Progress update on the New Worlded RTX 3090 Vision
Vložit
- čas přidán 12. 09. 2024
- Turns out GPU-z really doesn't put enough load on the card. 3Dmark cloud gate got the core to boost to over 2GHz so the core is fine.
My Patreon: / buildzoid
Teespring: teespring.com/...
The Twitch: / buildzoid
The Facebook: / actuallyhardcoreovercl...
#RTX3090 #NewWorld #Nvidia
Turns out GPU-z really doesn't put enough load on the card. 3Dmark cloud gate got the core to boost to over 2GHz so the core is fine.
Run Star Citizen on it :D It's also built on Lumberyard. Good stress test 😄
Just run Heaven no need for the benkmark, just open and run.
The card got "New Worlded".
You just causally created the best description for "failed power stage on RTX 30-series nvidia cards".
Btw some of the 30 series card models have shown a rather sluggish behavior during load transitions, with quite significant over/undershoot, which might be a contributing factor whenever or not the power stages blow up or not.
Bios issue?
man you've just saved the owner like 2000€ 😂
I would like to hear how you identified the blown vrm with them being parallel and all.
gg BZ! please, when the parts arrive, show us how you stick them to the card, how you repair it. for me at least, is really fun watching this kind of video. great work on repairing the card
if you enjoy board repairs like that, you might want to check out AdamantIT (If you don't know him yet). He has some really great videos doing and explaining it.
czcams.com/users/TechCemetery
czcams.com/channels/LqBEpeQPZTdhCd0nHWIf6g.html
NorthridgeFix is a good channel also
@@pcoverthink When he was yelling at JayzTwoCents in this video czcams.com/video/y_v7p3xEPDM/video.html , Louis Rossman was epic .
The Whole World: Must add more cooling to the 3090. More cooling to the memory. More cooling to the core. More cooling to the back.
Buildzoid: I like to cool my 3090 with a stock AMD Wraith Stealth cooler.
even hp likes to cool intel cpu´s with amd stock coolers, and vice versa^^
its not blowing up because of cooling. its blowing up because a circuit gets overloaded because amazon programed new world terribly. although you have a point any card with ddr6x memory should have better cooling. or just water cooling.... that shit gets hot....
@@majstealth no it's the other way around, Dell and HP like to cool AMD CPUs with Intel stock coolers
@@jackofallglass165 >amazon programed new world terribly
Unless the game is doing some undocumented API calls, it's 100% GPU's fault.
@@Vladislav888 You need to do some research before making broad statements. New world has been seen plenty of times to be pushing cards past max current somehow. In some cases over 20%. An overcurrent situation causes exactly this issue. Stay away from New world, problem solved.
really hope you do another update, i'd love to see the card hit full performance with all power stages. Awesome that you brought this card back to life
Ah Buildzoid performing gpu necromancy yet again
WOW, thank you for making this video.
With everyone being completely clueless on the issue, YOU are the one to bring actual, verifiable and technical information on the matter. You sir can have a medal.
No issue with my 6800xt on new world at all with the msi gaming x trio, which I trust for the power filtering quality anyways since your review, I do hope all those high-end cards don't start blowing up like fireworks in a few years when games start to put more heavy load on those as you said...
Keep up the insanely good work, cheers.
u mean msi right?
@@MRNo.8 arrr yeah I do I don't know what I was thinking I'm editing it thx xD
You must be talking about people like Jay’s who doesn’t seem to know a whole lot about a lot of different subjects
We are not clueless. This thing is absolutely obvious. The cards are garebage. Why? 3 reasons.
1. Why the hell is there no 12V a ground backplane. As it is now, some mosfets are feeding from one 8-pin, some from other and some from PCI-E and its UNEVEN. Aka one power connector has way higher power draw than the other (Iv tested this myself on the Evga one).
2. Who was the idiot that connected the mosfets to the controller in uneven configuration? What was the thinking there, the more phases the better for marketing, no matter if they can actually efficiently work in such configuration.
3. The clocks and voltages are way over the top, far beyond what is healthy. Those cards have 450W load and sometimes peaks over 600W. Its insane. So when a game comes that produces uneven load with spikes appearing often that the power-limit does not catch, bye bye the unlucky power stage. How about running that card reasonable like the A6000 which performs somewhere between 3080 and 3090, runs at around 1600Mhz and never goes beyond 300W and surprise, does not tend to die.
The funny part were the excuses like Evga with their "the card are fine it was soldering quality issue in the factory... bla bla. Lets face it their cards are crap, but generally the whole GA102 GDDR6X gaming lineup is like this. Core overdiven to insane clocks on insane power-draw (just so the RDNA2 cards do not beat them in bencharks). The only real way to keep those cards safe is significant undervolt. At around 0,8v and whatever freqency you hit stable on such voltage is what I would consider OK long-term.
would my xfx 6800xt run new world just fine u think?
2013: Can it run crysis?
2021: Will it survive New World?
well even if it survives it might still not run crysis!
Here's one thing I'm really curious about: The uP9512 supports automatic phase shedding. Could you probe the PWM signals to see whether phase shedding is actually enabled on this card? Maybe the uP9512 can't enable additional phases fast enough when a hard load spike hits so that a few phases get overloaded even though the total load is well below the VRM's rated output power.
my guess would be IF it sheds it's in idle mode only, anything else it'll be full phases.
After the last episode it was really stressful not knowing if the core got 12v sent to it or not. Glad it survived
Right! I'm wondering if the core being fine after a power stage failure like this is typical or just lucky?
I just like seeing that AMD stock cooler on everything you use
alright so i'm just going to check the turbo pressure on this 2JZ right quick, I'll run the coolant through an aluminum block and put my AMD cooler on it for the time being......
Go a wraith cooler on top of my 3090 FE - getting kick. It heats very good when F@H
Epower your 3090 to play New World. It's the only way! XD
I mean if the issue is with the alpha omega drmos parts, it might be possible to switch the IC, although the pinout seems to different depending on manufacter even with same packages
@@krazybonnie5523 Do you have more information on the pinout being different?
Evra 3090 have issues with the game because the gpu cheap but have we heard any faults with rog strix Asus tuf or founders edition 3090 playing to this game
@@shadowaussien7743 ???
I had my RTX 3090 Vision die in a similar way but not from New World. Gigabyte did honour the RMA but took, 3 months to repair it and send it back to me. Mine died whilst playing FFXIV with heavy reshade shaders running.
Thanks for doing these videos it is interesting to see why it failed etc. I guess from the fact there are quite a few reports of other gigabyte 3090s dying and mine doing it months ago no New World needed there must be some kind of flaw.
were you playing 4K?
Imagine 3 years from now WHEN EVERYONE IS OUT OF WARRANTY AND GAMES ARE SO DEMANDING/ GAMERS MOVE TO 1440P 240 HZ/270 hz they might blow up cards!!!
czcams.com/video/4DtUCwTXm9Q/video.html ................ I agree!
@@Multimeter1 Imagine that 1/10th of buyers got there card of ebay with no warranty at all...
im sure that never will happen.
Founders editions or Asus tuf are the only cards I don't here having complaints
Hehe funny enough I stoped using shaders in FFXIV because it would ramp up my fans to 100% after a couple of minutes, 3090 FTW3 here, after the new world open beta killed some 3090's I was glad that I stoped using reshade, EVGA did said that even if the New World code is trash and it DID killed the cards it was also because of a poor welding on their part which was worsen by the power drawn
So glad you did this. Thanks for the update was hoping to find out the outcome. Keep up the good work \m/
nice to see the card is alive, thank you for the update/share BZ!!
Motherboard power delivery generally doesn't blow up running prime95, yet gpu power delivery used to blow up all the time on furmark, and still would if they hadn't built driver safeties. Considering new world is blowing up gpus due to the work load, isn't it about time we asked the real question: why are we allowing gpu manufacturers to get away with using half assed power delivery? It should be a standard that gpus must be able to run extreme loads for hours without failure, especially now that gpu manufacturers are charging a premium.
Similar to what my strix did after 6 hours in new world it worked but would artifact and video outs would randomly go out. It wouldn't boost in games but would still run them with lockups and hitching and dip to 205mhz core and 410mhz mem in games. Had to RMA it. The repair notes said all display outputs failed in testing so they sent a new one.
Did you get your own card back or was it a new card/refurb? LHR?
Dragging down the power limit to 85-90 should do the trick. Still, this should never be possible to happen, software shouldn't be able to damage the hardware, there should be protections and limiters set in place.
Jay twocentz showed an issue with evga and undervolting the card specifically in New world the gpu would still spike well over the set limit. He was using afterburner with an evga card, not sure if precision would regulate it better... just double check if you do undervolt it that it is actually working.
like BZ said in his first Video, it is very likely that the Power distribution design is at fault here, that delivers most of the power through one "lane" of the two 8 pin connection and overload one of them with it, even if the Card isn't under full load. So that isn't something that can be fixed with better cooling or better Components. BZ would design a new Power delivery for this Card...not that he didn't do similar things before but i guess this High End Top of the Line Card is a bit to advanced even for him .
"New Worlded" 🤣🤣🤣
Side note: That afterburner skin is sick.
Thank you for sharing. Your skills seem impressive. Looking forwards to the final repair and analysis (Part 3)
"Can it run New World?"
What is New World - The new crysis mod you mean?
"Yes... but only once" :P
@@andersjjensen more like "how long does it run new world until it blackscreens and your room is filled with magic smoke?"
It’s interesting that Nvidia has taken steps to detect Furmark and aggressively throttle their cards to prevent damage. It makes me wonder if there’s something flawed about they way they handle power delivery on their cards that when poorly written software/power viruses run on it, it ends in hardware failure. I’d love to know exactly what the root cause of the issue is, because this is gonna be one of those problem that just goes away quietly. Nobody wants to own this one and be on the hook for their fuck up, so Amazon is gonna patch their buggy code (if they ever figure it out) and never put it into their patch notes (unless sales are tanking, in which case they’ll be advertising how they fixed the issue to get people willing to play again.) In addition to that, I’m sure Nvidia will patch their drivers to detect New World, and throttle it down like they do for Furmark, and OEMs will revise their boards or swap component vendors, so we won’t see it on any newer cards that are sold-so, we’ll never hear the actual end of it.
From the behavior I’ve heard people describing, these failures sound like the game is causing the GPU to quickly ramp the performance up and down, and if it’s causing massive voltage spikes, there may be some huge transients that are hitting it so quickly that the over voltage protection doesn’t have time to react and that’s what’s blowing these things up.
I’m not an electrical engineer or anything, but I’m an IT guy and I spent a decade studying audio recording and the way digital audio is recorded looks exactly like what you see on an oscilloscope, and audio engineers (like what I went to school for) encounter transient spikes that are pretty massive, and happen so quickly that unless you’re using a compressor or limiter than can look ahead, it can’t clamp down and attenuate the signal fast enough to catch them. On a loud enough transient on the exact wrong system, it could blow the speakers, which is why recordings that were mixed and mastered before digital was the defacto standard (and the software limiting and compression tools had look ahead functionality), the average volume on tracks weren’t as loud. Now that they’ve got tools that can anticipate when there’s a big transient spike, recordings have gotten ridiculously loud on average, and if you’ve ever experienced a TV commercial that sounded WAY louder than the program you were watching, then you’ve heard what overly compressed/limited audio sounds like.
Anyway, that digression aside, I was thinking that maybe once you get this card working as good as you can get it, that maybe you hook it up to the oscilloscope and see what those transient responses look like. If I recall, the issue when these GPUs launched on AIB cards was a similar one, where the types of capacitors being used in some cases weren’t as good at providing enough voltage quickly, and that was why some cards were having issues. I kinda wonder if they might’ve over done it with the fix for that problem, which is causing this problem, because if it’s too fast, and drawing too much, then the built-in over voltage protections aren’t going to have enough time to respond, and since they loaded these boards up with a ton of those fast little capacitors, the amount of voltage is probably more than a little bit nuts.
Man, as much as I like my 3070, I’m kinda bummed I couldn’t get my hands on a 6800XT like I wanted to when those launched. It just seems like Nvidia can’t get their shit together for the power delivery on a very power hungry card. I can honestly say I wouldn’t miss the ray tracing performance for faster rasterization and better efficiency, not to mention the fact that I have a monitor that supports FreeSync Premium, and regular ass Gsync just doesn’t look as nice.
Honestly? I'd say nvidia are saving themselves a headache doing it.
I remember some notorious maps in infamous2 which could cause the fans to go insanely fast - bearing in mind it had NEVER spun up to that extent before even in summer.
Poor code can easily wreck hardware. I don't have a better analogy, but it's like chucking a rock into a washing machine, where sure you can put the rock in, but it won't end well for the machine. Except here it's bad code and the gpu
I have no issues with my strix 3090.... none not in new world. It ocs like a champ
I'm wondering how engineers haven't yet thought to implement a hardware mechanism to detect an overload for any reason, no matter what software is being used. Just detect power usage spikes, amperes, temperature - whatever to throttle the card down and prevent burning. I would rather like my GPU to start throttling insanely in those dangerous situations, so that I know that something's not right with the game/app I'm using.
I mean - if our CPUs would burn out because of being overloaded in this way, we would be getting dead CPUs so much more often. I could run a simple code loop that loads all cores 100% and the CPU won't die. Why can't we have the same behavior on our GPUs?
That also makes me wonder - if I want a totally reliable non-overclockable GPU that favors reliability over performance, then where can I get one? I'm doing some gaming, but also experiments in Unreal Engine and neural networks, and Blender renders. I wouldn't want my GPU to burn just because I messed up my code.
@@camelCased read the original post. There are voltage and current protection mechanisms in place. The possible problem that OP raises is that the voltage or current may spike faster than the protective systems can engage.
@@martindinner3621 And that essentially means there is no effective protection against spikes and bursts. That's what I meant with "overload for any reason". CPUs seem to be much more protected against this.
So cool...good luck on replacing stage!
Damn, BZ needs to come out with more sequels to what hopefully will turn into an anthology series for this card! This is a journey!
This is really awesome of you buildzoid
Power stage failure (inductor good), input fuse replacement. Got it...Thanks for sharing.
< cue incoming flood of owners with popped cards > RIP BZ's inbox
I have a 3090 FE that hasn't had any issues in New World (I also limit my FPS to 60) but Satisfactory did expose a cooling issue within my system. The pushthrough design of the GPU cooler was causing my NH-D15 to hit 91c under heavy load and I'm assuming hitting thermal limits causing the system to restart. It never completely quit on me but I did get black screens needing restarts. After multiple restarts it wound up breaking a USB controller somehow causing certain ports to disconnect/reconnect randomly. I was able to fix both issues by replacing the motherboard and switching to a Corsair H115i. Temps are now in the 80's and USB is working fine on the new motherboard but I'm curious knowing what you know about the FE cards if they could potentially suffer the same issue.
I think its AIB cards and not the Founders.. At least I haven't come across anyone who had a founders and died.. I could be wrong tho.
From what I understand the issue that caused the cards to fail was software based not hardware and that was fixed. I think they said there was no framerate limit and in the menu it was going way to fast, but they have since patched that. So I don't think it's the hardware not being up to par, the problem is the software and firmware not limiting the processes so as to not overload the hardware.
How did you determine which power stage was dead? Also, is this really that widespread? EVGA claimed there were only about 25 cards that failed this way originally, which doesn't seem like very many, given how many cards they appear to be selling.
Also, WRT the 980 Ti, those seem to be failing mainly after many years of use. You could conceivably repair it and get another six years of use out of it, which isn't that bad a deal for an hour or two of work.
See his previous video for diagnosing.
This is a Gigabyte card not an EVGA
Hey! Where are you located? Impressed of your knowledges! Have 2 dead motherboards and both are the same asus crosshair
As an 980Ti owner, you got me really concerned. Tho I recently started saving money for a new maxed out rig, but that's gonna take me till summer, and at that point I'll probably just wait for 40s series.
Awesome work dude, your knowledge and dedication is amazing, keep it up
Thanks for the update. Very rewarding to return a device such as this to a functioning state.
Hi Buildzoid, do you think that this board design would benefit in anyway by having the large empty capacitors locations populated with tantalum polymer 2R5TPE470M9 2.5V 470U ?
It happens that I got plenty of those and a card that I'd be willing to put these on if any gain could be had.
At the moment the card seems to be starved of power on heavy workloads... The core clock lowers automagically when the DDR6X is OC'd and I can't get over 460W usage with a 500W|550W bios (thick pcie power cables are used, not getting hot at all under load).
With current air cooling mods, the temperatures are well within specs when testing; the core sits between 48-60c and the memory is most of the time lower than 80c depending of ambient.
Been following your channel for some time now and I must say that I really enjoy the unique content that you create. Thanks for sharing your knowledge with the community 🙂
I believe EVGA have released a Firmware update for their 3090s which can be downloaded and installed via their Precision X overclocking utility.
Jayz2cent showed how the game New world stresses the card while not actually playing but in the game menu, he also suggested to underwolt the cards until there is a solution, driver update or what not just wantet to say this, great video i just supcribed looking forward to watching more from you greetings from sweden
JayzTwoCents did some testing with new world he found the game makes the card pull over the set power limit it was as if the core wants more power the vrm says sure without checking the power limit, assuming blowing the power limit is the issue that would be a driver or firmware flaw, in Jay's testing some cards ignore it and other don't so i would guess the vbios has a bug
i saw that same video. it was weird how some followed the power limit settings and some didn't. but the power draw spikes were what? topping out at like 120 someodd percent? thats a LOT of power on some of those cards...
Bringing a 3090 back to live. Maybe we should think again about who to call "Tech Jesus". 😅
If you put the power stage back in and it works fine, then the best thing to do would be limit the power curve on core, and boost mem. Games are dangerous, use it to mine 😅
How did you work out that it was that particular mosfet? your previous video you narrowed it down to 5 of the mosfets. Did the capacitors shorted on the back also require replacement? I've got the same card and that same 12v line is shorted. But no fuses blown, i think the power supply might have protected it somehow. Gigabyte won't honor my warranty and I'm in Australia can't find anyone to repair it. I've never soldered smd before but it's kinda my last resort.
I lifted the inductors one at a time till the short was gone.
So other power stage that got mark in previous video did not broke?
Just pondering on your thoughts of the power stage.
Does AMD use the same chips?
Also pro tip if I may (been IT for over thirty years and still have my own computer business)?
When you turn off the PSU via it’s switch. Press the power on button on the case to discharge all the caps and any power left in the board before removing components 🙂.
Great Content :) 3090 BZ Edition on the horizon? ;)
Hey any chance you have a video of you doing the soldering and swapping the broken components out? I think that would be interesting.
Are you gonna send back the 3090 after you fix it or should we expect more 3090 videos?
He'll probably send it back. I mean the owner of the card probably payed thousands of bucks(scalper prices) for it and now the poor guy's probably watching paint dry or something.
it's getting sent back
@@ActuallyHardcoreOverclocking Mod the card before doing so with cap and voltage mods. The owner has no guarantee anyway, and if he pays up you get some more content out of it.
@@Airwave2k2 cap mods is a terrible idea. the caps put extra stress to the mosfets of the power stages - caps when not 'full' appear like a short and the in-rush current is pretty substantial. voltage mod is likely even dumber. The only sane mod is reducing the max power draw...
Good job
Next video BZ tops single card 3090 3dmark score on ln2 :P
Would you say from and educated guess, that lowering the power limit to say, 90%, while playing New World is probably recommended for these cards? Thanks.
Very interesting update and amazing fix so far :) , who'd think a game could cause a graphics card to blow its socks off like that! I did think about giving new world a try but I'm not risking my 2080,even if this issue is only related to 3090 cards i still don't trust it.
@Actually Hardcore Overclocking, Do you repair older cards such as a R7 360?,I believe there's an issue with the Schottky barrier diodes on it.
New World had nothing to do with the problem, it was the unforunate thing that happened to step on the landmine. The cards were dying because of crappy soldering at the factory, the cards were going to die, the house was already on doused in gasoline, New World just lit the match.
Some cards don't like to run properly with a missing power stage. It has something to do with output power calc miss mach and only custom bios would cure it (but changing a mosfet is simpler.
How did you figure out which power stage was broken?
What would be super interesting is if you were to breadboard the VRM circuit stand alone and attempt to load test it and see if it behaves like it should as per documentation or if it just goes pop!!
Could it be a stability/oscillation issue due to the way the stages are ganged up and not actually all driven by a controller as separate phases?
Just pondering how certain high frequency intermittent loads can cause certain frequencies of oscillation on the rails and theoretically lead to some sort of mismatch/overload/feedback on a stage. If it is the same one that always blows then it is plausible. If the load can be synced to any of the phases then I guess any could be the one to blow.
Most def a design flaw and also a firmware flaw, it should of prevented rapid cycling of power states (going from 60fps to 500fps, etc). So all 3090 etc are still at risk
I'm inclined to agree. In my opinion any card should be designed to withstand an absolute worst case scenario as we are talking about software. There should be no way for software to damage the card as long as it doesn't try to interfere with the power settings or frequencies on the card. That are things that software like Afterburner fiddle with, and it comes with warnings that you are touching things the user are not really supposed to tinker with so you are doing so at your own risk. A game however doesn't have any reason to touch anything like this. Just issuing draw primitives and accessing the options exposed through DirectX, OpenGL, Vulcan or whatever other API it shouldn't be possible to damage a graphics card even if you were to deliberately use the absolute worst kind of workload however unrealistic. That's what safety margins are for.
Lets look at heavy construction machinery. If you have a loader that are rated to lift say five tons then you expect it will do so all day every day without blowing up. If you try to lift six tons it will most probably do so, maybe even seven. But once you get high enough it will stall, overflow valves will trigger or it will somehow refuse to do more than try. This is a built in safety. It protects the machine from damage and also the workers from what a failing machine may cause. But the machine is actually not yet near the failure point. In some cases it's possible to override the safety features, but doing so means you have to manually disable them, and no one will do so without a really good reason because when a huge construction machine fails things can go really bad in really horrible ways.
Construction of any kind of graphics card should be a bit like that. If the card can normally draw a max of say 200W then the power delivery should be designed to provide at least 300W continuously and transients of say 400W. Now that is figures I lulled out of my ass so they may be way off. But the safety margin should never be less than 50% and more probably 100% to 200%. Some components age faster than others, and heat and high frequencies can accelerate that. A graphics card that costs north of $1K and often even $2K is way to expensive to have it's lifespan shortened by cheeping out on components costs.
I agree 100% had my buddy's 3090 ftw3 brick loading into apex, had my 3080 ftw 3 brick loading into zero hour (seige style tactical shooter). Got my new card in 4 days and I've undervolted since and underclocked immediately.
the hope is that a firmware will fix it to some degree, because this gpu will not be replaced by nvidia or gigabyte
Not a regular viewer this 2 part what happened had me curious glad to see you got it working. Please do a fixed vid as that would top off this very well. Thanks for your work on this. Daz from Australia.
New viewer as well it sounds like he will he just is waiting on the parts to come in. I'm kind of excited.
Great news to hear ,I thought you end it up as paper weight or a conversation piece
the gpu is borrowed from a subscriber, it is not his property, so will never be a conversation piece
Maybe the failed power stage wasn't making good contact with heat sink? Overheat maybe.
it is possible, but a good thermal pad should be installed over it after buildzoid replaces the parts
Well this video answers my question I posted on the previous video.
So is this fet tied to another power phase on the card that may be drawing on the rail at a higher current that this fet is able to operate at?
Do we know the theoretical limit of the wattage of the vcore rail vs what the fets are rated for?
These are probably "duh" questions but it's the stuff I think tickles my video interest.
So I'm taking it you don't believe the official release about the microfractures in the solder joints?
In previous video you found coupe of potential failed VRMs, how you actually distinguish the one which actually failed?
I would inject voltage and see what is becoming warm
Buildzoid, what do you think about the RTX A5000 for a workstation? I don't like that it has the power wires soldered on in such a jank way
How did you actually pin point the dead MOSFET. The previous video ends with a couple of them being marked, im curious how you found the real culprit?
Feels like it's link to the original black screen, where Nvidia had to release a driver to down clock the boost.
Card is just sucking up to much power.
They'll do it if it gets destroyed by various titles of games.
Just new world? no. xD
Unless Amazon ready to pay Nvidia a whole lot of money, because doing so (release a driver to reduce the boost or whatever)
will mean that NVIDIA agreed to fully take the blame, which I am pretty sure they aren't.
But then again Nvidia can say something like "None of the founder edition is reported, which means it's the AIB fault", but that will sour relation with AIBs.
My gigabyte gaming oc rtx3090 died slowly. Black screen + fans 100% became more frequent over time and then in idle it happened, the last time it posted
I like how BuildZoid doesn't care for clickbait. That shit bugs me, its very disingenuous and feels very 'last gen corporate trying to be one of the kids' I dunno bout you guys, but I scroll right past obviously clickbait titles/thumbnails. Buildzoid knows the views he gets, are from people that are genuinely curious are care about the actual content.
BZ at 2:10: "Well, no, we're not gonna' do that. Card's currently missing a power stage."
Me: My Sapphire RX 470 8GB runs, sometimes overclocked, on only three out of four power stages. (Because I don't have the skills to install the new stage.)
God daum doctor Frankenstein I'm gonna admit I did not expect this to work at all
Is there a reason your PCIe slot area looks like it's rusting on that motherboard? I'm sure it's not that, or I doubt the board would work :D rather curious about what it is tho
I think that is plasti dip (maybe some other insulator idk), he uses it for insulation when he runs ln2.
been watching your vids for a few years and your voice was so familiar, but i just couldn't place it until now. its kermit the frog lmao :)))
Hey BZ how does the Vision compare to the Aorus for VRM design?
Nvidia homepage states warranty issues if fürmark is used to test nvidia cards. But I full agree about fürmark as bad gpu stress tool by desgn.
Was just about to buy 3080ti vision oc … now what :D im planing to render on that gpu, so im bit worried it ll fry in months. Can u please test it in octane benchmark when its done ? just to be sure if there is any possibility for card to die just rendering in octane :)
Glad to hear the core is ok! Could you monitor power and transients when u test the repaired card? Jayz2cents has been seeing odd behavior from some 3090s violating their power limits, combining that with heavy transients might be blowing stuff! I’d love to know what this card is actually doing under load and if the phases are actually balanced effectively
Its because those UP9512's are the internet explorer of voltage controller.
Wanna put more load on GPU-Z? Select the full screen option, then press alt+enter and depending on the size of the window the load either decreases or increase.
ha ha power stage go boom
how do you notice a dead power stage ?, does all the pins on it short or something? i'm genuinely curious
Can you just install one new MOSFET and put the inductor back on? Is there no need for all mosfets being out of one lot?
In a previous video he said he might be able to get a schematic for that specific card because there were some leaked. He didn't mention it because either he couldn't get it, or it would be a time-wasting rabbit hole to evaluate the design of the power delivery circuit
He would probably need access to an R&D department's equipment/documentation to accurately analyze it and I don't know what equipment he has
I mean, he could probably do it himself, but how freakin' long would that take?
He'd have to do what gamers nexus did with the gigabyte power supplies. Have specialized equipment, either pre-made or equipment he made himself, to overdrive the power circuitry and see at what point it fails and how the circuitry handles the failure. That could cost a hundred thousand dollars easily just to do that. Basically, a crate of new cards to test with, or at least remove the gpu chip itself from the circuit and replace it with a simulated load with boxes of replacement power delivery components
That would be extremely time consuming and expensive
And replacing the gpu chip with a simulated load would be very difficult to make accurate
You can't just say, the gpu chip is a 0.1 ohm load so we'll just use a very high wattage resistor. It's also probably an inductive load at least and who knows how it behaves in total without killing a few just to find out
I think it can be summed up by saying, yes, new world is killing cards because it's badly coded, but it's the card's bad design allowing it to happen
So in essence, it isn't new world's fault, and every killed card should be replaced and any manufacturer that doesn't warranty all cards affected should die an a fire
I'm looking at you with a death stare, gigabyte
It could be possible there was a bad batch of components that are failing and Gigabyte, ZOTAC, and EVGA got them?
Wouldn't surprise me with the high demand/shortages with everything.... I'm sure you saw the GN power supply vids...
yes, it is a possibility, for sure, but seems to point more at a design failure more than a bad chip
My 3090 vision oc also broke a couple of month ago, whilst playing an old game (card not under full load). I sent it in for repair and received it back. I'm now kinda curious as to the reason why it failed. Gigabyte of course didn't mention anything, they just sent the repaired card back.
i played dark souls 3 with a fps unlocker mod and it fried it immediately i think i'm gonna frame lock globally to 144 maybe the surge in fps crapped it
Genuinely curious to find out what sort of shady nonsense was involved in this whole debacle. I refuse to believe a game can be blamed, my rx480 doesn't even get hotter than any other given game nor does it exceed its power limit (remember, these cards once did exceed their limits when they were new, and AMD patched it). And game code can't disable power limits. I think these manufacturers were fudging their power limits to have some sort of advantage in benchmarks and it's now being exposed.
^ This! Product is badly designed. It should be able to be used 100% without risk of damage.
Perhaps setting a max frame limit on the driver level would extend its life.
undervolt/powerlimit would do the trick
Both. Frame limits and undervolt/limit power and limit clock speeds
So...Get the GALAX HOF version instead of any other AIB model?
new world sounds like such a good game.
So
The only way to get a card is from a scalper
But if you buy a card from a scalper, the manufacturer will refuse to RMA it
Why isn't anything being done about this
Please buildzoid show full repair video i really enjoyed your explanation videos. I would like to see you get your hands dirty while you are repairing it
The heatsink top left of the thumbnail looks like it spells WEEB! lel.
I like how "New Worlded video card" is now a thing.
Check the near the pci-e gold contacts,, because i saw it crack on that area for gigabyte
This whole situation looks like transient load of 3090 are too high in New World and vrm design of some card models such as ftw3 cannot handle the load spikes because vrm switching frequency is not adequate.
I support that VRMs are not properly sized. It's not a software (Game) problem.
Would you say by dropping the power draw on the card would help in new world. Mine pulls 350 watt and I have dropped it down to pull 250 max.
Do you think with better cooling, the power stage would not have blown up?
I have a 3080 FE and considering to play New World, but didn't want to consider it until I put a water block on it _(which I already purchased)._ I figured it'd be safer with more than adequate cooling, but if faulty hardware, then maybe not.
Better cooling wont save this
@@effive7817 Why is that?
Breaking the power limit isn't a temperature issue. Some people limit power very low and fps cap, dunno how safe it is you'll need to research more reports I've only seen a handful.
@@effive7817 Quite a few 3090s were breaking from bad production issues as I understand it. There was no discussion on whether those people were breaking power limits and that's not what I was alluding to.
I simply thought that whatever the issue was, it might be temperature related due to big power demands by an app (game in this case). If temperature exacerbated the bad workmanship, such as it was rated for 100 degrees but the bad workmanship caused for failures to happen at 85 degrees, then watercooling might help avoid the issue.
If there are other problems, like bad grounding, EMI, faulty mosfets / VRMs, I guess that can cause other permanent damage, just not sure how when it only happens with certain types of loads / stress to the GPU.
Please, I'd really like to know, have any of the cards that died been founders edition cards? Is Nvidia's in-house build the best?
nvidia does in house gpus? afik evga makes their in house gpus, the coolers are made by cooler master
same for amd, in house model is made by sapphire
This is hard to watch, just because the current gpu market is nuts, and i just lost a 1660 playing new world, didn't think about the risk to my card because hey it's a 1660 and new world was killing the high end stuff. to be fair, after inspecting my bricked gpu i saw that the heatsink wasn't touching the vrms. so i imagine its a similar failure on my card to what you've shown.
I think when this fixed you should do a colab with gamers nexus get the news out there to possibly help others
Good job !
Many people say that the 3090 dying in new world is cause by a software issue.
I'm convinced that it's a hardware fault. 3090 push too hard on undersize components, or bad quality components. You prove it, thanks !
I think I am not all the way there… how can a 3090 with a powerlimit of 350-400W draw more than that limit? Or are the peaks the problem? I think they can’t be stopped by the software.
With my experience with one pascel card this was my guess. A hard transient thst blows the power limit way out of the watter before it gets corrected. And with these mixed phase/powestage count cards your probably playing the lotery that it doesn't happen on the single power stage part of the vrm
Is it mixed powerstage though? Under last video BZ corrected that reference PCB allows for 10 powerstage Vcore.
I don't know if this does mean that reference VRM is asymmetric or if 10 is correct number of phases for 3090.
To me - with this videos and what JayZ showed (whcih I don't entirely agree with him on - that it is VBIOS) New World acts like Furmark. On the other hand so many cards failed in other applications (heavy loads). Which probably means that there is a problem with at least QC in factories or design. But also - you shouldn't design your game to be a Furmark - this is insane. Furmark was DESIGNED to abuse GPU as much as possible - without any safeties.
I could be wrong about the mixed power stage. I did not see the update to the last video.
Although I woud argue this game is not like furmark. It may have a rapid intence operation but its very inconsistant. Also back when i was serious about daily stable ocs all my gpus passed a few hours of furmark. Mind you this was before gpu boost so you knew what you where getting in to and the card did not play arround trying to optomise and boost clocks when the workload let it.
Maybe the owner could run with the power slider at 90% for reliability. Not ideal but at least it may last a while that way. Or perhaps cap the frame rate in the driver.
it doesn't always work with new world unfortunately. some cards will still boost like crazy, or ignore the fps caps in the driver. This is a total mess between obviously botched game code revealing hardware and driver issues that should not be issues if the game was coded properly.
As Jay said, new world is basically more of a stress test than a game at this point
A vbios update will tone it down and it will probably survive right?
that is probably what will be done if you can't send the gpu back to get a replacement unit