I finally fixed the possessed PC! Here's what went wrong...
Vložit
- čas přidán 24. 04. 2023
- Remember the PC that was possessed? Well I finally fixed it! Here's what happened!
Sponsored Links
Check out the CableMod angled 12VHPWR adapters here - store.cablemod.com/12vhpwr-an...
Get your JayzTwoCents Merch Here! - www.jayztwocents.com
○○○○○○ Items featured in this video available at Amazon ○○○○○○
► Amazon US - bit.ly/1meybOF
► Amazon UK - amzn.to/Zx813L
► Amazon Canada - amzn.to/1tl6vc6
••• Follow me on your favorite Social Media! •••
Facebook: / jayztwocents
Twitter: / jayztwocents
Instagram: / jayztwocents
SUBSCRIBE! bit.ly/sub2JayzTwoCents - Věda a technologie
I had this exact same problem with the same graphics card a few months ago and it drove me crazy. The fix for me was to change the PCI slot speed from gen 4 to gen 3 in the bios. Another weird discovery I made is that it seemed to only be an issue when I was using the card in gen 4 with an nvme ssd installed. When I used an older 2.5 sata ssd the graphics card worked perfectly fine. Not sure if this will fix other people’s issues with this card but it worked on mine and I hope I it fixes someone else’s problem too.
I was going to suggest exactly this!
Please send me your specs. Branding/model.
could it be the lanes not beeing enough?
Honestly that would make sense considering it thinks it’s disconnecting from the system
I was just thinking this with how it was working with the 10 Gen Intel using PCIE 3.0 instead of 4.0 like all the others.
I spent 2 weeks recently trying to diagnose very similar random dxgi crashes on my system (Ryzen 7600X/Radeon 6750XT). After reinstalling windows, DDU, updating drivers and BIOS, rolling back drivers and BIOS, disabling XMP, undervolting, and disassembling the entire PC to ensure everything was seated correctly, literally EVRYTHING I COULD THINK of...I swapped the 8 pin power cables out and everything works fine. Sometimes owning a PC is stupid.
So technically, if the connection between the socket and the PCB isn't in good shape, that could cause similar problems.
Its as if it knows you are right on the edge of sanity, and all it took was this little fix. Yeah, I have been inches away from the insane asylums department for PC-builders many times...
@@CheapBastard1988 Possibly? That's beyond my skill/knowledge level TBH. I change da wire computer work now
I had similar issue with my 5700Xt. Lowering power by 3% solved the unstability. It didnt drop a single frame but crashes dissapeared.
"Sometimes owning a PC is stupid" is the most perfect thing I think I have ever heard my friend. Very well said lol.
Jacob was the best, man. He was basically the face for the queues and kept all of us gamers waiting on our emails with updates as often as possible. Bless that man for the work he did! And here's hope for a great future where he goes!
Who is Jacob? I don't remember
When you're comparing the GPU-Z specs of the "broken" card vs the replacement card, I noticed a difference in the Bus Interface data displayed between the two @6:24. The broken one shows PCIe x16 4.0 @ x16 2.0, whereas the replacement shows PCIe x16 4.0 @ x16 1.1 (I'm assuming that you're idling in both situations, but have different "resting" bus speeds). It just made me think, that if you try to run a game and check it in game, does the "broken" card ever get past 2.0? It should get to 4.0, but I thought I would bring this to your attention in case the card is having a hardware limiting issue (as in something wrong with the PCIe bus on the card) and that could be the potential fault. I could be wrong, but I'm only going by observation. People have reported 3000 series cards being stuck at 1.1 or 2.0 and either have horrible FPS, or game crashes.
Could potentially be a short somewhere within the card and it crashes when the shorted component attempts to run anything, seen something where a safety built into the card will allow it to run in a safe mode during idle but halt if something is truly wrong, the card going down to 16x 1.1 is normal idle while 16x 1.0/2.0 I believe is safemode
Makes perfect sense, it would bottleneck and crash if pushed beyond 2.0 bus speeds
Nicely done sir, good observation👍👍
I noticed that too! I’m don’t know enough to know the impact but everything being the same besides the physical graphics card makes me wonder why it’s reporting to gpu-z different.
Love how the channel has evolved, honestly been watching since around 2014, almost 10 years now. Good luck and keep it up!
I still would have updated/refreshed the BIOS on the broken card. I have done BIOS refreshes on cards that have presented as faulty before and brought them back. Glad you got it sorted.
Not sure if you saw this Jay, but At 6:27, when its showing the GPUZ screens, under Bus Interface, it does have different values. Not sure if that might be the cause - but maybe worth looking into.
That has something to do with the gpu being idle vs under load. Mine reads 1.1 at idle and 2.0 under load, and I think that's how it's supposed to be.
@@jon.wilson Fair enough. I don't play around with GPU-Z as much as I should, so I've never noticed that.
When high performance mode is off in nvidia control panel, the pcie slot will drop down to gen 1.0 to save power, and go back to 2.0, 3.0, or 4.0 when under load. In hwinfo it shows as GT/s ... 2.5gts (1.0 or 1.1) when idle, and 16gts when under load for 4.0 (8gts for 3.0 or 5gts for 2.0)
@@jon.wilson Nope if you are talking about a 3070 card, it should say what your gpu is working on that mother board.. if it is a gen 4.0 capable motherboard then it should say "PCIe x16 4.0 @ x16 4.0" if not then it should say "PCIe x16 4.0 @ x16 3.0". Assuming high performance mode is enabled on nvidia control panel
When you click on the question mark next to it then it starts a render test and it should show correct numbers; depending on the motherboard either 3.0 or 4.0.. that 1.1 or 2.0 or 3.0... means PCIe gen 1.1 or 2.0 or 3.0
I've noticed that too.. it might be a bad bus interface driver
If memory serves me, 10th gen is PCIe gen 3, where all the other affected systems are gen 4. Have you tested forcing gen 3 compatibility on the affected systems?
Edit: There's also the question of BAR causing issues. Either way, it would still be a defective card.
also no hybrid cores....
another commenter actually mentioned having the same issue and forcing gen 3 helped them. rebar doesnt give (at least) me personally issues on 10th gen
I think I have this problem with a Gigabyte RTX 3070 an a I7 6700K and a Z170 motherboard. :(
@robertjohansson3182 Do you have multiple NVMe SSD slots, or do you have any PCIe expansion cards in the middle slot of that board? Because then you may be running in PCIe x8 mode. Check the manual of your motherboard for the schematics of your specific board to see how they're exactly connected, but 6th gen only has 20 PCIe lanes. So 16 are for either the top slot or divided between the top and middle slot, and 4 lanes are connected to the motherboard chipset. If there is an expansion card in the middle slot, it will run your graphics card in an x8 configuration. And on 3rd gen PCIe, that's a low amount of bandwidth for a 3070. The same could happen on boards with multiple NVMe slots where at least one of them could be connected to those 16 lanes you need for your GPU. But that's very board specific, and you should check the manual if this is the case.
You could also try turning off ReBAR as it doesn't make any difference in performance on Nvidia GPU's. It does make a difference in performance on AMD GPUs, and it makes all the difference on Intel GPUs. But nothing noteworthy on Nvidia cards. If turning off ReBAR helps, you can leave it off or try a motherboard BIOS update if there is a fix for it.
This was my thought as well! All the others are PCIE4 or better, where the 10th Gen is PCIE3. I wonder if there is something off with that GPU and is very sensitive to PCIE4 (maybe the traces are poor or something) and thus dropping the PCIE4 down to PCIE3 helps with signalling?
I would love a (short?) follow up video where you teardown the bad card. Chances are there would be nothing visible, but who knows, maybe there is a cracked cap, slightly toasted VRM, bad/corroded pcie pin ...
Don't ya hate it when you know where the problem is but you don't know why the problem is there? Argh. But as we used to say in the programming halls... if you spend more than a few minutes on a problem, time for debug. 🙂 Which in this case is R&R the card 🙂
I suffered from that feeling not too terribly long ago, and I ended up finding myself re-installing Windows after hours of troubleshooting, and I'm tired, out of ideas
@@JesseGaming7593 it's painful but it's always nice once you've done it. Nice, snappy new system.
@Kholaslittlespot1 have you seen AtlasOS software? Linus tech tips did a video that showed it, its amazing! I'm still waiting on Windows 11 support, which the website says is "Coming Soon" , so I cannot wait
@@JesseGaming7593 just looking into it now, thanks. Looks cool!
@@JesseGaming7593 Try Linux next time ;P
I had this issue for about 3 months but it was with my 6800xt. Luckily for me I was able to fix it with a new driver update.
I love this content. I love that you found a viewers mystery, and worked through it and trouble shot it. A) helps me with my own trouble shooting B) feels like a nice connection with the community. I hope to see more content like this.
I had the exact same issue with my 3070FE, tried swapping systems, and a whole bunch of the stuff you guys tried. I was lucky that it happened after 9 months of owning the card so I just too the RMA option and was sent a new card that hasn't had any issues since (fingers crossed). I eventually assumed that I'd got a card that had only just scraped through the binning process or had some faulty SMD. It would have been good to know what nvidia found when they tested it though.
Had a similar issue with a friend's PC and we fixed it by plugging the GPU into the 8x slot instead. 16x sometimes locks up or crashes in games.
@@crashniels I could have tried that, except I have itx systems so had no x8 slot to try
@@AxR558 You can change how many lanes ya have in bios.
@@sedixmrboss5625 Well aware, just have no option to put it in a different physical slot to rule out PCIe slot being the root cause of the issue in the first place.
The two GPU-Z readouts weren't exactly the same though - the broken card was running at PCIe x16 2.0, whilst the replacement seems to have defaulted to x16 1.1.
Would be a surprise if that were somehow the root issue, but worth pointing out I feel
was thinking the same
That is not the issue that is just the load on the gpu if you run the load test the cards will perform the same.
Gpus lower their pci speed automatically depending on load at the time
My 3090ti will say pci x16 1.1 but as soon as i load a 3d app it goes to x16 4.0
@@nicekeyboardalan6972 Also if it would be the only program open? By that logic, shouldn't it show the same load?
@@dylan1234540 But if the broken card is stuck at x16 2.0 while idle, then something is wrong with it, because it should default to 1.0 for power savings at idle. That may be the problem with the card, that being that there is potentially a short somewhere that is causing it to run in "safemode".
Interesting note, from my long-ago days in manufacturing. Back in the early days of VGA we had to test multiple cards by a certain major manufacturer, with certain main boards that we designed and manufactured, in order to find one that performed properly in that assembly. The issues were related to bus timing and latency. The MB designers scoped everything and swore that our boards were within Intel's required bus specs, which the card manufacturer also claimed about their graphics cards. We had similar issues with another major manufacturer's Arcnet cards. This was in the 386DX/SX days.
love the kicks Jay !
Tear it down to see if you can pinpoint anything wrong with it. Maybe there's an SMD loose or missing, or maybe some other sign of damage/defect.
That 10th Gen System from the previous Video did not have PCIe4.0 and the GPU run at PCIe 3.0 speeds. Still does not explain why the new Card works and the old one didn't.
I was just thinking about this pc this morning! Glad to see the followup
I literally just finished the other video to come back and see this uploaded lol. Hooray! Closure!
This whole episode I was distracted by Jay's trainers, I need them!!
Another quality episode love being able to watch these troubleshooting vids
you can tell he tryin to flex on us.
Air Jordan 1 Zoom CMFT 2
No sh-t. Sick kicks.
His what? What do you mean by trainers?
@@Aremisalive Sneakers in Yank speak.
It's a big card. They might need an anti sag bracket. Check the memory ICs near the bottom of the board.
Part 1 definitely great video
I've been waiting for this video and been checking community post for this update
So glad you gave us a conclusion, even though the exact issue is still undefined. Would EVGA be kind enough to bench it to find the issue?
Us OCD people needed answers and conclusions! We needed them, Jay! Now, I can sleep....
I am so freaking happy you posted this video i have the exact same card with the same issue, now to look at replacement gpu's...
I could certainly imagine how a problem like this could occur. A slight manufacturing defect at only a certain point in the die causing pinpoint hotspot heating which can lead to component failure, but only certain patterns of use load that point in the circuit enough or frequently enough to push it over the edge. The management of temperature and heat dissipation within the silicon of modern chips is really wildly complex.
Thanks for sharing again your rabbit holes, not every tec channel is honest in this regard. Great stuff. Dont give up, try the bad card in every future build you do🤣 honestly half-kidding as i never give up and get mad about such unfixable issues and cant sleep anymore.... Dont recall, but did gen 3 and gen 4 make a difference?
I noticed that the PCI bus showed 2 different versions in the screenshots you captured. The new one was listed at 1.1, and the problematic one was at 2.0. Could this have possibly been the issue? As always, I love the content and amusement you guys bring to the channel.
Yea I had noticed this as well, and am curious if it was a slight revision to that card, but I'll be the first to admit I dont know much about the Bus Interface Versions and what it means, but that would be my hunch is that it was an experimental thing and discovered the issue and they revised it back to an older Interface. Just a thought of mine.
it doesn't matter .. if you are just on desktop and there isn't anything rendered that will be at 1.1 but if the board is needed, it will increase to maximum 4x in a game for ex.
An RTX card running at 16x 1.1 is basically the card on 'Idle' mode but only if it is on 16x 1.1, all Turing GPU's (including the 16 series) and newer have this feature, I believe if you get 16x 1.0/2.0 on these cards something has most likely shorted (rarely overheating) during post and the card is running in 'Safemode', it changes between 1.0 and 2.0 depending on how dangerous the short (or overheat) is to the entire card and if the short is severally bad (or the componentry on the card starts frying immediately) the card wont boot at all, in these instances where the card is able to boot and run the desktop it will appear fine since the desktop is super minimal but anytime the faulty component needs to work it falls apart and the card halts to save itself which causes the card disconnecting mid-session issue as it is literally freezing itself momentarily to stop drawing any power until it is safe to continue
@Celestin It does matter tho, pcie 1.1 is how cards throttle down, not 2.0. The broken card is at a higher speed than normal meaning it might also just be stuck there
@@Frozoken @Frozoken You could be right but you don't know what else is running in a system and because of that is x2. My card (a 3090) is either in 1.1 when nothing is processed or 4 where something minor is happening, like a 3d wallpaper. I tried for x2 or x3 with different scenarios but it seems to me that in my system is either on (4) or off (1.1)
You should buy those amazon usb microscope with articulated arm. I love it as I can zoom close in to the PCI,PCI-e slots and see
if any of the slots are bent or secretly dirty. Saved me tons on diagnostic time. The cool thing is they are cheap between $40-$120
Jay with the Js!!!
I love how CZcams makes computer building easy-peasy when in reality it's a hit-and-miss and mostly a miss. Yes, building your own PC can be fun and satisfying for some, but it can also be a nightmare for most.
Other than my first computer i have built all my own (10 complete builds or more, many upgrades), and never had an issue. You only get to hear the weird situations on tech channels, and it they make it so that you are aware that it may be an issue. It is not the norm.
@@michaelkaster5058 I've never had a PC build that didn't have some sort of a unique gremlin as every PC build is different.
Would love to know what was actually wrong with it - guess that's less likely to be figured out with EVGA not doing cards anymore.
Maybe Jay can convince EVGA to follow up finding and fixing the problem, heard they have one of the best repair centres for their products.
My daughter had the same problem with a Gigabyte 3070, so it doesn't look like an EVGA problem.
EVGA's repair crew hasn't closed just yet to my knowledge.
The problem is with pcie 4.0 when also using an nvme ssd
If you manually force the gpu to use pcie 3.0, the problem goes away
That is exactly why on 10th gen it worked perfectly fine as it does not support pcie 4.0
@@autoplanet4833 I'll try that thanks. I have a 3060ti, crashes only and only in BeamNG drive, and it's completely random. It can be fine for 7 hours straight, go to sleep, next day 3 crashes in a row after 10 minutes of gameplay. Thanks for the suggestion.
5:18 gosh I love those foreshadowings x)
Jay!!
Nice sneaks man!!!
Intel 10th Gen is PCIe Gen 3 at best. (You said the old rig was an Intel 10th gen)
Both the system where the bad GPU came from, and both of the test rigs that have crashed are capable of PCIe Gen 4 (or higher).
Something tells me the GPU is actually crashing/malfunctioning when it is operating in PCIe Gen4 speeds but the flaw doesn't surface when running at Gen3 speeds
I was actually having a very similar issue recently on my 3070 Ti. It actually ended up being a RAM issue. I ran the memory testing/checking software that comes on Windows, it said something was wrong, I swapped out the RAM, and everything cleared up.
i had the same issue. took me months to work it out. it was incompatible memory. I put in confirmed compatible memory and it hasn't crashed since. not even once.
@@TheRatlord74 I had a similar issue on my GTX 1070. I then replaced it with an RX 5700XT, for the problem to go away for a short while. It slowly got worse again until my ram completely shut up shop and died, so I replaced it and all my problems went away.
Fractal Design Meshify C... Love that case!
I've been waiting for this episode!! Pure hell of a troubleshooting issue!! :)
Krisfix Germany might be able to shed some light on this issue. Derbauer had a 4090 that got stuck in PCI-E 8x instead of 16x and he had to reball the GPU chip itself, I would think that this might be a similar issue with something going funky with the pins/pads underneath like a slightly cracked/oxidized soder joint or something wherever this api actually interfaces with the chip. Best guess though is it is the actual GPU die itself. Would be interesting to see resistance and voltage readouts from the defective card though even though i am sure they are all going to be normal.
Wouldn't surprise me if prolonged GPU sag is starting to affect solder joints or traces.
@@GenericPast mechanically it could cause some data pins to break or to have a weak connection which is why i hate that people support insane power requirements from card makers these days considering coolers will just get heavier and heavier
Exactly what I wrote, prior to seeing your comment ! I entirely agree. He should send the card to KrisFix Germany. Really hope that Jay sees your comment !
We never found out what happened to that card. We saw him apparently fix the card and it was sent back to derbauer, but it was still broken when derbauer came to test it. So either it wasn't properly fixed or somehow broke in transport.
@@LeJimster you are right i forgot about that. but reballing the GPU did work temporarily. I mean if you have to reball the chances are still low that it is going to be a fix. I remember watching something with kingpin where he talked about only being able to reball it a couple times before it just stopped working.
I wonder if he would have tried the card on windows 10 vs 11 if the problem would have persisted. Also I think the comment about changing the PCI version in the bios would also be a great troubleshooting step
Late reply but my daughter had the same issue with a 3070 from Gigabyte. We tried the card in her system running Windows 11 and mine running Windows 10. It showed the same problem on both versions of Windows. We were as stumped as Jay was with this issue. We tried everything conceivable. She eventually returned the card and got one from Asus and that has worked perfectly.
getting it done. that's a hell of situation
At 6:26 I noticed that there is a difference in the Bus Interface: old one is x16 2.0 while the replacement is x16 1.1; I don't now if maybe this can make such a huge difference to brake games but thats the only thing i noticed
Hopefully you can do a collab with GPU repair CZcamsrs who can take a look at this issue or at least can give you their insight about it. Looks like an issue on one of the components if not, its the GPU die itself.
Quite a few people, me included, pinpointed the issue in the comments of the last video: the orientation and physically handling it matters due to bad solder joints under one of the VRAM chips. It is a common issue caused by GPU sag with larger and heavier cards which are used unsupported like in this case. Running a MATS test should reveal the chip which is causing the provlem. I myself have had this same exact issue and it was repaired by swapping out the problematic chip with a new one
Then why would it work with certain other games?... Physical problems such as those I'm guessing would bring problems across everything you throw at the GPU. It even would barely function in windows, if at all.
@@3rdWorldGamer It depends on the load, memory allocation, temperature and the way the read and writes are done to VRAM. My card would also run just fine in some games and crash instantly in others. Underclocking the VRAM helped to alleviate the crashing but it did not solve it completely. Overclocking the memory made it crash more often and it even began artifacting in some titles (mainly 3dmark firestrike)
@@udatube and this you solved by replacing the chip? Interesting... I used to have RDR2 BSOD on me absolutely maxed out as in every single slider either on or at max under the Vulkan section and at 4k resolution... But only that game would cause a BSOD and turning only a few stuff to off such as tree tessellation and shadow resolutions to a tad below max resolved those BSODs while the card kept reaching pretty much the same temps.
Also no BSODs in any other games running also at absolute max.
@@3rdWorldGamer Yes the issue went away completely after I replaced the chip. The chip in question was the one next to the PCI-E slot on the right side of the GPU, which is typical for GPU sag damage (the PCB bends the most in that area). I confirmed the diagnosis with the leaked Nvidia MATS/MODS test tool which showed errors on this particular chip. The errors and crashing are now a thing of the past. But I do have to say that if you are experiencing BSOD's and crashing in one particular game only then I would not immediately suspect a VRAM issue. In my case there were no BSOD's by the way. The card crashed to desktop with various error messages just like with Jayz RTX 3070, and Windows was always able to recover without a reboot.
@@udatube yeah, I could rule out any problems related to the sag as my 3080 came with a support arm though. But what I still find interesting is the issue related to the difference in the instruction calls and ways the memory interact. Could still be something physical without having to be especially related to sagging.
The way I'm understanding it something like a mundane short in someplace very specific could also be causing my issue.
Very nice clean built btw...
It's either non-fatal static damage or a slight timing issue (which the former would also cause). A tiny timing issue would explain the difference in how it works. Static damage will spread through the damaged part in time. Cook the card for a week and it almost certainly get worse .
Hey Jay did you notice the bus interface was different at 6:31? One was ending in 2.0 and the other is 1.1.
Edit: not sure if this is important and showing something
If you'd press the "?" it will put a load on the GPU and it will show PCIE 4.0
i seen that as well
When gpus not on high performance mode in nvidia control panel, the pcie slot will go down to gen 1 to save power. On hwinfo it shows the GT\s 2.5gts when idle, 16gts when loaded (or 8gts for pcie 3.0)
Hey Jayz - I think your problem is peak power consumption browning out your 12V rail. Do you have an oscilloscope you can monitor the voltage with? If you can catch it dipping at the card connector, you'll know for sure. Otherwise, you could try supplying it with a massively overspecced power supply that will have huge output capacitors. And / or a better cable.
A card might be more prone to browning out the rail because of faulty capacitors on the card, or a dodgy connector. Anyway - if you get an oscilloscope and plug it into the 12V rail, and put the resulting squiggly lines on CZcams, you will feel super smart. And hundreds of engineers will tell you you're doing it wrong :)
thx for sharing!
What that sounds very much like is the SPECIFIC motherboard and the SPECIFIC card having problems passing data. Sort of sound like perhaps a soft memory error on the GPU that normally gets trapped and ignored / smoothed over but instead gets the mobo upset and triggers a crash. It's like a weird combination of each part being a little too sensitive about things and just giving up easily.
When you did the head to head showing of the bios etc i noticed that the bus on the old card was x16 2.0 and on the new card it read as x16 1.1 Could that have something to do with it crashing? (6:26 for comparison)
nope. If you'd press the "?" it will put a load on the GPU and it will show PCIE 4.0
@@jedenzet Correct, Modern GPU's slow down PCIe speeds to conserve power.
GPU-Z has a built in "benchmark" to max out the PCIe bus, hidden under the question mark.
Those cablemod adaptors get so hot and the male connector can wiggle. I know that's described in the QR code when you receive it, but I won't use the one I purchased.
What a great idea from cablemod....
6:25 shows the comparison of the two GPU stats. One difference noticed in the two cards is the bus interface. This actually could cause an issue. This may be due to some of the PCI pins being finicky. An issue with a pin could create a voltage surges or pulling too much power. It could be an error in a device register on the card as well.
Someone brought me their system many years ago that would say the graphics card needs to be replaced. They bought a new card only to find out they would get the same error. We are talking about Win 98. I look into it and found it was the power supply, replaced and no more trouble. Not saying this is your case but having low power can cause some real odd effects to a system. The original PS wen't bad after 4 years.
I've also thought it might be the PSU
@@LOW3beats well it crashed on 3 different systems now, so probably not PSU
I am surprised that when EVGA announced they were going to stop making graphics cards we didn't hear anything about AMD sending a rep to EVGA to get them onboard as an aib. I would have lovet to have seen what EVGA could do with modern or next gen Navi cards.
I'm sure AMD did just that but since nothing came of it there was nothing to announce. It may happen eventually but I'm sure Nvidia left a bad taste in the mouth of the EVGA CEO so he didn't want to jump into another GPU market.
Sapphire is kinda the EVGA of AMD GPUs. But yes, I'd love to have EVGA on AMDs side too.
I had issues similar to this that was eventually solved by changing the power plug spot on my modular psu. I had run two cables but inadvertently plugged them into the same 12v rail on the poorly labeled psu end, and only certain games would produce just the right sort of power demand surge to mess up the 12 regulation on that channel enough to crash the gpu but not the system.
“Linus give me money” 🤣😂😂 love watching the videos keep up the great work!
I noticed some slight number differences on the GPUz info ( I think it was the PCI bus version or something like that…one of them was a 16.1.1 the other was a 16.2.0…that’s the only discrepancy I noticed from the screenshot )
That's just a power saving thing I think. As long as they're both reading x16, it shouldn't be an issue.
It's most likely a power draw issue. I had the same issue until I realized that I had forgotten to plug the second PCI-E cable into the power supply. Most likely, the OCP is making the GPU shut off momentarily, causing the GPU to become "disconnected". Hope this helps.
Nah if it was OCP the entire Pc would shut off not error message.
@@CreamyI3eaver Not if it’s separated Vrails. I’ve had this problem IRL when I had only one plug connected into my GPU. It may be that Jay’s GPU is ramping up power demand too quickly (quick increase in voltage = high current through a capacitive load) and triggering OCP on the PSU’s Vrail for that cable.
@@joshuatyler4657 Except pretty much all power supplies now a day are single rail.
@@CreamyI3eaver All PUS have at least three voltage rails: 3.3V, 5V, and 12V. My PSU (EVGA G3 100W) has two separate 12V rails, one for the motherboard and CPU, and another for PCI-E. This way, the change in load between GPU tasks and CPU tasks doesn't stress a single rail simultaneously. This is an important safety feature when pairing a 300W CPU with a 350W GPU.
@@joshuatyler4657 I meant 12v rails but also Jay has probably installed well over 100+ GPU's you really think he plugged it in improperly multiple times?
Thanks!
6:24 the one difference I noticed is that the old broken card bus interface said "PCIe x16 4.0 @ x16 2.0".
Where as the replacement bus interface said "PCIe x16 4.0 @ x16 1.1".
So perhaps the issue with the old broken card was the PCIe bus interface was running too fast for it and it needed to be slowed down from 2.0 to 1.1 for it to function properly.
Hey Jay! I had a similar issue with my 1080Ti. The problem was related to the phase of power in my house. Turns out it was some sort of dirty power. Put the system on a UPS and it solved the issue.
6:24 the Bus Interface is slightly different. Both are x16 Gen 4, but one has 2.0 and the other has 1.1 at the end. No idea what that means, but it's the only difference between the two cards we can see on GPU-Z. Someone who knows tech, please explain, I'm curious!
If you'd press the "?" it will put a load on the GPU and it will show PCIE 4.0
I think alot of weird unheard of issues are coming to light now with the LHR cards, my Gigabyte 3080 Vision OC LHR doesn't work ATALL in my PCIe x16 slot yet works perfectly in my x4 slot. Yet it works perfectly fine in every other PC! The pain.
Love you man
I never stopped and thought about how nVidia sold cards to miners and then made LHR versions to combat them.
Creating the problem and the “solution” 101
i believe the miners hacked the firmware for hashrates anyway.
@@zaka8315 Well that’s even fuggin worse. Let me edit some quotes around “solution”
@@ChrisWijtmans Such a waste.
Jay: "Hey, guys! I fixed the card."
Everyone: "HOW?!"
Jay: "I got a new one!"
8:40
This may not seem like a big issue, but I have always had boot issues with my Ryzen 9 7950x. Hex codes 9C, 15, and just extremely long boot times in general. I narrowed down the post issues to the PC was trying to boot from my mouse and keyboard or really any USB I had plugged in so I just disabled all USB ports until OS in the bios. The only issue left was the long boot times and this just gave me clarity and relief.
Had the same issue with my gtx1080Ti two years ago. For me two things worked: limiting the power consumption with msi afterburner and (later on) replacing the PSU (might have been a „faulty“ cable or the whole PSU). So far I have seen SO MANY different fixes for it. But no go to fix.
I hate those situations, too. When you realize where the problem is, but can't find or figure out _why_ it wants to be a problem.
My daughter had this same issue with a Gigabyte 3070. Like Jay, I spent hours scratching my head trying to figure out why.
FINALLY!!!! I have been waiting for this video for so long because I have similar issues 😭
Sounds like you need to replace your GPU :/ Atleast you know now
@Tzunshun well the good news is I still have it under warranty. I'm still messing with mine since I believe it's just a discord setting after further research at this point. Rocket league and minecraft are the only ones that'll instantly crash or crash after a minute consistently, but it almost always only happens when I screen share on discord
Me too, but i have shappire nitro + 6700 xt.
jay your shoe game is on point today!
I have a EVGA 780ti classified that was supposed to go in my friends system as a very cheap upgrade. It will not post on that system (2600x on an Tuff b450).
Runs totally fine on my old Phenom II x6 system.
Nice kicks Jay. 👌🏽
Two lines down from the card bios version was the bus interface info. It's slightly different between the two cards.
Love these kind of videos Jay!
I used to have that same issue multiple times with my 1st gen ryzen 1500x and 1050ti. I never figured out what it was. Swapped for a 1660 and the issue went away. Moved the 1050ti to my wifes FX 6300 and its worked flawlessly ever since. 6+ years on it currently
I had my FTW3 replaced a few weeks after their announcement. Got a nice new replacement as well. Such a shame EVGA stopped making GPU's.
Pretty cool of you to help this person out 👍
Didn't he say he RMA'd the GPU before already? Do you think he got two bad cards with the same issue and why does the new one work? Is the new one a different model? Thanks for the videos Mr. Centz
yooo....fresh kicks Jay :D
digging the kicks :D
Had an issue like this once long ago. The first GPU I ever owned that needed a separate 8-pin. Lo and behold, either the 8-pin cord, or the receiver on the card was borked. Never did find out which it was for sure, as the card was well within warranty. RMA, and shockingly, got it back in about a week (I did pay for expedited, but not overnight, shipping).
Obviosly, I suggest checking if those API calls draw more power, and if said spike is hitting an electrical bottleneck.
A friend of mine had a very similar situation with a 3060 (Gigabyte I think), first we RMA's the new mb because this care worked fine in the previous system, then we swapped RAM (Still crashed) but the crashing took many hour to happen like 12 to 36 hours ! Finally after many Driver updates it suddenly stopped crashing !!! I felt so bad because I have built so many PC's but this was a build that I recommended to a gaming friend using a 5800X3D which lived too far to travel to, to help resolve this in person.
6:27 Only Bus Interface is different. Broken GPU: "PCIe x16 4.0 @ x16 2.0" Replacement: "PCIe x16 4.0 @ x16 1.1"
I got a 3070 Ti ,new in box with factory plastic wrap on the box . So I have the last one now J . Hehehe was a gift biuld I never got to build.
Problems like this can be heat related. The usage increases the heat which causes a solder joint to fail. A connector to the pci slot would be a perfect place for a joint like this. Sometimes called a 'cold solder joint'. If you had a way to run a heat gun or cold air this could help diag.
Ever tried changing voltage or clock settings? Every game has a different load and every GPU is binned differently. Sometimes factory OC is enough to let a game crash.
Just because the BIOS is the same revision doesn't mean that the BIOS hasn't been corrupted slightly. I admit it's a long shot (especially if the card has dual BIOS and switching BIOS didn't work already), but I think it's worth trying to reflash the BIOS.
Only thing I could think of as well
@natemauger9757 A different comment that describes how a bad cable caused similar problems for them makes me think it's more likely that there's something wrong with the power delivery of the card. But my OP remains valid as it's just better to make sure that the BIOS is good.
Good vid the kind of issues people have
Maybe this is something similar to the EVGA 3090 cards that had a bad component on them that the one MMO was causing to fail.
I was studying Computer science, and at the computer lab, there was a desktop computer that had some capacitors issues. When you turned on it shut down, I found a way to turn it on and unplug it immediately. Afterward you turned the computer it worked fine.
Wow, I had the exact same problem with the original GTX Titan card. I could run any burn in program with no problems. Most games worked with no issue. And then some had this problem, and it happened even when there was almost no load on the card. The store didn't want to RMA since "it worked for them when they tested it". Took me a long time to get it swapped (and it took me 1.5 months to be ALMOST certain that it was the card).
Used to have a GTX970 FTW ACX 2.0 from EVGA, I had a issue like this as well.
I'd play in example Black Desert, everything at high/ultra and no issues at all, smooth fps.
I'd launch Rocket League and idk what other game, and I'd suffer that same problem, constantly.
I feel for the guys with teh same issue trying to get a replacement. You just know its going to be a pain.
finally get an update, woots!!
I had very similar issues with a Strix 2070 Super and it was incredibly frustrating to trouble shoot. I stopped playing Microsoft Flight Simulator 2020 because of the issues I was having with it crashing to desktop, after the second or third time I had to reinstall Windows. All of my other games were working perfectly and the research I did led me down the path of memory issues or problems with CPU overclocking or many other things that were not actually at fault, I am thinking it was definitely the GPU to blame.
@jay Can you take the heatsink off the 3070 and take a series of close up shots of the board? If there is a broken trace or a cracked PCB, that may cause some of these issues.
Worst case, it is a cold solder joint under one of the chips, which would suck.