Gadgetory

today's topic has come up in the past when we reviewed the gigabyte gaming 7 motherboard for the 7700 K we found that its auto voltage curve ramp to the CPU and to unreasonably high temperatures for no performance gain when compared to competing options even manual options gigabyte later fixed that auto voltage and we now use the gaming 7 for GPU testing regularly and like the board a lot but we're exploring a similar scenario with the i9 7 960 X which also applies to certain 980 XE and 7900 X we'll be demonstrating how motherboard and BIOS revisions can heavily impact the interlinked thermal power and voltage values when using Auto settings before getting to that this video is brought to you by synergy the software that lets you share a keyboard and mouse between multiple systems if you have limited desk space and multiple computers to command synergy removes the need for separate peripherals or a KVM and works as over the network software use our link below to get 50% off the home or provision with SSL the point of this testing is to show well it's really to revisit a few things so we did all that deleting testing with the i9 CPUs and the testing there was it was either overclocked and over volted or it was 3.6 gigahertz so not really overclocked if at all depending on CPU and a fixed 1.15 voltage ID the 1.15 VI D as noted in those content pieces was higher than necessary but it was chosen because it would stay fixed at that voltage without moving at all and that's what you want when you're doing comparatives between two thermal interfaces you really don't want otto taking control of things because who knows what it's gonna do who knows if it's gonna respond differently under different thermal scenarios or if it just decides to do something different one time versus the other which happens basically every time you run a test without a voltages with the exception of the gigabyte board we'll be looking at in a moment but the point though is to revisit some of that with Auto settings not for purposes of showing thermals necessarily as the the main item we're looking at but more for purposes of showing how a motherboard and its auto settings with the voltage look-up tables can pretty significantly impact your performance in a lot of ways it can impact power consumption directly it can and will impact thermals directly as your power output in watts goes up so two goes up the thermal output or the measured thermals of the CPU because you have to dissipate more heat and it affects a lot of things so hopefully this educates a bit on doing some manual tuning yourself whenever you buy any processor because Auto voltages are often a bit excessive but it's to make sure they can support the speeds and the advertised specs across all CPUs and motherboards because there's some variants of course CPU to CPU so the boards were looking at we have five today well we have three boards with five configurations across them so the Asus Rampage six extreme was our primary board for the review the liquid metal burst Tim testing and everything else we've done so far with the 79 80 and 79 60 CPUs and then we also on that we have the zero eight zero two BIOS revision which is the latest EFI from Asus and this board was joined by the ACS prime X 299 deluxe motherboard a lower-end board and that uses both the pre 79 ATX e launch efi version and the newest 0 8 0 2 revision the gigabyte board the gaming nine joins this one and for that we're using the latest efi revision f6 and the launch 7900 x efi that was sent to press i don't have the specific number on that it was a press launch but the point of doing these is the 70 100x launch efi for the acs prime board and for the gigabyte gaming 9 board basically came out before the 79 ATX he really existed the public or was finalized and that means there's no profile tuning for that cpu the 18 core or the 16 core CPUs it was built for the tank or so we should see different results because the BIOS should respond differently based on if it has that voltage lookup table for that specific CPU we're just gonna be focusing on blender for most of these tests we did use prime 95 for everything as well it's a great test but prime does a lot of stuff that is kind of difficult to control for once you start getting into this kind of testing so it does power cycling which means that in some head-to-head tests you end up with occasionally way more amperage going through on one test past and the other based on a couple of different behaviors of the software so we did run prime but we ended up just going with blender for everything because blender once it ramps up and starts going nine times out of ten it's a perfectly fixed pretty much constant current going down the EPS 12 volt cables the load doesn't really change as long as the project file stays the same the other one out of ten times we will see really spiked loads or higher thermals but you just develop an eye for that and then eliminate those results as outliers as necessary for this type of testing so we're using the Intel I 979 60 X for these tests and we're using the liquid metal modern version can't really go back on the head for the most part so we're stuck with it and we're using it for these tests and we're gonna start with a voltage id behavior on auto settings between the three motherboards with the five different EFI versions voltage ID behavior directly influences everything else so using all auto settings with XMP one and max fan settings those are the only changes we made the first gaming nine tests with pre-launch BIOS pegs VI D at roughly one point zero one volt and that's again VI D for the entire test this never really changes it's a perfectly flat line for the most part the Asus Rampage motherboard fluctuates and bounced around between 0.91 VI d and 1.0 VI D generally sticking to zero point nine six to zero point nine seven for the entirety of the test if we're looking at the more frequent numbers we only started using this board with the seven 960 X launch and it had the latest zero eight zero two BIOS already so the board has a tune for the CPU and seems to behave as you'd expect the ACS Prime board with the original efi also bounced around a lot at least a lot more than the version with the efi version zero eight zero 2 and we averaged 0.89 to 0.96 a PID with the pre-launch version and 0.96 to 1.0 with the zero eight zero two the most noticeable jump is the gigabyte gaming nine EFI update to version f6 from the 1700 ex reviewer EFI version we've moved from one point zero one to VI D to a range of one point zero six five to one point zero eight voltage ID depend on which core is measured this has a more significant impact on both power and thermals than any of the previous tests for power testing we're measuring at the EPS 12-volt rails which means that the bulk of our power consumption is the CPU some other board components will also pull down this power cable but nothing close to the CPUs consumption that said there is some error involved in changing motherboards for power tests even when measuring a TBS 12 volt cables but not really that much comparatively the rampage 6 extreme on auto and 796 TX are measuring about 217 watts at the EBS 12 volt cables again this is under auto conditions and with a blender workload so it's not the most power we could pull but also not the least the power consumption will change based on the application used and as we're showing here the motherboard issues Prime board with the first EFI revision which ran over all lower voltages of 0.89 to 0.96 versus the average 0.91 to 0.98 of the ram-paige managed to consume about 206 watts on average updating efi to version zero 8:02 moved our VI d to 0.96 to 1.0 as the range which consequently shifted power consumption to 222 watt that's just from an EFI change on the same board the gigabyte gaming 9 with its original efi had a higher sustained voltage idea 1.01 - constantly which produced a higher power consumption and the gaming nine with its efi at version updated measured at 230 watts so that's now 30 watts more than the lowest board and lowest efi and about 15 watts higher than the previous three configuration average we ran two tests of the gaming nine with version f6 since the results were so much higher in power consumption than the others and we wanted to double check them the tests had some variance but we're overall much higher in power assumption anyway we were between 255 and 268 watts depending on which test and that's now 50 to 60 watts higher than the lowest board on the stack which shows just how much auto voltage tables can impact results as precisely why we fix voltages for things like comparative testing between Tim and liquid metal but obviously you'd still want to test Auto for the motherboard reviews or just to figure out what the CBO and board will do in a user scenario without tuning thermally voltage of course impacts results directly we have a tolerance of plus or minus 3 degrees Celsius for ambient and the thermal application differences here but can still see gaps in performance the ram-paige 6 cooled the best some of this is because of the spacing of the V RMS from the socket and landed around 40 to 44 degrees Celsius gaming 9 with launched EFI plotted at 42 to 46 degrees Celsius again all Auto settings with the ACS Prime testing at around forty six point five degrees Celsius for zero eight zero to the gigabyte gaming nine with the f6 EFI plotted at around 52 51 degrees C for a 10 degree gap versus the lowest temperature on the blot so that's pretty big deal that's what happens when your voltage lookup table when it's set to auto and the motherboard vendor is a little aggressive on the voltage table that's what you get higher wattage consumed higher thermals as a result and the performance is the same as long as all the cores can sustain their frequency it doesn't matter if you push more volts through it it's just burning power for the sake of burning power so there's an argument here that by taking a conservatively high voltage value VI D or V core whatever demand what you're working with a conservatively high value means that probably you're gonna support every die that comes off the line whether it's low quality or high quality sure but also there's kind of a limit here to what's reasonable so gigabyte once again is pulling a little bit more than they need to it's not nearly as bad as the gaming 7 Z to 70 motherboard a little bit higher but not terrible still though there's 10 degree difference there so if you get that gaming 9 board or already it we would recommend manually dropping the Vig setting until it stops being stable until you're either clock dropping or crashing and then increase it a little bit it should still be lower than Auto and you'll have lower thermals for it and lower power comes on it for it and both of those things are good especially for the life of your liquid cooler if you have one but the point isn't really to look at the motherboard specifically although would it I mean there's a note for gigabyte if you're watching this video just again look at the same thing as you looked at for Z 270 but that's not the point the point is motherboard changes matter a lot they always have but we just figured we'd show it on a CP that's particularly power-hungry because it exaggerated s-- the results a little bit and also kind of as a reminder of hey this is why when you're testing something like a CPU cooler a thermal interface or anything in between like a shim any kind of cooling dissipation test that does not involve a review of the CPU specifically should be done with a with the same motherboard every time with a fixed voltage a fixed frequency all manually set ideally a little bit conservatively high on the voltage like gigabytes doing here just to make sure nothing changes so that's why we do that for those tests but yeah depend on board you probably want to manually tune that voltage down to see where it's the it's still stable with the least amount of voltage go into the CPU if you're using it in the stock settings without any overclocking another note here so I want this quickly show some of our auto tests temperatures for the 79 60 X a little bit for this around 80 XE we we never did get to run all the liquid metal tests because I mean I ran them and then left to go to do Linus's thing but we've got those set to auto and before showing the charts didn't show these in the original review because again they were set to auto so the voltage changes a bit from tested test which means that it's not a fair or valid comparison if you're trying to test liquid metal advocacy versus Tim because the voltage properties are changing so you're not testing the interface material anymore you're just kind of testing the temperature at whatever voltage it decides to run out for that particular test pass and it will change pass to pass regardless of the interface material but we can show those charts so this is what the auto temperatures look like will show Prime we can show blender as well at some point basically it's a lot lower than with the fixed voltages and frequencies of course as you would expect but this demonstrates even still even with the variable voltage setting that's Auto and kind of bouncing around we're still seeing improvements with liquid metal if you overclock you should definitely put liquid metal on there because that's shown in the original content piece I once we were overclocking the 4.6 4.7 gigahertz with 1.2 2 to 1.2 for voltage ID 7 900 X I think was at one point one seven five is pretty pretty good chip even with those settings that were especially with those settings that's when you should start caring about either deleting and replaced with a better Tim more on that soon or deleting and replacing with liquid metal because dropping even 10 degrees which was about the lowest we dropped at that point with the overclocks even dropping to 10 degrees is a huge difference between your CPU being in the 90s 4 degrees Celsius on an open-air bench where ambient is 24 and being in the 80s and you might ask who cares if it's in the 90s or 80s well who cares is your liquid cooler and the manufacturer because they're not going to want to fulfill your RMA when it dies because the liquid was running at 60 degrees Celsius which is above or at spec that's who cares so it helps to bring down the temperatures it's unfortunate that you have to do it yourself if you're overclocking but if you're not overclocking you're fine you don't need to deal with the thing if you are and you're using a CLC that's either well so a few things here one either using a huge CLC like a 360 or a 420 because you need to deal with all that heat and you're like what temperatures are probably bit better and your fan noise is higher because you're gonna need it or using a smaller CLC and you're pushing the limit of what the pump realistically should be handling even if your CPU looks like it's okay so in both those scenarios you can improve things by D letting changing the interface and either you lower your noise levels and improve your cooling which is probable or you can step down your cooling solution to something cheaper maybe a 240 maybe a 280 with slower fans and still be better off than 10 but this isn't new as we showed this in the review and then the part one of the thermal thing but I mean the point again is motherboards impact things a lot do some manual tuning on it if even if you're not overclocking if you're not overclocking at all thermally you're gonna be in better shape than what we show when overclocking or went over bolting but even still even with that variable voltage we were plotting improvements with liquid metal it just doesn't look as big because the voltage is bouncing around so who knows what it's doing but that's all for this one as always you can check the article in the description below gamers Nexus thought that for that if you prefer a reading or we would recommend checking out parts 1 & 2 of this which we'll put in the end slate you can go to patreon.com/scishow helps that directly as always it is a big help at this point or store that game is next to Stein I just pick my shirt like this one which we've just restocked this is the graph logo and I think we have it in tribal ends as well at this point so thank you for watching subscribe for more I'll see you all next time