Gadgetory

hey guys builds lloyd from actually hardcore overclocking here today we're going to be taking a look at the rx Vegas Strix Edition from Asus this is the first of the Vega custom PCB cards and this one is surprisingly good which honestly it in some sense you know I shouldn't be saying that because Asus do make a point of generally not putting out subpar PCBs or at least not PCBs worse than the reference design before that this video is brought to you by thermal Grizzly makers of the conductor not liquid metal that we recently used to drop 20 degrees off of our coffee leak temperatures thermal grizzly also makes traditional thermal compounds we use on top of the IHS like cryo not and Hydra not pastes learn more at the link below with the sort of frontier Edition PCB and the reference PCB for all the Vega cards being so ridiculous because like that PCB is on par with say the 290x lightning I you know I was thinking it would be pretty hard or very expensive for manufacturers like Asus gigabyte and all the rest to actually manage to produce a PCB better than the reference design or you know if that like do they actually care to produce a PCB better than the reference design so asus have actually gone ahead and made made some improvement they have still kept it a 12 plus one phase vrm design so 12 phase V core and one phase for the HBM so exactly the same as the reference card the vcore vrm is actually made up of two blocks of phases so you end up with this kind of weird layout which if you've seen the reference cards then this isn't that surprising it's still an L shape except it's further away from the GPU core and it's not completely wrapped around it so you know it's slightly different layout it's a bit further that could increase a voltage drop across the power plane to the GPU core which you know the voltage controller can't compensate for that in which case it's just a little decrease in vrm efficiency because you're burning some extra power on the actual power plane itself the benefit though to this sort of more spaced out vrm design that the Strix is taking is that this should be a lot easier to cool as this vrm has a lot more surface area and a lot more copper to it as in you literally have just all this copper up here which you know it's it is covered in a bunch of PCB masks so it's not the best not the best heat sink ever but it's still better than nothing so and the same goes for the power planes themselves as well as you know just all the extra area that this vrm gets which I imagine was done by ASUS InDesign specifically to try improve vrm cooling for the card so that's the V Corps of erm layout now below that you find the HBM vrm which that's the single phase and that's exactly the same as the reference design as well in terms of how many phases it has you also have on the PCB these three minor rails and this should be the auxiliary so that's the memory controller voltage this sits typically at 0.9 volts then above that you find the which well I'm not actually sure which of these is what because I do not have the card in hand but I'm going off of what a frontier Edition or a reference Vega 64 Vega 56 PCB would have here and the layout because aces haven't really changed what is where as the vcore vrm is still in the general area of where the vcore vrm is on the reference card the HBM is basically in the same place relative same place as on a reference card so I assume these also didn't change places in which case this one up here is the VPP so this is a supporting voltage for the HBM it sits at 1.8 volts and it does basically nothing for overclocking you do not have to worry about that and above that you have display drive which again does nothing for overclocking on some past generations of AMD cards that the display drive voltage could help with black screen issues when running liquid nitrogen Vega has zero issues I have very low temperatures it actually really really likes them except for the part where the driver currently doesn't let you set a clock so high enough to max the card out at low perjures so yeah but nonetheless those are the three minor volt three minor VRMs on the board and they're not really that important I just point them out because people because more details more better obviously anyway let's move on to the details of the things that actually matter the V Corps v RM and Asus did actually manage to make a slight improvement over the reference design while keeping the exact same phase count and the difference is all down to the MOSFETs or in this case power stages the reference design uses International rectifier direct FETs a sixty eight ninety four and a 68 eleven this right here uses international rectifier power stages these are the IR 35 55 ms these are 60 amp power stages from international rectifier they're pretty much the second best power stage that international rectifier makes asus loves using these they use them on a bunch of other boards they use them on a bunch of GPUs actually they use them on like all their GPUs that are higher-end and there are really nice power stages so the end result of actually this choice of power stage on this card is that in comparison to the reference PCB for 1.2 volts output 300 kilohertz switching frequency and an assumed because again I do not have the card in hand I can't check but I assume this card runs seven volts gate drive because if it runs five volt gate drive then actually there's no improvement over the reference design at all but if it does run seven volts gate drive then this vrm is capable of pushing 200 amps at 18 watts of heat dissipation which is the same as a reference PCB 300 amps so this is where normally a Vega boils this is the normal Vega bias current limit so 300 amps it can push at about 25 watts of heat dissipation this is actually about nine Watts better than what the reference card can do for 300 amps current output 400 amps you're looking at about 42 watts of heat output which is again a bow 11 watts better than the reference design 500 amps this car does at 65 watts the reference design would do at about 75 watts so you know it is a small but constant improvement over the reference design in terms of vrm efficiency as it doesn't put out as much heat as the reference design would and this maxes out at a whopping 720 amps theoretical as long as you can keep the vrm at well below 125 degrees centigrade well you could go higher than that but really you shouldn't go above 125 so you don't want to measure the anywhere in the vrm area being over 125 degrees centigrade it could push 720 amps however the vrm would at that point produce about a hundred and thirty-five watts of heat which is quite frankly uncool about I mean even this 65 watt figure is going to be an issue to cool so that 135 one is purely theoretical but that is the theoretical limit if you put a big enough heatsink on this vrm and say a delta fan for airflow so yeah it is a is a you know it's not a massive improvement over the reference design and it's not necessarily gonna change the if you were running this card and the reference design on liquid nitrogen I imagine you probably wouldn't see any difference between either of them as this is roughly 10 watts more efficient than the reference design so you know it's not a huge difference there but it is it is a small improvement and you know the extra efficiency for basically for free is all I imagine a welcome thing for anybody using Vega even if it's like 10 watts compared to a reference card which really isn't such a huge difference so yeah the vcore vrm definitely I have no complaints about it's still a 12 phase and you know I could say yeah maybe they should have gone for a 16 phase except the thing is with the voltage controller that AMD requires Vega which is this chip right here and that chip is an IR three five two one seven this is a six plus two phase voltage controller maximum you can't actually run more phases than six plus two out of it so basically for any P via PCB while any manufacturer of GPUs using this voltage controller basically means that they can only do vrm designs like twelve which would be two times six as this vrm is obviously using doubling you can't buy a twelve phase voltage controller and in this case the doubling is done by these chips right here those are ir35 99s they're really not that like these aren't anything intelligent they literally just take a PWM signal and send every other PWM pulse to the alternating phase so you know one time one so in twelve okay you know what I think you got it the first time I'm not gonna bother trying to explain how the how it interleaves the signal but basically it also does cut the switching frequency in half so the thirty five thirty five to one seven would be pushing out say six hundred kilohertz to the 3599 and that would then push 300 kilohertz to the two phases in front of it so something like that for each of them but anyway the the rme setups basically are restricted to say 212 which is a 2 x 6 or a 16 which would be a 4 x 4 this would actually be problematic this could arguably end up with worse voltage a voltage regulation results than the 12 phase and worse efficiency than the 12 phase because if you're running a 4 because at that point the 3 5 2 1 7 isn't seeing a 6 phase vrm and seeing a four phase and it would be running all 16 phases like therefore so you could end up in situations where like 1 out of the 16 phases is taking way more load or very little load because the voltage controller has very little control over balancing them all ultimately any large you know doubled up phase design is going to run run into that issue as well but at least you'll have you know at least you'll have groups of four more groups of four balanced properly instead of just four groups of four balanced properly because it does end up just averaging all four phases together and sees them as one single big phase so that that's a bit of an issue so you're stuck with like a sixteen phase which is less than ideal or a twenty phase which that shouldn't be or a twenty phase which would be a four times five which wouldn't be that much worse than a you know that would are probably always end up being better than a two times six but that's twenty phases that's really expensive to implement and takes up a ton of PCB space and nobody's gonna bother with that it's just not worth it and then finally you could also go for a twenty four phase but you know if the twenty phase is expensive and ridiculous why would anybody do a twenty four so I'm really really not surprised we're probably gonna see all the Vega cards using twelve phase vrm designs I'm hoping that they all use twelve phases of the RM designs I hope nobody goes below that because that would be a problem but you know I really doubt we're gonna see a Vega with more than called phases just because the next step up that makes kind of sense with this voltage controller is twenty and twenty doesn't make sense it's too many phases so yeah that that's a kind of unfortunate situation forgive a guy if it wasn't stuck competing against a 10 atti I sure I'm sure that we could see something like a you know 20 or 24 phase of erm equipped Vega lightning edition or you know matrix additions I'm crazy ln2 overclocking card but as it stands the 1080 Ti is just better in the competitive overclocking scene so nobody's gonna bother with some or ridiculously overbuilt vega card not that that's much of an issue as the 12 phase will be almost like the current vrm for Vega the reference one is already plenty and even this marginal improvement just means more plenty so I'm not really that disappoint that we won't be seeing you know more overkill though twelve phases isn't really overkill it's just kind of enough as Vega does get ridiculously power-hungry once you start pushing the card the HBM vrm is a you know it's still a single phase this could be a two phase I don't see the pike I don't see a problem with just keeping it a single phase because the HBM - really doesn't pull that much power from this VRM anyway it's going to be in the range of maybe twenty possibly 30 amps at one point three five volts you know it's very low current it doesn't you know you don't need two phases and ultimately the HBM currently clocks great I've also tried extra capacitors on my own Vega cards for the HBM vrm they haven't made a single they like they didn't make a tiny amount of difference no improvement in overclocking range basically tells you that the single-phase hpm BRM is plenty in terms of voltage regulation to get the clocks as high as they can go at least without raising HBM voltage the only issue is raising HBM voltage is really risky as historically HBM has been very very sensitive to voltage and degrading very rapidly at higher voltages as one point 4 volts only took a few months to degrade at the HBM one on a few reacts so over volting HBM - i wouldn't really recommend either and so i don't see a problem with the single-phase vrm design here as you know there's not there wouldn't really be any benefit to a to face now for that 20 amps current output the RM would produce about 1.7 watts of heat again assuming 300 kilohertz switching frequency and seven volts gate drive and at 30 amps it would produce about 3.2 watts of heat so yeah yeah it's negligible amounts of heat you don't really have to worry about this VR I'm either no issues here so yeah that's that's the card in terms of the VRMs and you know good on Asus for actually making some improvement over the reference design because the reference design was already really nice this is slightly nicer it's a little bit more efficient not a huge amount but you know every little kind of helps and they did also make some nice additions to the PCB but not won over them the cards does still have a dual BIOS which which means if you buy a Vega 64 I mean Vega 56 Strix Edition flashing a Vega 64 BIOS will be pretty much breaks free as if you screw up the flash for whatever whatever reason you can use that BIOS which to recover very very easily from the from the you know failed flash now one addition that asus has made to the card which is kind of cool ignoring of course all the fan headers and I assume RGB headers and more RGB headers is these over-voltage points up here these are really like I'm not sure if they're enabled they probably require some extra soldiering elsewhere on the card as that's usually what Asus does for these to make sure that you know you can't really use these without some specific information from Asus about how to use these but ultimately these allow you to based on their descriptions raised core voltage memory voltage so that be the HBM and the VCO VCI so that should be the VDD CI so that would be the auxiliary voltage over here this is really helpful because as of right now the software voltage limitations for Vega are far too low for anybody running say well if you're running water cooling I think you're still probably gonna want to stay with with it under 1.3 volts but if you go on something more extreme like say dry ice liquid nitrogen you're gonna be looking at voltages like at least one point three five volts and they massively help the card like this card Allah loves the extra voltage even on air cooling and water cooling the card does benefit from extra voltage it's just a case of it gets really really really inefficient in terms of power consumption it just burns like you know everybody already says Vega is power-hungry at 1.2 volts once you start cranking up the voltage these cards will very quickly exceed say 500 watts power draw on the eight pins so yeah you know cool for extreme overclockers not really useful for any other usage so yeah that's it for this PCB breakdown thank you for watching like share subscribe leave a comment down below if you have any questions or well thoughts and please support gamers Nexus on patreon if you would like to see more videos like this you can head over to my channel called actually hardcore overclocking where I do PCB breakdowns and a whole bunch of other extreme overclocking related stuff that's it for the video and goodbye