Probing a Gigabyte RTX 3090 Vision that died when trying to run New World
Vložit
- čas přidán 25. 08. 2024
- My Patreon: / buildzoid
Teespring: teespring.com/...
The Twitch: / buildzoid
The Facebook: / actuallyhardcoreovercl...
#Nvidia #RTX3090 #NewWorld
Gigabyte: "We have no proof you purchased this card even though you have the card. You could have made a factory and built this identical card just to get a free repair. Nice try"
It wasn't that, it was that he got it through Ebay, so this just means that their warranty doesn't cover transfers. I do understand the frustration, though, because there are a lot of companies out there that offer warranties that cover the part no matter where you get it from or how many times it has changed hands.
@@OoDeucexoO No, people are building their own factories for free RMA.
@@OoDeucexoO "Doesn't cover transfers" is illegal in the EU at least. I think what he got was "must have original receipt" which it is, unfortunately, still legal to hit people with even for items that have serial numbers.
@@andersjjensen legal but utter BS. Not to mention horrible PR, especially when it happen to one of your halo products.
Yet another reason not to buy a scalped item.
@@Silverhks you could ask for a receipt before you commit to buy from the scalper, that way you don't get scammed by these OEM's (who are scammers)
> 19:47 - "that's it for the video"
> there's 9 minutes more 'till the end
ah yes, classic Buildzoid
Yesss
WOW I'm stupid the reference Nvidia PCB does allow a 10powerstage Vcore for the 3090. So yeah this is actually an Nvidia reference PCB with very slight upgrades.
Good to know. I know my Zotac trinity 3090 is the reference board. So if this reference board died playing that game, I should just avoid that game like the plague. Yikes.
amogus
out of curiosity, why didn't inject 0.7V to tests for the shorted component.
lol same day here you say 'i hope they fix their power delivery'... is the day red gaming tech now release a new video about future flagship card will be 450w tdp. with a new power connector. i find that highly amusing. but also not amusing at same time because its my only option to buy any gpu with my evga bucks being that evga are nvidia only shop for their graphics cards. only for the thing to be destined for the vcore to eventually blow itself up one day. wish they made more sense with the pcb design. or if just somebody could come along to explain why. like tin or ellmer or whoever
Crap. I had 2 launch FTW3s die so I just refunded the 2nd replacement (long before new world came out, PCSX2 killed both). Eventually got an ASUS EKWB 3090 down the line and I'm pretty sure its just a reference board like this with a water block slapped on it (minus those weird power connectors). It has had nutso inductor buzz since I put it in. Guess I better wind up the doomsday clock.
"if that sounds like insanity to you, join the club" 🤣
"let's mark it otherwise I will not remember" two seconds later "I wasn't actually paying attention to what I was doing"
😂
ADHD be like
Remember when "this game will melt your graphics card" used to be figurative, not literal?
And this is why I'm one of those who haven't even thought about playing New World. The second I heard the warning, I noped out.
@@katarh The game has settings in options to workaround this potential design flaw in cards. Let's be clear here. It is not the game's fault the card fails as it does not do anything outside of the norm and only uses DX calls. The problem is in the card design, it is a hardware fault. By default the game enables the frame limiter while in menues, minimized or unfocussed. Switching that off.. if your own responsibility.. Also.. the number of cards which actually failed is actually really small.. it's just that there was/is a lot of noise around this by mostly ppl who really are just .. making noise..
@@katarh apparently a couple of other games killing 3090s now! I'm scared with my 3080ti now.
@@Kholaslittlespot1 Also on a Gigabyte 3080 Ti. If it was just the issue of risking the hardware, it'd be one thing, but right now it's risking the hardware with literally no options for replacement, since Gigabyte is being poop about RMAs and there is still a massive shortage for everything else. I'm on a good PSU and my card has been waterblocked to avoid the VRAM overheating issues that people noticed last year, and until the issues that are triggering the dead cards are verified and resolved by Nvidia, I just don't see a point in trying a game I don't even think I will enjoy.
@@blazemonger1 _"It's not up to the game developers to make sure their product doesn't destroy your computer. It's up to you."_
Yeah, right. It's my responsibility to fix the game developed by some huge megacorporation. It's not their responsibility.
Awesome work as always Buildzoid. Your GPU necromancy skills and understanding of PCB's without schematics still continue to blow my mind.
And gigabyte continues to blow PSUs and at RMAs
@@micobugija6284 That's a lot of dmg! xD
If he fixes it, does the guy get it back?
@@Kholaslittlespot1 no clue, but probably he does
@@Kholaslittlespot1 ya, at a Cost. I can't believe someone would buy a used one on Ebay. Must have been dirt cheap.
Electrical engineering is a field where I don't know enough to be useful and I've always wanted to know more. It wasn't made as an educational video in that regard, but it was cool watching what you're looking at and what you're looking for.
It’s electronics engineering not electrical
@@waithony Ha Ha Ha
@@waithony No, it is properly called Electrical Engineering. That's what is written on the degree, and digital electronics is a subset of the field of electrical engineering. It could also include high voltage power transmission, for example. Nobody calls a power substation "electronics".
You can get some of that from louis rossmann's repair videos. Also eevblog. You may even want to start with their older videos.
@@jasonfullerton7763 nope it's electronics and electrical engineering. Different disciplines, different applications.
Gigabyte: “scalpers gonna scalp”
Also gigabyte: “well you bought this from a scalper so I can’t help”
K got it
gigabyte are the kinda of company that will be like "sorry, you bought this on a tuesday, we only service Monday, Thursday and May 17th Cards"
Gigabyte don't seem to care anymore, it's insulting and I will never support them with my money ever again.
@@rinner2801 your fault for buying from scalpers.
Never buy a GPU on ebay.
@@monkaSisLife My fault? I own a 3080 FE direct from Nvidia. I'd NEVER buy from scalpers, doing so just makes you part of the problem.
@@agzarah You are not wrong. My 3080ti died Their phone support tells you they are busy due to "covid" so no one is going to answer unless you wait a long time. They have an option to get a callback instead of waiting on the phone without losing your place in line. I waited all day and nothing. Their phone support doesn't pickup the phone anymore.
When he said he doesn't have schematics, I wondered if the schematics were in the leaks form the gigabyte hacks and ransomware attack?
He definitely referenced that nonchalantly.
I was wondering too, Louis Rossman gets lots of Apple Schematics and we know how tight Apple is.
there was a leak from your mother's house
I have fixed 12 Rtx cards due to new world, it defiantly a issue with power delivery and the gpu requesting more than designed. Most common is an internal short on a vcore mosfet. Hit me up if you want to know more.
Hi Christian, in your opinion what is the best way to prevent these cards from shorting out?
@@rhettro7236 best way would be to use msi afterburner or similar gpu control software and bring down the power limit to about 80%. I have seen that with the 3090 that has a TDP of 350 watts but I have seen them consume more than 850 watts of power. Now with further testing i thought that it might be "dirty power" from either the mother board or psu but it seems its an issue with the power delivery design, on the actual Pcb. Now I aways tell my customers that until either A Navida releases driver fix to tone down the power draw of the RTX 30 series, or Amazon hires better programmers for new world, to manually decrease the power draw of these cards to protect not only they cards but there motherboards.
I’ve been playing new world for a few days with afterburner open and rtss as a frame limiter to 161 , is this alright for protecting my card? The power limit is still set to 100% on afterburner. I’m wondering if I’ve been safe or lucky with mine surviving
@@xStaryyyHD i would say more lucky, you should 100 percent turn down the power limit to about 80 to 85 percent depending on the manufacture and revision number of the card. The issue is with the power delivery on the card, either its a component level issue with the shortage that happened before these cards came onto market which caused a last minute change due to lack of parts or it was an initial design flaw from Nvidia and not accounting for possible possible current overdrawn from a very power hungry gpu.
@@christianstout4383 i have a question that is quite separate, but im trynig to figure out the problem to. my rtx 3080 ti eagle power on, and fans spin, as well as seem to react to temperature by turning fans on and off at particular intervals, but it has no connection. it doesnt register in the computer at all. ive changed positioning, power cables, tested other gpus. i would just like to try and fix it if i were able to.
I appreciate an informed "don't know" far more than an ignorant "It's the fan controller hurr" diagnosis. So far evidence is pointing towards power delivery design skill issue, yes?
It has nothing to do with the mmo people are putting blame on the game instead of the gpu developers
@@ForeverHollowed ok Buuuuut.... If it were TRULY only a gpu issue, these cards would be dying in any game or stress test.
There is still something wrong with new world I think, which ends up killing weaker built cards that may have failed down the road anyways and whatever new world does, kills it faster.
@@ForeverHollowed there is something wrong with ‘the game’ and the ‘gpu’, mein kleiner Freund.
Love this type of video. Look forward to follow up videos about the repair.
Sincerely hoping for a part 2 of this video
oh no BZ got a 3090, RUN
Are you going to fix it now you somewhat know what's broken?
I am wondering too if BZ is going to repair this GPU and send it back to owner
@@DiJAndy if one of the power stages shorted itself (drain to source), the card is likely not repairable (sending 12v to the GPU). Other than that I'd expect him to have more than enough integrity and do that.
@@DiJAndy 28:25 He said he's going to see if he can
@@infernaldaedra if failed like a fuse... the fuse, itself, won't be dead. The most common failure mode for the mosfert is dead short, taking whatever they are supposed to switch along with them... and then the fuse.
Still there is a chance . No scorch marks, of course, as the fuse likely blew in 200ms or so.
@@stanimir4197 Yeah thats true. I've gotten quite lucky in the past with Mosfets just taking themselves out though.
hurts so much when You buy such an expensive card and it fails!
Like Louis Rossmann said : "Schematics or Die!"
Imagine your very first game causes the most expensive and overpriced cards with current global supply conditions to fail. 💥
Also imagine that you got frustrated waiting for one to pop up at MSRP and you "had to have the absolute best" so you paid $2800 for one from someone on eBay and as such voided the warranty.
That's why you don't bid unless original proof of purchase is included.
Conspiracy theory:
Amazon made a game that kills GPUs on purpose so that they could sell more cards.
Question is why would they do that when they sell out insantly regardless.
@@peppa1492 If they had any in Stock to begin with that is.
@@peppa1492 Alternate conspiracy theory: AMD has worked closely with Amazon to develop the game in such a way it destroys NVIDIA's flagship gpu while leaving AMD cards unscathed which leaves a certain level of distrust amongst consumers towards NVIDIA & more confident in AMD & likely to buy an AMD card in the future.
Amazon: Resells GPUs.
Also Amazon: makes game that bricks GPUs.
Not the game that bricks the gpus. Its just faulty hardware, bad conductors, bad vrm controllers. Code from a game cant kill a graphic card.
@@Startrance85 software can absolutely break hardware
@@nooneloveseveryone1 a game code in a game cant break a GPU, its faulty hardware and that have EVGA said themselves.
@@Startrance85 ah yes of course because EVGA manufacturing errors are absolutely the reason for a bricked gigabyte card....
@@nooneloveseveryone1 EVGA is buying powerstages and so on from a third source, have nothing to do with their manufacturing capabilities. its due to a bad batch of powerstages that they bought, the blowing up on EVGA and Gigabyte cards have nothing to do with game code. There is nothing these devs could have done to prevent this if the reason is faulty hardware on the cards.
Just ask yourself, why is this only a problem on a selected few 3090s but not everyone.
What production weeks and lots are those cards? In my daily job I see a lot of companies basically redesign PCBs on the fly to retrofit components of different specs and from different vendors because the whole parts supply chain is so fucked right now.
Basically if chip x of vendor y is currently unavailable due to global supply issues they try to use different parts that are available which sometimes requires new PCB revisions as well.
I’ve seen a lot of weird shit recently even automotive tier 1 suppliers fitting industrial grade spec parts because AECQ spec parts are unavailable.
So if boards of certain production runs are affected it might be because of the miserable supply situation and because everyone is scrambling to get enough parts to produce anything
From the news i have gotten, the EVGA3090 that died was all from 1 or more of the first batches. So i doubt this is changed from nvidias orignal spec/alowed design, it seems like some sort of a design flaw , OR the game triggering a flaw in the power supply/VRM of the original 3080/3090 design cards.
Huh which spec parts are better industrial or AECQ?
@@Operator588 aecq is an automotive set of requirements for parts and it’s more demanding than industrial spec. For example automotive parts have operational temperature ranges between -30 to +125 degrees Celsius and they have much stricter requirements as to EMI and EMC. So fitting industrial grade components on an automotive electronics component would be a downgrade
@@Numfuddle ah thanks for the complete explanation. i really hope those componant changes dont end up killing someone. where the parts that were changed possible to cause serious issues?
@@Operator588 they do this with the help of the vendors and the car companies OKed it as well. They also do additional qualifications and verifications on the parts. It sometimes happens even in normal times when a special part (like an LCD for example) is not yet available in automotive spec because it’s too new.
It’s just weird because usually without the recent supply issues this would not be done at all. You’d not even get quoted prices for non automotive parts.
Im more worried about consumer electronics because they also switch parts to those that are available but the testing and qualifications is usually less rigorous
You should do voltage injection 1.2 volt to the shorted side of the fuse. With a thermal cam, you can see witch power ic is bad.
Thanks for sharing. Get yourself some good needle probes (from Pomona Elec) and would of liked to see you troubleshoot to component level, then change the part and power it up,,,maybe next video. One trick i do is solder on a resistor lead to the shorted side of the fuse and another on gnd, then attach your meter to them (preferably mini grabbers) so you dont have to hole on. Then start hot air reflow and pulling the suspect parts off one by one and watching your meter when the short disappears, saves a bunch of time this way.
I watched the whole video and what I took from it was,
1) set phasers to stun
2) gold pen dots look nice on black.
Feed 1V at 1A into the shorted rail, get or borrow a thermal camera and see what gets hot. The gpu itself should be fine, but the shorted fet will get warmer than it’s surrounding. Yeet that mosfet, then the gpu should boot, just down a power phase. Gotta replace the fuse tho.
testing a gigabyte gc and thereby saying your power supply has its own protection ... yeah anyone seen the gigabyte power supplies blowing up^^
Guess we can say this one falls in line with expectations for gigabyte products by now 🤔
You need to use a high current low voltage source and a temperature camera or what i did some freeze spray or alcohol and see which part is heating up and boiling off the liquid first. I have found shorted capacitors this way in switches and stuff.
Can you livestream your work on repairing this card? I'd like to get into board level gpu repair (I think there will be great demand given how many people are buying from scalpers) about would greatly appreciate learning from your repair of this one.
check out the Tech cemetery youtube channel
@@ActuallyHardcoreOverclocking Oh sweet, thanks for the recommendation! I have always been interested by your modifications and GPU necromancy. I hope we will be able to see a zombie RTX 3090 sometime in the future.
In my mind this feels like performing MRI scan with fridge magnets but in the right hands this could actually be useful who knows
I very nearly went with a Gigabyte 3090 when building my new PC.
Glad I went with EVGA instead.
Gotta love a little forensics action. 🦋
Gigabyte being shitty with RMA as expected...
He bought it on eBay.... I'd say it's deserved
@@stardust9329 well considering it's one of the only places you could buy one for months you'd think a manufacturer could be a little accommodating.
@@timotmon but eBay isn't an authorized seller for these cards and ppl are paying scalped prices.... You take the warranty risk when you buy from eBay 🤷♂️
@@stardust9329 I get it. All you say is true. Just considering the fact that Nvidia and these other vendors having making an absolute killing on these products along with scalpers. Seems no matter what, the consumer is left in a precarious position here.
@@stardust9329 For such an expensive consumer product and not being very old, it's just a copout by Gigabyte.
And it's not like he can take the card to 3rd party repairer and fix it without a schematic?! (Well maybe he can, this time)
Now you know why "Right to Repair" is very big issue and not just an Apple/John Deere problem.
As someone who just bought a 6900 XT all of this has made me kinda worried about a VRM failure. This is my second high TDP card ever. My first was an EVGA GTX 980ti and one of the VRMs on it blew up.
I haven't heard of any Radeon gpus being affected yet.
What you want to do is make sure you have a good psu, if you want to lengthen the life of your GPU, buy some thermal grizzly kryonaut (or kryonaut extreme), some fujipoly 17w/mk pads and rip apart your GPU to replace the paste and pads.
You can easily limit your gpu from over working itself by setting an fps limit in the radeon software. This kind of stuff really happens when you run uncapped fps
I don't have such issues as I cannot afford not only top end high TDP cards, but more like any card whatsoever at current inflated pricing of ~3x msrp on average across the board 😂😂😂
@@mroutcast8515 assuming you're in the US and want AMD, keep trying the AMD drops. It took me about a month before I finally got lucky. I wanted a 6800 XT but they were out of stock when I got my turn. Had to settle for the 6900 XT.
honestly it's one of the first things I ever do before I buy ANY card, I look at the board layout, I hardly ever buy anything that I can't suss out, if it is reference, under spec'd etc. And I nearly always modify the cooling
About probing problems
Having sharp tips helps
Also put multimeter in to beeper mode, short tips and try to move a little, there might be break somewhere within probe
I always tend to make own probing cables because what they originally provide is garbage most of the time
Ot just wash tips in contact fluid, sometimes helps
Thanks for the video Rossmann.
Gigashit with 0 support! I will never buy a gigashit product again in my life!
Why the crickets from GN?...
Busy with moving things
My Gigabyte 3090 VISION OC-24GB died playing New World
Typical Gigabyte customer support - don't buy Gigabyte!
in some countries if you don't have the purchase receipt or you are not the original buyer, this is what happens, you get no warranty
this is the reason for most people to not buy second hand stuff
I can not believe Kermit the frog is doing the voice over for this video !
Wouldn't it be nice if video cards were modular mainboard-like sub-systems, with slots/sockets for memory and gpu chips or compute modules ...
the sockets create a lot of inefficiency that is torelable at the relatively low power levels that CPUs have but totally unacceptable for GPUs.
@@ActuallyHardcoreOverclocking
Reality ruining my dreams again
Or just make bga repair machines and training available to everyone lmfao. Honestly if we ever start getting provided schematics for gpus, that would be more than enough.
Strange, I could swear there was a card with modular vram at some point, Google says no tho.
@@cidmaria There were some really old video cards that had sockets that allowed you to add addition ram, but that was mostly before the era of modern GPUs. Also I believe intel made something called a Visual Compute Accelerator which was basically an entire system itself on a PCIE board with Xeon processors and SODIMM slots, but they weren't really a GPU and wouldn't be a card you'd want to game on. I think the Intel VCA was a predecessor of the Xeon Phi cards.
So glad you made this video!
Happened to my PNY RTX 3080 within the first 5 hours. Bought it from a scalper so I’m screwed on warranty. Gonna use this as a rough guide to probe mine
Might still want to try the warranty some times you get lucky
try the warranty first
I got my 3090 from a scalper back in November after launch and Zotac still let me register the card with their extended 3 year warranty (thank god I got it for $1600 compared to the $3,600 they're going for now). They even required an image for proof of purchase and I literally just used a screenshot off ebay and it worked. So there's still a chance. Just make sure if your card needed to be registered from the box. Why you should never throw out a GPU or motherboard box.
@@ActuallyHardcoreOverclocking I’ll give it a shot, however, The scalper pulled the card from a prebuilt system...
@@SwampyF4RT manufacturer might still warranty the card. did they say no already?
Very good channel, love your videos! Firm handshake from the Netherlands to you.
Every hardware component I ever bought from Gigabyte has had issues.. I don’t buy Gigabyte regardless how cheap it may be.
Why would anyone accept them turning down an RMA? Why let these manufacturers get away with not backing their products?
For what these cost they should be pretty bomb proof
Because warranties aren't transferable? ¯\_(ツ)_/¯
@@markomus1 a warranty does not affect your statutory rights tho
@@norfolkngood8960 moral of the story, stop buying on ebay especially from scalpers.
Any chance it's some weird high frequency current draw/ power saving voltage switching due to weird load patterns in that game?
Jay showed it might be more common on EVGA cards purely due to the cards improperly detecting their set power draw limit. He showed when set to 100% it could go as high as 128% and it would still pass 100% even at 70% on overclocking programs. Though he did show this is consistent even when not playing New World but New World seemed to trigger some other cards to occasionally do it too but only briefly.
@@Skylancer727 ...now you are going to say the voltage drop (power draw) changes with the temperature and no one accommodates for it.
@@stanimir4197 yes but considering this is one of the first parts in the chain that failed, no I don't think that would make an appreciable difference, especially not to this extent. The power draw values are also collected very early on the card which is why its worrying some cards like EVGA's are drawing more than they are rated for. Having a max overclock of 110% but using nearly 130% is a huge difference to those components. They only designed it for that spec, this is a fair bit above that.
Yes we do increase the power for VRAM over distance, but that results in exactly what is in spec. This is different. Nvidia and EVGA's specs were made under the consideration it will rarely pass the values dictated.
@Actually Hardcore Overclocking Please show how you fix this card!
Thanks for the vid. Just a heads up but it seems that New World is causing some OCP issues on older systems. 2 of my friends and myself have had new world trigger the overcurrent protection on a fairly regular basis. We have had this issue occur on systems with 750 and 850 watt psu in conjunction with a EVGAgtx 1080 a MSI gtx 1080ti and a Zotac gtx 1660. Issue only occurs in game while inside of a settlement no matter the graphics settings, even happens on lowest setting with frames locked to 30 fps. We have only been able to duplicate this in New World. We have been unable to trigger OCP in any other game or application, even Cyberpunk 2077 runs without issue.
i've seen this problem from several different manufacturers already. asus, evga, zotac. the zotac one was actually smoking when it failed.
He literally left it running for ages and was filming it. Should have prob turned the thing off.
@@Kholaslittlespot1 no card should fail in that short amount of time. PC parts are designed to be ran at high temps and power draw for extended periods of time. People have GPUs from over a decade ago still running fine.
@@12pandemon Yup. I'm one of them. I have 2x 8800gtx from my first sli build that work fine. Back then 768MB was a lot per card!
Im just saying half these other cards may have smoked if people had sat there with them powered on filming them.
first mistake that owner made was to buy from gigabyte
What do you mean, gigabyte refused RMA? That would be highly illegal in the EU!!!
That's why you never buy Gigabyte unless it's their mid to high end motherboards. And they are stingy on the RMA's.
He didn't buy it from an official source, he bought it used likely from a miner. So since it wasn't a new card they wouldn't register their warranty. Also Gigabyte was recently hacked so they just kinda haven't been accepting warranty requests currently as they're still working on getting their servers back online from a backup.
Buildzoid said it in the beginning: It was bought second-hand. In Germany you CAN transfer the legal warrenty (2 Years) even second-hand.
@@Skylancer727 And this is a reason why exactly? It is a card inside the warranty timespan, getting hacked is no reason to decline RMA of a card that has no possibility to be outside the warranty due to the fact that the production start time lies inside the warranty time...
no proof of purchase, you get no warranty
Would love to see the repair video
In the time you took explaining it you coulda fixed it. You sound knowledgeable enough I was looking forward for the actual repair.
Nontransferable warranties especially in this market are the worst. Unless you get really lucky with the newegg shuffle, or made it on the EVGA queue early, or lucky enough to live close to a micro center, it's impossible for everyday consumers to get a new gpu. The US is really behind in consumer legislation and not honoring warranties for purely the reason of not being the original purchaser makes no sense if there's been no modification to the product. There's no available gpu pcb repair shops so manufacturers are literally choosing to allow modern cards that can be fixed bc of fuses to become e-waste.
Also from a gpu seller standpoint, the people who buy 3090s are the whales of those companies so wouldn't they want to do everything they could to keep them loyal to their brand?
Zotac has a couple of cards in stock at slightly inflated prices. Their 3090 amp halo is like $2600 and 3080ti $1800 I think. At least it's direct sale so the warranty's intact.
Kermit be fixing computers now..
If you are having reading with the digital multi meter is either clean the tips of the probes or replace the probe leads as I ran into the same problem before as the tips sometimes get dirty or corroded I ended up getting the fine tips probes not had a single issue since.
For vrms testing you might need to test the gate pin to ground same with source pin and drain pin, the pins usually facing toward the power plugs it guaranteed to shows a short while testing with the digital multi meter. Another is using voltage injection to see which one will heats up with liquid nitrogen spray or use a thermal camera
0 dislikes, this is how it should be for you.
Always
You ruined it, now there are 2. :(
It's so weird to me that a video like this has dislikes. what here is disagreeable at all?
Dam gigabyte bad guys honour the RMA! Feel back for the owner.
on some countries, if you don't have the receipt and the original buyer, you do not get any warranty, specially if you say you bought it from some guy online
Hope u make a video reapairing that card... :)
People fixing graphics cards are doing the lords work. 😇
Hopefully the guy who bought it can contact the guy he bought it off Ebay for the original receipt for an RMA.
>Getting rejected for your RMA because you caved to ebay
Oh no.
Ah yes 4am and youtube recommends me the GPU necromancer
Awesome
For the love of all things electronic, if you're checking for super low resistance, please just adjust your range setting and go. Not all this wait for auto range to settle stuff. :) Cheers
I have the exact same card (Gigabyte Vision RTX 3090, 2 months in use), and had a series of temporary failures over the course of 2 weeks (GPU shutdowns, accompanied by fans on maximum, then reboot and all works fine for a couple days), followed by a permanent failure. For my card, the opposite 12V connector is shorted to GND, no fuse is blown (due to the PSU safety shutting things down), and almost the exact inverse of the power stages you found shorted are shorted on mine (with the ones you found shorted being okay). I'm not sure what that means, as far as which rails are involved. Assuming some of the MOSFET chips are shorted, I have been trying to find replacements, but I can't even find a datasheet on the Alpha Omega chips (AL001E1Z). I do have a good soldering station with hot air tips and vacuum (but, shaky hands that even caffeine can't steady). Any help finding replacement driver chips would be SO much appreciated!
Found a replacement (3rd video shows the datasheet): AOZ5312UQI. I'm going to try my luck and see if I can track down the shorted chip(s) with a current limited power supply and thermal camera, then repair with a hot air soldering station. Hopefully, the GPU chip isn't damaged.
Everyone chooses to blame the game for this issue, but imo it's the manufacturer's responsibility to harden their product against attack vectors. Even if the bug in the game gets fixed, the card will still be vulnerable. What if bad actors start writing ransomware that threatens to destroy your GPU unless you pay? The card ought to be able to resist such an attack. There's a saying among Linux kernel developers: "We do not break userspace". What this means is that even if a change to the kernel is an objective improvement, if said change causes a downstream piece of software to break because maybe it relied on some quirky behavior that it shouldn't have, then said change should be considered a bug in the kernel. It's more important to preserve the functionality of and compatibility with user mode software than it is to streamline things that are janky.
It is for sure manufacturers fault, but also the game triggers it, and there are much more GPU demanding games out there that dont do that. So something in new world is very wrong as well. Blame both.
This wasn't a New World problem, it was an nvidia problem. There should have been fail-safes built into the card just like every other card that prevents software from asking too much of that card.
🥤 🍿”let’s get to some probing”...
Needs a Schematics or Die tee from Rossman
Hi BZ is there a 60+ minutes Rambling video about AMD Dark Motherboard coming any time soon?
Luumi has a nice board review and he too was like "look for buildzoids board review" lol
@@fracturedlife1393 the only way I get my head around Vram is getting it in a form of Buildzoid Rambling
Curious, I thought there was one... Huh
Doesn’t the test current of your meter while measuring milliohms affect any components?
Test current depends on the range, with lower ohm ranges having a higher test current. The DMM sources a known current and reads off the voltage that the current source generates. According to the manual, this is at most 0.65 mA in the 300 and 30 ohm range. The highest possible voltage generated with this measurement is 0.65 mA * 300 Ohms = 0.195 V
0.2V is not enough to harm anything in the device.
However, 0.2V is enough to turn on diodes slightly, and it's definitely enough to charge capacitors. The former didn't show up as far as I could tell, and the latter was causing the resistance measured to change.
He was testing resistance not current. If you wanted to test the current draw you would have to pull something like the fuse and use that as where to put your probes for current detection. If you were just using it in the same places you would with resistance, not only would you be ruining the board, but it would also damage your multimeter.
@@Skylancer727 no, to test resistance, almost all multimeters source a known current through the circuit and then measure the resulting voltage across the leads. This also causes the behavior of capacitors charging up when you measure them. This is what the comment is asking about, as measuring the "resistance" of semi-conductors isn't straight forward, as they aren't ohmic devices. Applying a current across it can sometimes activate parts of the IC.
But I agree that measuring the current with the meter in place of the fuse will definitely fry the fuse of your meter! These gpu dies can take many, many, many amps at a low voltage, much more than the 10A fuse in most meters
FYI, my Gigabyte 3080 Ti just blew running Halo Infinite on Ultra settings. 12V PCI-E power on the top rail shorted to ground. The fuse did NOT blow on mine. Similar problem to this 3090, on the other rail, with shorted VRM caps and shorted 12V rail (PCI-E socket power is good), and no blown fuse.
sound like your PSU caught the short very quickly then.
thanks builzoid because I used this particular video to check my own card after the exact same new world scenario happened to me with forza horizon 5 I immediately "nose" checked my card after the shutdown and a heavy burned electronic smell was coming from what I didn't know was the inductors for Vcore VRM :/ the last 2-3 next to the power connectors (mine are on top of the card next to the power stages) my card still seems to work after that incident, anyway what I found out that in the menus, not running the game at ultra everything, in the "2d" menus the game can rise to 1000+fps with vsync unlocked and this draws spikes of 51.25A on the 12V O_O roughly 21A more than actual gameplay on maxed settings with RTX, this made my PSU trip because of the multi-rail OCP, I used an external PSU to diagnose that, set my real psu to single rail and that's when I saw the 51A :/ EDIT: wanted to record a video of this problem but honestly I don't want to blow up an unobtainable gpu so I just sent AHO and email he will hopefully receive
very interesting and im also very interested of a follow up of this case
Why not use a thermal camera and injecting power at safe levels for vcore and use a thermal camera to find your faulty parts?
I understand this cost some up front investment but you'll find your shorted parts on cards that sell for far more.
There is two hypothetical issues here, the caps if failed then the voltage rating may be too low for surges in power draw or if it's fets then fets are under sized for the current load and fail internally.
Lastly support Right to Repair so these schematics will be available for us all.
Finally someone who understands that the real problem is the cards, not the game.
I don't know why some youtube channels like JayzTwoCents spread misinformation saying that Amazon is to blame, when the real problem is the cards themselves. I've been saying it's the cards from Day-1, the second it popped up on all the youtube channels. But no, people wants to blame New World, but I guess people just don't know any better...
Sure, Amazon could put an fps limiter in there, but that doesn't doesn't fix the problem, it's just a workaround. If the GPU's didn't blow in New World they would blow in another game.
Most CZcams channels who talk about consumer products have zero clue what they are talking about, and should be ignored. It's like the marketing wank melts their brain.
Imagine paying 3 billion for a 3090 on eBay just to have it blow up... Damn.
imagine getting scalped for $2k on an ebay purchase, and not having a warranty as a result. Feel bad for this dude.
same applies for consoles sold by scalpers, this is why alot of people do not buy second hand stuff, the warranty matters alot
I have a Gigabyte Turbo 3090 with same pcb but with full poscap capacitors coverage front and back. I shunted the 6 shunts but still limited by Outputs around 460W total inputs in Port Royal. It would be nice if you could have schematics of these 3090s to understand why they fail. Great job as always 👍
17:43 Because otherwise it would not fail just after its guarantee time had gone out and Nvidia earn money when the costumer buy another card. And the thing repeats.
asymethrical VRMs seem like the kind of thing you need a very smart controller and powerstage to work right, so like "Hey powerstage 1 and 2 can take 120 amps but stages 3 through 10 can only do 80 amps, so I better set the powerful 2 to take on more load and monitor how much each stages is using in a very high speed to make sure one stage in particular is not under too much load". Seems bad, like the kinda thing that sounds good on paper but is too costly to make work.
Wait they said no? It blew up playing new world i thought they knew this was a problem.
But this was Gigabyte... They aren't exactly known for going above and beyond when it comes to responsibility. Check out Gamers Nexus' videos on their power supplies for a good reference point on how they handle it when someone methodically roots out the mistake and proves it to them....
The new owner didn't have the receipt = Go fuck yourself regardless of why the card failed.
I dont know if any of the brands do RMA for cards not owned by the original buyer.
EVGA has a form you can fill out when you sell the card so the new buyer get's the rest of the warranty (or they used to) but it needs to be filled out or no RMA.
ASUS will have no warranty on ebay cards
Gigabyte is same
idk the rest you can look them up.
@@andersjjensen everybody here has seen the gigabyte psu videos
on alot of countries you do need a proof of purchase from a seller, second hand does not apply for alot of manufacturers
they do this to avoid gpus that could have been manipulated or could be stolen
This is an invaluable video, thanks!!
how to craft fire. this GPU + gigabyte PSU + new world
Its hard to tell whos fault it is it could be the game software, driver software or even hardware all areas could have had preventative measures to stop a melt down.
From a gamer who has been looking for an upgrade for the past 18 months, this is sad.
Intro voice, KERMIT THE FRAWG
I think this GPU requires its PSU to be able to handle at least up to 700W power surges to call it properly sized. So it looks like just the onboard PSU wasn't properly sized to handle the load surges, especially if the cooling system is good and the GPU is cold (thermal throttling is off).
a problem I never had with a 1080ti but that clearly became one with a 3090, too high fps = insane heat (the 450+w 3090ti is scary) I mostly play vsync on a 120Hz monitor, I unlocked it in borderlands 3 1080p and was "woaw 240-350fps in badass that's stupid powerful 2-3x the fps of a 1080ti" suddenly I realized it was hot in the room very hot 36°C hot...enjoying the game I didn't notice but a massive heatwave was exhausted by the pc case, 3090+unlimited fps = the card runs at max power nonstop it's way worse than games "benchmarks" (garbage and irrealistic marketing tools) I only have a ref 350w one tried a 390w bios rolled back, the heat generated is insane we're at the state where you're talking +5 +10°C in your room for 10fps...without aircon obviously
hi kermit. glad you found a 2nd career.
you can get schematics for a Commodore 64 homecomputer, from the 1980s, bro.
Reduces my wattage by 25% during new world launches, works great, new world by default runs wattages higher then they should due to coding issues, simply reduce your wattage until they figure this out.. done :)
Would you be able to repair this card by replacing the affected power mosfets & filtering caps? Or does the chip itself get completely cooked when power stages fail like this?
I have a 6800XT and it's ridiculous how many game menus just let the FPS go crazy, pretty much the only time I hear coil whine is in some menu going nuts on the fps, it's genuinely frightening to know that a careless dev can brick my hardware.
An MMO that destroys GPU when Crysis didn't.
A new world indeed.
Or the Old World cutting up GPU's in the "New" Lumber Yard?
jeez, if you had access to that schematic it would cut your diagnosis stage down by like 2000% and change your confidence in complete repair from ~90% to 99.9%
hopefully there'll come a day when we can get access to schematics for personal repair..
I play New World on my EVGA RTX 3080 ftw3 Ultra and I've been monitoring the wattage draw and it's pretty low compared to many other titles I play so I'm still really confused about what exactly is the problem with New World. I limit the frames to 100 and the most I draw is around 320w-330w playing @ 1440p. While with older titles like Battlefield 4 I can draw 360w easy.
edit: wanted to mention I have 100 hours in New World already
Jay showed that for some reason a EVGA's cards were pulling more power than the overclocking software requested even in other programs than New World. But New World also being heavy on hardware makes it show. It's not the fact it's pulling power but pulling way more than it was rated for. It seems that New World's trigger might be the fact it swings back and forth similar to how stopping and starting a car is worst than constant highway driving it.
One way to test is using Dynamic Resolution scaling, which I believe you can set to 4K even if your display is 1440p, it will upscale the processing then downscale the output resolution to your native resolution. This was causing Jay's card to increase power consumption far over max power of the card on a FTW3 3090, but acted normal if he left it at 1440p.
@@racerex340 Yea most definitely upscaling to 4K will push your card to its limits. I wonder if there would be less or the same power draw if Jay had tested it again but with a 4K native panel.
@@r3conwoo I would practically guarantee it, although it might not be very much, but down scaling has to cost at least a little more GPU processing power
I also have an EVGA RTX 3080 FTW Ultra. I decided to install new world on my old system which has a GTX 1080 FTW. Runs just fine there at 1440p and no worries about the GPU.
My Asus Strix 3090 OC GPU also randomly died but playing Forza Horizon 5, system just switched off mid game, New World has never been near my system as it’s a game I'm not interested in, when I press the power button I get a split second of power to the card as the RGB on it flickers and to the rest of the system, RGB ram stays lit and the rest of the system powers off instantly. Once I took the 3090 out my system booted just fine to the desktop but as I use a 5950x and had no spare GPU at the time I couldn't see anything but I knew it as at the desktop as my RGB lights changed. Popped my card back in, even tried in a different slot and again only split second of power.
My GPU was only 11 months old, never had a single issue, no warning signs that something was wrong, GPU temps never passed 69c, VRAM temps were always around 80c to 85c at load so temps were well below the max temps (never mined on), happily playing FH5 and the system instantly shut off and that was the end of that card.
Managed to buy a new 1050 Ti to at least get me a display and even play some games temporarily, and so far my system had been up and running for 6 days straight without a single issue so the PSU was ok. My PSU is the Corsair AX1600i so way over powered even for a 3090 and 5950x, but the voltages according to icue are stable and rock solid.
Sent mine off for RMA last week and received it back in a total of 4 working days from Scan in the UK which was extremely impressive considering how hard GPUs are to get at the moment.
So far my new replacement Strix 3090 OC is working just fine, but makes me nervous now in case I have another ticking time bomb on my hands.
a gigabyte card now?
so its not evgas solder joints?
EVGA was taking the hit for Nvidia
this tells me that is a nvidia problem, they did not made the best vrm design
Amazon devs hiring process is so hard, that I can easily see them implementing the drawn calls using an overly designed and "optimized" implementation that works on paper but not in actual real world hardware.
The optimizations they made might actually be working too well. Lets say Nvidia made the assumption that the card will somewhat frequently be stalled out by memory access which would limit the peak power draw to safe levels. Now if a dev comes a long and implements an algorithm that basically never stalls achieving a much higher level of core utilization than Nvidia expected you end up with the card pulling way more power than the PCB is designed to handle.
EDIT : The never stalling on memory access is basically what makes Furmark so damn hot.
Amazon is a GARBAGE COMPANY do not use them for anything, do not buy!
Even the most optimized code shouldn't make a card blow up. I mean, I write simulations in C and CUDA and I actually use average power draw as a quick way to check if optimizations I make are worth it. More power means I got around a bottleneck. Any (stock) hardware that you can't trust to not blow up under full load is broken; it's just giving the illusion of working before it fails completely.
@@friendlyoctopus9391 I feel like a graveyard of cards killed by Furmark want to disagree with you.
@@rankcolour8780 Those cards were weak and deserved to die