Gadgetory

hey guys build good here and today we're going to be taking a look at the Vega frontier edition PCV and this thing packs the most powerful V core vrm I have ever seen on a reference card so we'll just start right off with that before that this coverage is brought to you by the core g21 enclosure from thermal take a $70 case with two four millimeters thick tempered glass side panels mesh ventilation in the front for breathability a rarity in cases these days and a power supply shroud with top mounted SSD sleds learn more at the link in the description below the VMO monstrosity is this giant l-shaped block around the card around the court and this L shape is actually like it looks really weird because we've never seen as far as I know I've never seen this on a another card but this actually is the optimum vrm layout for any scenario obviously it isn't very popular on GPUs because generally you don't want your GPU being you know really really tall but the reason why this is the optimal layout and why I am be used it is basically the distance from any of your actual phases to the GPU core because that's what it's powering this brick right here that's the GPU core these two down here are the HBM sacks but this vrm is pumping current into this block right here and so the closer it can be to that the less voltage drop you suffer which ever so slightly improves your efficiency and it also slightly improves voltage regulation because you have less propagation delay from basically the core to the VRM so this does have benefits they are very very minor though that's why it's you know it's like it doesn't matter a ton really the really user to hit diminishing returns on like high-end cards where they have the massive power planes going through the of through the PCB anyway where this would be more like where you'd really start seeing a difference is when you have vrm layouts which are really unab simal right like this might be the optimal layout but the difference between this and something less than ideal is not that bad well less than this is not that bad the difference between well cars where you'll see this actually be a major problem is dual GPU cards because they're the you basically end up with a situation where having massive power planes is just not an option because you need to fix so much stuff on those cards that you actually see voltage drop a cup from the vrm to the GPU core you know as high as fifty to even a hundred millivolts under load which is crazy so like your erm will be be putting out 1.3 volts by the time it hits the GPU core you're only getting 1.2 here you'll basically have like this phase right here would fit at 1.2 and by the time it gets to the GPU core you'd be at like 1.19 alright oops one too many ones but you'd be at like one point one nine volts so this is definitely like the perfect play how forever erm but it won't really like you know it's cool that it that AMD opposite for it it's cool that it's possible thanks to the HBM ultimately it won't probably matter that much but I did feel like pointing it out since it does explain why am he actually sort of went for this l-shaped because you know it's like why wouldn't they just use all of this empty space over here well they put the fan over there and this layout is actually just straight-up better than trying to put you know multiple than trying to cram a 16 phase DRM into this area so yeah that's kind of cool now let's talk actual details on on this erm why it's so powerful not you know funky layout obviously doesn't mean it's capable it's a one two three four five six seven eight nine ten eleven twelve phase vrm design so naturally it's obviously running a doubling scheme as you can't buy a twelve phase voltage controller which we need to go on the back of the card to check that out the voltage controller is this chip right over here and that is a eye are three five two one seven I currently do not have a datasheet for this thing as it seems to be so new that you can't look it up I don't think it's a rebrand or anything it is from the IR through is thirty five to 100-series but yeah I can't get a datasheet for it right now so there is a 35 201 no 211 there's a 201 there's also there's a bunch of other 35 200 series tips I'm assuming this is going to be just another one of those 30 well this has to be a six phase voltage controller just because we have one two three four five six these are IR there oops that's a be IR 3598 but nonetheless this is just you know it's going to be a six plus something voltage controller because it also controls the HBM vrm down here which we have one driver I see right there but this is a 35 200 series chip so it's going to it should do the usual 200 kilohertz to 2 megahertz switching frequency range not that the upper - you know the upper limits of that switching frequency range really aren't useful even here with the doubling scheme you could actually go all the way up to 2 megahertz without having this erm catch on fire for you but the problem is as you really push the switching frequency on vrm their efficiency just goes down the drain most MOSFETs are basically not designed to function above 1 megahertz anyway the ones on used on this card they will function above 1 megahertz switching frequency the IR 3598 can actually push more than 1 megahertz switching frequency output so you can push 2 megahertz into them but the problem is your efficiency is going to be really really bad so really that upper limit does not matter though pushing this up to say 1 megahertz switching frequency might see some benefit but with this being a 12 phase and with the decoupling job that AMD does on these high-end HBM cards I mean this is ceramic capacitor city this is more like just ridiculous amount of by decoupling capacitors right there and then we have high-end just these are SMD polymers these things are literally the best option for vrm capacitors and these are 470 micro farad so the chances of this vrm needing more phases just to get stable voltage regulation very very low I I'm gonna when I get my card I'm going to still try attach even more capacitors to it but I really doubt I will see any benefits and E is generally really really good about getting ridiculously stable voltage regulation on their reference cards and this just looking at the design we have here the you know the IARC voltage controller which incidentally this thing is digital so it will support software voltage control and then the IR 3598 doublers these are not any and they don't have any advanced features but they do interleave your two phases so this does run cleaner than a native six phase and the six days on the fury X was excellent and you know it was smaller so this is even a step up from the fury X erm it's pretty much double what the fury ax came with so voltage regulation wise nothing to worry about really like it doesn't get better this is very much like this this vrm in terms of control scheme excluding the fact that this uses a different voltage controller is exactly the same as what you would find on a 290x lightning for the vcore so I can't find any complaints to to have about the control scheme except you know maybe they could have gone for the 16 phase but that would kind of get really excessive that phase count is really rare for a reason it's generally pretty impractical so back on the front of the PCB let's talk actual power capabilities so the control of the VRM is good there's nothing to complain about there now the actual power capabilities each phase has a single low side mosa and a single high side MOSFET the low side set is IMD's favorite IRS 68 94 this is a international rectifier direct FET it is really easy to recognize these because they have the metal casing this helps with thermal dissipation as well as all well basically it's mostly a thermal dissipation thing these things are not cheap actually they come in well over air off the top of my head they're around 1.2 euros per thousand according to international rectifier zone website now the IR now the high side MOSFET is the usual IRF just because if you have a 68 94 this one always cut goes with it it's a 68 11 and that is also a direct fat much smaller though and worse thermal handling capabilities but nonetheless this combination of MOSFETs are like this is the same you would find on a 290x lightning and with each with two well with this set of MOSFETs in each phase this VR I'm assuming in operating conditions of 1.2 volts output 12 volts input obviously it's in a it's powered by an ATX PSU 300 kilohertz switching frequency and 5 volts gate drive which the 5 volts gate drive is very very unlikely on this card it might even be using 12 volts gate drive I am Not sure these MOSFETs can certainly run on 12 volts however the thing is 5 volts gate drive is the minimum operating voltage for these MOSFETs well really is 4.5 but 4.5 volts to 5 volts it's the same thing so I'm using the 5 volts gate drive voltage because I without the card in hand I can't check what the actual gate drive voltage is and this will give you the worst current handling capability figures for this erm 300 kilohertz switching frequency on the actual phases themselves I'm assuming based off of the fact that a.m. these past cards we're running 300 kilohertz switching frequencies so that would be 600 kilohertz on the controller there's a pretty good chance that the actual phase switching frequency is even lower than what I'm using here so I feel pretty safe in saying that any current figures I writ list as well as he outputs are worst case scenario certainly they're not going to be driving the MOSFETs as it's 5 volts that's no AMD card ever has driven with 5 volts they usually drive with 7 or 10 so yeah let's get into the current figure so 200 amps which I don't think the card will actually run on 200 amps just because we're looking at a 300 watt TDP and HBM does not use a lot of power nor do any of the other systems on the card 200 amps we're going to be looking at around 23 watts of heat output and on the MOSFETs not inductors and capacitors included just the MOSFETs will put out about 23 while so heat output so that's the thing then 300 amps you're looking at 43 watts of heat output now this I would say is slightly above what the card will normally use unless the card spends most of its sitting most of its time sitting at around one volt I haven't yet gotten my card I can't do checks on how the card runs under typical loading I generally when I do these analyses I like to run with numbers from cards with no power limit because then they don't change their voltage on the fly making it really easy to predict how they'll behave when you really really crank up the voltage instead of trying to figure out what the actual peak current rod that is like because basically if the card is power throttling right it'll drop the voltage it'll drop the clock but your current will still be up which is why it's still bouncing off the power limit and then if the current drops down it'll pull up the voltage and it'll pull up the clock because you're pulling less current but effectively your power draw is still the same but if I kept the voltage high and I didn't drop the clocks because the power limit was listed then your current would generally just shoot through the roof so I'm not currently sure how much power you would need for overclocking because overclocking you're obviously going to do everything in your power to reduce any kind of power based throttling or temperature based startling and so as of right now I don't have data on how much power this erm will actually have some handle when overclocked certainly for stock power consumption this thing is ridiculous overkill let's keep going up the current figures so 400 amps the CRM will put out 66 watts of heat that is starting to get into the problematic area the CRM is pretty spread out as you can clearly see a good amount of surface area on it so it's it's going to be you know it's not going to be the hardest thing ever to cool but sixty-six watts of heat is a good amount like that's a lot of heat so I know if you're hammering this card the arm temperature is our going to be work like you should keep an eye on them however the mall sets themselves can actually do 400 amps of current even at temperatures as high as 3 138 degrees centigrade so everything else on the vrm will complain at that point your your capacitors definitely capacitors generally don't come in ratings above 105 some of them comment 1 125 degrees rating but those are really really rare 125 degrees it's still less than 138 so the MOSFETs won't die if you run them really really haul everything else on the vrm will however suffer reduced significantly introduced lifespans so yeah in our short-term operation at very high like very high current load loads possible not recommended depends really on your VR I'm cooling the arm can go higher 500 amps you're going to be looking at about 92 watts of heat output and it can still go further as the I'm assuming that like in this is assuming you can keep the vrm at 125 degrees as MOSFETs do basically as long as you're not hitting the thermal Junction max of the MOSFETs you can keep watching more current once they exceed thermal Junction max they'll thermal runaway and blow themselves to pieces but if you can keep the vrm at 125 degrees even with the 92 watt heat output then 500 amps you can totally push that through it as well in fact you can go all the way up to 700 amps however at that point it'll put out the same amount of power as ACE was a GPU so this is basically it'll handle it in short-term bursts the heat output at those kinds of current levels is so high you're not going to be able to cool it however the good news is basically that this vrm right here is certainly very very capable in fact the mosfet selection the driving setup and everything but the voltage controller here and the inductors and the capacitor selection is the same as what you would find on a 290x lightning so power wise this VRM is on par with high-end custom GPUs of a few years ago so yeah it is by far the most powerful vrm we've ever seen like ever on a reference card this makes the GTX 480 SVR M look like a bad joke and the GTX 480 would be the only card I can think of that should be in the same kind of current levels because the GTX 480 came at extremely low stock halted due to its ridiculous power draw but yeah this erm is just like it's a thing of beauty but the side effect is that leaves the question of does this card really need 12 phases or is AMD just being very generous you know if we judge based on the length of our X 480 where the vrm was ridiculous overkill on the reference card like it wasn't exactly power efficient the MOSFETs really weren't the best for power efficiency but they were very very capable MOSFETs that card had a really liked that card out of VR I'm capable of delivering two to three times its stock power draw I will stall current consumption and what we're looking at here is you know it's really really is a really nice vrm the question is does the card pull so much power at stock that this is actually necessary or is md just doing what they did on the 480 where the stock we are I'm was like you could take a 480 in max like a reference 480 and max it out on liquid nitrogen and the vrm would not be the problem you've run into so that is the that is the question I'm still left wondering is just like I mean on one hand AMD does have a history of putting really nice crm designs on their cards even the fury acts it's like one point four volts the RM handles it no problem even the cooler on the fury ax will actually deal with one point 4 volts and more I've just not been brave enough to go higher yet but yeah it's just this is like this is literally twice the fury axis erm so does it run twice as much power at stalk as a fury ax I mean twice as much current at stalk or you know I mean it is $1,000 cards so maybe AMD just went the extra step and actually gave us a vrm that's worth a damn but kind of concerning kind of concerning and leaves a lot of questions to still answer at least for me now then let's move on to the minor vrm then when I say minor I mean less than one twelfth of what the main vrm is this right here is the HBM vrm it is a significant downgrade compared to the recurve erm it's a single phase we just have one choke here and that power is in both of the memory stacks and it uses a on semiconductor dual and fat this thing is nothing amazing it's a for cat6 and that's not the full model number but if you put that into google it'll find it or into on semiconductors website it'll find it so that's a dual line fat so basically the high side MOSFET and the low side MOSFET are integrated into the chip this thing is not that great the end result is that the CRM can do 10 amps at about 2 watts of heat output on the MOSFET 20 amps at about 4 watts of heat output which incidentally assuming one well again I'm assuming these operating specs for that MOSFET down here and the reason I'm assuming that 1.2 volts not a higher voltage is because HBM at least by SK Hynix spec is meant to run on 1.2 volts now in the past on the fury axe AMD was running 1.3 volts through the HBM but without a card in hand to check I have no idea what they're actually running it on I'm going to assume that this time around the HP m2 is actually working properly I hope I really really hope that at this point they can run the freaking memory at voltage spec voltage instead of over voltage just to hit the stock frequency that it's supposed to do and I'm assuming they're going to run 1.2 volts 300 kilohertz 5 volts gate drive again so assuming those settings this 20 amp figure 4 was heat output is actually 85% efficiency which is just cool because it confirms for me that my calculate V RM calculation matches exactly the calculation that they have on the datasheet well more or less exactly it's about 85% for both my calculator and there and the datasheet so 20 amps 4 watts and then 30 amps you're looking at 7 watts now I'm not going to go into any higher current figures because really the HBM on stock voltage should only be pulling between 10 and 2 I think around 16 maybe 17 amps on stock settings so it really doesn't need a lot of power and that's why we're looking at such a really small vrm so you know like such a small vrm for it now then moving on to the remaining minor rails we have the well these three guys over here and these are three different voltages I don't know which one is which I'm not sure how much any of these voltages pull the datasheet for this these MOSFETs is horrific ly useless in that it only gives me ambient temperature is ax am be like ambient air temperature ambient air thermal resistance which is useless for a GPU with a heatsink on the V R M that is the main problem it's like where like that 125 C figure right here that's MOSFET like if you put a thermocouple on top of the MOSFET that's the temp if that's the temperature of the MOSFET then this is how much current you can push through it before it starts to go you know thermal runaway on you obviously if the temperature goes up by one degree then you've just exceeded your current limit and bam it goes thermal runaway on you anyway but if you put a thermocouple on there this is the temperature that you'll measure and these are the current levels it'll do assuming you can keep it at that temperature the problem is I don't have the thermal resistance figure for these for case temperature so basically you put a thermocouple on these and I can't tell you anything about what they should handle ambient air temperature figures like I don't know what kind of actual operating temperature the MOSFETs will hit but it looks like from the datasheet any of these should be able to handle about 10 amps these are all dual n sets again and the voltages they're producing are going to be VDD C I so that's the memory controller on the AMD GPUs which if you notice compared to gddr5 AMD GPUs the VDD see i've vr m being one of these three is actually a huge difference because EDD CI on say something like a rx 480 is actually very similar to this so it's about twice to three times as powerful as what you're getting right here for DCI because the big difference with HBM you know you don't save that much power on the HBM own voltage it's not that different from gddr5 where there is a huge difference is the memory controller the memory controller for HBM is way wave mate way more power efficient and the end result is that we have a really really pathetic little VD DCI vrm as well as well it's not pathetic it's just you know reasonably sized now so you have a much smaller VDD see I've erm compared to what you would see on a lot of other cards now under well the other voltage these we'll be doing is VPP so this is a supporting voltage for the HBM gddr5 x also has a VPP rail basically this is a voltage that used to be internal to gddr5 ddr3 and all of that well ddr4 gddr5 acts I imagine whatever comes after gddr5 x h PM h vm - all of these have an external VPP and this was done because it improves memory power efficiency it used to be that the voltage would be created on inside the actual memory chips themselves here it's made external because this is the more power efficient way to generate it and then finally the third one of these will be the display drive ole tidge which I don't have a nice shortcut for so I'm just going to call it disp and that's basically that powers the some of the GPUs internal PLL's as well as the display outputs this voltage is very very useful when push when pushing AMD cards on liquid nitrogen because AMD cards on the quad nitrogen have a tendency to lose their display outputs as in your monitor will just go black the cards still running the display outputs just drop out and this happens specifically under load when pushing very high core voltages at very low temperatures the display just stops working for whatever reason and you can fix it by bumping up the display output voltage as that circuitry either gets too cold or power starved or something happens to it and it starts complaining so yeah but none of these voltages are well I think what man may or may not have VDD CI voltage control currently available VPP and v disp this one basically does nothing for overclocking this one is only relevant on liquid night neither of those going to be available in software so you don't have to worry about those really and yeah so that covers all the VRMs um you know vcore vrm could not like lovely i love it it could not be well i it could be even bigger but at that point i'd really start questioning like is AMD just like gonna start making extreme overclocking cards out of the box making all the non reference designs completely pointless or even worse because that's actually happened is like AMD's reference 7 970 PTV was just so nice and expensive i guess that there was 7 1970s we've downgraded p cds in fact there's RX for 80s with downgraded p cds so you know it's like it's somebody like on nvidia side it's like oh yeah we don't want the reference card it's awful on AMD side it's like sometimes the reference card is actually the best and you know that that's kind of starting to get ridicule so weak or vrm I really like I have no complaints about what I'm seeing so far I'm just hoping that it's not indicative of some ridiculous stock current consumption the HBM vrm it is a single phase and I will complain about that because the like I'd like to see a two-phase I it's just the HBM on the fury X I know that ran stock HBM over volted and it was just like you know generally a mess but that that card benefits from extra capacitors on the HBM vrm which basically tells you it should have had an extra phase because the voltage isn't stable enough on a single phase but here we do have a we do have a single phase and it really like it really depends I'm going to go and test if you know I'm going to obviously test if this single phase is enough I'm hoping it is but judging from the fury X it might not be so it's just going to throw that out there and then the minor ones is like you can't control VPP I don't care about even as an extreme overclockers I just don't care MIDI DC I really shouldn't pull enough power to be concerning and V this but never does that one has been 10-amp basically can amp fully like fully integrated but converter chip for ages and ages and ages and it's never actually pulled the 10 amps and it's basically completely irrelevant for everything except liquid nitrogen overclocking so the are my is the card is fine now let's look at some of the cool features that this also comes with this right here is a dual BIOS which which is standard for all high-end AMD reference cards you'd find this on a7 970 I think even by 5 7 nope 580 70 doesn't have it but 6 970 yeah 69 70 has had it to 7-9 70s had it - 90 - 90 s - 90s - 90 X's had it fury axes have it Vega now has it so you know bring that on I love BIOS modding and the BIOS which is really nice I just hope there's not going to be any more BIOS bio signature checks in the driver because that is those always just you know rain on your parade but it is nice that you at least get dual BIOS the boss there's one BIOS chip right here there's another one on the back of the card right on the basically mirrored with the BIOS chip on the front except tweaked by a little bit on the angle and then the last cool feature that has is the GPU tachometer right here which is basically a strip of LEDs it's I think it's really cool a lot of people think it's useless fluff like all the other RGB we see these days but this is actually useful because this delivers you know it's not system critical information like voltage or temperature but it's really useful for debug like figuring out what's wrong with like a crossfire setup because it will show you if you LPS is enabled it will show you if the cards aren't being loaded up properly and that's about it really it'll pretty much tell you it's like are the cards running the way they should be because this does indicate GPU load so very nice for you know if if you're running if you're doing what I do this comes in handy every so often which is a lot better than most of the RGB stuff where it's like oh like yeah it looks nice and also the cool thing is which a lot of people on the fury cards weren't aware of there's a little switch right here this is a two-channel dip switch you can change the color on the fury axes this was red in blue I assume it's red and blue again I would hope that it would have been blue in yellow because the card is blue and yellow and that would be color-coordinated but I'm going to just guess that it's blue and red anyway but you can turn change the color of the LEDs and you can completely turn them off using that switch so that's it for this video you know I hope you found it informative you can leave a comment down below if you have any thoughts or questions to share you can like the video if you like to dislike it if you disliked it please consider donating to the gamers Nexus patreon if you would like to see us do you know if you would like to support what we do here if you would like to see more content like this and more overclocking videos then I have a channel called actually hardcore overclocking you can go watch all kinds of hardcore overclocking related stuff like liquid nitrogen overclocking that kind of thing over there that's it for this video thank you for watching and see you next time you