Gadgetory

remember talking about once in a Content piece how we disabled multi-core enhancement for a CPU benchmark to make sure things were fair and one of the higher uploaded comments was about why did you do that that's unfair it makes the CPU better you hate the company that that makes the CPU well the thing is that I explained then and we're gonna do it more depth now molten core enhancement is not a CPU feature it's a motherboard feature and a lot of people don't even really seem to understand what it does when they criticize the disablement of that feature so we're gonna go over that today we're going through the coffee lake turbo boost table just because it's kind of a good time to do that and we'll be talking about the different features of aces and gigabyte what they do and when is it considered stock versus when is it kind of just cheating at the benchmark results before that this is brought to you by the EVGA 240 c LC which is a $120 MSRP closed-loop liquid cooler the EVGA 240 c LC has an RGB illuminated pomp plate uses a thermal probe within the lower pump chamber for liquid temperature monitoring and allows customization through software learn more at the link in the description below so multi-core enhancement then first of all it's called a few different things but basically means the same thing on every platform this is a feature that is either auto or disabled on ACS boards it's Auto enabled or disabled on gigabyte boards or some variants of that on MSI and asrock and the feature more or less just says lock the CPU and its maximum turbo state at all times on all cores regardless of whether the application is single dual quad or six threaded and in the case of the 8700 K to give you an example the turbo tables for this processor just to go through them if you're running six core applications six cores engaged that will output a 4.3 gigahertz speed per the Intel spec so that's that is as stock as stock gets now the word stock it gets a little confusing with CPUs and motherboards because the to directly influence one another as far as the CPU is concerned as defined by Intel the specification is 4.3 gigahertz six core 4.4 gigahertz for four to five core engagement 4.5 gigahertz for three core 4.6 for two core and 4.7 gigahertz for single core or single threaded engagement with the 8700 K and E has something similar they have XFR extended frequency range which more or less dictates that with the more thread limited applications you get maybe an extra one to two hundred megahertz out of the clock which is why with the 1950 X sometimes if you do an all core 4 gigahertz overclock performance is actually worse than out of box because out of box with a limited thread utilization application you end up with 4.1 4.2 gigahertz whereas your book gigahertz all core overclock is worse so that's where these come in just to get everyone up to speed the 8700 I seven CPU four to six cores 4.33 core is 4.4 gigahertz to is 4.5 single is 4.6 gigahertz the 80 600 K 5 to 6 core turbo is 4.1 gigahertz speed 2 to 4 quarter Bo 4.2 single core 4.3 and then down the line the 8400 cuz we also have that one that's five to six at 3.8 2 to 4 at 3.9 and single at 4 gigahertz the i3 CPUs just have one speed there's no turbo so that's the turbo table as distributed by Intel to its motherboard vendors the motherboard vendors can then take that information and change it it is up to the vendors to decide whether their motherboards auto multi-core enhancement defaults with auto meaning on or off and what it does is again locks the CPU to its maximum turbo speed regardless of the Intel specification this isn't breaking any rules with Intel as we understand it it's just that's they can do it if they want to and so then of course from morgen standpoint you disable that cuz it's not a stock out of box CPU behavior you were no longer benchmarking the 8700 K as the stock 8700 K you can't call it in your chart for example 8700 ka stock if it's got multi core enhancement on it because when you start running games that are let's say six threaded applications or something six core applications rather than running four point three gigahertz like it should it might run four point seven so at what point does that become disingenuous at what point does it seem like it's sort of cheating at the results and from a motherboard standpoint if as a vendor you realize you can push an extra three hundred megahertz with this multi-core enhancement then it makes your board look better maybe ten percent in some cases then the competitors because they might default to the Intel spec so that can get confusing wanted to point that out to explain a few things one if you see a disparity in some results for CPU benchmarks where for example we talks about our board hitting four point four and a lot of applications with the gigabyte ultra gaming Z 370 that's because that board follows the Intel spec I spoke with the gigabyte about this and they're pretty hardcore about being in line with the Intel spec and not playing any trickery with the multi-core enhancement or whatever it's called unless the user enables it so one thing that Asus does is a see also defaults this option to like it's technically Auto and if you enable XMP it will ask it'll prompt and say do you want to enable multi-core enhancement and if you're not paying attention or you just think hey that sounds pretty good and hit yes it'll do that 4.7 gigahertz all cores so that can well that does influence performance results which we'll go through in a moment so the two discussion pieces here are how is performance influenced in terms of a simple Cinebench score power consumption and thermals and then is it better to default to on or off from a user perspective and from a benchmarking perspective I'm just gonna tell you now off is better you benchmark stock stock to stock it does not change there's no stock modified division of CPU benchmarking either you call it stock or it's overclocked so that's kind of how I look at it but let's get into these so Cinebench first off gives us a pretty quick look at this we're gonna be focusing on the combined average results for enhanced multi-core on versus off with two different motherboards the difference from all quarter bow enhancements is immediately visible in Cinebench disabled with the clocks locked to the four point three gigahertz all core Intel specification our multi pass average sits at 14 48 CB marks enabling multi-core enhancement options boost that to about fifteen seventy eight marks for an eight point nine percent performance uplift as a result of the four point seven gigahertz all core turbo that gain isn't free though there are a couple of things attached to one for the next set of tests with thermals and power we actually wanted to use blender for something real world the problem is with the Asus z3 70 board using the auto multi-core enhancement setting we actually were crashing in blender so it would hit 4.7 gigahertz all core per the ACS setting but something was off in either voltage or just something wasn't quite configured properly when you let the motherboard do its thing and that meant blender crashed prime didn't but blender did and blender kind of has that effect lately we noticed that it's it's a bit more voltage intensive or perhaps a better word sensitive voltage sensitive than prime prime is intensive but it's not quite as sensitive as some of our blender tests so that means that enabling this straight away you're in a situation where as a user if you enable XMP and then say yes I would like this multi-core enhancement thing some applications might crash it's less stable which is kind of why it's better to ship with it default it to off which both ACS and gigabyte do though as a user it's pretty easy to to just hit yes to all the boxes that pop up so be careful of that I said the gain is not free in terms of that 9ish percent performance uplift by enabling multi-core enhancement less than stability the non-free aspect is the power consumption and thermals so we're measuring at the EPS 12-volt rails here this is not wall power and is representative almost exclusively of the CPU power consumption with the complete stock Intel specified configuration power consumption measures at about 102 watts during Cinebench testing with the Asus board enabling the 4.7 gigahertz forced all core turbo pushes us to 145 watts a substantial 42% increase in power consumption for about a 9% increase in Cinebench score this thermal test we're about to show uses prime95 28.5 with L FFT is to quickly burn in the CPU blender was again crashing so we couldn't use that as we wanted to and another note here this is after we've deleted an applied liquid metal to the CPU and just to address the question yes all the gaming and production tests were done and the stock configuration before deleting but now that we're into thermals obviously were deleted at this point this overtime plot shows the temperature behavior as power cycles ramp up and down enabled multi-core enhancement was placing us at around 57 degrees Celsius or 50 degrees Celsius with enhancement disabled meaning the 4.7 vs. 4.3 gigahertz difference was 7 degrees Celsius but it's not because of the frequency alone this also brings in a voltage change because clearly thermal behavior is influenced heavily by voltage tables assigned to each enhancement or configuration and ultimately it is volts times amps gives you power so with the ACS maximus board Otto voltage ID plots high in prime95 when using multi-core enhancement measuring at one point for two to one point four to six voltage ID again on the Maximus with enhancement enabled reverting to intel's stock out of box spec but still with XMP puts us at one point two six voltage ID average across all cores as for the three hundred megahertz lower clocks 300 to 400 anyway it's probably about four point three versus four point seven gigahertz and that's where our power and thermal difference primarily comes from so hopefully that explains the turbo behavior with coffee lake bringing this up again because clearly a 9% difference in Cinebench is pretty big you can definitely see that and for anyone who buys the cpu and feels like their performance doesn't match something they've seen online I would suggest checking that option first the next question though is with this option there does it make more sense to enable it by default or to disable it by default we're disabled might mean Auto and what we generally see motherboard vendors do now is default to an auto state which is off and then if you enable things like maybe change a multiplier that tends to basically just enable it overrides everything once you start playing with multipliers if you enable XMP on some boards it'll ask to enable it and what motherboard vendors want to know from you after speaking with some of them today is do you want this type of feature on or off by default I have my own opinions on it I will share them but just pause here and leave your comment below answering the question of enhanced multi-core as a feature boosting your clocks to all core turbo-bean maxed out given the data we've just given you should that be on or off by default by the motherboard vendor so my opinion of it is it should be off because this is a feature that in blender with the Asus board is clearly unstable we're crashing can't even really get the render going and yes it's faster it consumes more power but it's also crashing so it's not really a stable product and if it's on by default a lot of the people who buy these types of boards aren't going to go digging through BIOS and finding options and names things like Auto multi-core enhancement because why on earth would you disable that if it were on by default it just it isn't something that a non savvy user might think of and when i say non savvy i mean this is still someone who probably built their own system but it's it's really easy to lose sight of just how many people go into BIOS and it's not that many you get people who maybe enable XMP and that's about the extent and that's perfectly fine and perfectly normal and I was there too but because that audience is so big I would say you appeal to the stability of the Intel specification and run it stock the next point to make from a review standpoint although motherboard vendors don't make motherboards for reviewers specifically from a review standpoint you really want to test the CPU as it's made by Intel or AMD you don't want to test the CPU as it performs with the motherboard that's a motherboard review not a CPU review so it starts really screwing with result and makes it hard to compare numbers between your own tests if you change motherboards or vs. other reviewers and that leads to confusion in the market either with readers with reviewers just in general confusion so my vote goes to off but they were curious about your thoughts and so leave them below that said ultimately just a really quick thing you know it's it's 90% better performance with it on and Cinebench for like 40% more power and some instability so just if you're wondering how turbo works with Intel now that was the answer in the very beginning where six cores different from four to five different from three two and one they all different speeds Andy is doing kind of the same thing but less extreme they're just doing XFR and turbo which makes it a bit easier to understand Intel's doing it all the way down from I mean just every single core account basically has a different turbo and that means they can min max their performance more it also means it's a little bit more confusing so hopefully everyone can keep that in mind when you look at your CPUs performance and you're like why is this running at lower than spec speed it's because the spec is probably different based on what application you're using so that's all for this one as always you can get at patreon.com slash gamers Nexus 2 helps out directly store dock gamers Nexus dotnet to pick what shirt like this one subscribe for more I'll see you all next time