Gadgetory

multicore enhancement is an important topic that we've discussed before right after the launch of the 8700 k most recently it'll become even more important over the next few weeks and that's for a few reasons for one intel is launching its new B&H chipset soon and that will require some performance testing and validation and for two AMD is launching its rise in 2000 series chips on April 19th and those will include xfr - some x4 70 boards just like some X 370 boards will have MCE look-alike options but they're not exactly the same for Intel and AMD both enabling MCE or core performance boost means running a higher clock speed and depending on which setting you're talking about it could exit specification particularly the power specification and that mostly applies for low-end coolers the question is if any motherboard vendors enable MCE on Intel by default or silently because that's where results can become muddy for buyers before that this video is brought to you by EVGA and the X 299 dark motherboard for the Intel high end desktop CPUs the X 299 dark is one of the only motherboards on the market with proper vrm cooling we've tested this and found a significant performance increase over those without active cooling on the brms this board was used in our recent attempt to set a top-10 record in fire strike and you can learn more about the x-29 dark at the link in the description below there's been a lot of update since the last time we talked about multi-core enhancement but again the biggest one here is that we need to revisit it in preparation for B Series and H series chipset testing where we'll be looking at some of the previous Intel CPUs we've reviewed again except on motherboards that are more appropriate for them the i-5 non K CPU the i3 non K CPUs though it CPUs we've basically said from the start deserve to be on a non Z series motherboard and it's time to review those again re review those on the newer lower end boards so that we can get a realistic look at how they'll perform in a use case that makes sense so ramping into that then we wanted to revisit MCE part of revisiting imce as stated is also in preparation Rison - or the 2000 series rising CPUs which we'll be launching on April 19th and for that launch we'll be looking at some new performance options on X 470 motherboards that weren't fully present or present in the same fashion on the original rise in CPUs so we'll be looking at in r7 1700 today with poor performance boost it's not the same as a Massey II but just looking at the name of it it sounds the same that's an option in most BIOS versions 4 X 370 motherboards and then we'll be looking at the i7 8700 K with MCE multi-core enhancement on the Intel motherboards before ramping into a lot of charts there's a lot of data here some quick information on MCE and core performance boost if you're not familiar with either multi-core enhancement or MCE is basically a feature that's on some not all Intel motherboards and it's a manufacturer created feature this is something that asus or gigabyte or folks like that would have a hand in not Intel it's a non Intel feature and MCE is also a non stock feature multi-core enhancement by default is basically built to run the CPU at its maximum single core turbo except across all cores generally speaking actually always speaking this will also increase your power consumption so it will exit the power specification that is assigned by the CPU manufacturer core performance boost meanwhile on AMD motherboards is more to do with the precision boost and xfr features on Rison platforms and rise in cpus so it behaves a bit differently the clock speed is still impacted but it's impacted for a different reason rather than maxing out single core across all cores what's happening is you are applying xfr and precision boost across the CPU as expected by an these specifications it's get into the test platform now before the numbers so for testing we only use Cinebench this time and we tip we've actually replaced Cinebench with blender for all of our benchmarking for rendering application workloads that's mostly because people actually use blender as an application to make things and render things and also because we found blender to have a whole lot less variance we also build our own in-house blender benchmarks so we have total control over them which gives us a lot more confidence in testing Cinebench however is good for a very quick and comparable synthetic benchmark so we're using that today in the course of running Cinebench we ran it without any logging software at all for these Cinebench numbers we present and then when we start presenting things like frequency power characteristics things like that we use hardware info running while Cinebench is running but we're not average in the results from those Cinebench rounds while hardware info is running because it influences them a bit so that's the first thing you need to know secondly need to know is four components for the Intel systems what we're using are a couple of other words we have an ACS maximus X Heroes III 70 motherboard we have a gigabyte XIII 70 Ultra gaming we have an MSI XIII 70 I gaming Pro carbon I TX board for the Intel other components in that bench we used an 80 700 K it's deleted but that's irrelevant here and we also have four memory Corsair Vengeance lpx thirty-two hundred megahertz at sixteen eighteen 1836 for its XMP profile XMP I guess and that's 2 by 16 gigabyte sticks we're also using an EVGA 750 watt supernova g2 for the power supply and an EVGA 1080i FTW 3 for the graphics card which is irrelevant along with a Kraken X 62 cooler just to make sure we're not thermal throttling anywhere for the AMD systems we're using an ACS crosshair 6 hero motherboard X 370 on six thousand one beta bios we are also using an r7 1700 review sample we're using the guile 32 hundred megahertz CL 16 memory provided by AMD for the RF 5 review ages ago now and for the power supply and NZXT Hale 9 TB to GPU 1080i FTW 3 and cooler also a Kraken X 62 we moved on again from Cinebench for blender previously but now's a good time to reintroduce Cinebench for some of this testing this testing started out with an ASUS Maximus 10 or Maximus X motherboard which received after the initial launch reviews of the i7 8700 K our starting board was the gigabyte Z 370 ultra gaming it was a shame too that we received the Maximus later because we were able to overclock far better on the maximum ten and that was with the help of reduced V droop which we had on earlier visions of BIOS for the gigabyte ultra gaming board and better bios options also a wet way better erm on the Maximus motherboard in fact the ultra gaming was so problematic that gigabyte is now introducing an ultra gaming z3 70 V to motherboard v 1.0 that's actually it's actually the name of it it's the Ultra gaming v2 1.0 as opposed to the Ultra gaming v1 1.0 because we can't iterate 1.0 obviously and make it 2.0 but there is a significant difference between them so to be clear we're using the original one and for this first set of results look at the Asus board here's our test matrix for the ASUS Maximus 10 motherboard we ran tests for all the stuff in the charts here the first one is complete auto settings as reset from default not even with XMP MCE is auto second one auto settings with MCE explicitly disabled still without XMP next one do CP enabled the asus version of XMP presets and then answering no to the prompt MCE auto next one is do CP enabled and answering no to the prompts then disabling MCE explicitly next do CP enabled answering yes to the prom MC e auto and finally do CP enabled answering yes to the prompt MC e explicitly disabled as you can see there's a lot of testing going on here for just one board so we have a lot of data to go through because there's Cinebench frequency performance and power performance for each motherboard note also that ACS only offers auto or disabled for multi-core enhancement with the BIOS revision we're using which is exceptionally confusing for the end user we would strongly encourage Asus to add an explicit enabled and disabled option perhaps along with Auto as opposed to what presently exists which is you can either choose off or lol idk that's basically two options because Auto doesn't tell you what it's going to do so we'd like to see enabled and disabled explicitly but that's what we have right now after enabling XMP or do CP as Asus calls it Asus will prompt you yes or no on further changes we tested both options let's start with Cinnabon scores will have frequency and power next for the asus motherboard we logged scores as seen on the screen now with do CP on yes' selected an MC e set to auto we scored 1445 multi-threaded average for Cinebench or 197 single-threaded and that's after multiple passes averaged selecting no but leaving do CP on and MC e to auto we scored within margin of error for Cinebench which has a lot more variance than you'd expect and thus is why we dropped it for the more reliable and usable blender tests regardless we're at 14 43 and 199 here explicitly disabling MC e gives us another equivalent score 1439 multi-threaded 197 single-threaded these are within margin of error of the first few tests explicitly disabling MC e on selecting no when prompted we again land within margin of error 1441 and 196 going full stock meaning no XMP on the memory we landed at 14 32 multi-threaded and 196 single threaded the MC e auto results were within margin of error of the MC e disabled results after this we sent an email to asus asking what MC e actually does on this motherboard because it appears that on the version of bios we have which will mention and the article as well MC e does nothing right now but we haven't tested the newest version of bios on this motherboard because we're trying to see what performance looked like with the first public release as opposed to our initial testing which used a pre-release bios we'll revisit this topic with the asus board later once all of the finalized intel security patches come out and things like that so this might come down to the beta bios or whatever we're running presently being bugged but either way we'll revisit the topic when ACS ships a finalized BIOS version for upcoming CPUs we would assume it'll have 1 + or 4 security patches with Intel of course those are getting updates as well let's look at frequencies just for validation looking at frequencies do CPEs and MCE auto resulted in an average at single-threaded frequency of 4550 3.6 megahertz on the 8700 k multi-threaded was 42 99 this is expected and matches the Intel turbo boost table do CP no and MC e Auto resulted in a single-threaded frequency of 45 41 9 megahertz within margin of error of our averaging and test to test variance multi-threaded was 42 99 we're looking for differences equal to or in Access of 100 megahertz do see P yes and MCE explicitly disabled the US forty five thirty eight megahertz single threaded and multi-threaded was 42 99 do CB know MCE explicitly disabled to gave us 45 56 and $42.99 multi-threaded this again is completely within Cinebench is relatively wide variance we have an established range of about 23 megahertz single-threaded and basically 1 megahertz multi-threaded full stock settings with MCE off gave us 45 67 megahertz single or 43 20 multi the big difference here was disabling do CPR XMP resulting in a 20 megahertz change still this is not reaching those higher clock bins and that's under auto auto settings with MCE set to auto gave us 45 59 megahertz single or 43:19 multi we chalk this up to being a potentially buggy bios but we revisit this once bios is ratified and released publicly for the new cpus coming out and for security patches as for power consumption also for validation we observed approximately 104 to 108 watts consumption during multi-threaded workloads for all tests the stock tests without XMP were closer to 96 watts the range run to run was approximately 4 watts which is within reasonable variance a Seuss's Maximus 10 results with the bios used to seem to indicate that MCE is either bugged or non-functional with this set of testing we'll have to revisit either older BIOS revisions or upcoming BIOS revisions for more one final note we previously did testing with MCE on the asus board and noted that it was functional at least somewhat but it seems to have been broken in one of the later revisions and part of this is because in the earlier testing we noted that some of the MCE tests would result in blue screens I think that was with blender if I remember correctly and that was because the CPU was running voltages too low for the frequencies MCE was pushing thus causing an in stable overclocked unstable overclock or a crash and so we obviously recommended against using that and manually overclocking if that's what you wanted to move on to gigabyte gigabytes z 370 ultra gaming's next this one took some extra effort because we had to dig up old pre-release BIOS revisions for testing and these are BIOS revisions that we use during our review period and we're important to digging deeper into MCE in a heavier fashion here's a chart of the test matrix for gigabyte first BIOS f5 at complete stock MC Auto by OS f5 at stock MC e explicitly disabled by OS f5 with XMP enabled at MC e auto f5 with XMP enabled MC e disabled f5 with XMP enabled MC e enabled and the first release bios f4 at stock which came out publicly with board MC e auto pre-release bios f3 at stock MC e auto and initial bios f2 with XMP and no MC e option present here's where it gets interesting with that last option with bios version f2 which is what the motherboard shipped with for the review units or at least ours the gigabyte conversation we had previously we explicitly asked if MC e was present on the board and whether it was disabled or enabled by default and gigabyte had indicated to us that they believed in completely disabling MC e for stock testing and noted that it would be disabled for this board by default we also asked these questions because there was no MC e option in gigabytes f2 version of the bios so even if you suspected it to be present it wasn't something you could see or change from what we've seen with our review version of f2 and on that note took us a while to find version f2 because we saved it to an old USB key and fortunately still had it on there because it was a pre-release bios it's not publicly available and the later versions came out that were f4 and f5 a were used for other testing for game testing we used BIOS revision f4 or beta f5 a these were explicitly switched to for game benchmarking as we were told by gigabyte that f4 and f5 a would impact GT X 1080 and 1080i perforins on the ultra gaming we'd already run our synthetic round on BIOS revision 2 and we're told at the time that there shouldn't be any CPU only performance differences between the revisions outside of the GV performance issues so then we ran f2 for Cinebench pov-ray at all but ran f5 a or f4 for game testing let's get into the Cinebench numbers for bios version f5 multi-threaded Cinebench performance Landsat 14 23 points for f5 stock XMP often MCE auto there's an MC II option in f5 unlike f2 received 193 for single threaded at 5 stock MC disabled got us 14 21 points this indicates that on at least efi f5 MC e is disabled by default as gigabyte said one left to auto efi f5 x and p on MCE auto we score 14 27 points reasonable as memory is now faster f5 with XMP and MC e explicitly disabled further supports that MC e is disabled by default on version f5 f5 with XMP and MC e explicitly enabled gives us 15 forty five points multi-threaded finalizing that MC is disabled by default on version f5 and we then tested efi version f4 a complete stock MCE auto we scored 14 23 and 193 points and that again shows MC e is off by default with f4 ya 5 version f3 same thing same performance and then after digging through those old flash drives to find f2 bios revision we found the culprit here there's no MC option in this bios revision and although gigabyte did later embrace that MC each should be disabled by default the initial BIOS revision had no MC e option and MC e was in fact enabled by default silently and there was no way to explicitly disable it in that version of BIOS that shipped on our board anyway so then all the scores would be in the 1500 for Cinebench and still faster for single-threaded at 204 the same is true for XMP enabled with efi f2 efi revision f2 with MC e automatically enabled ends up with a performance uplift of approximately 7.7 27.8% over stock for Cinebench as we're reaching the max turbo frequency for one core but applied to all of course and for frequency here's what we saw for efi f5 all instances with MC e off or MC auto we had a frequency of 43 hundred megahertz all core the correct value given Intel's frequency tables that are publicly visible through X to you they're not hiding them but they're also not really pushing them anymore package power consumption was approximately 87 watts explicitly enabling MC e and F 5 resulted in 40 700 megahertz frequency this exit stock spec naturally gives a boost of 7.8% again we used f4 and f5 a for our gaming tests and the behavior is correct with MCE auto on these revisions for efi f2 the frequency matches efi f5 with MC enabled 4700 mega it's all for power consumption also goes up with the higher frequencies and voltages naturally we're at 87 watts average for MC disabled or 115 for n abled the next question is why why did gigabyte shipped like this with f2 enabling MCE by default and also not having an option to disable it well we talked to gigabyte back then and we also talked to a lot of other motherboard makers at the time and we've revisited the topic now but basically when we released our initial MCE benchmark video months ago at around the time of the 8700 you motherboard makers noted to us that they were anxious that their competitors other motherboard makers would enable MCE by default which of course pushes them to consider enabling MCE by default this is because the motherboard vendors particularly the higher-ups as we understand it don't want to look at charts of reviews of Intel processors and see that their motherboards are scoring lower than their competitors when there's really no good reason for that to be the case aside from BIOS differences well there's one good reason it turns out of course it's MCE so that's why you see some of the boards or some of the earlier BIOS revisions enabling MCE by default it's because the board makers were concerned that they would look bad in charts because their competitors ran CPUs functionally out of spec now fortunately it looks like some of these have been turned around gigabyte has corrected this and disabled them to e by default on later revisions of BIOS so we give the gigabyte credit for that at least but it would have been nice to have it properly done the first time either way it was just a day or two before the review went live that another BIOS revision came out f4 f5 8 and the revision also fixed some GPU performance issues so we use that for game testing the motherboard vendors here want to compete with each other it's not some Intel driven plot or conspiracy to make Intel CPUs look better this comes down to motherboard manufacturers and it's up to really everyone to make sure that MCE stays disabled by default as it should be for motherboards because from the folks spoken with it sounds like there's pressure internally for MCE to be enabled if any of the competing motherboard makers decide to enable it by default because they'll instantly make everyone else look worse if MCE isn't discussed in that equation so they're mostly competing that's what it comes down to we always aim the test with MCE disabled we were told it was disabled with the ultra gaming version f2 clearly that wasn't the case but fortunately the gaming results remain accurate on BIOS version f4 and f5 a which came out a bit later and included 1060 1070 1080 and up performance fixes that were present on BIOS version f2 as for what we think of MCE here we're a bit disappointed that motherboard makers even think about enabling MCE automatically by default at all because in some testing we've done again you can experience blue screens instability and crashes and if you're an end-user who maybe knows nothing or a little about BIOS that's not an option you'd think to disable there is zero reason as an uninformed user you'd think my system is blue screening I'm gonna go into BIOS and then you'd see an option called this makes everything better and disable it there's no reason you would do that because multi-core enhancement doesn't sound like something that causes a blue screen it sounds quite good and if it's on auto why would you change it so we strongly believe it needs to be off it's unfortunate that some of the vendors have defaulted it to on but it does look like most of them are actually moving to default by off by now so later as e3 Sony ultra gaming revisions for example were part of that and all this comes down to voltage tables LLC tables as well under auto setting is not always the best as we've discussed before you'll have to keep a careful eye on this for B and H boards but more importantly for rise in two thousand CPUs with XFR to something we'll be looking at and we have some data on that that we can share later of course the last Intel board we tested was the MSI gt70 eye gaming Pro carbon which has EFI version 7b for three version eleven installed we only needed two tests for this one so it'll be quick with stocks that ends our Cinebench score was fourteen thirty point seven points or 196 single threaded enabling XMP gave us an expected uplift to 1439 that's all we needed to test we did not see MCE option on this board but some manufacturers actually do Auto enable MCE with XMP or have in the past so we wanted to at least make sure nothing similarly sneaky was going on with these III 70 I and XMP this one seems fine moving on to the ACS crosshair six hero this is an AMD risin motherboard and this has an r7 1700 in it we tested with the following configuration first we stock with CPB auto performance by is disabled note that asus calls the feature core performance boost for AMD boards but as core performance boost is to honda fit in charge were calling at ctv we next tested XMP with CPV auto and performance by is disabled then XMP with CP be disabled explicitly and performs by is disabled finally XMP with CP be enabled explicitly in CB 15 bias the results for Cinebench are on the screen now with core performance boost set to auto stock settings without XMP and performance by is disabled we've measured 1400 points for multi-threaded 145 for single threaded enabling XMP but leaving the other options auto and disabled we measured 14 32 points and 148 points boosted from higher memory speeds enabling XMP and explicitly disabling core performance boost we measured leap dot seven points multi-threaded or 123 for single threaded finally enabling core performance boost and Cinebench bias we measured 14 72 points multi-threaded a substantial uplift and 151 points single threaded here are the frequencies with core performance boost enabled we ended up at 3154 multi or 28:15 single with a disabled we're at 3000 multi or 2662 single we always test the core performance boost enabled on AMD Rison platforms but it's not the same as MCE for AMD core performs boost seems to trigger precision boost and xfr-s noted previously which are both stock features of AMD Rison by default and so it makes sense that the motherboards enables CPB by default this is a different feature from Intel CPUs and how they interact with mze we do however leave the biases off as those are motherboard one-offs power does exit spec as defined by TDP but remember that TDP is not equivalent to actual power consumption well you know this power chart what we see is that we measured about sixty eight point seven watts close to the 65 watt TDP again not really the same thing but either way and 26 watts for single threaded this was with CP be disabled with CP be enabled the default option measured seventy nine watts multi thirty-three watt single and just a quick recap for TDP its more measure of how much cooling power not really the best word but how much cooling prowess we'll call it is required to keep a sieve you within spec as opposed to how much power is drawn so the two numbers are not technically the same however they are very close tends to be the case that TDP is within single digit percentages of the actual power consumption so recapping here why does it happen motherboard vendors are competing to try and keep up with each other some of them have in the past enabled MCE by default they're probably some boards out there that still do it but we haven't run into any yet and on the motherboard vendor side it's just not wanting to look bad in comparison so everyone needs to kind of make sure that MC stays off it's really on us because what happens is once a motherboard vendor enables MCE by default as gigabyte did with bios version f2 before they later changed it and thank you gigabyte for doing that what happens is other vendors will see those results freaked out about why their performance looks bad when all they do is make a mother all they do quotes make a motherboard and then enable MCE on their own so that's kind of the concern here it's not an Intel hidden secret narrative it's not Intel trying to control the motherboard makers its motherboard makers independently competing with one another so and ultimately executives at the top looking and seeing their boards lower on the charts and not understanding why for future testing for our own testing we're going to standardize a couple of new procedures to ensure that motherboards and manufacturers aren't unintentionally or intentionally pulling one over on us so a couple things here one we're going to begin taking note and detail in which efi version we use for the CPU and motherboard in the review we tend to do this internally but it's probably a good time to start putting that external for validation reasons and because really the the BIOS makes a huge impact on performance so that's an important thing we now believe that it's basically just as important as listen what GPU we use so we're gonna start putting that in the test tables which will help isolate problems in the future that our BIOS related to going to begin exporting the BIOS when possible and saving it somewhere so that we can ensure we have permanent access to pre-production BIOS revisions because the review BIOS is important to keep around because we need to see if they did anything tricky or different or weird in that BIOS and make sure we have for later fortunately we still had all the old biases from previous flashes but we're gonna save them internally explicitly now to make sure of that for all future biases as well and three we usually conduct a fully automated test suite from synthetics and production workloads including blender we'll keep doing that because it is a slightly accurate you can repeat stuff really easily but we're also now adding in some additional hardware info logging so this is going to be a separate run as stated previously hardware info can impact performance results we are not going to be including hardware info for the test pass used in the results averaging but we're going to be using them to average or to validate rather CPU frequencies voltages and cache memory behaviors things like that which will just be a completely separate isolated run to give us a bunch of data that we can then crawl through validate make sure everything is performing to spec and not doing anything tricky one more thing we're gonna go back and correct some of the old Cinebench charts on pov-ray that use the gigabyte f2 bios and make sure they're reflected with the new data based on what we found here today so that's I think that's about the deepest dive you'll get on MCE so far for the specific motherboards we're looking at today we're gonna keep a close eye on this for the B and H motherboards you could assume coming soon based on the weeks that have been out there and for the rise in 2000 CPUs you could also assume coming soon and we've got a couple of other really cool content pieces coming up this week so make sure you subscribe for those as always go to patreon.com/scishow it helps out directly and go to store that gamers nexus net to pick up one of our new mod mats thanks for watching I'll see you all next time