Gadgetory

although coffee lake and intel six core CPUs have undoubtedly been planned since before aizen's launched it appears that AMD woke the sleeping giant this year after some prods from risin and from thread ripper intel's i7 8700 k launches today it's a 6 quart 12 thread part that targets high frequencies and that launch is accompanied by a target MSRP of $360 about a 10 to 20 dollar price increase over kb lake this is the competition part of the market intel moved up the X 299 launched to compete with thread Ripper but what we'll never know is whether Intel adjusted its target price to compete with risin at least on the coffee leg CPUs either way the i7 8700 K is here now and we've got a densely packed review covering most aspects of the new CPU lineup this video is brought to you by the be quiet dark Bass Pro 900 white Edition the DBP 900 marks a return to full tower cases equipped with ample harddrive support effective noise damping foam high performance fans and the option to be inverted into an alternative layout the tinted tempered glass window and Qi charger at a high untouched to an already well-built case learn more at the link in the description below so lots of things to know for this review first of all we got the worst-possible 8700 K sample that could be made the overclocking head room on ours was much worse in terms of voltage versus frequency than what many of our peers got so while I am envious of the other CPU samples out there it at least gave us an excuse to step through a couple hours worth of trying every trick in the book to get the thing the clock a little bit higher we ended up stuck at about 4.9 gigahertz but kind to hit 5.0 and the thing is I've got a screenshot from their Bower I can share on the screen he is giddy at the CPU he got I mean der bauer he has a couple of them and that one in particular does 5.2 gigahertz it's something like one point three seven five volts or somewhere around there the guy is lucky that's a crazy overclock ours shows just how the Silicon lottery works why companies and websites like Silicon lottery are in existence our CPU does four point nine to 5.0 at about one point four to one point four to volts depending on the test and at 5.0 we can't pass blender even with one point forty volts so a little disappointed there but we still did the deal eighteen fainted liquid metal all that stuff plenty of stuff to do still it overclocked to four point nine or 5.0 where it held and we haven't tried overclocking since applying liquid metal so there might be more room in there now because we've got more thermal Headroom to increase the voltage further we've reduced power leakage things like that doing everything that we can to push the thing further and we'll do that in the follow-up content piece along with tomorrow's content piece for the next day which will be the i-5 review from Coffey Lake today we're focusing on the 8700 K these tests will be a mixture of legacy tests from our test suite so we've got a couple old ones that were slowly phasing out and we have new tests coming in the reason we have both is because we ran them both in parallel so 8700 k went through the old test and then the new tests and that's because the new tests don't have ten years worth of CPUs on them in terms of I mean over the past year we've tested CPUs going back ten years and so those are not present on the new benchmarks but they are on the old one so we ran both and parallel that way you get a sense for scaling so we're gonna pop up the timestamps for this video the table of contents whatever you want to call it and that's because this is densely packed and all the disclaimers aside now we'll have thermals and elating overclocking blender premiere gaming all that stuff and you can jump around if you want to though we'd obviously encourage you to watch all of it because it all comes together in the end and paints a pretty clear picture of how this cpu performs but it is dense in terms of charts and we'll have all of those in the article linked in the description below if you wanted all on one page for easier reference in the future we'll start this one with thermals and D lidding since that's been a popular topic for us lately running these tests with fixed frequencies and voltages as that's the only way to properly control for fluctuations and behavior we used rocket cools @d LED kit to remove the IHS which will link below this proves trivial to use and works flawlessly which for a forty dollar kit is pretty damn good we can recommend this one for coffee like dee lids the previous model for KB Lake also works here and D letting the 8700 K is easier than the sky like xcp use as it lacks the dual substrate layout deleting the CPU was it matched with removing the silicon adhesive from only the IHS we left it on the substrate for now and then we applied a thermal grizzly conduct or not liquid metal will link that in the description as well if you're interested in it and you can check some of our previous d-league content for more information on performance with skylight gags let's start with the more exaggerated results when testing blender with a four point nine gigahertz frequency and one point four core voltage our 8700 K with Intel Tim and an end exp X 62 landed at 76 degrees Celsius average core temperature with a 10-second peak of seventy six point six C and a liquid temperature of 40 point six five C the liquid metal version at the same frequency and voltage measured at 52 point five nine degrees Celsius that's a reduction in average core temperature of nearly 24 C liquid temperature measured about the same as before at thirty nine point eight to see looking into this further we realized that measured at the current clamp on the EPS 12-volt rails the 8700 K with Tim was drawing about 10 to 20 watts more power at the same voltage and frequency we're not yet positive but our present theory is that this outcome as a result of reduced power leakage on the CPU as a result of energy transfer efficiency improvement from the die to the IHS by way of using the conductor not liquid metal which is about 73 watts per meter Kelvin thermal conductivity versus something like three so that would explain part of the power leakage reduction testing with prime95 28.5 on the cpu lock to 4.7 gigahertz core 1.3 5v core we found the delta much closer than in the earlier over volted tests the Tim test plants us at sixty two point six degrees Celsius average core temperature with a 10 second high at 63 point eight two liquid temperatures effectively matched to the liquid metal version at thirty nine point six verses thirty nine point HC and that's within our test error and resolution so we can just call them effectively equal using tinactin aught we're at fifty two point five nine C for a 10 degree reduction in average core temperature which is a pretty reasonable gain though the previous gain was much more impressive this isn't as big of a deal as with the sky like act CP is when we were constrained by thermal limits with overclocking the coffee lake CP is run much cooler we think I've consulted with some folks about this we think that there's potentially better Tim on the coffee like CPUs which would coincide with some hints that Intel's given but we have no way to really properly validate that we can't scrape and test the thermal conductivity of the Tim it's just not feasible for us so that said coffee like actually does run reasonably cool all things considered it starts getting a little warm once you have four point nine gigahertz and one point four volts but up until that point it's really not bad but the thermal scaling at the higher voltages and frequencies does ramp pretty aggressively it's not a linear increase so keep that in mind a 10 to 20 degree reduction though is no small feat and does mean you can potentially reduce the rpm on your coolers lower noise things like that and it's good for anyone pushing higher over clocks as well where you're gonna run into potential thermal limits with the lower end or smaller coolers so for the kits we used I'm gonna go ahead and give a shout again to rock it cool because they sent us that kit a while ago for KB Lake we finally used it and it worked well so if you're interested check them out conduct an artwork has worked well for us also and then if this stuff is scary to you gonna give a quick shout to Silicon Lottery calm because I've heard they're going to have been two CPUs within one to two weeks of launch and we worked with them previously on a KB Lake CPU so those services are all options if you want to either deal it yourself deal it yourself probably don't do that but if you want to deal with the CPU yourself and apply liquid metal it's a 10 20 degrees pretty damn good so consider it whether its DIY or not but it's definitely not necessary unlike skylake acts you will not be thermally constrained on the overclocks with coffee lake on a reasonably sized cooler as opposed to the 79 60 X where we were actually 200 megahertz higher by adding liquid metal so that isn't as big a change as sky acts but still a pretty good change next set of tests we have to streaming and recording tests prior to the game in workloads for the first test we're benchmarking that live streaming capabilities as encoded on the CPU using OBS and x264 with the faster preset outputting to YouTube at 10 megabits per second 1080p 60 the second test is done with local recording captured at 15 megabits per second so Internet's not an issue and using x264 fast for the preset so it is a better encoding option for higher quality the second test is far more intensive versus the two and is more likely to stress the CPU into a point of dropping frames even if somewhat of a synthetic test we'll start with the easier workload streaming dirt rally to YouTube at 10 megabits per second and with the faster preset we end up encoding 100% of the frames out of both the 8700 K and the r7 1700 with effectively no dropped frames on either we didn't overclock either because it just wasn't necessary we are at or below 0.1% drop frames on the r7 1700 but that's within test variance and margins so effectively zero at this quality setting the to produce the same viewer experience for the stream the 7700 K drops 44 percent of its frames prior to the process prioritization tweak that we did with the overclocked variant dropping about 30% of its frames manually assigning processed priority allowed the 7700 K to deliver 100% of its frames dropping zero but did require manual tuning and had another hit to performance that we'll see in a moment the 8700 K the coffee lake CPU and the r7 1700 both avoid this prioritization requirement for this particular test and they were perfectly fine at delivering the full experience to the viewer so they both pass this one the next side of the coin is player experience shown in FPS the 8700 K delivers a baseline performance of 136 FPS average with no stream the output with 1% frame times measured at 109 or 98 FPS for 0.1% lows streaming drops us down to 122 FPS average really not a bad drop with 1% lows at 87 which is also not bad 0.1% lows however have fallen down to 37 FPS which seems a trend for all of our streams outputs thus far there are seven 1,700 places at 108 FPS average base line with its streaming output at 91 or 65 FPS for the one percent lows and 37 for 0.1% lows this comes down to threads versus frequency as an argument Rison has more threads at a lower frequency and the game wants frequency but the stream wants threads it's able to keep up with both reasonably but has a hit to game performance whereas the Intel alternative 8700 K but again it's got the extra couple thread so give and take as needed but that's the last generation the 8700 K adds two more cores and four more threads while keeping a similar frequency this significantly improves on the 7700 k's former position and manages to keep up with these 1,700 of these settings when streaming with a faster frame rate when gaming this is more of a worst-case scenario meant to stress CPUs to a point of showing differences we're recording locally at 15 megabits per second and using the fast preset done for both dirt and dota 2 with dirt we deliver 54.6% of frames to the recording dropping 45.4% the r7 1700 however manages to deliver 57.8% of its frames dropping 42.2% extra threads are helping in the encoding process here and managed to push the r7 1,700 stock CPU into lead over the 8700 k stock cpu where the r7 outperforms by a few percentage points and delivered frames to the stream it does technically deliver them with more variable latency the 8700 K delivers its 54.6% of frames with at ninety point eight percent of them averaging the desired 16 point six six seven milliseconds so that be your 60fps refresh roughly 4.6 percent of these are above these 16.7 millisecond range and about 4.6 below that rain the r7 1700 cpu delivers its percentage of frames 57.8 so higher with seventy five point six percent of those frames averaging 16 points seven milliseconds just under 11 percent are faster than sixteen point seven and just under 14% are slower than sixteen point seven milliseconds as for fps the 8700 k average is 136 fps baseline without any capture interference and the r7 average is 108 FPS average baseline neither CPU drops very far in its captured performance we fall to 126 FPS average on the 8700 K and 92 FPS average on the r7 1700 which amount to a seven point four percent reduction and fourteen point eight percent reduction from baseline respectively both CPUs have room in player side FPS to improve capture side delivery there's not much point in spending all of those CPU resources on fps for the player to get above 100 FPS when considering that the recording just can't keep up the 8700 K has more room to play with in this particular title in this manner and has greater consistency despite a slight deficit in total frame delivery with dota 2 under the same conditions the story changes a bit the r7 1700 captures 85 percent of its frame successfully driving 14.9% in this torture workload the i7 8700 K captures 68 percent of its frames successfully playing these files back side-by-side the result is obvious in this particular title the r7 1700 does provide a better capture output than the 8700 K though neither is ideal for our torture workload you'd want to drop settings a little bit on the 1700 to recoup those 15 percent of dropped frames but you'd have to drop significantly on the 87 Hart K or start tuning with priorities here's a look at the frame rate chart this give us a better idea as to where the 8700 K is power is going baseline performance is 158 FPS average for the 8700 K or 80 FPS for 1 percent load this is significantly bolstered over the 110 FPS average m56 fps 1% lows of the seven 1700 which is already known to drag a bit in dota 2 beginning the game capture the 8700 K drops 36% of its frame throughput to 101 FPS average with these 1,700 dropping a similar 35% of its frame throughput but delivers more of its frames to the stream or the recording capture where the 8700 K significantly now performs the r7 1700 in player side frame rate it is significantly underperforming and capture frame rate for this particular title giving process priority to OBS would solve this problem largely as would an overclock that's exiting out of box territory and just because these sort of synthetic torture workloads make it so that people often lose sight of the original real world scenario both of these CPUs are perfectly fine for streaming and capture at 1080p 62 YouTube at 10 megabits per second you can even go a bit higher than that especially if you start tuning the x264 preset and you're really not gonna have to do much with priorities unless you're trying to do something like a fast or a medium preset then you might need to give it will be as process priority let's move on to power testing and we'll have more power tests in the article linked below but we can start with just a few for now starting with blender the i7 8700 K pulls about 96 watts down the EPS 12-volt cables this is not power draw from the wall it's measured with the current clamp this is the stock configuration and permits the CPU to complete the render in twenty six point six minutes our seven 1700 poles about 80 watts in this configuration completing the render in around 29 minutes more on that later overclocking the 8700 k puts us up to 130 watts when at 4.9 gigahertz and one point 4 volts on our magically awful chip this is right around where our overclocked r7 17-hundred lands for power the i7 77 100k stock CPU measures 74 watts for this test marking at about 20 watts lower than the 8700 K a lot of this has to do with motherboard and BIOS as well as always so these numbers would change based on which board you're using and how much voltage it pushes for the fire strike physics test - the 87 100k a plots at 68 watts under its stock configuration or one six overclocked for comparison the 1700 draws about 55 watt stock 95 overclock and the 7700 case it's around 50 watts with our gaming 7 motherboard on the latest efi this is the important part for the gaming tests going forward so for these we're starting with our legacy benchmarks first these are conducted completely differently from the rest of the game test that we're using moving forward and will stand as a long-term support option that allows you to compare against a lot of CPUs but it's being phased out because it's becoming less accurate with time that's because the legacy tests that we have done for the past year now are now outdated in a few ways video card drivers windows version and the graphics card we're starting to bump into GPU limits in some games going forward we're using a 1080 Ti FTW three for the GPU rather than a 1080 FTW one non TI and we've updated our memory time in timings we've updated our Windows version we've updated our graphics drivers so the tests are not comparable between them just to make that clear starting with two legacy charts in battlefield one the 8700 k-chart tops at 151 FPS average placing it a few percent ahead of the 7700 K as for scalability it goes like this the Intel i7 930 Nehalem CPU runs at 96 FPS average and is about 10 years old with the overclocked variant at 118 FPS average the i7 2618 FPS average would be 4.7 gigahertz version performing at 132 FPS average that's roughly a 23% climb in stock to stock performance we skip Ivy Bridge here and jump to Devil's Canyon 4790k for the i7 operating at 140 FPS average the i7 6700 K is at 141 FPS average and the 7700 K is at 146 FPS average part of the reason for our new tests is this one we're bumping into other limits here clearly so we'll soon be moving on to the new GPU and new games in our legacy watchdogs to test the 8700 K performed nearly the same as the 7700 K held back in some ways by the more limited boost clock we noticed that our gigabyte or too often only turbos to 4.4 gigahertz on the Z 370 board for all core turbo during gaming scenarios this is 100 megahertz lower than the 7700 k's all core turbo and is sometimes reflected in games by a slight frame deficit that said this is a legacy test and we'll see if that changes for the new one despite the increased core count Intel here is facing a similar scenario to AMD with Rison they have heavily multi-threaded capable CPUs but they're facing adoption challenges on the gaming front in the future now that both vendors are pushing for higher thread count it'll happen but four games out today some will benefit from threads and somewhat better benefit from speed we've been told that some of the Asus boards boost to 4.7 gigahertz all core in stock settings which would net a higher fps let's move on to the new game tests which use all the new testing methodology that we've kind of already pointed out but is also in the article linked in description below along with extra charts for games that won't be shown here starting with Civilization six we used the AI benchmark to test a different metric time required to compute AI turns as FPS is rather useless here the turn time is about the same at 1440p as it is at 1080p that we did test both average FPS actually goes up for worse CPUs because the time spent idling on the screen is longer so just to reiterate that and make sure it gets through the reason we're not using FPS as a metric for the AI benchmark is because as you plot it with worse and worse CPUs FPS goes up why does that happen it happens because in an AI benchmark where you're testing turn time a longer turn means more time spent staring at these same non moving elements on the screen whereas a faster CPU is going to jump around two points on the map a lot more frequently and will fit in the times benchmark more so that means the FPS will be worse on those hence we're not including FPS benchmarks because there's no point sims not really a game where you do that so back to the x then the 5 gigahertz overclocked to i7 8700 k holds the fastest average turn time at about fifteen point four to fifteen point five seconds per turn to give an idea for range this is a twenty six point six percent reduction from the slowest time although a five second average turn time is not huge keep in mind that this is per turn so a five player game would benefit from a 25 second reduction in total time per total turn to get to your next turn across the lawn play periods this can add up but the relevance of it is really up to you anyway the 8,700 case stocks CPU completes its turns in about 16.1 seconds or about 4% slower than the overclock the 7700 case the oxy View completes its turns in sixteen point five seconds with the four point one Giga at 1600 X not far behind as you can tell by looking at the 1600 X stock and OC numbers B 1700 stock and OC numbers and the 8700 ka stock ANOVA see numbers frequency matters in this game to a point that's perhaps greater than thread the 7700 cavers is 87 100k it also indicates that frequency is of at least slightly more important than the thread count for total Wars a new benchmark we measured the 8700 K is a chart topper at 1080p with 176 FPS average and lows at 110 and 95 fps we experienced a bottleneck at 1080p with the overclocked not providing any additional performance versus stocks or stuck at the GPU here the 7700 K is 7.4 percent slower at 163 FPS average with the r5 1600 X at 4.1 gigahertz running a 147 FPS average and is 16% slower than the 8700 K Total War favors the frequency advantage of the 1600 x over the stock 1700 clearly at 1440p we equalize some of the distance with GP limitations but still see differences I 780 700 K is clearly still bottleneck on the GPU operating a now reduced 153 FPS average for each skew this is 7700 K runs at 143 or 7% behind and the stock r7 1600 X runs 20% behind the 8700 K as shown here project cars at 1080p has the 5 gigahertz 8700 K at 127 FPS average benefiting from the frequency focus of the game the 4.4 gigahertz all core operating frequency condemns the 8700 K to perform about the same as the 4.5 gigahertz 7 100 K and our testing both that around 108 to 110 F he has average are 7 1700 further demonstrates the frequency focus of this game plays in at 78 point 5 fps stock but 87 FPS at 4 gigahertz the result is a staggering 45% difference for the 5 gigahertz 8700 K versus the 4 gigahertz 1700 or 43% versus the 4.1 Giga at 1600 X stock to stock the different strengths to 27% versus the higher clock 1600 X at 1440p the 8700 K manages 118 FPS average overclocked 106 fps stock and with our all core 100 megahertz deficit to the 7700 K producing the expected favored 4 KB lake Rison performance remains more or less exactly where it was for the 1080p results as a result of being CPU bottlenecked we have a few more game benchmarks in the article linked below if you want GTA 5 and some others it'll be there but this video is getting pretty long at this point so we'll leave them to the article get into some production workloads now we'll start with our legacy blender test using two point seven eight a then move to the updated renders on two point seven nine like the game in section we're doing this to give an idea for scaling against a year's worth of CPUs but also to provide important modernized information with a more limited data set with version 2.7 8 a and our in-house monkey heads render the 8,700 case stock CPU takes twenty three point seven minutes to complete the render which is about a 44% time reduction from the 7700 K a stock CPUs 42-minute under time compared to r7 CPU is the 8700 K takes about 11 to 13 percent less time to render them the 3.9 gigahertz overclocked r7 cpus including the 1700 x @ 3.9 gigahertz and 3466 maegor's memory speed the 8700 k takes 28% less time to render than the r7 1700 and that's thanks to its high frequency and core account combination for the performance turns out intel also has version 2.7 8a to thank for version 2.7 9 we're using the test shown on the screen now the monkeyhead render takes twenty six point six minutes to render on the 8700 k compared to twenty eight point eight minutes on the 1700 both with 32 hundred megahertz CL 16 memory 8700 k completes the monkeyhead render seven point six percent faster than the r7 17 and that closes the gap in this case overclocking both produces a twenty four point one five-minute render on the 8700 k or twenty four point seven minutes on these 1704 Adobe Premiere and other benchmarks not shown here check the article linked in the description below I'm gonna leave it there you've got enough data to really piece together your own conclusion at this point I don't need to spend five more minutes talking about it the only commentary I have to offer right now is pricing and availability we know that Hamas RP is we don't know what it's gonna cost when it launches because we film this before it's available on retail so if the price is MSRP the 8700 K definitely has a place in the environment for the right users apparently there are power delivery optimizations on Z 370 it would still be nice to see the CPU work on Z 270 but it might not be that simple and until frankly we won't tell us anything further than what we already know as for Z 390 that is supposed to carry a performance improvements Z 370 is somewhat of a stopgap between now and Z 390 and Intel clearly move this launch a little bit forward to better compete because they just launched cable a commit 7700 K in January so if you already own something like a 77 heart k unless you have a very specific reason for this 80 a 700 K I'd say don't upgrade but for older CPUs obviously there are reasons to upgrade to either Rison we're Intel at this point you've got a whole list of data pick through it find your use case and decide from there because it's very use case dependent at this point I'm not going to put a blanket statement on it if you should buy it or not availability in some regions will allegedly be low but we're just not sure yet those are still rumors so if it's low I'm sure I'll be complaining about it in an episode soon enough so subscribe for that if it happens otherwise the i-5 review comes shortly this was enough for now patreon.com/scishow us before this is what we do so help us out if you like it I'll see you all next time