Gadgetory

hey guys build lloyd here and today we're going to be taking a look at EVGA stake on a high-end GTX 10 80 TI PCB this right here is the FTW three and so a little bit of backstory to some of the design decisions that EVGA made with this card the first 10 series FTW cards the 1070 FTW and the 1080 FTW got quite a lot of negative coverage when a few people had the vrm explode on them and the first thing everybody sort of jumped at and started to blame was vrm temperatures because EVGA those cards the vrm would sort of run between a hundred and maybe a hundred twenty degrees and you know most people don't know that VR ends are basically built to run at 125 that's why I do all my temperature rent ratings at 125 mall sets can withstand 125 degrees operating temperature absolutely no problem you know the data sheets often have ratings that assume no active cooling whatsoever so no heat sink on the MOSFETs at all admittedly those ratings are much much lower in terms than compared to ratings where the temperature of the MOSFET is actually something reasonable like say 125 degrees or 100 degrees centigrade and so basically everybody panics over the like everybody first blame the the vrm temperatures that eventually like EVGA have figured out what what it was that was causing the issues unfortunately I you know a lot of the details are not for public like and under NDA however basically what happened EVGA got components that were faulty from a supplier and EVGA being in GPU manufacturer not a component testing facility went and use those components assuming that they actually passed the quality assurance that the manufacturer should have done on them and well some of the vrm has exploded due to manufacturing defects that you know EVGA had no way of ever finding like realistically there's nothing EVGA could have done to find out that this specific component was causing the issues because you know if they were like if you think of think about it it's like where does the testing for faulty components stop do you do you stop at like the complex things or the not so complex things like that's the thing and so EVGA really just got really unlucky that they were sent components that had a tendency to blow up and whoever ended up with those who components was equally unlucky as EVGA basically but because of all those sort of the initial you know negative coverage of FTW vrm temperatures EVGA is really really overcompensating on the vrm cooling here I mean look at this thing this is a copper contact play a copper contact play a thermal pad for the vrm right there so this is all the MOSFETs imprinted and there you have your drivers and doublers and printed as well so we have the copper contact plate then a heat pipe and then this massive aluminum plate which if you see the the teardown video up on the channel then there's like on the other side of the this plate there is a lot of surface area so EVGA have gotten kind of insane and they've also gone in like there's a thermal pad for the inductors which generally like those are one of the most heat tolerant components out there they really don't need cooling and then there's a thermal title the capacitors which does actually sort of make sense because capacitors are generally the most temperature sensitive component of an entire vrm however most manufacturers don't bother with putting thermal pads on capacitors at all and then there's another thermal pad back here for the capacity for risk for these capacitor banks and then you have this little l-shaped thing which goes over oops which goes over these components right here we our fan controllers which like I have no idea why those are getting cold like I could understand the capacitors and the inductors but fan controllers really high either way EVGA is going you know absolutely insane with the cooling on this card mostly because of all the past contra past negative coverage of vrm temperatures so this time around the temperatures should be low very very low actually so now let's actually take a look at the PCB which is probably what you all care came for before that this content is brought to you by our patreon backers you can go to patreon.com/crashcourse alves out directly which helps us with this in-depth reporting because it takes quite a bit of time and that direct support is necessary you can also go to store dot gamers nexus botnet if you prefer to pick up one of our three new shirt designs this is a really nice card let's just start it off with that but before we get into why I like it we're going to go over all the minor little all the all the various minor VRMs that are necessary for a 1080 teait function so starting off with this thing right here that is a voltage regulator that converts 3.3 volts into 1.8 volts for the BIOS chips and some of the internal PLL's of the of NVIDIA GPUs and there are two BIOS chips here you do get dual BIOS functionality on this card which i think is a really nice feature to have as if you're messing around with bios is for whatever reason having dual BIOS makes your life so much easier because if you flash a BIOS and you know you flash your BIOS you recent you try to restart and you get a black screen you just have to flick this switch right here then you can recent try restart again this time around it should boot up unless you flashed both BIOS chips for some inexplicable reason and then so now you boot up you get into Windows you flick the switch again and you flash the bad BIOS chip and your card will work as if nothing ever happened to it which is why the dual BIOS feature is like you find it on a lot of overclocker cards and it's really really nice to have in my opinion really for me this is one of those features where I'd actually choose this is one of those features where I actually prioritize it over a lot of other things like assuming there's two cards with the same vrm and everything else if one of them has a bio storage I'm just taking the one with the bottle switch I don't really care about the rat about any more of the details so that's a nice feature that the card has then below that we find the 1.8 volts for the gddr5 X this is a supporting voltage for the gddr5 X chips they basically needed to function it doesn't really do anything for overclocking but it is worth knowing that it's there so yeah down here we find the well actually it's on the back of the PCB but we do see part of it this is the 1 volt PE x or PLL voltage whichever you prefer to call it and this power is the PLL's and internal PLL's of the GPU core without this the card will not function on air cooling and water cooling adjusting this voltage through physical modifications to the card doesn't do anything for overclocking on liquid nitrogen you can gain extra stability and better overclocks by tweaking this voltage up so yeah now we finally get to a vrm I'm kind of excited about this is the vcore vrm the thing that gets all of that cooling so a sock voltage for the gtx 980ti is 1.06 volts and video limits you to a maximum of 1.09 3 volts when overclocking so you know that's not really that great a voltage range and Pascal cards do scale somewhat with extra voltage not a ton but somewhat you can pick up a few tens of megahertz by going past and videos voltage cap using physical modifications or there's a few cards out there where you can do it through software the problem with actually exceeding the 1.0 93 voltage is your power consumption just goes through the route like absolutely shoots through the roof and you only really pick up something like you know fifty to maybe even a hundred megahertz at best so not really worth it on liquid nitrogen however voltages as high as 1 point 6 and 1 point 6 volts vcore have been used for benchmarking purposes because more voltage generally means higher clocks assuming that everything is working as it should so now that's the week or VRM and then up here we find the memory vrm so that power that provides one point three five volts all right memory provides one point three five volts to all of the gddr5 X chips around the GPU core so then with those out of the way let's take a closer look at the actual make up of say the V Corps v RM first so at first glance this V core vrm looks like a one two three four five six seven eight nine ten phase and it is in fact a ten phase as we do have one two three four five six seven eight nine ten driver ICS as well as one two three four five doublers because you cannot actually buy a ten phase voltage controller the voltage controller on this card is located right on the other side of the PCB right on in this spot and it is a MCP eight one two seven four this is an eight phase voltage controller capable of going up to one point two megahertz switching frequency and we're here it is running in five phase mode so you basically have five PWM signals go into the doublers like so and then they split the signal between two drivers actually i've mists that one goes like that this one goes like this and so the so the vrm basically is controlled in this setup and the actual doublers and driver is used on this card the well starting with the doublers those are NCP CP 8 1 1 62 and these are cool because they do actually feature load balancing between the two phases which is a sort of more advanced doubling feature that some other doublers like you can get some doublers which literally just move the signal between switch the PWM pulses between the two phases which does mean that the phases can get really out of balance in terms of current and by that I mean you could have one phase with like 50 amps going through it and another phase with say 10 amps going through it and the actual doubler can't do anything about it it's not within its capabilities however the doubler EVGA is using here the NCP eight one one sixty two does include current balancing so it will actually if one phase is significantly I have significantly higher load on it than another phase it will actually skip a signal for the phase that is pulling more current so basically dropping the current through that phase going through that phase down very very quickly it's not the best like it still doesn't be like say the voltage regulation achievable on a true a phase design with a high end voltage controller but as far as doubling schemes go this is the next like this is the next best option so yeah and the reason why you want actual current balance between your phases is basically gets you better life span for the vrm you know the term thermals are better spread out across the MOSFETs the other benefit is it does improve voltage regulation a little bit as and it also puts less strain on your 12 volt rail so your power supply will actually last longer if you have a vrm that balances the current through the phase is better because one if basically when a phase turns on your PSU see that is a massive current spike coming on which is filtered by this capacitor bank here and here and these capacitors as well as these chokes so it's not like that's not suppressed your PSU actually will never see a a massive amount of noise that evita that a gpus main vrm produces but it is better if the vrm just produces less noise to start with as that does put less strain on the capacitors behind it as well as the chokes so yeah that's sort of the benefit of the NTP a1 162 now the drivers are here also important because the NCP 8:1 158 driver is not actually capable of hitting the turn on and it will fall rise and fall times of the maus dual n fats used inside the CRM so EVGA here has opted for dual NFS so you basically get two packages per phase like that so you have one here and one here and these are both alpha and omega semiconductor e 69 30s and dual NFS basically means that there is a high side masa and a low side MOSFET built into one chip as well as the diodes necessary for them to run and the spec on these is your high side masa in one of these and your low side set in one of these are seven million ohms high side and one milli ohm low side at 5 volts gate drive and the reason why I'm spiking the 5 volt figure here is because the NCP eight one eight one 108 one 158 the drivers here do not support higher gate drive voltages and 5 volts they run on 5 volts and they drive the vrm with 5 volts so that does mean like higher gate drive voltages get you better perform a set performance lower gate drive voltages put less strain on the actual drivers and these sort of balance the efficiency of the vrm differently at lower loads having lower gate drive voltage generally as benefit will generally be beneficial for your efficiency as you're not spending so much power on just turning a phase on where is that very very high loads when you have a ton of current going through the VRM you would prefer a driving voltage because your main losses at that point is the current going through the VRM not the act of turning it on and off so they're trading driver efficiency for vrm performance is a better idea but here with this vrm which is incredibly overkill which we will get to very soon EVGA has resulted four five volts a drive which actually most manufacturers just go straight for 5 volts gate drive on GPUs so yeah that's that now interesting little issue with the the other reason why the drivers here are important the NCPA 1 158 is these respects to rise and fall a MOSFET a high side MOSFET in 16 nanoseconds and 11 nanoseconds for the full time the issue with that is is that the e 69 30s under perfect conditions can achieve four nanoseconds fall I mean rise and three nanoseconds fall in three nano so yeah and basically that means that the MOSFETs here are kind of under driven so they can't actually achieve those switching those speeds however the end result for the actual vrm efficiency here is not really like it doesn't really impact in the AR m in a negative way as you still end up with a vr m that is capable of doing 1.09 3 volts at 125 degrees centigrade 300 kilohertz switching frequency on the actual MOSFETs so that would be 600 kilohertz on this line right here that would be running at 600 kilohertz from the voltage controller then you get 300 kilohertz on the coming out of the doublers so assuming 300 kilohertz switching frequency 125 degrees operating temperature and 101.9 3 volts output voltage this erm can do out 250 amps with only 22 watts of heat loss this is better than basically every other 10 pti out there slightly better it's not a huge difference but it is a small improvement over other ten ATT eyes except maybe some of the ones using the international rex except say the strikes and the extreme for the Auris extreme card from gigabyte which both of those using uh both of those use power stages which do have a very high efficiency that's the main reason why those things exist so 250 ohms 22 watts that's a very nice like you know rating here a completely achievable with the ridiculous overkill cooling system the EVGA packs this card with and this is actually higher than the average current draw that a 1080 TI will draw its stock and even even when overclocked you will not be going much past this as Nvidia does limit you to say 350 watts whole card draw so you're not really going to be going much past this figure now if you were to remove all restrictions and hard mod your voltage you might actually be able to achieve save for uh and I can't draw over the screw hole 400 amps you know what I'm just going to put it down here 400 amps current draw at which point this vrm would produce 45 watts of heat still completely sustainable on this cooling system you'd still be able to stay under the 125 degrees and in fact 400 amps is doable on these MOSFETs even at much much higher temperatures as a 125 degrees you can also go up to 600 amps at which point this vrm would produce 840 watts and this is going beyond like liquid nitrogen power consumption on a 1080 Ti no card will ever pull this no 1080 GI will ever pull this much current but if by some but if you were to say cut this vrm off and hook it up to something else that did need 600 amps it can do it with 84 watts of heat output at 125 degrees and this is also the same amount of current and roughly the same amount of heat as what you would see on say the Asus 1080 TI Strix and the ACE and the gigabytes extreme card the Auris extreme oh really this matches a lot of the like the the top and GTX the other top-end GTX 10 ATT is in terms of power capabilities now if you absolutely want to hit the maximum of this V RMS capability 920 amps going past 920 amps at 125 degrees centigrade will blow up the most sets and if you are at 120 amps it will produce 164 watts of heat so we will be running very very hot and that's mostly why this rating right here isn't actually realistically achievable as you will not be able to dissipate that much heat and if the temperature goes up a little bit or the current goes slightly beyond that 920 amp figure there's a very high probability that one of the high side MOSFETs in this vrm will die and you know when that happens the whole vrm stops working basically so yeah the vcore vrm here is really really nice I have nothing to complain about it's good for liquid nitrogen it's fine it's totally overkill for any kind of normal usage this card is going to be seeing the memory vrm is the same law sets and saying similar driver restrictions the memory vrm will never really need to put out more than 30 amps and at 30 amps output it will be producing about 3 to 4 watts of heat so really not worth worrying about most cards actually don't most other 1080 tea eyes don't even bother cooling the memory vrm at all there's no thermal pads and no heatsink on it EVGA has however opted yet again to you know heat sink everything in that we are and that's the 12-volt capacitors there the chokes aren't heatsink 12-volt capacitors don't get heat sinks the MOSFET well 12 volt capacitors do get heat sinks the MOSFET get the MOSFETs get heat sinks and even the output capacitor bank gets gets heat sinks so yeah you know cooling on absolutely everything maximum current rating on the memory vrm would be about 92 amps and we are talking about the same 125 degrees 300 kilohertz just different voltage 92 amps and and 16 watts of heat so yeah that's where you're looking at there yeah you know it's a very nice grm it's a very nice card it's kind of sad that you know you'll never actually get to use the full capabilities of the ridiculous overkill VR and that EVGA ships it with but it is you know it is the second highest and 10a DTI that EVGA is going to be making as we will probably be seeing the 1080 TI king thing in the kingpin Edition sometime soon that thing is going to mean I'm like I can't wait to get to you know see that things and that's going to be amazing if by any chance you do want to go and at least try to fully utilize the vrm if you want to lift the power limit on your to eight pins these shunts right here if you short them with liquid metal thermal paste you will basically trick the GPU into thinking it's pulling less power than it actually is and then your power limit essentially to end up being higher if you try to short them out by Sol during it's very likely that you will run into an video safety restrictions which will detect if the power consumption by the caro-kann the power consumption reported by the GPU is too low the card will go into safety mode and you will no longer be able to run at full 3d clocks so that's why I recommend using liquid metal now if you don't want to go you know have liquid metal on chunks the other option is that there are these three capacitors around the actual chip that monitors the currents going through the card which is this thing right here that's a Texas Instruments current monitoring chip if you put 12 for the reference 10a DTI it's 10 Oh resistors on - well if you put ten ohm resistors in parallel with these capacitors on a reference 10a DTI it cuts your power consumption sumption into 1/3 so yeah that that's the option for those of you handy with a soldering pen and who don't care about your warranty I mean technically modding the shunts would sort of fall into the territory of modding you're like you know voiding your warranty but the thing about the liquid metal is it's very easy to remove it's not going to stay there if you don't want it to so yeah just going to point that out I'm not going to debate about doing it or not so that is it for the card I think I've made a plenty obvious I am impressed by what EVGA has done here you know I would have no like sure if you're considering getting this card go right ahead it is quite high in terms of MSRP it is at around seven hundred and eighty dollars however a lot of the other high-end 1080 T eyes are in sort of the same price bracket and most of them don't have a BIOS which which personally for mean that makes this card worth it compared to the other ones but uh if you don't need the BIOS switch then yeah it's your money do whatever that you want so thank you for watching like share subscribe please do consider supporting us here at gamers Nexus through the patreon link down in the description below and if you would like to see more PCB breakdowns and other overclocking related videos I have a channel called actually hardcore overclocking you can find that down in the description as well where basically it's more of erm videos and overclocking stuff thank you for watching and see you next time you