The PCB Breakdown: Uncovering the Secrets of Nvidia's Favorite Voltage Current and Voltage Monitoring Chip

As we examine the PCB of this high-end GPU, it becomes clear that Nvidia's favorite voltage current and voltage monitoring chip is designed for power consumption management. This chip is incredibly easy to trick, and its functionality can be exploited by overclocking enthusiasts looking to push their cards to the limits.

One of the most striking features of this chip is its ease of modification. With a simple desoldering process, it's possible to bypass the normal monitoring circuit and feed in fake voltages or adjust resistances to alter power readings. This can be used to remove power limits entirely, allowing overclockers to push their cards to extreme levels without worrying about overheating or voltage instability.

Another method of modification involves using liquid metal or conductive ink pens to draw across the shunt resistor. By doing so, the electrical resistance of the shunt is lowered, and the chip's ability to monitor power consumption is severely impaired. As a result, the driver will not be able to accurately track the power being pulled by the card, allowing it to clock higher and higher without any limits.

The PCB itself appears to be designed for high-performance applications, with a 3-to-1 ratio of current monitoring. However, it's worth noting that the chip's functionality can be greatly impacted by modifications made to its internal circuitry. The presence of additional resistors, voltage dividers, and shunt resistors all contribute to the chip's ability to accurately monitor power consumption.

One interesting feature of this PCB is the loss of unoccupied solder pads, which may indicate that it's an engineering sample or a low-production-volume component. Additionally, there are several instances where pins appear to be missing, but these can likely be attributed to the design process rather than any significant issue with the component itself.

In terms of performance, this chip is undoubtedly one of the best in the business, capable of delivering incredible power handling and voltage regulation capabilities. However, it's worth noting that Nvidia may have intentionally chosen not to implement a 16-phase VRM system, instead opting for an 8-phase configuration. While this may not make a significant difference in terms of performance, it's an interesting design choice that could potentially impact the card's overclocking potential.

The PCB is indeed cramped and features a large number of unoccupied solder pads, which can be seen as both a positive and negative aspect of its design. On the one hand, this may indicate that Nvidia chose to prioritize other aspects of the component over power delivery, such as overclocking headroom or thermal management. On the other hand, it's clear that the designers have intentionally left room for modifications and customizations, which could be a major selling point for enthusiasts looking to push their cards to extreme levels.

In conclusion, this PCB breakdown has provided a fascinating glimpse into the inner workings of Nvidia's favorite voltage current and voltage monitoring chip. By understanding how these chips work and how they can be modified, overclockers can unlock incredible performance potential from even the most powerful GPUs. While there may be some limitations and trade-offs involved, the rewards are well worth it for those willing to push their cards to the limits.

Supporting information:

* A detailed guide on how to modify this chip's functionality, including tutorials on soldering, liquid metal application, and shunt resistor modification.

* An in-depth analysis of the VRM system used in this component, including its design decisions and potential implications for overclocking performance.

* A comparison with other high-end GPUs, highlighting their unique features and strengths when it comes to power delivery and overclocking capabilities.

**Additional Resources:**

* "How to Overclock Your GPU: A Beginner's Guide" - Learn the basics of overclocking and how to apply them to your own system.

* "The Ultimate VRM System Guide" - Dive deep into the world of VRMs, including design decisions, component selection, and optimization techniques.

* "Overclocking Live Streams" - Join our live stream events for the latest news, updates, and expert analysis on overclocking and GPU performance.

"WEBVTTKind: captionsLanguage: enhey guys build Zoid here from actually hardcore overclocking and today we're taking a look at the Nvidia Tyson V PCB so you know without any delay let's get right into it we'll start with the various VRMs that are located on the card and then we're gonna go into some of the details for the vcore vrm the HBM vrm and also how you could lift the power limit on this thing because well it chokes on the power limit pretty hard out of the box even after you increase it you know to the limit that the software allows that this card is still very very very very power limited as in video opening allows you to increase the stock power limit by another 20% before we get into those this coverage is brought to you by EVGA and their 1080 tee is c2 which we've recommended fairly highly for its build quality and the icx sensors which are kind of fun to play with you can check our full sc2 review for the 1080i if you're curious to learn more or you can click the link in the description below to find the product page for the 1080i sc2 anyway let's get into the VRMs starting with the very oddly laid out V core vrm so you get one set of phases for v core right here and then you go across the card and get another set of phases for v core on the other side this vrm layout is you know as pretty much optimal it gets the vrm as close to the load which would be the GPU here as is possible for a you know PCB format of this kind and the reason why getting the vrm as close to the GPU core as possible is important is that well the less distance the current from the vrm has to travel the less voltage drop you get across the power plane and the better you get an better transient response you get because obviously if there's a big change in power consumption from the GPU core that change in current demand has to take some time to propagate so the longer your power plane is the bigger the delay between the current draw from the GPU core and it getting to the actual capacity banks and the inductors which results in slightly worse voltage regulation so this vrm layout is definitely really really good behind that we find the V the HBM vrm now the HBM vrm has kind of where the V curve erm sort of got layout PCB layout priority the HBM v RM is actually behind the V Corps V RMS one of the V Corps Buffay you know parts of the V Corps v RM this is really awkward for me to deal with because most cards you just have one big block even if that block might not necessarily you know like Vega has an L shape but this is two separate blocks but the HBM erm is located behind one of the sets of ecore phases and this does actually negatively impact the voltage drop from the HBM v RM luckily the HBM isn't super you know high power draw or anything so it's not you know it's not as critical to have the HBM v RM as close to the card as possible but it is noticeable as the voltage coming out of the HP mprm is pretty high considering what actually ends up getting to the HBM stacks that you have on the GPU so you know that's the HBM v RM and then you get the 2 minor v rms up here so this one and this one we will not be going into the details of these i'm not actually sure which one does what but basically this card requires two more voltages in order to function one of them is the PEX voltage this is typically between 0.9 volts and 1 volt this powers the PCI interface as well as some of the PLL's inside the GPU core this voltage is completely useless on air cooling I mean obviously the card won't work without it but worrying about tweaking this voltage is pointless it doesn't it won't help you overclock any higher the other world did you get is the 1.8 volts VPP as well as bias voltage so Nvidia cards have a the bios system on and video cards runs on 1.8 volts the HBM also has a supporting voltage of 1.8 volts this is similar to gddr5 XO is not a huge difference from what you would normally see on an nvidia card so those are your two minor rails I'm not sure which one does what but there's go these have to be present on this card to function another interesting thing you kind of find on this PCB is well is these two groups of four MOSFETs right here I am NOT one like I do not have confirmation on what these actually do and unfortunately the card is in the US and I'm in the I'm in the you know in Europe so I've not had a chance to probe this myself and well trying trying to get you know long range measurements just does not work for these kinds of things so I'm not a hundred percent certain what these do but based on the fact that there's an inductor right in front of them and that they're well you know between the PCIe eight pins and the V curve erm I have a sneaking suspicion that Nvidia these might be actual sort of sub V RMS which would be used for boosting vrm efficiency because when you have well basically we are converting doing DC DC voltage conversions from say 12 volts to anything less than 1.2 volts tends to be pretty inefficient due to the very low duty cycle that you end up running and you know the G V 100 here it runs on voltages anywhere from 0.75 volts at full load to 1.09 3 volts full load so it's very much in that sort of it's outside the sweet spot for your 12 volts down conversion and Intel for example deals with this on their mult very high core count Zeon's which run on very low voltage very high current they deal with that by having a fully integrated voltage regulator into and video doesn't have one so what I assume they're doing is there this this set of MOSFETs right here and this inductor basically would step down 12 volts to something like 8 volts and that would slightly boost the actual efficiency of the main V curve erm without really costing you a lot of power because this is a low current erm high voltage high voltage output so this can be very efficient without too much difficulty and it could significantly improve the actual efficiency of the conversion of the of the B curve erm but without the card in hand I I can't say for sure that that's exactly what it does but you do get two of these which would nicely line up with the fact that you have to you know power connectors so yeah just that's just my theory on what they do I'm not 100% certain so with that out of the way let's get right into the actual sort of meat of this which is the details of the vcore vrm and the HBM erm so the vcore vrm is an absolute monstrosity it may look like a one two three four five six seven eight nine ten eleven twelve thirteen fourteen fifteen sixteen phase V core vrm design however it's not actually a sixteen phase at least not with the way it's driven you do have sixteen inductors you do have 16 power stages you do not have sixteen interleaved PWM signals back of the card here you can actually see that the card has mounts for PWM doubler chips but they're unoccupied so this card how it deals with the vrm is that assume this voltage controller right here is 4 v core unfortunately Nvidia decided that the voltage controllers on this card are two over the same chip so I'm not sure which one does what because either it could drive either vrm they both have the right phase counts for it but either way this is a monolithic power systems which I'm just going to abbreviate as MPs which is their company logo this is an MP 2 8 8 8 I'm not sure about the feature set of these chips I do know they do go up to 8 phase v r8 phase output and they do switching frequencies in excess of 300 kilohertz however there is no public data sheet so I can't really give you any details on what other features these offer but either way on this card they are running you know one of them is running eight phase-- mode 300 kilohertz for the beaker of erm the other one is running two-phase mode for the HBM but basically the card is essentially a massive eight phase because of the lack of doublers so normally if you had a doubling scheme you basically would have a PWM signal is getting multiplied into sixteen by the actual doublers and here what you get is you basically have eight PWM signals and well if we have like say we have one PWM signal go here pw1 one then that signal also goes to want one other chip so essentially why this looks like an eight phase-- is that you have two phases turn on at the same time however this does still have benefits the huge number of power stages and inductors means you get very good thermal performance because the vrm is absolutely like it has a huge amount of surface area the other thing is the actual power stages are extremely high-end and you basically cut the current load on each power stage in half by basically having two of them turn on at the same time so the this vrm is it ends up being ridiculously powerful because these aren't any old you know any old power stages now these are Fairchild Semiconductor my C is terrible so these are Fairchild Semiconductor FD mf3 one seventies and the thing is these are 70 amp smart power stages so just SPS for short and the reason why these are well so they're 70 amps max output well they won't actually shut down at 70 amps they shut down at 80 amps so that's part of the smart power stage functionality they have a an overcurrent protection of 8880 amps and they have a built-in over temperature protection of a hundred and thirty six degrees centigrade so basically once they hit one hundred and thirty six degrees centigrade they raise our flag that hey the vrm is kind of overheating here you might want to shut it down the OCP on the other hand shuts down the power stage the moment the current value is exceeded for the time period specified in the datasheet so basically you don't have to worry about this vrm you know destroying itself or anything because it has built-in temperature monitoring built-in current monitoring and built-in protection features right on the power stages themselves they are capable of sustaining up to 70 amps output however at that point their heat output is about 14 watts each which you know that's just unsustainable so if you had one of these chips and a big enough heatsink for it yeah you could actually push 70 amps through it in a 16 phase vrm design on a cramped PCB like this that is not gonna work luckily the Titan V doesn't really pull that much power it does do it very low voltages but it really doesn't pull that much power on the stock power limit you're only looking at about 200 amps coming through the vrm because I'm assuming there's another 50 watts of power dedicated rest to the rest of the PCB and you're looking at about you know so you're looking at about 200 watts going into the GPU core which around 1 volt works out to about 200 amps current output obviously at the lower voltages it would actually be more current output on the with the power limit max out so with plus 20% power limit you're gonna be looking at the whole card pulling about 300 watts I'm gonna assume that is still pulling the same amount of power for the fan and the HBM because we're only looking at the core power consumption here because the route I'm that's an assumption but is it probably a pretty good one in general that works out and then with the lifted + 20% power limit you basically be looking at something like 250 amps output for the core vrm which for these monstrously overpowered smart power stages is really not a workout which is really really good because it means this vrm ends up being ridiculously efficient and in fact it ends up being a ridiculous overkill I think this might be because the this PCB is probably used for tests laws as well that go into outright data centers and well data centers want tip-top reliability so you know anything less than massive overkill is just kind of not acceptable there so the end result is that at these typical current figures both that around one volt output because really the on stock the voltage output and fluctuates so much it's really hard to say and on on the plus 20% power limit actually ends up hovering around point 0.99 three volts well I'd like to say at 1 volt you know for these current outputs I have actual heat dissipation numbers for these power stages unfortunately the datasheet for these power stages does not specify how output voltage impacts efficiency so the datasheet is entirely SPECT at 1.8 volts however at 1.8 volts 200 amps output you'd be looking at 19 watts of heat output for the entire v core vrm so spread across this much air surface area that's going to be nothing and I would like to point out that these power stages should get more efficient at lower output voltages because this is uh they're optimized for ten percent to fifteen percent duty cycle that 1.8 volts output is on the high side of that ten to fifteen percent so realistically you're gonna be looking at less than nineteen watts of heat output at stock and four plus twenty percent power limit you're gonna be looking less than twenty four watts heat output because again it's 1.8 volts 300 kilohertz switching frequency for the whole vrm which I did mention earlier so really really efficient unfortunately it's hard to say how this compares directly against some other top-end you know say gtx 980ti VRMs like say a hall-of-fame because for those i actually have these sheets that work at lower voltages and yeah it's hard to do the comparison when these are only rated at 1.8 volts they should be right about in line with the other power stages that are used on say a gtx 10a DTI hall of fame or a gtx 980ti kingpin edition in fact this VRM is more powerful than what you would find on a gtx 980ti kingpin edition and probably more powerful than what you'll find on the hall of fame - just because these power stages do max out at 70 amps instead of the 88 instead of the 60 amps that the International rectifier is 35 75 and the 35 55 used on the Hall of Fame and the kingpin max out at so basically NVIDIA has for the first time built a card where I'm gonna say you don't need to upgrade the vrm on this you're gonna need to strap a pretty big heat sink to the vrm maybe get a lot of airflow so very high rpm fine if you're gonna take this on liquid nitrogen which I doubt anybody will do because you know if the card is three thousand dollars but on the off chance that you know if you want to take one of these on liquid nitrogen you don't you shouldn't have to worry about the vrm and if there is a vrm problem on this thing there's not really an alternative unfortunately they don't make a 32 Phase II power that I'm aware of at least so there is not much you know there's not much room to upgrade the VR and this is pretty much the peak of GPU vrm power capability certainly I think it would be better if it interleaved and maybe ran a higher switching frequency after interleaving but at least in terms of raw power output this thing is an absolute beast and to put a real number on that well let's say you wanted to push 600 amps through this monstrosity at again that 1.8 volts figure it would only produce about 90 watts of heat which is in line with every other top-end 1080 Ti as well as any other top-end extreme overclocking GPU that I've ever seen in terms of erm heat dissipation so yeah this thing is amazing absolutely amazing in terms of the B curve erm so nothing to complain about unfortunately except you know it could actually get 16 phases I think here 16 phases would be justifiable compared to like 780ti is in you can you can get away with 10 you can get away with 12 or 16 but here I it's liking those those doublers don't cost that much there's footprints on the PCB already for them why aren't they there I have no idea but power-wise I I've got no complaints now then let's move on to the HBM VR I'm the HBM vrm as you can clearly see is a 1 to phase affair the third phase took a vacation I think along with the HBM stack that is actually disabled on the Titan movie as the GV 100 chip and it's full implementation this would come with for working HB m stacks this version only has three so I assume that one of the phases got taken away because well you're not running all four stacks of HP m you're only running three so you can you know you can get rid of that extra phase the power stages are exactly the same as what you had for the vcore vrm so again massive freaking overkill for a assumed output of about 35 amps for the HB m which i think might be a little bit high or maybe I'm overestimating the vcore vrm output currents for 35 amps at well again that 1.8 volts cuz the same power stage with the equally limited power datasheet you'd be looking at the power consumption below 40 watts for the CRM so in terms of at least efficiency there's nothing to complain about is it's a two-phase memory vrm so you know that this is more phases than a few reacts had for its HP m1 this is more phases than a Vega has for its HBM I can't complain about this again very very impressive controlled by the same monolithic power systems MP 2 8 8 8 voltage controller but this time configured for only two phases I imagine there's some versions of this card where it has the full three phases I assume that would be the tesla version so yeah again just Nvidia like where did this come from I'm super surprised cuz I'm used to seeing you know reference and video PCBs and recently they've gotten a lot better but most of the time it was a case of like this this is kind of impressive too especially on some dual GPU and video cards it was actually outright concerning like how long do you expect this vrm to last I imagine you know no overclocking allowed is is a policy that actually for example a GTX 590 doesn't support overclocking at all specifically because the vrm on that was so shoddy but this is just like this is up there this is up there with the Hall of Fame this is up there with the kingpin edition arguably this has the better vrm layout than basically any other extreme overclocking card because as I said before the you have the V curve erm flanking either side of the GPU core which there was some ln2 like there was some extreme overclocking GPUs and high phase count V RMS that we've seen on an exam 1080 T eyes 980 T is and a few other cards where they would actually like put two rows of phases behind each other which like it works out great for your power capabilities but you end up in a situation where you're wasting power on on basically just getting power from the furthest the sets of phases to the GPU core so this thing is just like amazing absolutely amazing I I'm not I mean I I've never had any doubt Nvidia could design a good vrm I never thought they would want to but they have this is awesome now then let's go and take a take the chains off and lift those power limits so the cardia has an interesting power monitoring situation there's actually way more shunts on this than you would expect there's only three 12 volts inputs you have plus 12 volts on the PCIe you have plus 12 volts on the six pin you have plus 12 volts on the eight pin but for some reason there's three current monitoring shunt resistors so you have one down here which would be for the PCIe slot this one up here which is probably for this expen this one down here which is for the eight pin and you have an i na three to two one Texa this is a chip from Texas Instruments Nvidia's favorite voltage current and voltage monitoring chip for basically monitoring power consumption so this thing is really easy to trick I've done it on a 1070 you can actually do all kinds of if you feel like doing a lot of soldiering you can basically de-solder the monitoring circuit that this normally runs on the actual power inputs and you can replace it with well you can set up a voltage divider from plus 12 volts which you can source from the PCIe slot or the eight pin or the six pin you set up a voltage divider and you can feed they're completely fake voltages into the actual monitoring inputs that will remove all power limits you can also use resistors varying sizes to reduce power readings by up to five times and the last option which is probably the option most people will opt for is well most people there's not that many people with this so it's gonna be like what three people on earth they're gonna do this but the last option is that you put liquid metal across the actual shunt resistor or you can go with with conductive ink pens and you can basically just draw across the shunt resistors this lowers the electrical resistance of the shunt and the way shunts work is the well by a three to one just monitors the voltage drop across the shunt resistor and from that voltage drop it calculates the amount of current going into the card and also since it's monitoring the voltages it does current times voltage e were called power consumption and that's how the in-video driver knows how much power the card is pulling and if it needs to lower the core clock or if it can just go wild well if you tweak the resistances on the shunts or you know just completely fake all the measurements for the ia three two to one the driver will not know how much power the card is pulling and as long as it thinks it's pulling less power than the maximum than the power limit it'll allow the card to clock higher and higher and higher and let you push basically as much power as you can possibly want so that's the shot mod there's also this shunt down here but I would recommend that you basically don't touch that one because I can't figure out what it's hooked up to and why it's there if I had the card in hand I could probably figure out what it's for but as it is right now I have no idea so I would just not bother with that one for safety's sake so yeah that that covers the PCB the only other side note I have about this thing is there is a loss of unoccupied solder pads on this it almost looks like an engineer like is it kind it just kind of looks like an engineering sample PCB but I guess this thing is so low production volume or maybe it's just since it's a professional card it might be because they want to be able to check for any damages better but yeah this thing this PCB is really cramped and there is a lot of unoccupied solder pads which I find really odd but there's no negative there's no real downside to that there's also the missing a pin up here but we've been seeing that on a lot of a high end in video cards for a while now so I don't find that surprising even remotely but yeah this thing you know if you're buying a $3,000 GPU especially this one you can be at least certain that for once you're getting a vrm that it's more or less worth the $3,000 asking price on the card I still say I would like to see doublers but you know power capability wise this is this is amazing but yeah I'd like to see doublers because with more interleaving phases you basically get better or vrm ripple voltage ripple but going from eight phases to 16 phases tends not to make as much of a difference as going from say three to eight so and video probably calculated that past this phase account it doesn't really make a difference for their usual applications and decided that we're just gonna keep the efficiency of the 16 phase monstrosity that we built and you know that's that's fine but I'd still like to see a 16 phase and that's it for this PCB breakdown thank you for watching like share comment leave a comment down below or any questions if you would like to support we do here at gamers Nexus and we have a patreon you can find a link down in the description below for that and if you would like to see more overclocking related videos I have my own channel called actually hardcore overclocking I do other PCB breakdowns for GPUs motherboards and I also do various modification guides experiments overclocking live streams and all kinds of other overclocking related content thank you for watching and see you next timehey guys build Zoid here from actually hardcore overclocking and today we're taking a look at the Nvidia Tyson V PCB so you know without any delay let's get right into it we'll start with the various VRMs that are located on the card and then we're gonna go into some of the details for the vcore vrm the HBM vrm and also how you could lift the power limit on this thing because well it chokes on the power limit pretty hard out of the box even after you increase it you know to the limit that the software allows that this card is still very very very very power limited as in video opening allows you to increase the stock power limit by another 20% before we get into those this coverage is brought to you by EVGA and their 1080 tee is c2 which we've recommended fairly highly for its build quality and the icx sensors which are kind of fun to play with you can check our full sc2 review for the 1080i if you're curious to learn more or you can click the link in the description below to find the product page for the 1080i sc2 anyway let's get into the VRMs starting with the very oddly laid out V core vrm so you get one set of phases for v core right here and then you go across the card and get another set of phases for v core on the other side this vrm layout is you know as pretty much optimal it gets the vrm as close to the load which would be the GPU here as is possible for a you know PCB format of this kind and the reason why getting the vrm as close to the GPU core as possible is important is that well the less distance the current from the vrm has to travel the less voltage drop you get across the power plane and the better you get an better transient response you get because obviously if there's a big change in power consumption from the GPU core that change in current demand has to take some time to propagate so the longer your power plane is the bigger the delay between the current draw from the GPU core and it getting to the actual capacity banks and the inductors which results in slightly worse voltage regulation so this vrm layout is definitely really really good behind that we find the V the HBM vrm now the HBM vrm has kind of where the V curve erm sort of got layout PCB layout priority the HBM v RM is actually behind the V Corps V RMS one of the V Corps Buffay you know parts of the V Corps v RM this is really awkward for me to deal with because most cards you just have one big block even if that block might not necessarily you know like Vega has an L shape but this is two separate blocks but the HBM erm is located behind one of the sets of ecore phases and this does actually negatively impact the voltage drop from the HBM v RM luckily the HBM isn't super you know high power draw or anything so it's not you know it's not as critical to have the HBM v RM as close to the card as possible but it is noticeable as the voltage coming out of the HP mprm is pretty high considering what actually ends up getting to the HBM stacks that you have on the GPU so you know that's the HBM v RM and then you get the 2 minor v rms up here so this one and this one we will not be going into the details of these i'm not actually sure which one does what but basically this card requires two more voltages in order to function one of them is the PEX voltage this is typically between 0.9 volts and 1 volt this powers the PCI interface as well as some of the PLL's inside the GPU core this voltage is completely useless on air cooling I mean obviously the card won't work without it but worrying about tweaking this voltage is pointless it doesn't it won't help you overclock any higher the other world did you get is the 1.8 volts VPP as well as bias voltage so Nvidia cards have a the bios system on and video cards runs on 1.8 volts the HBM also has a supporting voltage of 1.8 volts this is similar to gddr5 XO is not a huge difference from what you would normally see on an nvidia card so those are your two minor rails I'm not sure which one does what but there's go these have to be present on this card to function another interesting thing you kind of find on this PCB is well is these two groups of four MOSFETs right here I am NOT one like I do not have confirmation on what these actually do and unfortunately the card is in the US and I'm in the I'm in the you know in Europe so I've not had a chance to probe this myself and well trying trying to get you know long range measurements just does not work for these kinds of things so I'm not a hundred percent certain what these do but based on the fact that there's an inductor right in front of them and that they're well you know between the PCIe eight pins and the V curve erm I have a sneaking suspicion that Nvidia these might be actual sort of sub V RMS which would be used for boosting vrm efficiency because when you have well basically we are converting doing DC DC voltage conversions from say 12 volts to anything less than 1.2 volts tends to be pretty inefficient due to the very low duty cycle that you end up running and you know the G V 100 here it runs on voltages anywhere from 0.75 volts at full load to 1.09 3 volts full load so it's very much in that sort of it's outside the sweet spot for your 12 volts down conversion and Intel for example deals with this on their mult very high core count Zeon's which run on very low voltage very high current they deal with that by having a fully integrated voltage regulator into and video doesn't have one so what I assume they're doing is there this this set of MOSFETs right here and this inductor basically would step down 12 volts to something like 8 volts and that would slightly boost the actual efficiency of the main V curve erm without really costing you a lot of power because this is a low current erm high voltage high voltage output so this can be very efficient without too much difficulty and it could significantly improve the actual efficiency of the conversion of the of the B curve erm but without the card in hand I I can't say for sure that that's exactly what it does but you do get two of these which would nicely line up with the fact that you have to you know power connectors so yeah just that's just my theory on what they do I'm not 100% certain so with that out of the way let's get right into the actual sort of meat of this which is the details of the vcore vrm and the HBM erm so the vcore vrm is an absolute monstrosity it may look like a one two three four five six seven eight nine ten eleven twelve thirteen fourteen fifteen sixteen phase V core vrm design however it's not actually a sixteen phase at least not with the way it's driven you do have sixteen inductors you do have 16 power stages you do not have sixteen interleaved PWM signals back of the card here you can actually see that the card has mounts for PWM doubler chips but they're unoccupied so this card how it deals with the vrm is that assume this voltage controller right here is 4 v core unfortunately Nvidia decided that the voltage controllers on this card are two over the same chip so I'm not sure which one does what because either it could drive either vrm they both have the right phase counts for it but either way this is a monolithic power systems which I'm just going to abbreviate as MPs which is their company logo this is an MP 2 8 8 8 I'm not sure about the feature set of these chips I do know they do go up to 8 phase v r8 phase output and they do switching frequencies in excess of 300 kilohertz however there is no public data sheet so I can't really give you any details on what other features these offer but either way on this card they are running you know one of them is running eight phase-- mode 300 kilohertz for the beaker of erm the other one is running two-phase mode for the HBM but basically the card is essentially a massive eight phase because of the lack of doublers so normally if you had a doubling scheme you basically would have a PWM signal is getting multiplied into sixteen by the actual doublers and here what you get is you basically have eight PWM signals and well if we have like say we have one PWM signal go here pw1 one then that signal also goes to want one other chip so essentially why this looks like an eight phase-- is that you have two phases turn on at the same time however this does still have benefits the huge number of power stages and inductors means you get very good thermal performance because the vrm is absolutely like it has a huge amount of surface area the other thing is the actual power stages are extremely high-end and you basically cut the current load on each power stage in half by basically having two of them turn on at the same time so the this vrm is it ends up being ridiculously powerful because these aren't any old you know any old power stages now these are Fairchild Semiconductor my C is terrible so these are Fairchild Semiconductor FD mf3 one seventies and the thing is these are 70 amp smart power stages so just SPS for short and the reason why these are well so they're 70 amps max output well they won't actually shut down at 70 amps they shut down at 80 amps so that's part of the smart power stage functionality they have a an overcurrent protection of 8880 amps and they have a built-in over temperature protection of a hundred and thirty six degrees centigrade so basically once they hit one hundred and thirty six degrees centigrade they raise our flag that hey the vrm is kind of overheating here you might want to shut it down the OCP on the other hand shuts down the power stage the moment the current value is exceeded for the time period specified in the datasheet so basically you don't have to worry about this vrm you know destroying itself or anything because it has built-in temperature monitoring built-in current monitoring and built-in protection features right on the power stages themselves they are capable of sustaining up to 70 amps output however at that point their heat output is about 14 watts each which you know that's just unsustainable so if you had one of these chips and a big enough heatsink for it yeah you could actually push 70 amps through it in a 16 phase vrm design on a cramped PCB like this that is not gonna work luckily the Titan V doesn't really pull that much power it does do it very low voltages but it really doesn't pull that much power on the stock power limit you're only looking at about 200 amps coming through the vrm because I'm assuming there's another 50 watts of power dedicated rest to the rest of the PCB and you're looking at about you know so you're looking at about 200 watts going into the GPU core which around 1 volt works out to about 200 amps current output obviously at the lower voltages it would actually be more current output on the with the power limit max out so with plus 20% power limit you're gonna be looking at the whole card pulling about 300 watts I'm gonna assume that is still pulling the same amount of power for the fan and the HBM because we're only looking at the core power consumption here because the route I'm that's an assumption but is it probably a pretty good one in general that works out and then with the lifted + 20% power limit you basically be looking at something like 250 amps output for the core vrm which for these monstrously overpowered smart power stages is really not a workout which is really really good because it means this vrm ends up being ridiculously efficient and in fact it ends up being a ridiculous overkill I think this might be because the this PCB is probably used for tests laws as well that go into outright data centers and well data centers want tip-top reliability so you know anything less than massive overkill is just kind of not acceptable there so the end result is that at these typical current figures both that around one volt output because really the on stock the voltage output and fluctuates so much it's really hard to say and on on the plus 20% power limit actually ends up hovering around point 0.99 three volts well I'd like to say at 1 volt you know for these current outputs I have actual heat dissipation numbers for these power stages unfortunately the datasheet for these power stages does not specify how output voltage impacts efficiency so the datasheet is entirely SPECT at 1.8 volts however at 1.8 volts 200 amps output you'd be looking at 19 watts of heat output for the entire v core vrm so spread across this much air surface area that's going to be nothing and I would like to point out that these power stages should get more efficient at lower output voltages because this is uh they're optimized for ten percent to fifteen percent duty cycle that 1.8 volts output is on the high side of that ten to fifteen percent so realistically you're gonna be looking at less than nineteen watts of heat output at stock and four plus twenty percent power limit you're gonna be looking less than twenty four watts heat output because again it's 1.8 volts 300 kilohertz switching frequency for the whole vrm which I did mention earlier so really really efficient unfortunately it's hard to say how this compares directly against some other top-end you know say gtx 980ti VRMs like say a hall-of-fame because for those i actually have these sheets that work at lower voltages and yeah it's hard to do the comparison when these are only rated at 1.8 volts they should be right about in line with the other power stages that are used on say a gtx 10a DTI hall of fame or a gtx 980ti kingpin edition in fact this VRM is more powerful than what you would find on a gtx 980ti kingpin edition and probably more powerful than what you'll find on the hall of fame - just because these power stages do max out at 70 amps instead of the 88 instead of the 60 amps that the International rectifier is 35 75 and the 35 55 used on the Hall of Fame and the kingpin max out at so basically NVIDIA has for the first time built a card where I'm gonna say you don't need to upgrade the vrm on this you're gonna need to strap a pretty big heat sink to the vrm maybe get a lot of airflow so very high rpm fine if you're gonna take this on liquid nitrogen which I doubt anybody will do because you know if the card is three thousand dollars but on the off chance that you know if you want to take one of these on liquid nitrogen you don't you shouldn't have to worry about the vrm and if there is a vrm problem on this thing there's not really an alternative unfortunately they don't make a 32 Phase II power that I'm aware of at least so there is not much you know there's not much room to upgrade the VR and this is pretty much the peak of GPU vrm power capability certainly I think it would be better if it interleaved and maybe ran a higher switching frequency after interleaving but at least in terms of raw power output this thing is an absolute beast and to put a real number on that well let's say you wanted to push 600 amps through this monstrosity at again that 1.8 volts figure it would only produce about 90 watts of heat which is in line with every other top-end 1080 Ti as well as any other top-end extreme overclocking GPU that I've ever seen in terms of erm heat dissipation so yeah this thing is amazing absolutely amazing in terms of the B curve erm so nothing to complain about unfortunately except you know it could actually get 16 phases I think here 16 phases would be justifiable compared to like 780ti is in you can you can get away with 10 you can get away with 12 or 16 but here I it's liking those those doublers don't cost that much there's footprints on the PCB already for them why aren't they there I have no idea but power-wise I I've got no complaints now then let's move on to the HBM VR I'm the HBM vrm as you can clearly see is a 1 to phase affair the third phase took a vacation I think along with the HBM stack that is actually disabled on the Titan movie as the GV 100 chip and it's full implementation this would come with for working HB m stacks this version only has three so I assume that one of the phases got taken away because well you're not running all four stacks of HP m you're only running three so you can you know you can get rid of that extra phase the power stages are exactly the same as what you had for the vcore vrm so again massive freaking overkill for a assumed output of about 35 amps for the HB m which i think might be a little bit high or maybe I'm overestimating the vcore vrm output currents for 35 amps at well again that 1.8 volts cuz the same power stage with the equally limited power datasheet you'd be looking at the power consumption below 40 watts for the CRM so in terms of at least efficiency there's nothing to complain about is it's a two-phase memory vrm so you know that this is more phases than a few reacts had for its HP m1 this is more phases than a Vega has for its HBM I can't complain about this again very very impressive controlled by the same monolithic power systems MP 2 8 8 8 voltage controller but this time configured for only two phases I imagine there's some versions of this card where it has the full three phases I assume that would be the tesla version so yeah again just Nvidia like where did this come from I'm super surprised cuz I'm used to seeing you know reference and video PCBs and recently they've gotten a lot better but most of the time it was a case of like this this is kind of impressive too especially on some dual GPU and video cards it was actually outright concerning like how long do you expect this vrm to last I imagine you know no overclocking allowed is is a policy that actually for example a GTX 590 doesn't support overclocking at all specifically because the vrm on that was so shoddy but this is just like this is up there this is up there with the Hall of Fame this is up there with the kingpin edition arguably this has the better vrm layout than basically any other extreme overclocking card because as I said before the you have the V curve erm flanking either side of the GPU core which there was some ln2 like there was some extreme overclocking GPUs and high phase count V RMS that we've seen on an exam 1080 T eyes 980 T is and a few other cards where they would actually like put two rows of phases behind each other which like it works out great for your power capabilities but you end up in a situation where you're wasting power on on basically just getting power from the furthest the sets of phases to the GPU core so this thing is just like amazing absolutely amazing I I'm not I mean I I've never had any doubt Nvidia could design a good vrm I never thought they would want to but they have this is awesome now then let's go and take a take the chains off and lift those power limits so the cardia has an interesting power monitoring situation there's actually way more shunts on this than you would expect there's only three 12 volts inputs you have plus 12 volts on the PCIe you have plus 12 volts on the six pin you have plus 12 volts on the eight pin but for some reason there's three current monitoring shunt resistors so you have one down here which would be for the PCIe slot this one up here which is probably for this expen this one down here which is for the eight pin and you have an i na three to two one Texa this is a chip from Texas Instruments Nvidia's favorite voltage current and voltage monitoring chip for basically monitoring power consumption so this thing is really easy to trick I've done it on a 1070 you can actually do all kinds of if you feel like doing a lot of soldiering you can basically de-solder the monitoring circuit that this normally runs on the actual power inputs and you can replace it with well you can set up a voltage divider from plus 12 volts which you can source from the PCIe slot or the eight pin or the six pin you set up a voltage divider and you can feed they're completely fake voltages into the actual monitoring inputs that will remove all power limits you can also use resistors varying sizes to reduce power readings by up to five times and the last option which is probably the option most people will opt for is well most people there's not that many people with this so it's gonna be like what three people on earth they're gonna do this but the last option is that you put liquid metal across the actual shunt resistor or you can go with with conductive ink pens and you can basically just draw across the shunt resistors this lowers the electrical resistance of the shunt and the way shunts work is the well by a three to one just monitors the voltage drop across the shunt resistor and from that voltage drop it calculates the amount of current going into the card and also since it's monitoring the voltages it does current times voltage e were called power consumption and that's how the in-video driver knows how much power the card is pulling and if it needs to lower the core clock or if it can just go wild well if you tweak the resistances on the shunts or you know just completely fake all the measurements for the ia three two to one the driver will not know how much power the card is pulling and as long as it thinks it's pulling less power than the maximum than the power limit it'll allow the card to clock higher and higher and higher and let you push basically as much power as you can possibly want so that's the shot mod there's also this shunt down here but I would recommend that you basically don't touch that one because I can't figure out what it's hooked up to and why it's there if I had the card in hand I could probably figure out what it's for but as it is right now I have no idea so I would just not bother with that one for safety's sake so yeah that that covers the PCB the only other side note I have about this thing is there is a loss of unoccupied solder pads on this it almost looks like an engineer like is it kind it just kind of looks like an engineering sample PCB but I guess this thing is so low production volume or maybe it's just since it's a professional card it might be because they want to be able to check for any damages better but yeah this thing this PCB is really cramped and there is a lot of unoccupied solder pads which I find really odd but there's no negative there's no real downside to that there's also the missing a pin up here but we've been seeing that on a lot of a high end in video cards for a while now so I don't find that surprising even remotely but yeah this thing you know if you're buying a $3,000 GPU especially this one you can be at least certain that for once you're getting a vrm that it's more or less worth the $3,000 asking price on the card I still say I would like to see doublers but you know power capability wise this is this is amazing but yeah I'd like to see doublers because with more interleaving phases you basically get better or vrm ripple voltage ripple but going from eight phases to 16 phases tends not to make as much of a difference as going from say three to eight so and video probably calculated that past this phase account it doesn't really make a difference for their usual applications and decided that we're just gonna keep the efficiency of the 16 phase monstrosity that we built and you know that's that's fine but I'd still like to see a 16 phase and that's it for this PCB breakdown thank you for watching like share comment leave a comment down below or any questions if you would like to support we do here at gamers Nexus and we have a patreon you can find a link down in the description below for that and if you would like to see more overclocking related videos I have my own channel called actually hardcore overclocking I do other PCB breakdowns for GPUs motherboards and I also do various modification guides experiments overclocking live streams and all kinds of other overclocking related content thank you for watching and see you next time\n"