Gadgetory

let's get straight into it today we're reviewing the r7 2,700 x and showing numbers for the r7 2,700 both of which are refresh is on the existing AMD r7 1700 product line we acquired our CPUs from third-party sources rather than from AMD but we've held and iterated on our data over the past month which gives us some interesting information just from going through the whole process for that lawn today is the day to walk through a small portion of what we've found we're releasing several videos surrounding rise in 2000 split apart for organization and time this one is your review video recapping all the core performance metrics before that this video is brought to you by thermal grizzly makers of the conductor not liquid metal that we recently used to drop 20 degrees off of our temperatures thermal grizzly also makes traditional thermal compounds we use on top of the IHS like cryo not and hydro not pastes learn more at the link below so again we've had these Seabees for about a month because we got them outside of embargo normal restrictions and things like that but we respect an embargo anyway we have a ton of information as a result including X 470 vertex 370 r5 2600 X and 2600 4 ones 2700 performance we have more vrm thermal testing on the new X 470 motherboards we have PCB analysis from builds IDE we're splitting all these into multiple videos because otherwise they would be very long and also subscribe immediately because at least one other video will go up today and you're gonna want to see that and check back later today to catch that so getting straight into it then we are only overclocking one CPU in each family so the r5 2600 r5 1600 whatever we're only overclocking one of them only overclocking one cv for the r7 2700 or 700 family because or one 7 X 1/8 X whatever because it's the same thing so our sevens are all the same and the our fives are mostly all the same and they are the same for the new series so for that reason we're overclocking one because 4.2 gigahertz on one equals four points you get your heads on the other one that stated we're gonna start with talking about some overclocking numbers rather than all the gaming stuff because this is where it gets really interesting Verizon 2000 the main differences for the new CPUs actually come from the minimum voltage required to maintain a target frequency so at 4 gigahertz on Rison mm versus 4 gigahertz sunrise in 1,000 we are able to operate at a significantly lower voltage and that is where the primary advantages are derived not from raw performance because in terms of raw performance you're talking hundreds of megahertz at most this isn't a huge jump in terms of frequency or IPC or anything like that we're sort of at some level looking at Intel sized steps in raw gaming performance you're talking single digit percentages in some cases low double digits and others so we're not looking at big jumps it's kind of like sixty seven hundred seventy seven hundred K CPUs however we see big jumps and things that aren't performance numbers so what you need to pay attention to is not just the raw game in Florence because if that's all you pay attention to when you watch other reviews it's gonna be really boring but if you pay attention to things like the efficiency the voltage required to sustain a fixed clock that is where it gets interesting so let's start there we found that at a given frequency of four gigahertz our r7 2,700 x held stable at one point one seven five volts input at LLC level four on the ACS crosshair seven hero also acquired separately which equated to one point one six two volts v core @ sv TFN the result was stability in blender and prime95 with torturous FFTs while measuring at about 129 watts power consumption in blender measured at our current clamp for this same test our seventeen hundred at four gigahertz required a one point four to five volt input at LLC level five yielding a one point four to 5 V core a two hundred one watt power draws so 70 watts higher and pushed thermal to 79 degrees TDI prime95 produced similar results you can find those in the article below but the takeaway is that lower voltage for a given frequency also means lower power consumption for the same frequency again 70 Watts difference in that one test to some extent this is Bening but most of the large Delta seen here is from improvement of the products clock efficiency at the quote old high clocks of 4 gigahertz this is somewhat exponential so 4 4 gigahertz not a big deal basically 1.2 volts and below on the CPUs that we have and for 4.2 get your Hertz that's where it changes quite a lot to get 4.2 stable on our 2700 X our 2700 our 1600 X 2600 X rather all of them for pointy stable required 1.4 2 volts and above in some cases with high LLC's and max Allah sees in some cases and we were not able to get 4.3 stable on basically anything that we tried we worked on it but close we got was basically 4.25 sort of asterisk it crashes sometimes so that's kind of the takeaway here basically the overclocking Headroom not exciting the Volt frequency curve very interesting and job well done Andy for improving that in very significant ways so that's the takeaway I wanted to highlight now again one last thing here is that in order to step up the frequency a couple hundred megahertz it requires a ton more voltage that's what I'm saying next thing to talk about is memory overclocking also very interesting and in height of discussion with the rise in to rumors there were some rumors w CCF tech that earlier said something like 40 100 megahertz would be achieve of Hall memory overclocking on rise into 2000 series and to their credit several months ago we heard from vendors at some trade show that they were expecting 4000 megahertz to be doable not really this is what it turns out to me so memory overclocking hasn't changed much for Eisen plus and X 470 platforms initial reports again suggested pretty high we found that both of our platforms were capable of handling 30 600 megahertz XMP for our g.skill Trident z kit which is the one that we took to 4000 mega Hertz with our hashtag rip Ltd build so it can handle more than 3600 and it's not a big deal to boot at 3,600 cold but it's a bit better than early rise and launch memory training and things like that so it required trainings get from 3600 to 3666 we could not boot 3666 cold we the start at 3600 and then we could not take the memory higher than 30 666 this Samsung B died it's a really good Triton Z kit that we are positive does 4000 megahertz plus so we know it's a good kit this comes down to the board to some extent to the CPU i am c binning to some extent and just horizon so as from memory scalability we did some testing on that as well and this gets very granular in that you need to be very careful about controlling timings because otherwise frequency looks like there's a zero scaling whatsoever this test was conducted with completely controlled memory timings meaning we controlled every single time and presented in UEFI that includes sub timings and tertiary timings not just the first four or five when you control x 100% and change only the memory frequency plus whatever unpresentable fied the result for x pi is what you see on the screen now x pi is highly memory sensitive and so it's a bit of an extreme case we end up with 36 66 megahertz potting 9465 points and uplift of 0.6% over 3600 megahertz which is 0.4 percent over 3466 Macker you get the idea there is absolutely scaling here but it's incremental at 3600 versus 3200 baggert Hertz we observed a 3.5 percent difference given the price difference in some regions you may as well stick with 3200 megahertz the difference between 32 and 2400 megahertz is tremendous though marked at 12.5% that's with x 5 where memory is sensitive here's where it gets interesting with the crosshair 7 hero turning the frequency down a bit will result in an automatic tightening of sub timings which results in increased performance overall if you were to manually in for 16 18 18 36 timings and allow all other times to be auto which most users do you'd have the results we've had on the screen for a moment here and that's if you set the frequency manually we have everything else the same you end up with this chart the result is much tighter than we saw previously we suddenly have a range of a couple hundred points rather than 1400 points instead of 80 70 to 9,500 we're at 90 100 to 90 300 that's because the timings automatically tightened to a point of improving beyond the frequency improvement alone this is mostly asus is doing and job well done aces for doing that basically if you're serious about getting everything out of performance that you can you've got a lot of work to do on tuning the times some other boards like the Aces crosshair 7 hero do a lot of that for you and that's where the difference in motherboards comes from that's why not all motherboards are the same it's stuff like that so job well done again on that but this illustrates how testing can also be completely invalidated if you don't pay close attention for example if you're not changing those buried lower-level timings and only the top ones you'd see a couple hundred points difference looks like there's zero scaling difference at all that's what we started with before looking at those timings which are being automatically tuned so what this does teach us though is that if you go with a slower kit in frequency but you can get tighter timings as a result of being a slower frequency it kind of evens out you really might as well go with thirty-two hundred megahertz with a tighter set of timings than something like thirty six hundred with looser times and there are applications where maybe that's not the case you hopefully know what they are if that's true but for most people we'd recommend thirty-two with good timings you'll be fine you're not missing much performance if anything at all and one more thing times by is of course very sensitive to this so here's an example of fire strike which doesn't use the same API as times by we observed nearly zero scaling between memory frequencies even we controlled for timings Cinebench also demonstrated at nearly zero scaling difference at least not until twenty four hundred megahertz twelve into game streaming benchmarks next we're presently showing pub G gameplay at ten megabits per second and faster on each side of the screen one is a 2700 X the other isn't 8700 K neither is overclocked will reveal which these are in a moment but we want to hide them from you for now so that you can view them without bias and then maybe if you're on you'll see what bias you have later these benchmarks will be described in greater detail the test methodology section that's linked in the article in the description below as a quick recap we're testing live streaming over a gigabit fiber connection using OBS logging for network dropouts which there were none and x264 encoding our baseline test for a good quality stream is considered 10 megabits per second upload with x264 faster enabled that's what we use for our live streams are close to it if all CP is past that we stress them with the x264 medium and 12 megabit per second uploads our goal is to upload 100% of frames to stream and draw zero ideally all near 16.7 millisecond delivery times we also want to lose as little FPS from baseline non streaming gameplay as possible in other words we care about the scaling of how each CPU handle streaming for both the player the streamer in this case and the viewer our goal is efficiency in delivering a 60fps stream zero drops which would be considered a perfect stream while also maintaining a high player side fps so that the streamer can have a good experience too we're starting with pub G and streamer side FPS so this is what the streamer the player would see not what the viewer would see both are important parts of the equation for the streamer we're observing a baseline FPS of 132 FPS at average for the r7 2,700 X when stock which decays to 101 FPS average when streaming at 10 megabits per second faster h.264 the one percent lows are relatively proportionate to 57 FPS average at the bottom end at 12 megabits per second and stress with a rather unrealistic medium encode speed but used as a synthetic test we fall to a still fully playable 83 FPS average and 47 fps 1% lows the streamer is perfectly happy in this instance the 87 hardcase baseline no stream at all that's what baseline means its baseline is hitting the a frame rate cap and average is at 141 FPS average it'd be higher without that ceiling we're at comparable low values to the 2,700 X streaming at 10 megabits per second faster brings us to 124 FPS average or 107 fps for medium as for which was which we can pop those labels up now on the 8700 K and 2700 X from the earlier footage technically 8700 K is in the lead for a streamer side performance but that's only again half the equation for viewer side things change a lot and quickly first off both the 8700 K and 2700 X stock CPUs deliver 100% of frames to the stream at our 10 megabyte per second faster preset with 98% 99.9% of frames delivered within our sixteen point seven millisecond target this is excellent performance for each CPU and either is fully capable of gameplay while streaming there's functionally no downside to either for this preset while streaming for the viewer for the player there's a bit of a difference yes notice however that the 8700 cake crumbles when pressed with h.264 medium and followed megabit per second encoding we end up at thirty two point four percent of all frames delivered meaning that the stream drops 70% of frames or there abouts will show those videos side by side now we overclock the 8700 K to see if performance could meet the stock 2700 X and found that it's still under delivered at eighty four point two percent of all frames this disparity is partially a result of the two fewer cores and partially a result of scheduling where resources were spent on the game than OBS with Intel this can be rectified with manual affinity and priority tuning for OBS you can correct for that law somewhat but it's extra work for the streamer it is not out of box easy performance to configure here's dota 2 and the player side performance for dota 2 in terms of raw FPS the 8700 K easily chart topped as Intel often does in this game this game is heavily dependent on one to two threads and that's why we see the frequency advantage they are capped at one to two threads pegged nearly a hundred percent we're at 181 FPS average which is 149 FPS average with low is reasonably comparable and within DotA's variants at 10 megabits per second those numbers are 138 and 108 FPS average 12 megabit per second and medium puts a huge strain on each CPU versus baseline producing large losses from baseline as resources are reallocated to the stream here are these streaming results at the good quality setting we're at 100% of frames encoded on both the 8700 K and 2700 X with frame time consistency both good at 12 megabits per second we lose 60% of our frame delivery on the stock 8500 K at about 7% of our delivery only 2700 X here's a comparison of the 12 megabit per second playback video on both Intel and rise and put next to each other this is again a thread and scheduling advantage Rison has more threads to throw it encoding and medium configuration which matters far more than frequency for this task where Andy is behind an FPS streamer side it is far ahead and viewable stream quality in Dallas Kellan AMD and Raw framerate streamer side definitely but they're getting killed in viewer side playback in this workload which is again somewhat synthetics because it's so abusive the viewer side playback matters no overclocking intel helped tremendously but not enough to come close to the 2700 X again tuning with priority and affinity would largely help with this but that's asking for a lot of extra work and it doesn't perfectly fix the problem the end of the day more threads is more better provided a reasonably similar frequency for this type of workload you could also cap game framerate to further relieve and tells resources to say 60 FPS but the point is that the 2700 X is natively more advantage than Intel here we didn't need to overclock the 2700 X to sustain reasonable quality so we did it closing out the streaming section here are 2 10 megabit per second faster clips next to each other from the 8700 K and the 2700 X they're basically the same so the takeaway here is that yes both the 8700 K and the 2700 X are perfectly capable of delivering a high quality viewer experience and the 87 arcade does have an advantage in player side experiences that said if you try to increase the quality of your stream beyond what you're seeing right now then AMD holds an advantage for the viewer which is easily arguable as the most important aspect of streaming basically with the 2700 X you have a bit more headroom to increase your stream quality over the 80 to 1 RK even though both can do it just fine if you have the same slightly lower quality for some production benchmarking we're using blender two point seven eight and two point seven nine for in-house render benchmarks the newer blender 2.7 9gn monkeyhead render passes the r7 2700 X towards the top of our charts requiring 23 minutes to render when stock using X 370 and basically the same amount of time when using x4 70 in stock settings running in 4.2 gigahertz all core doesn't get as much considering how acts of r2 works and pushes us to 22 minutes again consider that this is functionally pre overclocked anyway hence the limited difference between manual overclocks for comparison the threader for 1920 x stock CPU completed its render in 17 minutes about a 25% time reduction the i7 8700 K at 4.9 gigahertz finishes its render in 24 minutes marking a stock 2700 acts as 4% reduced in time requirement a good thing go into 5.0 gigahertz almost ties the two with an overclock on the 2700 X pulling it about a minute ahead of the overclocked 87 hard K F I've gigahertz when overclocked 8700 K is punchy equal in performance stock however the 8700 K ends up at twenty six point five minutes allowing the 2700 X stocks to reduce its time requirement by 12.5% the 2700 X stock CPU is also approximately 13.5 percent reduced and time requirement from the stock original 1700 X that we retested on the X 470 motherboard our GN logo rendered test is the most abusive and stretched the 2700 overclocked to a point that we could not achieve stability at 4.2 gigahertz and it wasn't worth going to four point one since that result in equivalent or worse performance than stock with X of R 2 for this one the 1920 exit leads the 2700 X stock CPU on export 70 by about 28% time reduction a substantial change for anyone serious about CPU rendering lady 700k overclocked at 5 gigahertz lands at twenty eight point six minutes allowing the 2700 X 82% time reduction over the high end Intel CPU of course note that we don't have any intel h EDT parts on here presently as those have been busy with rip LT t streams and other ongoing content for time spying other synthetics check the link in the description below for the full article moving on to game benchmarks assassin's creed origins is new to our test bench for this one the AMD are seven twenty seven hundred x stock cpu performs at 107 FPS average with x4 70 or 112 FPS average when overclocked to 4.2 gigahertz those are reasonably tightly timed in the 80s for 1% and 70s from 0.1% want consistency of frame x and we got that here the stock r7 2700 performs at about 103 FPS average marking the stock 2700 X as 4.2 percent ahead at 107 FPS average the art 720 700 X performs about 7.5% ahead of the r7 1700 X stock CPU on X 370 overclocked in the r7 1700 to 4 gigahertz gets it to 106 FPS average with low as reasonably behind this places it below the stock 2700 X with the overclocked 2700 X about 5.7 percent ahead the 8700 K stock CPU operates at 120 FPS average with lows at 91 and 80 thus landing the 87 our DK about 11.6 percent ahead of the stock 2700 X the i-580 600k is also a bit ahead including ruff equivalency and frame time pacing and consistency of the lows overclock in the 8700 k pushes our GPU limits and hits a bottleneck landing us at 125 FPS average or 12 percent ahead of the 2,700 X overclocked for comparison here's Assassin's Creed at 1440p most of the difference is vanished thanks to the creation of a hard GPU bottleneck that prohibits further scallion on the cpu still the Intel CPUs hold a bit of a lead averaging about 10 FPS or around 10% of a lead well hop songs to at 1080p is next for this one the 2700 x stock CPU and overclocked 2700 perform equivalently with thanks to XF r2 is frequency boosting 2700 X stock CPU ends up 6.5% ahead of the stock 2700 non ex but keep in mind that overclocked in the 2700 that's the same performance they're the same thing basically 2700 acts at one 11.8 fps also ends off about 10.4% ahead of the stock 1700 acts although overclocking the original r7 CPUs closes that gap partially as indicated by the r7 1700 at 4 gigahertz the 8700 k stock cpu at 1:30 2.5 FPS maintains a lead of 18.5% over the stock 2700 x at 111 FPS 0.8 the lows are also advantage somewhat significantly on the 8700 k mark and its overall frame time consistency has superior in this test overclocking gives us extra performance but not much at 1440p we observed similar performance to what we saw in Assassin's Creed origins our numbers plot and these CPUs as limited to about 92 to 98 FPS average where Intel CPUs are limited to about 1 or 2 to 1 or 6 FPS average project cards at 1080p no it's the 2700 X stock CPU at 111 FPS average with overclocking boosting to 114 IPs average this lends the 2700 X stock CPU 10% ahead of the 2700 9 X which overclocked to equivalent and about 14.4% ahead of the r7 1700 X with X 470 the fact that the 8600 K at 5 gigahertz passes the stock 8700 K but comes close to the 5 gigahertz 8700 K suggests that project cars favors frequency to thread that beware our rather sizable difference in data emerges with the stock 87 hard K is 135 point eighty FPS average landing 22% I'll be 2700 AK stock CPU overclocked we end up at 152 FPS average 122 fps 1% lows and so forth which is 34% ahead of the overclocked 2700 X clearly there's still something to be said for frequency in some applications we have a couple of other games we test it as well we'll leave them the link below otherwise if it'll be too long let's move on to power testing next this is where we need to put a note here Andy's version of TDP isn't comparable to Intel's they've measured them differently and so the numbers are not something you can't say 95 watt TDP 8700 K is less than 105 watt TDP I'm 2700 X Kent it doesn't work that way so not for just the power drawn anyway that's not a power draw number it's basically how much cooling is needed to keep the CPU under a certain spec so numbers are each in different ways for AMD TDP is calculated by subtracting 42 from 60 1.8 and then dividing it by 0.189 those numbers if you're want written are derived from what AMD claims to be the optimal tks temperature 61 point eight degrees Celsius be what they claimed to be optimal Inlet temperature for the heat sink fan at 42 degrees Celsius and then 0.189 basically degrees per watt of the heat sink so for power testing Cinebench multi-threaded positions the 2700 x stock cpu at 143 watts consumption when left to full auto settings or 192 watts when overclocked to 4.2 gigahertz at one point 4 volts the 1700 x stock operated at 1:13 at watts multi-threaded single-threaded the 27 X operated at 46 watts on the export study board so we're on the single thread chart now were 37 watts on the X 370 board these 1700 X ran at 43 watts x4 1737 Watts X 370 you get the idea for 3dmark physics we measured 80 watts for the stock 1700 X 110 watts for the overclocked 17 hour Dax and 102 106 for the stock 2700 X overclocking pushed us to 130 watts on the 2700 X and that's why I wanted to point out the TDP difference because in Cinebench again stock no changes just acts of r2 we're approaching 150 watts which is quite far from 105 watts that's because it's not measuring the same thing so when you're buying a heatsink just make sure you're careful of what you buy and that it can actually sustain up to 140 150 it's for a stock 2700 acts basically that's what you're looking for so yeah conclusion we're working on some major volt frequency content with thermals you'll want to check back for that that's going to be really cool and we have a 2600 ax review coming up momentarily as for what you should buy it's like before except more stable fewer blue screens almost zero now so round of applause Andy that's actually a really serious improvement to have nearly zero blue screens as time quite happy with it when we do get blue screens that comes from timings on the memory most of those have been fixed with UEFI updates in the final hour here so it's pretty stable at this point it's nearly it's it's nearly perfectly stable now overclocking doesn't get you much don't bother in most cases X if r2 does a fine job there's not much Headroom for you until you exit ambient cooling and then for the rest by the non ex versions of the chips if you can overclock you can spend five minutes of your time that's where overclocking matters over logging 2700 X 2600 X not exciting we're talking at 2700 to become a 2700 X that's where you get your money's worth with overclocking so to be clear I'm not contradicting myself by saying don't overclock but overclock then don't bother overclocking the x-series overclock the 9x series save 30 bucks or whatever it is and be happy that you did if you absolutely can't overclock for some reason I guess by the X trees now as far as comparison to Intel we gave you plenty of charts look over them figure out what's best for you this video is quite long if you want more of a hard conclusion of anthe versus Intel I can try and guide you through it but I'm gonna do it in the article link the description below on the conclusion page where I have some more space to think about it so that's it for now subscribe for more we have a lot more coming go to patreon.com/scishow stop directly go to store that gamers nexus net it's by one of our mod mats they're on backorder you guys keep buying and we sell out but we have more coming now so they'll be here within a couple of weeks and shipping again thanks for watching I'll see you all next time