Gadgetory

if you think Cinebench is a reliable benchmark here's a histogram of all our test results for the Intel I $9.99 hundred K at presumably stock settings the scale is shrunken and non zero as the results are so tightly clustered we had to push them together but you can still see that we're ranging from 1970 CB marks to 2300 which is a massive range that's the difference between a heavily overclocked to r7 2700 and an overclocked 7900 X so 4.2 gigahertz eight cores versus 4.6 gigahertz 10 cores except this is all on a single CPU it's all in the 9900 Kate stock the only difference is that we used five different motherboards for these tests along with a mix of auto XMP and MCE settings the discussion today focuses on one it is considered quote cheating to modify CPU settings via bios without the users awareness of those changes the most common change is to base clock where bios might report a value of 100 point zero zero but actually produces a value of 100 point eight or point nine on the cpu this functionally pre overclock sit but does so in a way that is hard for most users to ever notice before that this video is brought to you by thermal grizzly and their high-end thermal compounds Thermal grizzly makes cryo not paste for high thermal performance and conductivity without being electrically conductive so you don't have to worry about shorting components Crona is particularly good for replacing stock GPU pastes as cryo not is a non curing compound learn more at the link in the description below that's what we're looking at today we're looking at the differences in BC LK and how it affects baseline performance so this is a problem because running tests sometimes they'll come across reviews or user benchmarks where it's not really controlled environment where they might not know that BC LK is different we had this issue with MC e last generation where one of gigabytes board had boards at MC e enabled by default in the original pre-release bios this was later corrected for the released bios and all later iterations MC e was disabled now what is MC e you might be asking if you don't already know multi-core and hasn't we did a video on previously but it's just this feature that's in the boards now you can enable it whatever and it will auto boost all of the cores to the highest single core frequency Intel bins it scores in a way that you've got 5.0 gigahertz on the best core for a single core boost whereas you might have 4.7 gigahertz all core for the 9900 K it has an example and with MCE on it would be used all of them the 5.0 and that's an overclock now that's not an issue this time fortunately MCE is basically gone it's disabled by default on all five of these boards and the only thing now is that where MC e is gone some of the vendors primarily one have become a bit tricky with base clock instead which is another way to pre overclock the CPU so for the boards we have today that we're testing with the ACS Maximus 11 hero we have the gigabyte zi 390 master we have the MSI godlike board the EVGA Z 390 FTW and the asrock Taichi ultimate that's what our test platforms are for today you've got an idea of MCE base clock effects everything so it is what the CPUs multiplier multiplies against to get the core frequency that's a big deal so even a 101 base clock can have a big impact on the frequency because you end up 50 megahertz ahead if you're going times 50 for example and then it also impacts PCIe bus speed it impacts the ring bus the system agent lots of stuff gets affected by B CLK memory frequency as well gets affected by B CLK and we also talked to companies about why they chose the BC LK settings they did so base clock our firm belief is that as all of these boards should be 100 point zero max and ideally that's it just 100 point zero period but that's not what they all do and we talked to the companies that don't do that asked why and got some ads we'll talk about those at the end but before those let's go through the component testing where we look at just Cinebench this time cinder bench isn't that creative a task for reliability and you'll see that in here but it's good enough for some quick runs and it's good enough to enumerate clock so we can see how it behaves and multi-threaded for single threaded so good for that situation we just Hardware in photo log everything constantly and we did our Cinebench runs without any login software then did them again with logging software just to make sure the results for performance were isolated from the logging of the frequencies in the base clock in case there's any interference there we have over 325 runs of Cinebench conducted on these boards so it's a lot of data and we have all that to look for the differences in base clock and when is it considered cheating to increase your base clock this testing primarily looks at BC LK differences as this is a sneakier way to impact performance of parts without leveraging maxed out ratios via MCE something a lot more obvious this generation all five of our boards ship with MCE disabled so that's great news BC L case should be set to one hundred point zero by default for all motherboards for the most part the board's fall within this range or they're close enough at ninety nine point nine nine nine or something EVGA operates at around ninety nine point eight and although this might sound like a small reduction of only 0.2 megahertz it does mean that all core frequency ends up at forty six ninety instead of 4700 mega it's not a big deal but it shows you just how much small changes to be CLK can impact me final speeds remember this is the base number that multipliers work against ergo the difference multiplies this isn't really that big of a deal again ninety nine point eight is close enough and worst case it's undershooting not overshooting so there's no stability concern ebj has some performance overhead they could calab AK if they wanted to though msi really does some screwy stuff here msi maintains a 100 point 8 bc LK when XMP is enabled in either version of BIOS or 100.5 when the board is left to complete Auto settings ie out-of-the-box configuration this means that the CPU runs at about 40 megahertz higher natively at all times that Ram is also overclocked PCIe is affected rain bus and system agents are affected and so on MSI has boosted the numbers in a somewhat noteworthy fashion and appears to be the only vendor doing so in such a sweeping way for this generation well come back to em assign a moment but let's get through the rest of the boards first gigabytes master motherboard runs dead steady at 100 megahertz in all of our testing averaging thousands of cells of data it's about 99.999 865 so we call that one this is what the frequency should be as one hundred megahertz times 47 is 4700 megahertz or 4.7 gigahertz the asrock board is also at around one hundred megahertz and all tests between both the old and new versions of BIOS with XMP either on or off doesn't matter Asus is the next of the boards that shows some questionable behavior but only briefly with XMP on or with settings manually adjusted Asus operates at 100 point zero megahertz PC LK which is exactly where it should be with the full default settings complete auto ACS operates at 100.5 BC LK what starts impacting performance results in funny ways sometimes fortunately this is only for the full auto settings so if you're worried about reviewers who use this board like us it wouldn't affect any of them for the most part or at least it doesn't affect us because we always enable XMP and then configure the timings if necessary this eliminates at most concern of data contamination a good portion of users probably don't know to enable XMP though and so we'll be running pre overclocked unwittingly 100.5 times 47 is 47 24 megahertz the ring clock runs at 43 20 records instead of 4300 system agent runs at 1005 instead of 1000 and single core runs at 50 27 megahertz instead of five thousand mega hertz memory speeds are also affected here we'll come back to this more in a moment but to better illustrate how these BC LK changes impact performance here's a chart of multi-threaded Cinebench scoring across all configurations we should first point out the outlier result at twenty one ninety five points gigabyte z3 90 master motherboard is leveraging MCE to max out all cores to the single core turbo of 5.0 here's the thing and this instance MCE is okay because it was a manual change made by us gigabyte does not try to trick MCE into being enabled under auto or x and b configurations this is the correct way to do it we think they're offering an option but not sneaking it into a state of being on and so we commend gigabyte for this arrangement asus does sort of the same thing for MCE where it disables it completely unless you tell it explicitly to be enabled at the low end we have the acs maximus 11 hero with XMP enabled and no selected when prom for enhancement options after enabling XMP to performance is actually worse than depicted here because the CPU is hitting a snooze power limitations on ICC Macs on occasion we saw a power throttling from 10.8 amps to 8.8 amps or 130 watts to 105 watts under some Cinebench multi-threaded workloads when running some passes it doesn't always happen but a power throttle is enough to cause a lower average score and the worst AC score here because of the power throttling wasn't 1976 points just for reference if we highlight the a C's Maximus 11 hero with Auto settings you'll see that it actually outperforms the board with XMP and no selected to the prompt if Asus is following proper power throttling to match Intel's ICC Mac specification they should probably also follow it under complete Auto settings just with X and P on we find this choice of behavior odd to say the least selecting yes to the prompt results in a score of 20 56 points which is markedly higher than the 1976 to 2030 range of extreme variants that we had with the no selection option and roughly equivalent with the auto settings that's because the auto settings are using a 100 point 5b CLK so it's able to achieve the same performance as the memory under XMP finally msi clearly holds an unfair lead Emma sighs BC LK of 100.8 is resulting in a 2087 score in Cinebench multi-threaded which is significantly ahead of other boards that use lower base clocks the next closest board is gigabytes of master motherboard which runs a be CLK of 100 point zero or zero zero if you want to be really specific power consumption is another major part of the story and this one doesn't carry the questionable variance and accuracy of a Cinebench in general for these tests we were looking at clamped power draw down the EPS 12-volt cables for each board EVGA runs the lowest power consumption a consequence of also running lower voltages the EVGA FTW operates a 120 watt EPS 12-volt draw during both the auto and XMP tests gigabyte is adjacent at 125 watts with all of the asus tests falling within the range of 130 watts let's again knock out the outlier and address the one 96 watt draw number that's from enabling MCE explicitly by the user on the gigabyte board which again is done manually you could do this on the Asus board as well if you wanted to power consumption goes way up to support the 5 gigahertz all-court overclock as expected and single core power draw also increases as you can see here MSI draws significantly more power than EVGA gigabyte and A's use motherboards borrowing the MCE results pushing upwards of 140 watts with a baseline configuration this is a result of higher voltages sustaining a higher base clock higher system agent clock higher rainbows higher PCIe speeds and still also contributes to temperature variance in results we've talked about that part in the past you can see previous videos for that even still as rocks board is aggressive in its power pulled down the EPS 12-volt cables with a 100 bc lkm the word clocks than msi asrock is blasting voltage unnecessarily and boiling 145 watts to 160 watts during tests the updated bios is better by 12 watts but still overkill when compared to everyone else and that includes msi which is running a higher b CLK than anyone else either this board should be closer to EVGA and gigabyte for the asrock board that is and it's over drawing power in an unnecessary way after all that then here's the thing msi is the only one that really steps way out of line but that's a pretty big jump when you consider it's multiplied by whatever the multiplier might be and it's not huge in so far as performance it's a higher performance result for sure 2080 cv marks first is everything else below it but it's not game-changing it is however enough where Emma sized board looks artificially better in ways that it doesn't really deserve now we're not saying which board is or isn't the best here but what we are saying and we have different content for that we are saying is that Emma size board these ether 90 God like maybe it is better in some ways but the ways in which it is demonstrated as better here with higher points in Cinebench are not because of the board it's because the CPU is functionally overclocked and so that kind of creates an hour view an unfair stack where looks like the point structure the point distribution favors MSI's board even though it's actually just favoring a higher clocked CPU that's really the difference so that's a point of concern and again it's not a big difference not like we're looking at 2500 versus 2000 but it is a difference and motherboard vendors are constantly struggling to try and prove that they are providing a value over their competitors so that's a big point that should be made where if someone's not at 100 if they're not at spec or at least not at what is expected to be spec then yeah they look better than everyone else and maybe not for reasons that are resultant of the motherboard quality there are other concerns with this too because so there's a good reason to do a higher base clock maybe an MSI we talked to them about this maybe you want better support for high frequency memory and so MSI so that they set their base clock higher we wanted to talk to them about it figure out why are they doing this and I'll just bash on it without understanding why so MSI's says they saw their base clock higher to more easily support higher memory frequencies like 40 to 66 megahertz for example which is very hard to sustain so there's there's good basis there but here's the thing the trouble is the math doesn't really work all memory is divisible by the same dividers and it can be multiplied by the same multipliers so against PC okay so if you got 42 66 megahertz memory and you divide that by 41 point 33 then that equals a 103 base clock not 100.8 if you do it the other way let's say you do it forty to sixty six divided by forty two point six six well that equals 100 so neither of those is 100.8 so we're not really sure where they're getting 100 point eight because it just it doesn't work for this math the clock gen we understand might have something to do with it but the end of the day there is good science I guess to saying that increase in your base clock increases your high in memory support because you can reduce the memory multiplier where you can run 103 times again 41 point 33 or something instead of 100 times 42 point 66 that might be more stable for you but that's not what's happening here and it should be done probably on a per kit basis if they can store all that in the board's flash so that was homicides reasoning they said they use it for qvl guidance for qualified vendor list guidance for Z 390 at least they have a reason we don't think that's a good reason they also said after we kind of pushed them on it a bit that well it's better performance and so of course our response is right but that isn't that doesn't that seem wrong is that unfair to achieve better performance that way and it didn't seem like at the people we spoke with at MSI really either understood that line of thinking or agreed with it one or the other but it seemed more like a a disconnect and what we were concerned with versus what they were concerned with so to MSI's credit yes they are correct it does increase performance and that's the goal that's fine we just think it should be openly declared somewhere on the product page on the Box in BIOS BIOS it says 100 point zero right now even though it's 100 point eight so BIOS isn't it's kind of misleading to so we it's okay it's a pre overclocked product if you want to they do it for GPUs all the time it's just it should be openly disclosed and we mentioned that to MSI and they're looking at updating their website and their pages for the product to state that so that's good that's what we want ultimately is for it to be openly declared so it's not like it's a secret that people just don't notice because why would you notice 100.8 most people who aren't reviewers will never notice that and so then their power consumption is suddenly way higher than what all the reviews show and they're confused why so that's a kind of big deal with motherboards where these small changes in base clock that shouldn't be there greatly affect numbers so what I want from you all is I want to know post your comments below try not to let too much of the musings influence your thoughts just go by the numbers we showed what I want to know is this post a comment and say if you think they should all be 100 base clock or if you think it's completely fine if they pre overclocking because that's basically that I mean that is what it is it's not basically what it is it's literally what they're doing is pre overclocking the base clock so if you think that's fine that's totally cool just please let us know why and if you don't let us know why as well because what's gonna happen is we're gonna send this all the motherboards manufacturers features in this video and they can read through it and see why you think one way or the other and not have to hear it from me they can hear from however many tens of thousands of you instead that would be far more impactful than just coming from me because I just complain all the time anyway so other reasons we think that this might be a problem it's kind of a slippery slope you get back on that slope towards MCE being on by default that's a bad thing we don't want that that influences performance it's cheating and people don't notice it and the power consumption is higher and if you have PSO DS we've seen it because of pre overclocks like that not with 100.8 BC okay probably not gonna happen but it's a slippery slope argument and and one which could actually become true because it has in the past for this situation for Ace uses part a CS agreed that base clock should be 100 megahertz on auto settings and weren't quite sure why it wasn't so they are looking into that I believe is my understanding now so that's good too we think this is a problem because manufacturers will go into an arms race by pre overclocking means that they probably shouldn't it could be bad for stability if there's power consumption numbers way off and ultimately you end up with reviews where now your different reviewers you watch using different motherboards if they don't notice this stuff and that's totally fair because it's kind of sneaky some of it then the results will be different we didn't experience that fortunately we use the AC board we selected XMP so our base clock was 100 and all those tests but I mean look back at for example the principal technologies testing they used em aside boards those red 100.8 so all the results can be boosted by another 40 megahertz on the clock and that is somewhat substantial in some tests so anyway that's it for this one let us know what you think this is kind of it's hard to blow it up and make too big of a deal out of it because the numbers don't look super inflated it's just it's a really small incremental way to create what we think are unfair differences between motherboards when it's actually the CPU doing the work I mean it effectively be like if you put the CPU in that board set it to five points you gigahertz and say that it's better because it was preset to 5 point 2 Giga it's not really the motherboard doing the work there so that's it for this one let us know what you think subscribe for more we go to stored I cameras access think of one of these shirts this is the limited edition shirt we're running low on stock now so we probably have enough to get through the next livestream or two and then that'll be it they're gone we're not reordering more so grab one on the store if you want one and subscribe as I said thank you for watching I'll see you all next time