Gadgetory

today we're going to show you how to overclock a CPU to death so basically what we're going to be looking at here is how increasing SOC voltages or graphics voltages as it's called with ap use can result in voltages that are deadly or at least could degrade the integrated memory controller or other components on the CPU this is something that's very easy to do we talk often about how overclocking is of course at your own risk well there's a reason for that and even with guidance posted online by overclockers by media outlets like us by AMD or Intel official companies even with that guidance it's still possible to end up in a scenario where you input a voltage number that was suggested in one place but on your particular BIOS on your particular motherboard whatever it could end up being damaging without you even knowing it so we're gonna talk about that today two things we're focusing on are showing how the SOC voltage through the socket through the back of the motherboard is actually often different from what's measured in software like hard R and fo and of course if you were at the max of what's safe and hardware info but this is higher reading it off the real board that's where problems could arise before that this video is brought to you by thermal grizzly makers of the conductor hot liquid metal that we recently used to drop 20 degrees off of our temperatures thermal grizzly also makes traditional thermal compounds we use on top of the IHS like cryo not and hydronaut pastes learn more at the link below let's start with a basic example one of the first things that anyone asks including us with overclocking is what's a safe voltage so for most consumers you might go to Andy Intel Nvidia or come to us we go to people like their Bower or build Zoid or any of them and then hopefully between them and between the companies we can figure something out but at the end of the day it's all kind of just you're relying on hoping that the number is correct because without really testing it you don't necessarily know where things die and of course you know if you pump 1.5 volts into a CPU it will be bad but the question is more of is one point two five okay is one point three too far because these are only 0.05 apart but there could be degradation over a period of six months for example which is of course very bad and you don't want that so with the Raven Ridge ap use as an example conventional wisdom with Rison suggests that a safe SOC voltage is one point two here's where the question comes in is that one point two as in the number that I type into the interface rise in master bios or is it one point two actually going through SOC on the board and through the SOC vrm with Raven Ridge a new suggested safe SOC voltage depending on which company you asked could be as high as 1 point three volts but again if we're talking one point three typed into the interface versus one point three in reality those numbers can be very different to give you a quick example we have three tables that we built took about 16 hours we'll put them on the screen briefly and then we'll go back through them very specifically later and in these three tables one thing we did was check the different input voltages SOC and APU graphics voltage versus what's measured on the back of the socket which you can measure with a simple digital multimeter and in some instances putting in a number of 1.3 which is a new supposed potentially safe SOC voltage again depend on what company you asked putting that number in in some cases can equal about 1.3 in other cases it could equal 1.4 if it's 1.4 and especially if it's sustained you've just condemned your integrated memory controller to death or more likely degradation it's not like it's gonna up and die at 1.4 but what might happen is over a span of six months or it could be 10 months over a period of time there will be degradation in what frequency be in this case and integrated graphics chip can sustain at its current voltage as it degrades you might have to increase voltage to a point that it is no longer safe or even less safe or you decrease the frequency and these are things we don't want so go through a couple more notes here before starting I'm going to show you how the discrepancy appears between the back of the socket and software hardware info arising master and then we're going to talk about some of the tables that we spent a whole day creating with just three motherboards only three and how much they differ one to the next in terms of V droop or over voltage or under voltage or whatever so some very quick things your V group can require higher LLC or load line calibration levels to stabilize the input voltage closer to the configured number this is one of the easiest ways to ensure the number you type in is what you're getting as opposed to maybe some auto LLC's will actually really under deliver on that voltage we've had this problem to pass you type in 1.4 and reality are getting 1.3 so load line calibration can help with this if you're not sure what that is and you want to know more than what I've just told you we have an entire video on that done by a hardcore overclocker if you use this allen tears names build Zoid so we'll link that below and that's that's one of the things we need to know LLC tables on boards can differ so one board to the next it's not gonna be the same that means if you follow a guide for a board you don't have you could kill something damage it or you might just end up in the best case with it unstable overclock because voltage isn't high enough to things specifically to Raven Ridge here before we start apu graphics voltage is a number that you'll see rise and master has it for example so rise and match recalls it AP graphics voltage they also have an SRC voltage these are more or less the same thing so ultimately AP graphics voltage and SRC voltage are going to the same place they're controlled by the same vrm it's the SOC BRM it's typically the well it's always the more limited phase count v RM on the board of the two main V RMS for AMD boards so when you adjust graphics voltage you are adjusting SOC voltage the same rules of safety apply to each even though for example you could type in a very high number it doesn't mean it's same so that those two numbers are basically the same now we have a kind of a hypothesis on what the difference is why there would be two different inputs for them and based on some limited testing on a couple of motherboards our current theory is that AP graphics voltage on some boards is instantiated specifically when 3d clocks are called whereas the soc voltage is instantiated when 2d clocks are called so if you have numbers of 1.14 SOC and 1.2 for graphics then you launch 3d mark 1.2 is what should be being referenced however if you're doing something without 3d clocks then it should be that as I see voltage however this theory only really worked on one line of motherboards from one vendor because graphics voltage actually doesn't work at all on one of the other boards we test it it does nothing and then on the third board it seems to stack with the SRC voltage so you would do 1.2 SOC 1.2 graphics and you'd end up with one point 4 volts SOC measured at the back of the board so it all depended on the vendor but let's get into the testing now this is going to start with a demonstration on a gigabyte board this is the gigabyte gaming k5 it's one of the board's we tend to prefer for rise and overclocking however it does have some limitations and one of the things that gigabyte commonly does is they tend to push a bit more voltage through everything than what you've asked for now that's not necessarily a bad thing other vendors push too little so at the end of the day all that matters is that you were aware of the behavior of the particular board so the fact that it pushes more isn't inherently bad unless you're not aware of it then it could be very bad so I'm gonna do first we just have everything set up auto right now I'm gonna launch 3d mark on the host system so for this first demonstration on the gaming k5 we're gonna use the auto voltage settings in BIOS and just these rise in master for sake of ease and demonstration and all we're gonna do is set a super achievable 1,300 megahertz clock so we're not really overclocking it right now but that's not the point so we'll go 1,300 it does in increments of 50 and we're going to set a 1.1 SOC and 1.1 graphics voltage so both of those are 1.1 and we just need to put it under some load so for SOC voltage right now and hardware in 464 we're reading about 1.0 volts point-9 to 1.01 volts and this is with a one point one configuration and risin master for both SOC and the APO graphics but what we need to do is check the real voltage at the back of the socket so we can do that with a multimeter pretty easily and all I'm going to do is stab a a molex ground which if you don't know where those are our mod mats have handy diagrams that show where they are and then I'm going to just hit the correct corresponding SOC cap which we've already determined 1.2 so we are at 1.2 volts going into the socket as opposed to on hardware info 1.0 ish 11.0 basically so in this instance you might think that you're only pushing about a volt into there and at max 1.01 which is quite a bit different from 1.2 that is a massive difference so we're getting a bit of over voltage versus what's being reported here a part of that is some of the resistance between the pin and the pads and the socket as we discussed elsewhere in this video but part of its just gigabyte behaving a bit differently from the other boards so if we want to further this maybe you didn't check the back of the socket you're a normal end-user you're just looking at hardware info and rise master you think your voltage is still pretty low so at this point you might go ok I'm gonna push the 1500 megahertz and you know what just maybe I had some stability issues at this voltage so let's go all the way up to 1.25 and this specific number is really meant to just illustrate a large change we wouldn't expect you to and wouldn't recommend that you do increments that large but the point is to demonstrate something not to provide a guide of how to overclock so this number specifically some vendors have recommended 1.25 as a safe SOC voltage in the past some have a very hard recommendation of 1.2 as a max SOC voltage and then AMD has apparently also mentioned day 1.3 SOC voltage or APU graphics all digits same exact Thayne these two things they go through the same VRM so let's just apply 1.25 and see what happens so now we're seeing in hardware info one point one five one point one four so we're we're under one point two and over one point three it looks like one point one three excuse me so yeah one point one five is about where we are now versus one point two five for the configuration so it still looks pretty good you might just kind of leave it here you can even push it a little higher if you're assuming one point three is safe you can push it higher if you're assuming one point two safe doesn't really matter if you listen to one vendor or the other if one says one point two one says one point three either way we're below both 1.2 and 1.3 right now so let's check the socket side for its voltage one point three for quite high getting dangerously high in fact so with the folks we've spoken to about this one over 1.3 likes a sustained one point three six volts is getting kind of rough for an APU for the SOC specifically you could start damaging it degrading it or burning it out with a sustained over 1.3 volt input to the CPU but of course rise master and hardware are giving us completely different much lower numbers and this isn't necessarily a fault of one board or the other or even a fault of hardware info it's that they are technically we were reading two different things one we're reading at the back of the socket you can't get much more direct than that and the other one is through sensors through the CPU after resistance and aggregate data and all that other stuff so that's what we've got for basics now we could do one more example here and let's say that you know one point three five certainly is inadvisable however you might still be able to argue it let's say you ended up pushing out I was sixteen hundred sixteen fifty something like that and in the process of doing so you went with maybe a really high well comparatively high voltage of something like one point two eight seven five let's go with that because if you're listening to guidance from the board vendors or from AMD where sometimes they say one point three is say if once again there's difference between typing in 1.3 here and getting 1.3 into the CPU and we need to determine which one they're talking about when they say that you need to understand what they mean when they say 1.25 or 1.3 or whatever is safe because they're different things so right now we're getting one point one seven five we're not quite at one point two yet and you could see this as potentially an LLC issue because higher LLC if we could set it with this board would potentially stabilize out of it but either way we're almost at one point two sometimes let's do one more measurement back the socket so one point two sometimes for hardware info versus one point three seven to one point three eight that's bad that will degrade the integrated memory controller over time it is definitely not a good voltage to push into there so one last number something we've determined is pretty stable on our sample is sixteen fifty megahertz on this motherboard at only one point one eight I think one point one eight one two five so we determined this to be stable that number is pretty damn low compared to the guidance of 1.2 to 1.3 as a safety but again it's because there's a difference versus load line calibration V group stuff like that where the number we're reading out versus putting in our two different things so one point one for hardware info looks pretty good we're stable at sixteen fifty megahertz one point one eight for AP graphics which is SOC voltage and then at the socket one point two eight perfectly well it's hard to say exactly but some vendors have told us this is perfectly acceptable and then guidance for rise in speaking with some of the XO c guys the extreme overclockers we've heard one point two is the safety guidance and some other board vendors as well however Raven Ridge is a different products than Rison so it's possible it's changed but either way compared to one point three seven one point two seven is significantly lower that's a big deal so that's what we're talking about and make sure you know which voltage people are when I say people which voltage menu actuaries or overclockers are talking about when they say the number 1.2 is safe for the number 1 point 3 is safe because the difference in point 1 there is enough to seriously degrade or damage or destroy an SOC over time because a point 1 difference is one point three two one point four one point four definitely starts to create your IMC one point three I'm a bit nebulous on I've heard from multiple different people that one point three is fine I've heard from more people that one point two to one point two five is that they suggested the most reliable source that I've spoken with so far has suggested that peaking at one point three is perfectly acceptable for the Raven Ridge SOC and the same source has suggested that peaking at something like one point three six and holding there is bad so that's kind of your your margin for error if we go with that specific source but again this is all practical hands-on stuff so until things start dying it's hard to say precisely what's acceptable if you want my recommendation for a very safe guidance 1.2 is a good place to sit for the actual voltage going into the CPU not necessarily what you type in 1.25 I would also feel somewhat comfortable with with Raven Ridge and one point three is where you're just you're you're just kind of betting on who said what because again some people have told us one point three is fine some have said the opposite so it all depends Raven Ridge as I understand it should be fine at one point three sustained but obviously lower is better so maybe shoot for one point to one point two five let's quickly recap some of our findings on the three boards we tested we observed to the following what's the MSI motherboard the be 350 tomahawk we noticed that APU graphics voltage the option and rise and Master did not seem to do anything that's broken we ended up using SOC voltage via bios instead to adjust the stability of it all and LLC options also didn't seem to do very much on this particular be 350 tomahawk board level six LLC resulted in lower voltage than auto and level eight is apparently somewhat equal to auto or slightly lower depending you're looking at and this seems to have V group issues and under-voltage vs. target at times as well so that's the MSI Board's characteristics again we're not necessarily saying what's bad or not all we're saying is be aware of those things the Asus Board behaves a bit differently first of all we tested an Asus at B 350 M - E this is not an overclocking motherboard for from it and I mean you could tell that by looking at the lack of heat sinks on the DRAM so that said we wanted to test it for a specific reason which will be our second video and we'll talk about that later but either way the MOSFETs can easily breach 120 degrees Celsius if you're pushing high voltages because it's really not meant to handle that kind of voltage however we did find that we could reasonably hit 15 50 megahertz with a 1.1 volt SOC input and 1.2 volt graphics input the result was 1.25 volts via multimeter which seems acceptable MOSFETs still run pretty hot but not bad considering what a cheap board it is and conversely from that with Auto LLC 1.2 volts SOC and 1 point 3 volts graphics which are settings that might actually be okay on some boards but not this one what you end up with is 1.3 7 volt to be a multimeter it was capable of 1650 megahertz but you'd be roasting the MOSFETs and at one point three seven volts you'll be damaging the SOC as well gigabytes board we couldn't find any SOC LLC options for now users will need to be aware that while the msi board is often biased under volting the targets the gigabyte board is often biased over volt in it a 1.1 to 5 volt input is sufficient to stabilize at 1550 megahertz and hardware info and the gaming k5 seems under report voltage somewhat notably versus reality to the effect that hardware info will show 1.0 seven volts SOC but the multimeter will show one point two three seven which could lead to users killing the IMC if they are just applying some blanket one point two number and assuming it works because one point two can equal one point three two one point three four or higher in reality so now we're gonna get into some SOC and LC tables that we put together from hands-on testing with everything that was rude anyway I was saying with the tables what we're gonna be showing you is just some hands-on testing and we've got that for the same three board so let's go through those now here's a table of our issues SOC checks the color coding is based on frequencies when we changed frequency we changed row color at one point one volts SOC and one point one volts apu graphics input we were measuring one point one two to one point one three volts with a multimeter or 1.05 vi 1.08 one with hardware info this was with a low frequency of 13 or megahertz going to fifteen fifty megahertz had a one point one volt SOC and one point two volt graphics reading out as one point two to three via digital multimeter or one point one five to one point one eight one via hardware info that's with high LLC which was required for stability here's the dangerous one at 1600 megahertz and with extreme LLC we configured 1.2 volts SOC at 1.2 volts graphics and read out one point three five volts SOC via DMM if you're relying on hardware info you'd think that you were only at one point two nine to one point three which again depend on who you ask are sort of acceptable though pushing it and not really acceptable on this motherboard mind you given the heatsink limitations but conventional wisdom would suggest that 1.25 volts is ok for most ap use according to some of our contacts this board isn't really meant to push this high but that's beside the point the point is that these voltage is 1.2 on each to grade the IMC over time that was with extreme LLC though let's dial back to Otto since that's what most people use even with Auto LLC a 1650 megahertz clock was held with a 1.2 bolt as our CN 1 point 3 volt graphics voltage in reality these numbers equaled one point three seven volts and up in other words hope you don't need a memory controller for very long the msi be 350 tomahawk is next this one sometimes had V droop but not always and we also noticed the APU graphics voltage didn't really seem to do anything on this board it was all driven by normal SOC voltage going for 1600 megahertz with a 1.2 volt SOC and 1.2 bolt graphics allowed the frequency to hold using Auto LLC and the DMM output was one point two four five volts while hardware info reported one point one eight to one point two volts increasing graphics voltage to one point three did nothing here and did not change the voltage readings at all going to 1650 megahertz 1.3 volts SOC had us reading 1 point 3 5 volts via DMM but hardware info was reading one point to 7 to 1 point 3 volts finally the gigabyte gaming k5 seems to push voltage more heavily than other boards this isn't inherently a fault with gigabytes of motherboard it's just that users need to be aware of the behavior on this board and how it's different from MSI and Asus which means that following a guide for a different motherboard could easily have you inputting unsafe or unstable voltages at 1500 megahertz with a 1.1 volt SOC one fights evil graphics we have BIOS we measured a 1.3 volt output via digital multimeter harder info read 1 point 1 go into 1 point one volts SOC and 1 point 3 volts graphics gave us at one point three nine one point four volts SOC which is dangerous to the health of IMC this is where we realized that gigabyte was being more aggressive than MSI again not a fault of either just the behavior we dropped to one point one five volts graphics input voltage which resulted in a 1.25 volt DMM reading as you can see in this table 16 50 megahertz was held with an input number of one point one volts SOC and one point one eight one two five graphics resulting in a one point two eight ball SOC voltage as measured at the socket but one point one volts via Hardware info the takeaway here is that the gig white board only required us to input one point 1 8 volts for graphics to get 1 point 3 out whereas the msi board would require 1.3 volts input to get roughly the same out the Asus board would require 1.2 to 1.25 to get 1.3 out and these are behaviors that you need to be aware of on the motherboards and they are somewhat unique to each board as for why the numbers are different between a physical measurement and software some of it is of course that software is relying on what the CPU tells it what it's gauges are telling it internally and going from the back of the board to the socket or to the CPU die level we enter an area where the resistance of the connection of pins to pads could be throwing off those numbers and it is quite clearly so ideally the board vendors get things to a point where LLC isn't crazy aggressive when Auto and that voltages are close to what you put in as an input but as long as you're just kind of aware of everything and read what you can on the boards you should be fine the biggest thing is that typing in a number if MD says or gigabyte or ace use anyone if anyone says something like 1.25 is safe 1.3 is safe you have to know what their what they're talking about because if you're saying 1.25 is safe is that 1.25 read at the back of the board in reality or is it 1.25 as in the number that you type in to get things stable because depend on what their perspective is they might mean the number you type in or they might mean in reality and they're talking reality but you type in a high number you could end up far over what they consider to be safe and that just comes down to communicating with the manufacturers some of them have websites dedicated to overclocking so check those but all this kind of made sense to do now just because Raven Ridge and Verizon in general but Raven Ridge is pretty easy to overclock and D although we don't like software overclocking at all Andy's done well to make rise and master at least kind of work for end users to some extent most the time so it's not great BIOS is better if you can do it but in terms of accessibility to the end user Rhys masters pretty accessible and it lets you change things that would be potentially bad for you to change if you don't know what you're doing whereas a lot of software solutions don't give you that control it tends to be in bios so yeah so worth talking about because Raven Ridge is pretty easy to overclock and most this information applies generally as knowledge to everything else so if you like this type of coverage as always you can subscribe to us for more we'll be talking out how to kill your motherboard next this was sort of how to kill your CPU and that'll be on the mosfet thermal level go to patreon.com/scishow stop us out directly or store that gamers nexus net to make a shirt like this one which we've just restocked it's the graph logo shirt thank you for watching I'll see you all next time you you