You might like our Titan V Hybrid results: ua-cam.com/video/OZcxBP8rNCY/v-deo.html Shunt Article: www.gamersnexus.net/guides/3185-how-to-short-shunts-on-gpu-for-better-overclocks-titan-v Modmat: store.gamersnexus.net/modmat
So application of the liquid metal has potential to damage parts on the pcb. If one was to use tin/solder to short the shunts would there then be an issue of sending too much power to the pcie?
Steve - my concern with liquid metal here is what unknown amalgams it will form with today's lead-free solders since it's likely a gallium based alloy/mixture. Now's a great time to get smt rework skills! You might want to retouch the pads with real solder to dilute any remaining liquid metal and ensure longterm reliability. For shorting shunt resistors, the easiest / cleanest way is buy a roll of "solder wick" - a copper braid coated in dry flux which, when heated, is for desoldering. Heat the wick with a soldering iron (preferably Metcal brand but anything will work) and feed it some solder from a roll to soak up. when cool, cut off a length of solder-soaked wick equal to the resistor to be shunted and tack solder it to both sides (usually just touching it down with a hot iron works for short term use). To remove, grab in the middle with a fine tweezer, heat each end, and lift away. No nail polish needed. To make it more permanent, you can add some real solder. Liquid metal isn't a great bypass for low impedance parts in the milli-ohm size because internally, they have a larger cross section of metal than the film of liquid metal applied on top. By comparison, solder wick is huge.
Just a small suggestion. You can solder an equal value resistor in parallel with the existing one, and theoretically, you will be doubling the card's maximum allowed input current (Ampere) while at the same time remove the liquid metal problem from the equation. Soldering 2 SMT resistors in parallel is a relatively easy process and would probably allow for a more precise control of input current. Thanks for another great video.
I'm guessing i'll be showing my age here, but... We used to do this with high graphite pencils, if all we wanted to do was lower the resistance, and soldering on a wire, if we wanted to bypass the shunt completely. Pretty much no risk involved in the first, and very little in the second. Also, if soldering is to scary, there is actually proper "trace repair" paint, that is a lot more safe then gallium compound...
Amazing content as always. Thank you so much for going down the rabbit hole with us and dedicating valuable time to these niche kind of projects. It's awesome.
Wouldn't it be safer to solder a low value resistor (Say 0.25 Ohm) in parallel with the Shunt resistor to lower the overall resistance just a bit? It would prevent the dangers of liquid metals and also not trip any protection circuitry... If the value is not low enough, maybe buy a packet of these shunt resistors and 'piggy back them' to half the resistance value... might be worth the while (I just read more comments and found one pretty much like this... my bad)
I guess it comes down to how confident you are at soldering electronic components. It would be simple to remove the additional resistor/s if you wanted to revert it back to the original configuration. (But I am the type of guy that has soldered a PSU's 12V and Ground wires directly to a GPU to power it haha)
But in that way you have more control on the resistance of the shunt. If you change the electrical value of the shunt, you can calculate the power draw. And what if you solder a second shunt on top identical (physical and electrical) to the first shunt? It would be also be reversible, no?
Precisely, but I still think that you need to be confident enough with using a soldering iron before attacking a GPU of such cost.. You could even look at using a variable wire-wound resistor and make changes on-the-fly (would be awesome to see the results in real-time). But I understand GN not doing this. Not saying Liquid metal is a safer or wiser option but due to convenience it can be wiped off with isopropyl alcohol and a tissue pretty easily with minimal risk of permanently hurting the hardware.
Next level stuff here. I am not a hardcore over-clocking enthusiast, but as a moderately informed and life-long consumer and gaming/PC enthusiast I have yet to see this method used. It is hardly common knowledge. I imagine you will be bringing this method to light not only for most of us viewers, but to some of the other PC UA-cam gods.... I predict we will be seeing a lot more about shunting from other sources thanks to Gamers Nexus. Truly awesome stuff, guys
Screwing around with a gallium based thermal paste in order to affect change in resistance is the silliest shit I have ever seen... simply buy more of the resistors the board is using and stack a second one on top of each existing one with proper solder. This will exactly halve the resistance and double the current capacity.
Soldering on, even stacking, an SMD resistor that huge and then removing it again is child's play to anyone with the skills to be messing with component level modifications on a GPU in the first place... As for why I won't do it personally, see Steve's video on ROI. I don't have a platform to get any ROI on buying a marked up professional GPU.
No beginner should be fucking with something this dangerous, you need to entirely understand the component level workings of PCBs first. Once you have the skillset to safely do what's shown, you've got access to superior techniques.
I wouldn't midn seeing a volt mod for this card, but I really think at the end of the day, even with more voltage we won't see massive gains. It really looks like Pascal and Volta already operate at the best voltage and power level for optimal operation. We will start to see diminishing returns very quickly. If you are going to mod for either architecture, it seems thermals make the most difference, but yes I'm sure Voltage will allow for more clocks, how much though, probably 100-150 MHz at most.
These videos make me want to look up possible mods for Maxwell and give my 970 one hell of a send off. Maybe even duct tape a 512mb DDR2 stick on it to make it an actual 4Gb card.
Reason 1 for watching a ton of tech channels: They can do things to similar cards I'd get, and do things that I'd be too chicken to try out if I paid for my own expensive product :^)
Great Video GN! For those who don't know, thicker the metal, lower the resistance. Hence he got higher amps when he applied more because it was able to lower the resistance even more. The shunt and liquid metal act as parallel resistors hence even a thin layer lowered the total resistance to get 1, 2 amps more. (Physics FTW!) BTW there is a continuity test on most Multimeters with a triangle and a line as a symbol, the diode symbol for those who know. It will beep if the points are connected directly without any components between them. That helps more when trying to find connected pins, a definite beep.
Earned a sub! From a simple BIOS flash to upgrading a GTX690 to double memory by adding bigger surface mount memory chips and a custom BIOS, these mods are awesome (unless you brick your card)!
73/5000 Why do you risk liquid metal? You just can remove by desoldering The SMD resistors . Not that it is particularly small resistors such as 0402 . Remove the resistors And soldering shortened wire . 2 min work Max
It gets to me that people lose their minds when a little LM gets spilled around a delidded cpu die. But no one bats an eye when it gets painted on gpu shunt resistors.
I'm also confused by a non heatgun method? I can't see these 800 other comments? and it's not in the arrticle... Gallium spreads like mould, it could ruin the card.
+David Az I completely agree. Nothing more to add. Either short it out properly or set a desired resistance by changing the resistors to different ones or at least by soldering a second one on top...
Thin layer of liquid metal will not change resistance much too And resistance is not measured isn't it? This is not a test that I will trust for sure, It's terrible
For future reference, you can heat each side of the resistor with a soldering iron until you can push it off its solder and, with a little luck, you can drop on a 2.5 mOhm resistor on there and reheat the solder and you're done. Worst case, you need to add a little solder to the solder pads prior to adding the resistor. If you need need to go up or down you can jist change out resistors. This is the safe way to perform a shunt mod. It is difficult to judge the resistance of a 'pool' or LM. Sweet content. I don't see other techtubers doing things as crazy as this.
You better start using soldering iron. You could solder small gauge of copper wire on the shunt and so reduce the resistance. If you will find proper gauge+lenght, many other people can do same and get same power increase easily.
The issue of gallium and tin reacting if it's long term, why not just use solder to short the shunts if you are doing it long term? After all liquid metal is liquid and could over time loose contact whereas solder would be solid. Unless of course the shunts get to like 150°c or so that I highly doubt
Liquid metal has a much higher resistance than solder. If you just bridge the 2 contacts using a wire or solder then the card activates some protections and significantly restricts the clock speed.
@@fwmtbSoldered permanently all three patches on my old rtx2070 ,base clock is now 1410 ,boost 1900 and unbelievably stable only problem is completely lack control of power limit, but I can control temperature limit. No any drops clock,FPS in all games jump around 30 percent .I don't know how long it will works but now it is completely different card, comparable to my other 2080
as always, the detail in which u go in to stuff is awsome! and whats crazy is so much detail usually = borring. yet you keep it super intresting! nj! keep it up man.
The steadiness of the clock graph is amazing! This is probably my favorite GPU mod ever made, also the reason I subscribed. Please, please attempt an overclock higher than 205mhz, something that makes use of this wonderful cooling and vrm.
Nice one. But it's always recommend to soldering this shunts. The bad thing is also, that with Liquid metal the smd components will fall off with the time. And soldering probably gonna give you also better results because there is just better contact when it's soldered... ✌️
Even though I'd never do this, Your instructions are very clear and easy to follow. Good Video. and ohh btw the Mod Mat seems like a really good idea and very useful. 🤖
in my first "gaming" pc. I had a Intel core 2 quad q6600. base clock I believe was 2.4ghz. I was able to overclock the chip by using a very small piece of tape on one of the fsb pins. this increased my fsb from 1066mhz to 1333mhz, giving me a 3.0ghz core clock. very fun learning experience
I use liquid metal on a delidded 4790k IHS and die side, and on 2 980GTXs for the shunt mod and on the die. This is a overclocked long term gaming PC. In my experience the liquid metal wets onto whatever surface you apply it to and sticks really well. It will not drip off unless you have a ridiculous amount and vibrate it or you did not rub it in to allow it to wet onto the surface. The amount of liquid metal you apply will be so small that gravity is not a concern. Also i used silicon conformal coating instead of nail polish. I had to do multiple applications on the die before i got the hang of applying it, but now i get good results first try. I've done a 1080GTX shunt mod also. IMO the risk is tiny and It is totally worth it to get the performance for my money, but if you aren't fighting for lower temps, lower voltage and higher overclock definitely don't bother.
Excellent content Steve and crew!! It's stuff like this that set's you guys way apart from all those other run-of-the-mill channels!! Keep up the great work!
By the way. If you want to make sure you halve the measured current over a shunt, all you have to do is put an identical resistor in parallel. The way it's sensing the current is via ohms law, which says: "Voltage = current * resistance" so from that "Current = voltage / resistance" In this case your resistance is 5mOhm or 0.005 ohms, and the voltage over that resistor is what you are measuring, so say at 100 Amps, the voltage over the resistor would be 100*0.005 = 0.5v = 500mV, so the electronics would measure 500mV and calculate the current like so: current = 0.5v / 0.005 Ohms = 100A Now the reason the parallel resistor will halve the measured current is that resistance of multiple resistors in parallel is: 1/(1/R1 + 1/R2 + ... 1/Rn) now, that's the general way to calculate it, but for halving it's easy. if you have two resistors of equal value, and you put them in parallel (just like parallel water cooling), the value of the two parallel resistors will be half of their value (2.5mOhm) If we replug that in ohms law you get: 100A *0.0025 ohms = 250mV, however, the software/hardware on the card still thinks that you're using a 5mOhm resistor, so it'll measure the 250mV and find the current like so: current = 0.25v / 0.005 ohms = 50A You have effectively tricked the card into measuring exactly half of what it's designed to, without entirely disabling current sensing. Now to find a suitable resistor is easy: go to something like digikey, and look up resistors. Here's a link to their 5mOhm surface mount selection: www.digikey.com/products/en/resistors/chip-resistor-surface-mount/52?k=Current+Sense+Resistors&k=&pkeyword=Current+Sense+Resistors&umin2085=5&umax2085=5&rfu2085=mOhms&FV=ffe00034&mnonly=0&ColumnSort=0&page=1&quantity=0&ptm=0&fid=0&pageSize=25 You'd want to pick a size in package / case, but that's easy. If you go by the imperial name, then something like a 1206 is 0.12 x 0.06 inches, if you go by metric, the same would be called a 3216, and be 3.2 x 1.6mm Man that turned in to a wall of text quickly What I'm trying to say is, just meassure the dimentions and tadah.
To everyone suggesting soldering copper wire, you would need a few sizes smaller than 30awg to have adequate resistance so the protection circuits aren't tripped. Piggy-backing another shunt on top of the existing one (or just replacing it outright) is the best way, but wouldn't work well for a proof-of-concept video like this where he will want to revert the mod easily for testing.
You can mitigate the liquid metal drip problem by using a liquid metal that's liquid only at higher temperatures. It's called "solder". Use it to connect a 0R resistor across the existing shunt. It takes a few seconds and has less risk of damage. Also a 0R resistor costs much less than gallium based liquid metal...
Steve, graphite is also conductive, and I doubt it would react at all with anything on a board. Although I have not tried it myself, I'm thinking actually simply using a soft-enough pencil to "draw/drop" some graphite where needed, and then using nail polish to not let that fall off or be blown away, might be a working option. Another option, instead of using liquid metal, could be to actually solder a little loop cable from one end of the shunt resistor to the other. So actually soldering; somewhat riskier of course
I would be concerned about gallium in contact with the aluminum at the ends of the shunts. Over time aluminum crumbles like clay. I remember say 15 yrs ago someone produced a voltage or shunt bypass that was built out of a spring loaded surgical clamp and had a resistor connected that was made for the shunts on older video card. That device was patented and maybe still around, the maker called himself "Voodo" and he was from Canada. I would like to see a something come to market.
This is easily reversible. Also if you drop the resistance too low the card goes into safety mode. Liquid metal is not conductive to trigger that safety.
because you can take liquid metal off easily. lets say you remove the shunts or hard bypass them, if something goes wrong its a lot harder to go back and make the card functional again.
I know that you’re probably sick of the topic already, but by my estimates, to reach a resistance of similar value to the shunt resistor you would need roughly 81um of gallium (assuming pure gallium without solvents and contaminants). By my very rough estimates based on a very rough guess of 10um perfectly applied, you had a resistance of 41um on your first coat, which indeed wouldn’t change the resistance much.
Knowing about computer hardware is different to basic electronics this video proves it. This mod will not last, that gallium will eat that solder up nom nom nom. Stacking a resistor on top is the way to go.
I would suggest using something like mg chemicals conductive pens that will not alloy with the low lead solder like liquid metal will. They are effectively superglu doped with silver or nickle that will achieve the same result and can be removed with acetone.
I would recommend to calculate the parallel resistance and solder a identical resistor package with the calculated value for a more accurate % resistance bypass. since you are putting glue on the board, RMA or warranty won't be an option anymore.
Careful with just caking on conformal coat, in this case nail polish. Thick coating can have detrimental effects on long term solder joint thermal cycling durability due to CTE mismatch. Combine that with large non j leaded smt resistors (2512+) have notoriously bad thermal cycling performance to start with and you can possibly be creating an early failure point. I cannot share any proprietary info I have, but NASA-STD-8739.1B is a great reference. Table 10-1 recommends 1-5 mil (0.025-0.127mm) for acrylic or urethane coatings.
Thank you for that! You wouldn't happen to know of any standards that spec when underfill is recommended vs. damming the perimeter of the BGA on conformally coated boards? I've tried IPC-6012D, IPC-6012DA, and IPC-9701A but to no avail. Thanks in advance.
Gotta say, thanks for doing this. But this investigation proves that indeed, Pascal and well this Volta card operate at their optimal position from stock when it comes to power and voltage. Thermals is really the bottleneck for Pascal and Volta it seems, and this makes the largest difference to how a card performs. All in all, if you really want your card to improve, this shunt mod probably isn't worth it. What is worth it, is putting your card under water, as that is what will make a major difference for the money. At best, you stable your overclock with this shunt mod, but with better cooling, whether that be water or LN2, you can increase your clock potentially and make it more stable with both architectures. Thanks for this journalism Steve and GN, now we know the best way to approach overclocking these two architectures.
Would it not have been better to solder in a jumper cable and it would be very easy to take it off when finished. It is also better for long term use if you so wished to do due to those reasons you mentioned in the video. i.e. liquid metal dripping and shorting other components.
For what's it's worth, you need to desolder and bridge those shunts for this to really work. The liquid metal is still pretty resistive, you'd have better results with a solder bridge.
Louis Rossmann recently had a broken Titan Xp with resistors that would fall off from the card with little force after a shunt mod. Looks like the LM is reacting with the solder, softening it down.
Did you do a temp check on your wires and traces? Wire and traces can get so hot it can damage them. Fun fact the leading cause of home fires is from wires getting so hot, it starts fires. Not saying this will cause a fire. But it's possible over time your wires and tracings will get damaged. Love the vids, keep it up.
Nice experiment. You might want to consult an electronics engineer for this one. There is no proper way to do this mod with a liquid metal. You need some real soldering if you want to do it right. Remember, that shunt resistors are in series with the load. As such they need to be able to pass the entire current without dissipating excessive amounts of heat. The liquid metal is not a good choice. You need to solder and actual bridge 0R resistor on top of the existing ones. This is the only way to ensure you've got a low resistance path for the current to flow!
The sensible side of me wants to listen to your advice on this. ... The MANIC GEEK side of me wants to do this to ALL of the cards! I haven't looked at my EVGA GTX 980 FTW recently, but do you think a mod like this could be done on Maxwell-era cards as well? Hell I'd do it just to see if it does anything significant at all, it'd be cool to see my card hold a steady 1500 MHz clock or better for once x-)
If you guys want to do this a bit more safely you can user conductive ink pens, they do not give you quite as much extra power though but for allot of people I think that is a good thing.
That mystery 5mR shunt looks like it's connected to the HBM2 VRM supply. The INA3221 is a three-input device so I wonder which shunt is not being monitored by the INA current monitor ... unless it is multiplexed with the PCIe +12V rail.
jiust use god dang HB pencil...job done. used that mod many many times going back too the good ole G80 an gave a duron the full cache of the athlon to make a full fledged athlon xp barton core,...an a wirepin mod to spank 2.5v thru it
Why use liquid gallium? Just solder a wire across the resistor, done! You don't have to worry about gallium shorting out other components and you can put a switch or a potentiometer inline with the wire to enable and disable shunt bypass or adjust the resistance
You should have tested the shunt mod in Furmark! These heat/power-bug benchmarks reveal a huge difference when you use a shunt mod. Furmark pretty much throttles every non modded card.
Nvidia has a reputation for deliberately limiting their cards performance in furmark (which makes the cards appear to use less power than they actually might)
4:25 look, after 8-pin connection there are 8 places to which there is nothing soldered, and looks the same as what is soldered 8pin, would it be possible to solder an additional connector on it??
If you have time and another card you would like shunt try "MG Chemicals Silver Conductive Pen (842ar-p)" it : Volume resistivity: 0.0001 ohms·cm Typical trace width: 0.9 mm Dries to touch in minutes Adheres to plastics, epoxy, copper, aluminum, ceramics, wood, and most electronic substrates as well as ABS, PLA, and other 3D printing plastics
Controlling the amount of LM helps to control how much extra power headroom you have. We're used to working with LM and it's easy, so easier to use LM and modulate the amount.
What about for a more permanent use case? Would you still recommend (assuming someone really wants to run their GPU like this for daily use) LM over something like conductive tape?
The safest way to perma-mod is to buy a few of the exact same shunt resistors your card uses, and solder them right on top of the existing shunts, so they are in parallel. This will cut the power sensed in half, so effectively you will have +100% power limit. If you want less than +100%, you can use a higher value resistor.
BackyardChemist has the right idea. You can only go so hard with that before tripping 2D clocks, but yes, that would theoretically work and be "safer" (once past the soldering process). It would certainly be more permanent! We haven't done anything like that, though, so you are venturing out of our territory. Our friend at Actually Hardcore Overclocking would be a great fit for you: ua-cam.com/channels/rwObTfqv8u1KO7Fgk-FXHQ.html
Just solder in a lower Ohm resistor instead of risking liquid metal spilling and shorting stuff out, its not like you have a gaurentee anymore when you do extreme overclocking, far less risk and more repeatable.
I love that you're willing to do this kind of test and methodically report the results but I've got to agree with everyone who said that using gallium to bridge a solder joint pointlessly dangerous when there are plenty of safe, "reversible" alternatives that would have been *slightly* more effort. Actually they would have probably been less effort because you could have been much more sloppy and not worried about contamination when this is all over.
Using LM to short the shunts isn't an awfully good solution. It works, but it's a bit risky with the gallium and the fact that you're adding an electrically conductive liquid to a PCB's power lines. The amount of bias you get on the shunt is also quite hard to control. If you changed the shunt for one with a lower resistance, or added a second shunt to each existing shunt in parallel, you could drop the resistance to a more controlled value.
What if you apply an epoxy resin over the liquid metal after application? Wouldn't this alleviate any concerns of dripping or altering the solder joints? Perhaps even applying epoxy to the board and joint prior to application of the liquid metal and then after to cap it off.
Wouldn't soldering a small wire across the shunt be much safer and more effective? When i heard gallium was being used i had a brief "oh gawd no, that's a horrible idea" moment. There has to be a better way to do that.
Not related to this but what''s the empty 8 pin power connector for? Can you push more power to the card if you solder an 8-pin connector and power it?
the pcb is made to accommodate both the titan V and probably some (unknown) tesla card. The tesla version would be more efficiency focused, so could be powered from a single 8-pin, and since these cards are made for datacenter usage, "top" mounted power connectors are a bad thing (they take too much space) so a single 8-pin on the end would probably be used in this scenario. whether or not you could push more power using it? no, not really, because you are bottle-necked by the soft power limits anyhow, the only thing it would help with is not melting your connectors when pushing extreme amounts of power trough them, but then you run into other issues) (I have to say, part of the job for the soft power limits is to keep within the spec of the 6 and 8-pin connectors, at 75W and 150W each, but those numbers are extremely conservative and you can easily pull double that over a quality cable/connector)
I personally think that you could use just solder with short copper wire and that would be that. Better results than with that metal ..or get shunt with 1/2 of the value and you are fine..
You might like our Titan V Hybrid results: ua-cam.com/video/OZcxBP8rNCY/v-deo.html
Shunt Article: www.gamersnexus.net/guides/3185-how-to-short-shunts-on-gpu-for-better-overclocks-titan-v
Modmat: store.gamersnexus.net/modmat
What about using pencil instead of liquid metal? In the olden days you traced circuits with pencil to overvolt GPU's.
So application of the liquid metal has potential to damage parts on the pcb. If one was to use tin/solder to short the shunts would there then be an issue of sending too much power to the pcie?
Steve - my concern with liquid metal here is what unknown amalgams it will form with today's lead-free solders since it's likely a gallium based alloy/mixture. Now's a great time to get smt rework skills! You might want to retouch the pads with real solder to dilute any remaining liquid metal and ensure longterm reliability. For shorting shunt resistors, the easiest / cleanest way is buy a roll of "solder wick" - a copper braid coated in dry flux which, when heated, is for desoldering. Heat the wick with a soldering iron (preferably Metcal brand but anything will work) and feed it some solder from a roll to soak up. when cool, cut off a length of solder-soaked wick equal to the resistor to be shunted and tack solder it to both sides (usually just touching it down with a hot iron works for short term use). To remove, grab in the middle with a fine tweezer, heat each end, and lift away. No nail polish needed. To make it more permanent, you can add some real solder. Liquid metal isn't a great bypass for low impedance parts in the milli-ohm size because internally, they have a larger cross section of metal than the film of liquid metal applied on top. By comparison, solder wick is huge.
Why not just short them with a wire and some solder? easier for a viewer to do?
Gamers Nexus excellent content as always
Just a small suggestion. You can solder an equal value resistor in parallel with the existing one, and theoretically, you will be doubling the card's maximum allowed input current (Ampere) while at the same time remove the liquid metal problem from the equation. Soldering 2 SMT resistors in parallel is a relatively easy process and would probably allow for a more precise control of input current. Thanks for another great video.
I'm guessing i'll be showing my age here, but... We used to do this with high graphite pencils, if all we wanted to do was lower the resistance, and soldering on a wire, if we wanted to bypass the shunt completely. Pretty much no risk involved in the first, and very little in the second.
Also, if soldering is to scary, there is actually proper "trace repair" paint, that is a lot more safe then gallium compound...
I was thinking this the entire time but these days who knows. I feel out of tune with these newer cards
That's how I unlocked my Athlon XP CPUs. 🙂
Amazing content as always. Thank you so much for going down the rabbit hole with us and dedicating valuable time to these niche kind of projects. It's awesome.
Thanks for supporting them! They're lots of fun, kind of new to us, and teach us a lot.
Buildzoid: Don't touch it.
GN: ok *touches anyway*
That power draw increase is actually scary big. Especially for such a small gain.
"MOOOOOOOOOORE POOWWWWWWWWEEEEEEEEEEEEEEEEEERRR!!"
~ Jeremy Clarkson and Steve
"I see, you are a man of culture as well."
~ Last Samurai
While the Titan V is like a cheese sandwich, shunt mod is like a cheese sandwich with prawns
SlavjanA "The fastest carD, pause, in the world!"
POWEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEER - Terry Crews
MOOOORE POOWER - guy with drilldo on efukt
Wouldn't it be safer to solder a low value resistor (Say 0.25 Ohm) in parallel with the Shunt resistor to lower the overall resistance just a bit? It would prevent the dangers of liquid metals and also not trip any protection circuitry... If the value is not low enough, maybe buy a packet of these shunt resistors and 'piggy back them' to half the resistance value... might be worth the while (I just read more comments and found one pretty much like this... my bad)
It would be irreversible, and that was the whole point -- doing something a user would actually do, and something that is reversible.
I guess it comes down to how confident you are at soldering electronic components. It would be simple to remove the additional resistor/s if you wanted to revert it back to the original configuration. (But I am the type of guy that has soldered a PSU's 12V and Ground wires directly to a GPU to power it haha)
But in that way you have more control on the resistance of the shunt. If you change the electrical value of the shunt, you can calculate the power draw.
And what if you solder a second shunt on top identical (physical and electrical) to the first shunt? It would be also be reversible, no?
Precisely, but I still think that you need to be confident enough with using a soldering iron before attacking a GPU of such cost.. You could even look at using a variable wire-wound resistor and make changes on-the-fly (would be awesome to see the results in real-time). But I understand GN not doing this. Not saying Liquid metal is a safer or wiser option but due to convenience it can be wiped off with isopropyl alcohol and a tissue pretty easily with minimal risk of permanently hurting the hardware.
How reversible is that nail polish?
"Obviously you're going to want to watch multiple sources, not just me."
I'll wait for the Verge shunt mod tutorial.
Next level stuff here. I am not a hardcore over-clocking enthusiast, but as a moderately informed and life-long consumer and gaming/PC enthusiast I have yet to see this method used. It is hardly common knowledge. I imagine you will be bringing this method to light not only for most of us viewers, but to some of the other PC UA-cam gods.... I predict we will be seeing a lot more about shunting from other sources thanks to Gamers Nexus. Truly awesome stuff, guys
Screwing around with a gallium based thermal paste in order to affect change in resistance is the silliest shit I have ever seen... simply buy more of the resistors the board is using and stack a second one on top of each existing one with proper solder. This will exactly halve the resistance and double the current capacity.
Agreed.
deimosian Buy that 3000 dolar worth card and do it properly please...
Soldering on, even stacking, an SMD resistor that huge and then removing it again is child's play to anyone with the skills to be messing with component level modifications on a GPU in the first place...
As for why I won't do it personally, see Steve's video on ROI. I don't have a platform to get any ROI on buying a marked up professional GPU.
deimosian He did this as a demo for beginners, not for people who know their shit
No beginner should be fucking with something this dangerous, you need to entirely understand the component level workings of PCBs first. Once you have the skillset to safely do what's shown, you've got access to superior techniques.
1. Delidding CPU
2. Shunt modding GPU
3......................................???
4. Epic Extreme Enthusiast build
Next is small volt mod
I wouldn't midn seeing a volt mod for this card, but I really think at the end of the day, even with more voltage we won't see massive gains. It really looks like Pascal and Volta already operate at the best voltage and power level for optimal operation. We will start to see diminishing returns very quickly. If you are going to mod for either architecture, it seems thermals make the most difference, but yes I'm sure Voltage will allow for more clocks, how much though, probably 100-150 MHz at most.
These videos make me want to look up possible mods for Maxwell and give my 970 one hell of a send off. Maybe even duct tape a 512mb DDR2 stick on it to make it an actual 4Gb card.
luuzeri that 4 gb joke though XD
no point there is a know BIOS editor for Maxwell already; BRA
john doe I can still get creative with cooling. Actually my mate works in a machine shop, hmm. Copper scoop I could fill with snow anyone?
970 has gddr5, totally different from your ddr2 and even ddr3/4
Dual Scimitars really, that's your biggest problem with attaching memory with duct tape?
Gamers Nexus brings us real and informative content, instead of run-of-the-mill comedy content.
You guys have an awesome tech channel. Good work.
Reason 1 for watching a ton of tech channels: They can do things to similar cards I'd get, and do things that I'd be too chicken to try out if I paid for my own expensive product :^)
Great Video GN!
For those who don't know, thicker the metal, lower the resistance. Hence he got higher amps when he applied more because it was able to lower the resistance even more. The shunt and liquid metal act as parallel resistors hence even a thin layer lowered the total resistance to get 1, 2 amps more. (Physics FTW!)
BTW there is a continuity test on most Multimeters with a triangle and a line as a symbol, the diode symbol for those who know. It will beep if the points are connected directly without any components between them. That helps more when trying to find connected pins, a definite beep.
Earned a sub! From a simple BIOS flash to upgrading a GTX690 to double memory by adding bigger surface mount memory chips and a custom BIOS, these mods are awesome (unless you brick your card)!
I am subscribed, but I don't watch every video from all channels I am subscribed to. I do with this channel.
73/5000
Why do you risk liquid metal?
You just can remove by desoldering The SMD resistors .
Not that it is particularly small resistors such as 0402 .
Remove the resistors
And soldering shortened wire .
2 min work Max
Please read the other 800 comments about this. Thanks. 0/10.
It gets to me that people lose their minds when a little LM gets spilled around a delidded cpu die. But no one bats an eye when it gets painted on gpu shunt resistors.
I'm also confused by a non heatgun method? I can't see these 800 other comments? and it's not in the arrticle... Gallium spreads like mould, it could ruin the card.
+David Az I completely agree. Nothing more to add. Either short it out properly or set a desired resistance by changing the resistors to different ones or at least by soldering a second one on top...
Thin layer of liquid metal will not change resistance much too
And resistance is not measured isn't it? This is not a test that I will trust for sure, It's terrible
Nice video my 2080 ti work over 2 years with this shunt mod + volt mod 👌
For future reference, you can heat each side of the resistor with a soldering iron until you can push it off its solder and, with a little luck, you can drop on a 2.5 mOhm resistor on there and reheat the solder and you're done. Worst case, you need to add a little solder to the solder pads prior to adding the resistor. If you need need to go up or down you can jist change out resistors. This is the safe way to perform a shunt mod. It is difficult to judge the resistance of a 'pool' or LM. Sweet content. I don't see other techtubers doing things as crazy as this.
You better start using soldering iron.
You could solder small gauge of copper wire on the shunt and so reduce the resistance.
If you will find proper gauge+lenght, many other people can do same and get same power increase easily.
2risky4me
3expensive5me
4spoopy6me
The issue of gallium and tin reacting if it's long term, why not just use solder to short the shunts if you are doing it long term? After all liquid metal is liquid and could over time loose contact whereas solder would be solid. Unless of course the shunts get to like 150°c or so that I highly doubt
Liquid metal has a much higher resistance than solder. If you just bridge the 2 contacts using a wire or solder then the card activates some protections and significantly restricts the clock speed.
@@fwmtbSoldered permanently all three patches on my old rtx2070 ,base clock is now 1410 ,boost 1900 and unbelievably stable only problem is completely lack control of power limit, but I can control temperature limit. No any drops clock,FPS in all games jump around 30 percent .I don't know how long it will works but now it is completely different card, comparable to my other 2080
as always, the detail in which u go in to stuff is awsome! and whats crazy is so much detail usually = borring. yet you keep it super intresting! nj! keep it up man.
I like that you guys get to sponsor your own product.
aww. I kinda wanted to see how quickly a 3000 dollar gpu would melt
The Mad Salvi instantly lmao
I like how your videos go from "First look on ..." to "We found a 10th way to boost the card".
Literally shit I didn't know you could do to a card, and think someone had the Cajones to do to something that expensive! Take my like good sir :)
Lol, i wish Buildzoid or Der8auer or someone would be at you Studio to just crush the number one spot on the list.
The steadiness of the clock graph is amazing!
This is probably my favorite GPU mod ever made, also the reason I subscribed.
Please, please attempt an overclock higher than 205mhz, something that makes use of this wonderful cooling and vrm.
been waiting for a demonstration thank you so much
Nice one. But it's always recommend to soldering this shunts. The bad thing is also, that with Liquid metal the smd components will fall off with the time. And soldering probably gonna give you also better results because there is just better contact when it's soldered... ✌️
Even though I'd never do this, Your instructions are very clear and easy to follow. Good Video.
and ohh btw the Mod Mat seems like a really good idea and very useful. 🤖
I would solder leads to each shunt, then put them on a multi-gang switch!
Then you have a single on/off switch for the mod.
You learn new things everday. Thx for these great videos, Steve.
in my first "gaming" pc. I had a Intel core 2 quad q6600. base clock I believe was 2.4ghz. I was able to overclock the chip by using a very small piece of tape on one of the fsb pins. this increased my fsb from 1066mhz to 1333mhz, giving me a 3.0ghz core clock. very fun learning experience
Great video, detail oriented, technical, very interesting and cool! Thanks Steve!
damn thats a nice mat! getting one for sure :)
Likewise. Might be overkill for me, but it looks professional. Just need it by March when I do my next build.
Pretty cool snunt mod explanation
finally overclocking that gets me excited like in the old days
I use liquid metal on a delidded 4790k IHS and die side, and on 2 980GTXs for the shunt mod and on the die. This is a overclocked long term gaming PC. In my experience the liquid metal wets onto whatever surface you apply it to and sticks really well. It will not drip off unless you have a ridiculous amount and vibrate it or you did not rub it in to allow it to wet onto the surface. The amount of liquid metal you apply will be so small that gravity is not a concern. Also i used silicon conformal coating instead of nail polish. I had to do multiple applications on the die before i got the hang of applying it, but now i get good results first try. I've done a 1080GTX shunt mod also. IMO the risk is tiny and It is totally worth it to get the performance for my money, but if you aren't fighting for lower temps, lower voltage and higher overclock definitely don't bother.
Excellent content Steve and crew!! It's stuff like this that set's you guys way apart from all those other run-of-the-mill channels!! Keep up the great work!
By the way. If you want to make sure you halve the measured current over a shunt, all you have to do is put an identical resistor in parallel.
The way it's sensing the current is via ohms law, which says: "Voltage = current * resistance" so from that "Current = voltage / resistance" In this case your resistance is 5mOhm or 0.005 ohms, and the voltage over that resistor is what you are measuring, so say at 100 Amps, the voltage over the resistor would be 100*0.005 = 0.5v = 500mV, so the electronics would measure 500mV and calculate the current like so: current = 0.5v / 0.005 Ohms = 100A
Now the reason the parallel resistor will halve the measured current is that resistance of multiple resistors in parallel is: 1/(1/R1 + 1/R2 + ... 1/Rn) now, that's the general way to calculate it, but for halving it's easy. if you have two resistors of equal value, and you put them in parallel (just like parallel water cooling), the value of the two parallel resistors will be half of their value (2.5mOhm)
If we replug that in ohms law you get: 100A *0.0025 ohms = 250mV, however, the software/hardware on the card still thinks that you're using a 5mOhm resistor, so it'll measure the 250mV and find the current like so: current = 0.25v / 0.005 ohms = 50A
You have effectively tricked the card into measuring exactly half of what it's designed to, without entirely disabling current sensing.
Now to find a suitable resistor is easy: go to something like digikey, and look up resistors. Here's a link to their 5mOhm surface mount selection:
www.digikey.com/products/en/resistors/chip-resistor-surface-mount/52?k=Current+Sense+Resistors&k=&pkeyword=Current+Sense+Resistors&umin2085=5&umax2085=5&rfu2085=mOhms&FV=ffe00034&mnonly=0&ColumnSort=0&page=1&quantity=0&ptm=0&fid=0&pageSize=25
You'd want to pick a size in package / case, but that's easy. If you go by the imperial name, then something like a 1206 is 0.12 x 0.06 inches, if you go by metric, the same would be called a 3216, and be 3.2 x 1.6mm
Man that turned in to a wall of text quickly
What I'm trying to say is, just meassure the dimentions and tadah.
"And we're trying it first with a 3000 dollar card," said as if it's business as usual. Now this is how you set the tone for a channel.
Real enthusiast-level content right here
To everyone suggesting soldering copper wire, you would need a few sizes smaller than 30awg to have adequate resistance so the protection circuits aren't tripped. Piggy-backing another shunt on top of the existing one (or just replacing it outright) is the best way, but wouldn't work well for a proof-of-concept video like this where he will want to revert the mod easily for testing.
You can mitigate the liquid metal drip problem by using a liquid metal that's liquid only at higher temperatures. It's called "solder". Use it to connect a 0R resistor across the existing shunt. It takes a few seconds and has less risk of damage. Also a 0R resistor costs much less than gallium based liquid metal...
Steve, graphite is also conductive, and I doubt it would react at all with anything on a board. Although I have not tried it myself, I'm thinking actually simply using a soft-enough pencil to "draw/drop" some graphite where needed, and then using nail polish to not let that fall off or be blown away, might be a working option. Another option, instead of using liquid metal, could be to actually solder a little loop cable from one end of the shunt resistor to the other. So actually soldering; somewhat riskier of course
I would be concerned about gallium in contact with the aluminum at the ends of the shunts. Over time aluminum crumbles like clay. I remember say 15 yrs ago someone produced a voltage or shunt bypass that was built out of a spring loaded surgical clamp and had a resistor connected that was made for the shunts on older video card. That device was patented and maybe still around, the maker called himself "Voodo" and he was from Canada. I would like to see a something come to market.
I feel like Snunt is gonna become a meme now. i.imgur.com/zvGFt4X.png
Snunt up.
WHAT'S THE OUT OF THE BOX SNUNT THO
Out of the box meme score for "Snunt" is pretty high!
why dont you just use solder? why it has to be liquid metal?
This is easily reversible. Also if you drop the resistance too low the card goes into safety mode. Liquid metal is not conductive to trigger that safety.
First of all, "just" makes that sound a whole lot more palatable than it is. The point is to reverse the mod and show something that anyone can do.
because you can take liquid metal off easily. lets say you remove the shunts or hard bypass them, if something goes wrong its a lot harder to go back and make the card functional again.
Come on that component is huge not that hard to desolder and replace.
That's just short circuiting it entirely, doesn't have the desired effect.
I know that you’re probably sick of the topic already, but by my estimates, to reach a resistance of similar value to the shunt resistor you would need roughly 81um of gallium (assuming pure gallium without solvents and contaminants). By my very rough estimates based on a very rough guess of 10um perfectly applied, you had a resistance of 41um on your first coat, which indeed wouldn’t change the resistance much.
Knowing about computer hardware is different to basic electronics this video proves it. This mod will not last, that gallium will eat that solder up nom nom nom. Stacking a resistor on top is the way to go.
I would suggest using something like mg chemicals conductive pens that will not alloy with the low lead solder like liquid metal will. They are effectively superglu doped with silver or nickle that will achieve the same result and can be removed with acetone.
I would recommend to calculate the parallel resistance and solder a identical resistor package with the calculated value for a more accurate % resistance bypass. since you are putting glue on the board, RMA or warranty won't be an option anymore.
Hey, Buildzoid has taught Steve how to dot it himself. Cool.
Careful with just caking on conformal coat, in this case nail polish. Thick coating can have detrimental effects on long term solder joint thermal cycling durability due to CTE mismatch. Combine that with large non j leaded smt resistors (2512+) have notoriously bad thermal cycling performance to start with and you can possibly be creating an early failure point.
I cannot share any proprietary info I have, but NASA-STD-8739.1B is a great reference. Table 10-1 recommends 1-5 mil (0.025-0.127mm) for acrylic or urethane coatings.
Thank you for that! You wouldn't happen to know of any standards that spec when underfill is recommended vs. damming the perimeter of the BGA on conformally coated boards? I've tried IPC-6012D, IPC-6012DA, and IPC-9701A but to no avail. Thanks in advance.
You guys have way to less viewers. I liked your Video as always.
Gotta say, thanks for doing this. But this investigation proves that indeed, Pascal and well this Volta card operate at their optimal position from stock when it comes to power and voltage. Thermals is really the bottleneck for Pascal and Volta it seems, and this makes the largest difference to how a card performs. All in all, if you really want your card to improve, this shunt mod probably isn't worth it. What is worth it, is putting your card under water, as that is what will make a major difference for the money. At best, you stable your overclock with this shunt mod, but with better cooling, whether that be water or LN2, you can increase your clock potentially and make it more stable with both architectures. Thanks for this journalism Steve and GN, now we know the best way to approach overclocking these two architectures.
Would it not have been better to solder in a jumper cable and it would be very easy to take it off when finished. It is also better for long term use if you so wished to do due to those reasons you mentioned in the video. i.e. liquid metal dripping and shorting other components.
Just buy a 0.003 ohm resistor from digikey or mouser. Like 1.00 dollar.
I'd rather go for 0.005. You won't need more than twice the power.
Or did you mean you'd desolder the old one?
I"d go for a second 5mOhm one too, and solder it on top of the existing shunt. Seems much more robust
I assumed he meant he would desolder but stacking in parallel makes way more sense, and would be easier for most people.
Tadas Nanartonis It might not work that well, the additional wiring and relay contacts would add additional resistance making the mod less effective.
jort93z yeah, desolder and replace.
Damn this was cool. nice vid man
For what's it's worth, you need to desolder and bridge those shunts for this to really work. The liquid metal is still pretty resistive, you'd have better results with a solder bridge.
UV activated plastic glue stuff works great in place of nail polish
Louis Rossmann recently had a broken Titan Xp with resistors that would fall off from the card with little force after a shunt mod. Looks like the LM is reacting with the solder, softening it down.
Did you do a temp check on your wires and traces? Wire and traces can get so hot it can damage them. Fun fact the leading cause of home fires is from wires getting so hot, it starts fires. Not saying this will cause a fire. But it's possible over time your wires and tracings will get damaged. Love the vids, keep it up.
Nice experiment. You might want to consult an electronics engineer for this one. There is no proper way to do this mod with a liquid metal. You need some real soldering if you want to do it right. Remember, that shunt resistors are in series with the load. As such they need to be able to pass the entire current without dissipating excessive amounts of heat.
The liquid metal is not a good choice. You need to solder and actual bridge 0R resistor on top of the existing ones. This is the only way to ensure you've got a low resistance path for the current to flow!
As an alternative to liquid metal, you could use rear defrost repair kits. This should be safe on solder as far as I know.
The sensible side of me wants to listen to your advice on this.
... The MANIC GEEK side of me wants to do this to ALL of the cards! I haven't looked at my EVGA GTX 980 FTW recently, but do you think a mod like this could be done on Maxwell-era cards as well? Hell I'd do it just to see if it does anything significant at all, it'd be cool to see my card hold a steady 1500 MHz clock or better for once x-)
Could always solder a small wire across the shunt to bridge past it's resistances and that would eliminate the inexperience with liquid metal.
Nvidia Titan V Ultra Instinct
If you guys want to do this a bit more safely you can user conductive ink pens, they do not give you quite as much extra power though but for allot of people I think that is a good thing.
That mystery 5mR shunt looks like it's connected to the HBM2 VRM supply. The INA3221 is a three-input device so I wonder which shunt is not being monitored by the INA current monitor ... unless it is multiplexed with the PCIe +12V rail.
Seemed like it was connected to the same 8-pin as the other shunt, as we did ultimately have to short it. Thought the same, though.
you could just soder a jumper cable in parallel to the resistor, might give you even more overclocking headroom cause of the lower resistance!
AWESOME VIDEO!!! I am an electronic tech III. Perfect explanation of the currents and voltages. I want that mat ya have for sale!
jiust use god dang HB pencil...job done. used that mod many many times going back too the good ole G80 an gave a duron the full cache of the athlon to make a full fledged athlon xp barton core,...an a wirepin mod to spank 2.5v thru it
quick question/tip. if your looking for 0 resistance why not test for continuity? the multi meter will beep and it'll be a lot easier
Why use liquid gallium? Just solder a wire across the resistor, done! You don't have to worry about gallium shorting out other components and you can put a switch or a potentiometer inline with the wire to enable and disable shunt bypass or adjust the resistance
You should have tested the shunt mod in Furmark! These heat/power-bug benchmarks reveal a huge difference when you use a shunt mod. Furmark pretty much throttles every non modded card.
Nvidia has a reputation for deliberately limiting their cards performance in furmark (which makes the cards appear to use less power than they actually might)
4:25 look, after 8-pin connection there are 8 places to which there is nothing soldered, and looks the same as what is soldered 8pin, would it be possible to solder an additional connector on it??
If you have time and another card you would like shunt try "MG Chemicals Silver Conductive Pen (842ar-p)" it :
Volume resistivity: 0.0001 ohms·cm
Typical trace width: 0.9 mm
Dries to touch in minutes
Adheres to plastics, epoxy, copper, aluminum, ceramics, wood, and most electronic substrates as well as ABS, PLA, and other 3D printing plastics
What would happen if instead of liquid metal, you used a copper wire?
You would probably trip the protection circuit, which would lock us to 2D clocks. It would also be irreversible.
Gamers Nexus oh that doesn’t sound good.
maybe swap the shunts for a lower resistance one or stack another ontop?
Stacking another on top would work and be a more permanent solution, yeah.
Conductive pen?
You can also solder another resistor with same value in parallel so you know that the actual power consumption is double of what the card thinks it is
Would love to see the results on a more mainstream card like a 1070. Mostly because I own a 1070.
You don't want to do this. Trust me!
gamers nexus balling out of control over here
Would it not be smarter to use conductive tape to short the shunts instead of liquid metal if you're worried about damaging the solder joints?
Controlling the amount of LM helps to control how much extra power headroom you have. We're used to working with LM and it's easy, so easier to use LM and modulate the amount.
What about for a more permanent use case? Would you still recommend (assuming someone really wants to run their GPU like this for daily use) LM over something like conductive tape?
The safest way to perma-mod is to buy a few of the exact same shunt resistors your card uses, and solder them right on top of the existing shunts, so they are in parallel. This will cut the power sensed in half, so effectively you will have +100% power limit. If you want less than +100%, you can use a higher value resistor.
Oh right, that makes perfect sense. Thank you! I will definitely look more into this, I love all these semi niche mods :)
BackyardChemist has the right idea. You can only go so hard with that before tripping 2D clocks, but yes, that would theoretically work and be "safer" (once past the soldering process). It would certainly be more permanent! We haven't done anything like that, though, so you are venturing out of our territory. Our friend at Actually Hardcore Overclocking would be a great fit for you: ua-cam.com/channels/rwObTfqv8u1KO7Fgk-FXHQ.html
Just solder in a lower Ohm resistor instead of risking liquid metal spilling and shorting stuff out, its not like you have a gaurentee anymore when you do extreme overclocking, far less risk and more repeatable.
When you think your a good overclocker/enthusiast then you watch a Gamers Nexus vid
You should really try overclocking using the voltage frequency curves. You can usually get higher clocks that way but it does require more tweaking
Why would you use liquid metal and not simply solder a wire on the top to make a consistent bridge?
I love that you're willing to do this kind of test and methodically report the results but I've got to agree with everyone who said that using gallium to bridge a solder joint pointlessly dangerous when there are plenty of safe, "reversible" alternatives that would have been *slightly* more effort. Actually they would have probably been less effort because you could have been much more sloppy and not worried about contamination when this is all over.
Why you don't screw around with gallium: ua-cam.com/video/jeghGhVdt9s/v-deo.html
Ohms Law. That is the term you were looking for at 2:15.
Using LM to short the shunts isn't an awfully good solution. It works, but it's a bit risky with the gallium and the fact that you're adding an electrically conductive liquid to a PCB's power lines.
The amount of bias you get on the shunt is also quite hard to control. If you changed the shunt for one with a lower resistance, or added a second shunt to each existing shunt in parallel, you could drop the resistance to a more controlled value.
What if you apply an epoxy resin over the liquid metal after application? Wouldn't this alleviate any concerns of dripping or altering the solder joints? Perhaps even applying epoxy to the board and joint prior to application of the liquid metal and then after to cap it off.
Nice shunt mod..i am not going to do it on my 1080ti (but more power to you) :)
Wouldn't soldering a small wire across the shunt be much safer and more effective? When i heard gallium was being used i had a brief "oh gawd no, that's a horrible idea" moment. There has to be a better way to do that.
Not related to this but what''s the empty 8 pin power connector for? Can you push more power to the card if you solder an 8-pin connector and power it?
the pcb is made to accommodate both the titan V and probably some (unknown) tesla card. The tesla version would be more efficiency focused, so could be powered from a single 8-pin, and since these cards are made for datacenter usage, "top" mounted power connectors are a bad thing (they take too much space) so a single 8-pin on the end would probably be used in this scenario.
whether or not you could push more power using it? no, not really, because you are bottle-necked by the soft power limits anyhow, the only thing it would help with is not melting your connectors when pushing extreme amounts of power trough them, but then you run into other issues)
(I have to say, part of the job for the soft power limits is to keep within the spec of the 6 and 8-pin connectors, at 75W and 150W each, but those numbers are extremely conservative and you can easily pull double that over a quality cable/connector)
I personally think that you could use just solder with short copper wire and that would be that. Better results than with that metal ..or get shunt with 1/2 of the value and you are fine..
store.gamersnexus.com/modmat is the new lttstore.com
correction: seeing the upload date, it's actually the original lttstore.com