shows battlefield one's memory performance and running 1080p with ultra settings on the test platform the main one and we're seeing largely unfettered performance with the four sticks of 3200 mega Hertz corsair dominator Platinum memory pushing a 161 FPS average about 128 one percent lows and 117 fps 0.1% blows this is followed next by the ddr4 2,400 memory so we've clocked down the Platinum Series kit to 2400 megahertz that's at 158 FPS average or about four fps behind the 3200 mega Hertz kit with it lows at 118 and 107 fps a step down with 1600 megahertz ddr4 memory which basically doesn't exist and does poor access times anyway we're at 150 FPS average now 11 fps lower than the 3200 mega Hertz kit and now again no one's really buying 1600 megahertz memory and ddr4 kits but it was a worthwhile test if we create another unlikely scenario 1333 mega Hertz memory that's operating at 140 FPS average that's a full 20 FPS lower than the 3200 mega Hertz kit this could be compared in some ways the single channel performance if you remember our video from ages ago we didn't see much impact from multi channel platforms with ddr3 and our Gaming tests a couple years back but game development has changed and so has memory with single channel platforms memory speeds are effectively halved from the advertised rate so these 1333 and 1600 results give a look into single channel performance with bf 1 and as a quick PSA keep in mind that there's no such thing as quote unquote dual channel memory or single channel memory the memory itself is not in charge of how many channels there are it's the platform here's a chart showing percent scaling as offset from ddr4 2400 verse a fairly standard 20 400 megahertz kit of ddr4 memory we're seen scaling of about 2.5 percent gains when moving to 3,200 bag Hertz and we're seeing losses of about 5 percent when stepping down to 1600 megahertz and the jump from 2400 to 1333 is about 12 percent again remember that as we begin restricting performance through other variables like increasing the resolution basically taxing the GPU more taxing the CPU more the memory impact will become less significant it's a small part of the performance overall but it is actually showing a change which is cool it's not really that common but when we increase resolution is something like 4k which is maybe more appropriate for a 1080 ftw these results do vary to the point that with 4k resolution ultra settings were basically seen zero scaling with memory at 32 hundred megahertz versus 1866 megahertz not a big difference there in performance I think we were seeing about 67 FPS average for both tests both the x11 only variable changes the memory and that's just because the GPU is getting taxed so heavily by the rest of the stuff in the pipeline like the increased pixel throughput that the memory becomes less relevant so keep that in mind but we're still seeing some scaling which again is a cool thing now dx12 also shows almost no scaling from memory changes even at 1080p and that's because the game performance is too erratic already from poorly optimized dx12 so we're not seeing I shouldn't say dx12 poorly optimized it's battlefield one is poorly optimized on dx12 it's all in the developers hands for the most part and it's just too spotty and erratic to see a meaningful gain from memory changes alone just for good measure and to cover another architecture let's throw AMD's FX 83-70 into the mix here we've tested memory at 21 33 megahertz and 1600 megahertz all ddr3 kits and we can see a difference of about 5.5 to 6 fps in the average framerate performance between the two that's about 6 percent change considering that ddr3 kits are priced pretty equally between 1600 1866 and 20 133 megahertz these days especially with manufacturers basically dropping 1600 from production entirely it makes sense to spend the extra $1 for something that's 6% faster use-case and at this point it's actually a meaningful change because we're limited down to around ninety eighty FPS anyway it also makes sense to run two sticks for battlefield one if operating on a dual-channel platform especially with these lower end CPUs as for capacity this is harder to measure for a number of reasons memory isn't instantly saturated so it's easier to test with something more heuristic we tried playing a 64 person multiplayer match for about 40 minutes while running logging utilities to track system memory consumption and checking res Mon for commit memory versus working set of memory memory usage by the application goes up quickly but the working set of active memory being used by the game the application never exceeded six to seven gigabytes even when I had 32 gigabytes in the system we played with one stick of eight gigabytes of RAM as well just for another 64 player match and that was done to see if any visible stutters or popping issues occurred and the answer was no not really we just still see the popping issues that are normally there but nothing exaggerated by the memory change and it's all the same as what you'd normally see with battlefield one nothing specific to memory so the capacity seems like you'd want definitely eight gigabytes but the only immediate gain from having a larger capacity of memory is if you want to run background applications like Chrome or something like that that eats memory for its tabs and not have to close them so that's it for this benchmark this one's pretty simple memory has a bigger impact in battlefield 1 than we've seen with other games and that's always a cool thing it's something that we can use for memory benchmarking in the future and even something we can use for SSD benchmarking because this is one of the things I haven't talked about yet in any of these videos SSDs in battlefield 1 actually make a pretty noticeable difference especially when loading multiplayer maps the change from using one of our nvme SSDs to using a low-end SSD alone is fairly substantial and putting it on a hard drive would would definitely be noticeable but in terms of memory we are seeing a difference with the lower end CPUs it's definitely probably more critical that you actually just spend the extra one or two or five dollars to get memory that jumps from 1600 to 2133 megahertz if you're on in ddr3 because that price is not bad couple bucks and there's actually a meaningful change now with this particular game again five to six percent between those two numbers 1621 33 so that is significant capacity just recapping that quickly it pushes the 8 gigabyte barrier of the game but it's never really even when just testing it heuristic aliy playing the game properly for 40 minutes or so we're not seeing any worst performance that's visible other than the framerate performance as I've already described for the frequencies we're not seeing any performance degradation just from the capacity change that we don't already see somewhere else like 32 gigabytes of RAM so memory is not a huge deal in terms of capacity you probably want a bit more if you do tend to keep a lot of applications of the background 16 gigabytes isn't a bad idea but it is plenty for this game if you're just playing the game and killing everything else in the background so that's all for this time is always patreon link in the post roll video if you want to helps out directly links in the description below for more information subscribe for more I'll see you all next time you