Gadgetory

hey guys build Zoid here from actually hardcore overclocking and today we're gonna be taking a look at the RT x 2070 reference PCB from well EVGA here but this is the same PCB you would find on like a Nvidia founders Edition card because while this is the reference PCB as much as nvidia likes to pretend like reference like the founders cards founders edition cards aren't reference cards they very much are you're not gonna find a card with a worse PCB than a founders Edition it just so happens that founders editions are generally a lot better than what Nvidia used to put out so it kind of makes sense to to you know hold them in a higher regard compared to say like a reference I don't know GTX 580 PCB gtx 680 PCB yeah definitely much better than like a gtx 680 reference PCB before that this video is brought to you by autonomous AI where you can buy they're automatically raising and lowering smartdesq tables and table legs if you'd like to build a raising and lowering table similar to our set table the DIY kit is $250 and is easy to assemble to a service of your choice if you'd prefer something pre-built the full tables can be had from 300 and up you can learn more at the link in the description below for autonomous AI anyway today we're taking a look at the 2070 and so this is a twenty series and you know and video cards so of course we have the usual set of voltages and so let's start going around and identifying all of them starting of course with the most important voltage regulator for of course the V chord which powers the GPU core which is why we call it V core and I've written that really badly so let me just slightly tweak that that's slightly better up above that in which is kind of a normal position for this we have the memory of erm so that's an e mem and that of course powers all over the micron GDD are six memory chips micron is currently the exclusive supplier of GD d are six for all of the geforce cards samsung also makes GD d are six but you're not gonna find it on a on a sort of consumer card you might find like i think they're reserved only for like tesla cards and quad rose yeah micron G DDR six on this as usual down in this area I'm assuming we have the pecs rail I am not 100% certain about that but this just seems like a reasonable place to put the pecs rail because that powers the PCIe interface and some of the internal PLL's of the GPU core this right here might be and this is where it gets messy cuz this right here I assume as the BIOS chip because there's not really anything else that chip could be considering where it's located but the thing is we have a whole bunch of regulators in this area and at this point this is gets becomes a bit of a mess because we need a 5 volts rail for the V R M so you need a v RM to power the v RM and i'm not sure where that is it could be this but this could also be 1.8 volts which would be a supporting rail for the GDD r 6 memory chips as well as powering the bios chip and then this could be 5 volts but the thing is this also has a USB C port so this over here could be the USB C power rail so yeah and unfortunately all of these like two of them they do seem to share like the the componentry between them seems to be the same for a couple of them so that makes it even harder to figure out which is which but those are the rails that are present here I'm just not sure which one is which one because I don't have the card in hand to check for sure if I had the card in hand it'd be relatively easy so that's kind of the situation with the minor rails and now let's get into actually looking at the V Corps v RM since it's so important in a bit more detail so we have a 1 2 3 4 5 6 phase so nothing crazy right there controlled by a chip on the back of the board because of course there's not like the board's not long enough to cram it behind the V curve erm proper and that's this chip down here which is a that's a u p95 12 so this is pretty standard for the Nvidia cards this goes up to eight phases output of course here it's only running on six so not running at 8:00 it's running six-phase it goes up to two megahertz switching frequency it has a SM bus interface so you can in theory do things like control the load line on it well configure a whole bunch of functions of the chip and like control the voltage and monitor the voltage and output current and operating temperature of the vrm assuming Nvidia actually allows you to access that last last I've heard or at least as far as I'm aware you can't actually get access to this because the Nvidia driver won't let you and I think it might even be straight-up disabled so it's kind of like like the chip has it it's just not necessarily being used now for the actual voltage control this being an Nvidia chip and video card it of course uses the PWM vid interface which basically is a PWM signal from the GPU core to this chip which is used to then set the output voltage of the vrn which is a nice really like simple way of doing voltage control at like very very rapidly so that's that's why Nvidia goes with that because that's a really it's a really elegant way of solving your voltage control it's just kind of unfortunate that that means you also need to have like you need to go through the Nvidia driver to control that though it is easy to physically modify but software wise you need to go through the driver anyway back to the front of the board so it is a six phase and as each of the phases runs one of these power stages here and these are on semiconductor NCP 3 1 0 255 and these are 55 amp power stages or dr moss so they're not so that that stands for driver MOFs and basically these are these are power stages they're just not very intelligent compared to say like smart power stages or but there's a whole bunch of other flavors of power stages that integrate things like temperature monitoring current monitoring over current protection these don't really do any of that these are just a 55 amp power stage where you get a dry it you get a driver and you get to power and you get your high side and your low side MOSFET in one convenient package and these are respect to be well the datasheet says these go all the way up to fifty five amps but all of the efficiency curves only go up to 40 so that makes things a little bit difficult because basically we can't actually look at current outputs you know in that 55 like perfectly like how much heat each of these power stages makes at 55m five amps output is not actually available in the datasheet even though they are marketed as being 55 amp port which is always fun well that's the same thing as all of the 60 amp power stage is only going up to 50 amps in the data sheets except for some exceptions but most of them if it's spectat 60 it only spec aspects the heat output up to 50 amps and then it's just like and normally that's done because honestly you're not gonna go to the maximum output of a power stage just because it would produce so much heat that it'll be uncool about regardless of the fact that you're aware that it's going to be producing too much heat so anyway looking at the output for at the vrm heat output for a 1 volt output so that's nice compared to say all of the like on the other cards so far on the 20 series it's always like really high-end 70 M for smart power stages from like Fairchild or something and those are always SPECT in 1.8 volts well these are expecting 1 volt so that's very nice 500 kilohertz switching frequency and of course running on 5 volts because they don't run on anything but 5 volts you're gonna be looking at sort of the following efficiencies for the following output currents so 50 amps output which that would be roughly around stock so a stock stock current output would be somewhere around 150 amps average and that that now looks like an a but there that's a 1 yeah anyway so stock would be around somewhere maybe slightly over that 150 amp figure there you'd be looking at about 15 watts of heat output which is really just fine like that is not a problem whatsoever then once you start overclocking assuming the power limit goes high enough which last I've heard it and you could be looking at something like 200 amps output and then the vrm would be producing about 24 watts of heat which still really isn't that bad immediately that's 24 watts of heat concentrated on a much smaller surface area than say some of the other GP like VRMs we've looked at recently right like it's worth considering that this is a six phase so while it might be producing only a little bit more heat than some of the eight phase and ten phase and twelve phase and much higher phase count VRMs at 200 amps output this is a lot denser so it will still end up being hotter even though that like it'll be much more hotter which that's great English right there but it'll be like more hotter than you would necessarily expect from the very slight decrease in sort of efficiency right that's not a huge loss like the efficiency isn't that much worse but it's also denser so that's worth keeping in mind and then the absolute maximum current output that the datasheet really allows you to go up to easily without while extrapolating where the where the efficiency curves would go you're looking at about 240 amps producing about 32 watts of heat which is still manageable like that's admittedly you will need like of erm heatsink on some airflow but that is still not a ridiculous amount of heat but yeah this this is not exactly a super-powerful vrm especially compared to say like a twenty eighty or even other GPUs in this price category this video room is really not not impressive because well honestly if you if you actually look at even cheaper cards like vega's these days are cheaper than one of these the the Vega reference PCB is like ridiculous compared to this so you know but obviously this doesn't need an insane vrm it's just worth considering that yeah this this is pretty it fits the purpose but it's not exactly incredibly impressive and it's not really it's it's also a bigger step down in terms of the vrm design compared to like the the 2080 or the 2080 ti like the the the downgrading of the vrm going from like a 20 atti to a 20 80 - this thing is pretty significant even if the well the price the price differences are also so kind of makes sense but it is worth noting that this isn't super powerful but it's not super terrible and honestly I don't see why in videos shipping these cards like I would happily shove 240 amps through this like even on air cooling as long as the air cooler has good vrm cooling I would totally just go for and shove 24 240 amps through this vrm it shouldn't really have a problem with that this isn't so much heat that it's uncool about especially considering that there's like you know well there there's a lot of cards in the past that would do well hell of a lot worse than this right here and those were fine so the this is yeah this is a nice enough erm it's not super overkill but it's fitting for sort of a mid-range card it's kind of unfortunate that apparently the power limits are super low the memory of erm is in typical memory vrm fashion massive overkill it's like every time somebody needs to power memory they go for ridiculous overkill though here we actually see a pretty big difference compared to some to the say the 20 80 and the 20 atti because for once we're not using a new sort of NVIDIA voltage controller the voltage controller for this lovely two-phase right here right one two is AUP 1666 and this is the same voltage controller you would find on a loss of the 10 series GPUs and this thing this thing is really dumb this does not have any digital interface whatsoever it just has a PWM vid input and every all of its functionality is configured by resistors around the chip so that's that's kind of how that works right there but it is a 2 phase voltage controller so this is a two phase vrm and for the actual MOSFETs because these are power stages these are straight up Moll so that's while dual N fats because you have one chip but there's two MOSFETs built into that chip and then the drivers for the MOSFETs are actually built into the U P 1666 here so basically this is hooked directly to the two MOSFETs in each of those packages over there so these right here are Fairchild Semiconductor fdpc can't see my notes properly 5 0 1 8 SG dual infants and they have a 1.6 million low side which is really really good and if I remember correctly in video is using these for some reference card on the 10 series as v core power so these are totally out of place for powering a like a couple G DDR 6 memory chips this is really really overkill the end result is that basically you're looking at sort of doing one point three five volts actually let's put that where it normally is so you're looking at these doing like 1.3 5 volts output because that's what G DDR six runs on 500 kilohertz switching frequency and I'm assuming a gate drive voltage of 12 volts so there because that's just the easiest way to power them since you have 12 volts available on the card anyway so 420 amps output on these you'd be looking at about 2 watts of heat which is in line with what you would get with like the like the 20 80 and the 20 80 TI V RMS actually get about the same efficiency even though those of you arms are like much nicer it's just they're massively like the components there are super over SPECT and you can't really like there's a limit to how efficient your vrm can be also this isn't factoring in the power that this chip uses which skews these results a bit better because the power stages do include their driver losses which this doesn't because the drivers are built into that and I'm calculating just the losses on the MOSFETs themselves anyway going up from that 20 amps output to 30 amps these start actually getting a bit worse and producing about 3.5 watts of heat and then 40 amps output you're going to be looking at about 5 watts of the heat so definitely like these these are definitely czar definitely weaker this is definitely a weaker memory vrm than what you would have on like a 2080 or a 2080 ti it's just not a massive difference because you're really not going to be going very high current output anyway GD d r 6 especially with just the 1 2 3 4 5 8 chips that you have here they're really not going to be exceeding that 20 to 30 amp output range so this should be perfectly fine running passive as long as there's not too much heat just sort of soaking into the PCB from everything else because obviously these produce heat that produces heat this produces heat and with the PCB being as small as it is the PCB zone sort of thermal dissipation capabilities kind of suck because it doesn't have a lot of its own surface area which is actually like a valid way of getting things cooled if you can't like if you don't want to have an actual heatsink having a lot of unnecessary PCB space can actually do a lot of that for you but obviously this doesn't have that this is as short as it can possibly be so yeah but the memory of your room here again just completely like definitely fitting for the fit for purpose basically they're nice solid memory vrm completely acceptable v core nothing incredible but definitely acceptable and the last thing left is of course the usual 20 series power management stuff it's probably the best term to use for it and there is as usual a ton of it so that right there is of course our favorite up7 6 v 1 power management chip which is used of course for balancing the 12 volt inputs of the card because you have 12 volts coming in from there and 12 volts coming in from here and this might have a power limit of like 150 and this might have a power limit of a 75 watts and I now realize why NVIDIA has such a low power limit saying for these cards of course it only it can only pull a maximum of 25 watts without exceeding PCIe specs yeah that actually makes a lot of sense yeah that pretty much explains it the stock TDP has to be like 190 or something yeah yeah that makes sense anyway so you have like a 150 watt limit here 75 watt limit here and unfortunately you don't really get to choose how current goes through your phases well the the you p95 12 can actually current balance all of your phases but you don't have perfect control over where you're gonna be getting your power from so you could end up in situations where you know this has a 150 watt limit but you're gonna try to pull 200 Watts through that and only like 25 Watts through the PCIe slot which is like you're technically still under that 225 limit that the card has total but your you'd be power throttling off of the PCIe connector so the job of the U P of the up7 651 here is to basically go and switch some of the phases of the vrm from pulling power on the 8 pin to pulling power on the PCIe slot or if the PCIe slot is being overdrawn taking whatever phase is over pulling the PCIe slot and moving it onto the 8 pin reducing the power throttling because you're not going to be bouncing off of the power limit for each of the connectors as much as you'd normally would and you can really like actually like max out the total board limit instead of just whatever power connector you're maxing out so that's a new feature of the 20 series owners personally I think this is completely pointless there's a lot of cards in the past where they just didn't bother with worrying about where the power came from it was just kind of a case well the card has a power limit of 300 watts where those 300 Watts come from that's a completely different matter and is not actually considered by the GPU whatsoever it'll just kind of pull them wherever it can get them which can in theory like lead to some issues but you know it's worked fine up until now I don't I don't really see a reason why you would you know bother with something like this but Nvidia felt like it was necessary so that they went ahead and did it and to make sure that this chip can do its power manage as quickly and as effectively as possible you obviously need something to actually monitor that all of the power going through the card and that's done by this chip over here which is of course the on semiconductor NCP forty five four nine one and unlike the rest of the twenty series so far there is only one of those because normally you have two of them here you have yes indeed you have just one and it's on the back of the card so that's how the power management is done here you know you have the NCP for forty five four nine one for monitoring all of your power draws and this for balancing them and while balancing your to power inputs the voltage controller itself actually balances the current through the phases but which phases are on which power connectors the voltage controller really doesn't care about which is why you can run into that situation where you'd be pulling more from the eight pin then you'd be allowed to without actually maxing out the PCI uslaugh and then power throttling which is why you need the extra circuitry to manage where the phases actually get their power from because the U P 9512 really doesn't know and it really doesn't care so that pretty much covers the power management stuff except for the shunts which you have one here and you have one here and those are being monitored by that NCP for forty five four nine one chip and you also have a shunt over here so I guess that might be for the USB C port may be kind of intra there a third shunt okay well if you're wondering about how they're wired up you can always just use a multimeter to check which ones connected to what I'd have it as a guess that this one's for the eight pin and this one would probably be beat for the PCIe slot so that is the RT x 2070 reference PCB it's not the most impressive thing ever but it's definitely like for for what it has to do it's you know completely adequate you got a solid little six phase and little is actually the right term to use because that is compact and then you have a nice you know the nice little to faith well that's not actually a little memory VR I'm you can get much much smaller memory power systems but you get a nice two-phase memory power and yeah not really much else to the card than that so that is it for the video like share subscribe leave any comments questions suggestions down in the comment section below and if you'd like to support gamers Nexus there is stored on gamers nexus net where you can pick up shirts as well as the mod map which you can kind of see in the background of the picture here and there's also the gamers Nexus patreon if you'd like to support us that way I also run a channel called actually hardcore overclocking where I do a whole bunch of other overclocking related stuff like removing and video power limits because I think power limits are stupid so yeah that's it for the video and goodbye