Gadgetory

welcome back to harbor unbox today we are checking out how the intel core oh nine ninety nine hundred k performs when paired with ddr4 four thousand memory this is a test that quite a few of you have been requesting for some time now I've been a bit slack but I've got my hands on some four thousand memory and it's around here somewhere I'll find at the moment doesn't really matter but anyway I've put it to the test I've actually done a bit more than just test with just the four thousand speed it's more of a suppose you call it a memory scaling benchmark this one I've tested a number of games and applications at memory speeds ranging from twenty four hundred which is just below spec of course twenty 666 which is the spec right up to four thousand there's about eight memory speeds in total so yeah lots of interesting results to go over but before we do today's video has been sponsored by asrock and then you phantom gaming range of zero ninety motherboards the Zed 390 phantom gaming six and nine include a blazing fast 2.5 gigabits per second network interface offering gamers and content creators two and a half times the bandwidth compared to standard Gigabit Ethernet for more information please check the link in the video description ok so over the past month I've noticed quite a few viewers asking us to run some tests on the nine ROK using ddr4 4000 memory I've also noticed quite a few people interested in memory scaling as they can work out where the sweet spot is in terms of memory frequency so I thought why not just do both at once kill two birds with the ones to commemorate as a say G skill recently sent over a 16 gigabyte kit of their Trident Z royal ddr4 memory I haven't actually taken the plastic off these ones anyway this is or these modules rather are rated at the 4000 speed using CL 17 timings that's probably shining all over the place I should be careful where I aim these anyway so for this test I will be taking as Rox Zed 390 Taichi ultimate gigabytes RT X 20 80 TI horas extreme the night are ok of course and then G skills bullying fest memory that could potentially blind people depending on the angle that I aim this stuff on anyway we see how that impacts performance at a range of clock frequencies in total there are eight memory configurations all of which use the same CL 17 timing so we're really just looking at the difference that the memory speed or memory frequency has here the 99er okay isn't overclocked but that won't really impact the results and yeah I think that's about everything so let's get into those blue bar graphs before we get into the games of which there are just a few let's go over some memory bandwidth results and then check out a couple of applications looking at the I 264 read bandwidth results we see that the increase in memory speed translates to a similar percentage gain for example ddr4 2666 is clocked 11% higher than ddr4 2400 and we saw a 12 percent boost in memory bandwidth the margin of error is likely responsible for that extra percent then we see from 20 666 to 3,000 it's a twelve and a half percent frequency bump and the bandwidth is increased by eleven and a half percent this kind of scaling is pretty much seen right up to the 4000 spec it's not quite as efficient but some decent scaling is seen at the high end typically I test Intel processors with ddr4 3200 memory though it's l14 stuff and that allows for a bandwidth of roughly 50 gigabytes per second in this test so somewhere between CL 17 ddr4 3400 and 3600 then still with ddr4 4000 there is a 12% memory bandwidth increase to be had so it will be interesting to see what kind of impact this has on performance before we jump into the application benchmarks here's just another quick look at some memory bandwidth results though this time I will be using sigh soft sound of 2016 memory bandwidth test a 264 reports the peak bandwidth whereas I soft reports sustained bandwidth overall though despite the numbers being lower the margins are much the same as we found in the past memory bandwidth doesn't impact 3d rendering applications all that much and once again we see that here with Cinebench r15 going all the way from ddr4 2400 to ddr4 4000 boosted the score by just 2% a measly 2% increase there it's a similar story when testing with blender here the ddr4 4000 memory reduced the rendering time by just one point four percent shaving just 10 seconds off the completion time when compared to ddr4 2400 so despite an almost 60 increase in memory bandwidth less than a trooper settlements boost was seen the merchants are ever so slightly better in Corona but even so for those running rendering applications the increased bandwidth really isn't required this is one of the reasons why the thread group of twenty nine ninety WX does so well in these rendering applications typically this is a CPU that's bound with staff with all caused active but in the rendering applications that bandwidth intensive it does do very well that said we also see when encoding that the extra memory bandwidth isn't that useful though it is substantially more useful when compared to what we just saw in the rendering applications ddr4 4000 offers a seven percent boost over the twenty four hundred stuff far from impressive but like I said it's better than a two percent boost that we saw in the rendering applications for those of you who regularly compress archives or files the increased memory bandwidth will be greatly appreciated we basically see perfect scaling here for almost sixty percent increase in bandwidth we see an almost 60 percent increase in performance when going from ddr4 24 hundred to four thousand still as good as the compression performance was when it comes to decompression work the increased memory bandwidth is next to useless we're back to single digit gains here okay so enough applications time for a few quick game benchmarks and first up we have Strange Brigade a well optimized title that mostly taxes the GPU it's not at all a CPU demanding title so I wasn't surprised to find that the increased memory bandwidth had little impact on performance the one percent low frame time performance is slightly improved as the RAM frequency is ramped up but it's certainly nothing to write home about and with just a 3% increase in performance from ddr4 3,000 to 4,000 you just get the cheaper stuff that said trying out any more CPU demanding title such as the NPC heavy hitman 2 provides us with slightly more favorable results for the higher clocked memory that said even here ddr4 4000 was just 14 percent faster than ddr4 3000 we're looking at frame time performance I'd say ddr4 3400 looks to be the sweet spot here and I'd say the CL 14 ddr4 3200 memory and typically use provides marginally better results one of the most demanding tiles of 2018 we shut off the Tomb Raider and here we find pretty solid gains right up to ddr4 thirty eight hundred at least when looking at the frame time performance overall I'd say ddr4 3400 is the sweet spot while the 3600 result would be comparable to the low latency 3,200 memory the real performance it comes when using ddr4 2400 which is below the official spec while 2666 is the official spec in fact anything slower than ddr4 3200 using sales 17 timings really does leave a lot of performance on the table at least 29 Tech's 20 atti at this resolution using these quality settings and I'll address that in a moment okay so here's the last gaming benchmark when I look at battlefield 5 and what we have here is another CPU demanding title here at 1080p we continue to see performance gains as the memory frequency is increased though ddr4 3400 does once again appear to be the sweet spot while frame time performance I really did drop off with ddr4 3000 but so far I've only tested at 1080p with an RT x20 atti so what does memory scaling performance look like at what I suppose you would call a more realistic resolution for such a beastly GPU well at 1080p we saw a 34 percent improvement in frame time performance when going from ddr4 2400 to 4000 and now at 1440p that margin is reduced to just 7% so yeah quite a change there we really are reaching the point of diminishing returns with I'd say ddr4 3200 then if you plan to play at the 4k resolution with the RT X xx atti well you don't really need to worry about memory speed it basically has zero impact on performance here you're entirely GPU limited even in a taxing title such as battlefield 5 ok so there you have it memory scaling performance with the Intel Core i $9.99 hundred K in short for most applications in games memory speed as it'll impact I granted you don't want to be running below spec with ddr4 2400 memory we just included that out of interest I'd say the sweet spot somewhere between 3,000 to 3,400 3,500 would certainly be sufficient even with looser timings for 16 gigabytes of ddr4 4000 memory you're looking at having to spend at least $220 us meanwhile ddr4 3000 memory starts at just under half that value and for one hundred and twenty dollars u.s. you can land yourself a CL 16 ddr4 3200 kit if you're a game of the battlefield five 1440p results are probably the most telling this is the kind of scaling you can typically expect to see for example if you have a gtx 10:17 you're playing at 1080p you'll see very similar memory scaling results then outside of gaming it becomes very application dependent and even then and what you're doing if the application can also determine how useful the extra memory bandwidth really is it seems like for 3d rendering and encoding work the extra bandwidth goes largely unused but again you will want to research how memory bandwidth impact your particular workload overall as I said just a moment ago I recommend ddr4 3000 to 3400 as that stuff really is going to provide you with the most bang for your buck while anything faster as we saw is likely going to end up being a waste of money under realistic gaming conditions maybe some scenarios where it's beneficial but I think for the most part and most of you won't be able to take advantage of it typically for around a hundred 240 dollars you're looking at cell 16 a CL 17 memory again we tested with cell 17 stuff in this video and as was the case with frequency spending big to get low latency memory you know CL 15 CL 14 possibly even lower it's not really going to be worth it I will just quickly before I wrap this up a bit of a fun look back at the start of 2018 I made a video series called why building a gaming PC right now is a bad idea it's quite a depressing time there for PC gamers and part one talked about ddr4 memory prices which along with pretty much everything else was sky-high at the time so back in January the cheapest 16 gigabyte ddr4 kit that I was able to find online cost a hundred and seventy dollars u.s. so today we're looking at basically half that fear so yeah it's fair to say prices have dropped they may not have dropped as low as we would like but they've definitely dropped and I did predict in January of 2018 that it would be about this time that we would see memory prices start to drop back down towards what we'd seen in 2017 so yeah seems I've got a bit lucky on that call this means you probably shouldn't expect ddr4 memory pricing to fall anytime soon certainly not too much later in the year the construction of new is underway to help the strained supply but they won't be ready for mass production until 2019 at the earliest that's predicted by Gartner that DDR for pricing will crash in 2019 history would suggest this is likely to happen as it's the sort of cycle we go through every few years with memory pricing and with that I am going to end this one I hope you enjoyed the video if we did please hit the like button subscribe for more content just like this and if you appreciate the weird red Horan box then consider supporting us on patreon thanks for watching I'm your host Steve see you next time