Watch what this man has discovered about AMD 6900XT after running stress test
ฝัง
- เผยแพร่เมื่อ 18 พ.ย. 2023
- This video will get many dislikes due to shocking revelation at the end of this video.
For repair, please contact me by following the link in the channel ABOUT page.
Buy me a candy at paypal.me/tonynameless
Tools, schematics, boardview files etc are available here
drive.google.com/drive/folder... - แนวปฏิบัติและการใช้ชีวิต
i like how at this point he is just pulling core and re-installing them like its nothing. he even does time lapse of this.
For us its mindblowing, for him it is nothing special, just daily work.
yea i noticed it too lol !, like "well lets try first a core swap just in case" :p
I used to work in those PCB assembly house. It's GPU reball & reflow process, but I'm amazed he has 1of those machine, & more importantly the skill to reball & reflow that kind of size GPU.
The more I learn, the more I realize I don't know. I may have to stop learning before I know nothing. Thanks for the Video-Man!
Haha, fun fact, but 1000% true. The deeper you dig, the more clueless you become :)
I had this, buying an oscilloscope helped a little.
Back in 1979, our school's computer room had a poster of an Orangutan. The text read: "The more I study, the more I know. The more I know, the more I forget. The more I forget, the less I know, so why study"?
all scientists and professors that keep studying, keep digging, reach this point... none of the top ones are Athiest lol
This guy is bad. Don't "learn" too much from him
It's about the many patients you were able to save. Not about the few where it didn't work. Absolutely brilliant work.❤
And here i was terrified re-doing the thermal past on my gtx 1080. This guy is ripping through it like a 80s action hero disarming a bomb!
Back in the late 1990s and early 2000s hardware generally shipped with healthy margins (some exceptions existed of course like the Pentium III 1.13 GHz which had to be recalled). Nowadays all products seem to be pushed to within a percentage point of their performance limits in their stock configurations, pushed within degrees of the danger zone in thermal terms, and pushed within 10s of mV of 'absolute maximum ratings' under controller transients. They are marvels of optimization but they are also basically what we used to call 'heavily overclocked' products right from the factory. Oh, and the board partners are squeezed HARD on margins so they are pressed to shave pennies off the reference design so good luck everybody.
Race to the bottom...
It´s even worse, considering the massive neglect on AMD side with the drivers. While i was using 6600XT, some drivers were able to push the GPU to 160W power draw, despite GPU itself being locked to 135W. I also got artifacts once, when i removed and later inserted back the GPU into the same slot - drivers went haywire for some reason and had to be reinstalled. Latest "quirk" was mix of green/black screen, which turned out being yet another defective driver.
Whole thing runs HOT, especially the power stages (75°C in load all the time).
Meanwhile, if i install ANY nvidia GPU into the same system, it´s all smooth sailing.
Recently, i had to replace Ryzen 3400G in parent´s PC. iGPU died for absolutely no reason.
With Nvidia, you´re mostly buying fully supported product. With AMD, you´re buying a PROMISE, that maybe a year or two later, the drivers will be stable enough to make the GPU usable. Despite the cultists (they don´t deserve to be called a "community") claims about ’fine wine’, this is the undesirable outcome. If i buy a GPU, it should work from day 1. Wine gets better over time, because it´s properly stored, aka nobody is using it.
AMD keeps doing this for years, obviously not caring about its customer´ss experience the single bit. And it´s one of the reasons, why Nvidia keeps laughing and spitting into customer´s faces all this time.
It seems for me this specific AMD model is already too much tweaked on stock settings. I've seen in the Internet people having similar problems with this card, even with the new one. The workaround is to lower vcore, limit refresh rate and FPS which only means this card is already overstressing on full load which makes the GPU malfunctioning. I'm only wondering if BIOS firmware update could fix those artifact issues. Maybe some options need to be adjusted so that this card can work without any problems. If so, then this specific model has either manufacturing flaws or bad stock settings.
This explains why I always end up with better performance on my AMD hardware after slightly reducing voltage. I do this on my GPU's and CPU's from AMD. I get cooler temps, more consistent performance, and I would like to believe I get a longer lifespan. Undervolting is the new overclocking.
@@Morpheus-pt3wq That's a windows issue. It has done it with my RTX 2080 ti's in S.L.I. I went to enable S.L.I & got a black screen. I thought both my cards died turns out windows just decided to update the drivers after I pulled one card out & started windows then re-slotted the second one after shutting down & restarting.
When I received my Reference 6800, the temperature delta was around 20-23c, which seemed rather high for the power usage. So I used a special phase changing thermal pad, PTM 7950, to close the gap. Afterwards, it sat closer to 10-15c delta, allowing my clocks to stay more consistent. Did the same thing with my 7900XT reference, and yielded similar results.
Many of the cards are failing due to the cheaper thermal interface on the core. I always change my thermal compound with any card I get to improve their lifespans. Just because AMD or Nvidia states the cards can handle a certain threshold, does not mean we should test those limits. Keeping the card cooler, longer, will help you get the most from it. Just an FYI for anyone curious.
For those of us that will have to get a tech to do this, is there anything else you can suggest to extend the life of the gpu?
@@ZinfandelRed1914 Downclocking and undervolting. Makes cards much colder, with minimal performance losses.
@@alexturnbackthearmy1907 Depends on the card though. But for most yes.
Man, reballing cores has become routine for you, like me, checking my oil in car.
I am really happy for the growth you got on this channel!! I remember when you were under 1000 subs. Now, you are killing it. Congrats!!
I wish 😁
view count is low given the number of subscribers
@@northwestrepair I am not surprised, as a GPU repair video is something to view occasionally and not content that most want to regularly consume.
I love your videos and i am watching them in one of two scenarios. One is when i've checked on updates from what i regularly watch, and the latter is in the evening. You have a very calm and relaxed voice. I used to watch Rossman daily, but he has gone from fixing stuff to being all chatty about things that's wrong. It's nice that people care, but that's not why i subscribed to him. I also love your video format, where you speed up trough reballing, but still show it. I instantly subsrcribed to your channel, and understanding how YT works, i like every one of them. You've spent time recording it, the least i can do is like it. Oh, and you don't nag about like and subscribe in the beginning of every video like some. 🤣
Best wishes for you and your canal, from Norway.
i'm watching cause its foon to delve into modern tech and control your spendings according to quality of the components..even that is irrelevant at some point consider how owners treat their cards..it is art to watch you reballing the core and listen music that can soothe your mind..priceless💪
@@Grievous- Noo, I´m loving his Video´s! I´m looked ALL of his Old/er Vid´s in 3 Day´s and waiting every Day for a New One...Super it is...
I really enjoy watching your videos. It's remarkable how much knowledge you have and how talented you are. I hope you keep posting. If my 1070 ever bites the dust, I know the first place I'm going to see about paying someone to rescue it. Thanks for the videos and helping us all to learn.
WTF, I was waiting for "But wait" and then a successful conclusion. Guess you cant win them all...
Nah, once they start artifacting like this there's nothing you can do but RMA. Almost aways a dead core issue.
@@exoticspeedefy7916 RMA suggests a warranty, which would've been void the moment it was opened up. but the point is the video title and then what we actually discover, leaves something to be desired. not that there might be something else going on with the card or it's just a dead core because of persistent artifacts, but how the video is presented as "Watch what this man has discovered about AMD 6900XT" and then it's just not that interesting or deep or finding a root cause, just discovering something that points to the conclusion that it's a dead core. just kind of click bait title but it's still interesting to watch even if i didn't know what we were discovering or why or how
@@ivanjakanovyou cannot void warranty in US and most countries now by just opening the card
Him soldering wires onto the card though probably instantly voids warranty though.
I really like the design of this card, simple yet aesthetic. Great content as always!!
Isn’t it dangerous that the core is like on the outside of the card
I'm surprised you didn't just turn down the clock speed on the core to see if that helped, in the past i get those artifacts when pushing the chip to hard, probably just has some degredation to the core, a slight undervolt and downclock might of made it useable ?
I was thinking this. Undervolt/underclock and posssibly a reflash of the gpu bios.
Slow objects not cause the deformation but fast ones. Could wrong addressing in VRAM cause this? I believe there is pixel draw test or something so you can see which values returns most error
a vram test was imposible? how do you know is the core without testing vram? i ask because i want to know
Usually, I test the cards before working on them. I run the benchmarks and I run a game. This is to know how the situation was before me working on a card compared with the situation after I was working on that card. A card that gives flashes means the main chip is damaged.
@@Dandan-tg6tj thanks
if you're sensitive to disgusting noises *DO NOT WATCH WITH HEADPHONES!*
RIP my GPU :C
T_T
does it run on lower fps ?
Haven't got to test it yet, gotta get it shipped back first lol@@DeepakKumar-lv4te
have you tried limiting your fps to like 80 and undervolting gpu?
Have a question… is the core visible on the outside of the card and isn’t that dangerous?
Out of curiosity, was the memory not a potential issue? I know that can sometimes be a source of artifacting
I had defective VRAM once, it didnt look like this, more like watching a video file that had some data corrupted. Like a digital satellite TV signal in bad weather.
The artefacts we see here are clearly the GPU itself having issues with certain kinds of transparent effects mostly, which points at the GPU not working properly.
You can clearly see that the GPU is doing that to certain effects, is not like everywhere and at all time
I thought this kind of artifacting was a RAM problem?
Graphics memory is more likely cause. With some more weeks of use, the artifact may start to look more like dead vram. Or statical electricity.
I thought it was the memory, in the past when i overclocked the memory a little to high it does this sort of artifacting.
You are an artist!! Thank you for the video!!
this problem looks like driving the memory speed to hard with a too weak power circuit or insufficient shielding of the data lines. the same artefacts occur also on cpu build in graphics if you stress the memory controller too hard. this mostly happen with ram that running over specs (overclocking) or running ram thats clocking higher as the motherboard is designed for. the signal integrity is influenced by noise of the power stages and nearby signal and power lines. this why overclocked ram is stable in idle but under cpu load it collapses. you may noticing something weird if you go to overboard with ram speed but for the most part it happen under load. the same happen if your ram voltage is to low. some motherboards only allow Low power ram and normal ram works some times but the signal is not strong and may collapse under load. the same can be applied for gpus but if this happen something is wrong with the power stage for the gpu ram or your firmware is modified or corrupted or the data line between gpu core and gpu ram is damaged or the one of the ram is damaged or what i said before you have overclocked the gpu ram.
one of those things are the problem. but power seams fine, mainbord is undameged, cpu core works at first glace so may the memory controller is damaged or one of the ram chips.
i would say it might be fixable but is it worth? probably not. for the fact you say it happens on decoding workload so i think something is not right in the gpu core in that area. you need tho look up a die diagram who is what located and this might tell us what i expecting. it is near the memory controller so yea the gpu core is the problem maybe a crack forming.
Ya I was going to say that I saw another video with artifacting fixed by replacing one of the vrams
How in the world did you learn about this stuff. I thought I was great with computers just from growing up using them, but this is on another level of technicality.
How do you get to Carnegie Hall? “Practice”
@@ToniHiltunen1980that reference might go over some heads due to the generations. But it's nice to see it still being used!
Engineering Degrees are the main way.
Hardware certifications + OTJ learning is another.
Going nothing but what is called 'basement experience' is the longest road, and you might not ever get hired anywhere based on that alone (with no schooling at all) but if you have 30+ years of studying manuals, tearing down and repairing things over that time, etc then you might be able to do a repair shop startup someday on your own with that knowledge.
@@Munenushi Engineering degrees won't teach this.
so which brand has the best built?
Are there many higher models from RDNA 2 that have such problems with the core?
check your title brother
"Watch what this has man discovered about AMD 6900XT after running stress test"
I see problem no here what mean you?😂
i also see so problem. Don't get confuse with the thumbnail 😂
I thought thumbnail was in the section descriptive? 😂
@@666Daheretic he changed it ..
@@TechoWeeds he changed it
Is it possible that something is missing from the board. Check the time 4.02 chip lebeled 470 and there was one below of that?
Sad to see such a powerful card dead, thx for sharing lost cases too.
This makes me wonder how much alteration was done by Sapphire with their Nitro+ SE version of the 6900XT (this is the card that I run). I wonder if they changed the phase to a true 16 phase instead of an 8 Phase + Doubler?
The second i saw the flickering i thought it was one of the memory modules
is it possible could be driver issue? I sometimes also faced this artifact with the driver version Nvidia
Can' be a faulty memory chip?
I'm sure he did a mem test and just didn't show it in the vid, not like we have seen it done a bunch of times...
its amd card not nvidia, im a technician and its pain in the ass
No reballing music today? 😢
no music is good
I've heard that this Infineon dongle costs a leg and an arm while it looks like a cheap FTDI based USB to I2C adapter which in many other DIY sites costs around 15USD. But I suppose it's the software that comes along which really matters here, right ? As for the issue in this card, it's the first time I see you dim it dead so fast without checking anything else. Could it be an unstable voltage somewhere else (like a linear PSU) or a mosfet ?
its like 71$ original
That 3-pin extension cable you made , what is the size/number of the connector?
That looks to be about the right size and pin count I need to make an extension cable for a project I'm working on.
Thank you. I hope this holiday season goes well for you.
It looks like a typical 0.1 inch (2.54mm) pin header. Try searching for "DuPont connector" or "jumper wire".
If this is an extension cable you plan to unplug and replug a lot or subject to mechanical stress, you might want something more durable, e.g. a 3.5mm audio jack, the RJ45 connector used in ethernet cables, or even a USB type-A connector.
@@wereoctopus Thank you, and I appreciate the advice.
i've been having my poor 1060 throw random grid-patter black squares on my screen under peaks in load, but not continous load... is that just a driver issue or is it dying? (im sorry idk where else to ask this)
WOW! You're badass! You're only the second person I have seen to be able to remove the cores from cards like that...That has to be a PITA!
Hey, sorry if this is random to ask here. But i find that in many titles paired with my i5 10600k on a z490 a-pro my red devil 6900 xt tends to always be 50-110W for power consumption and 30-40 utilization and giving me low frames on titles that it shouldnt.. on passmark dx9 and dx12 it gives 60 fps but on dx 11 it gives figures comparable to other models "289". I'm not sure whats going on, i then try super position and in 1080; extreme will see 5900 ish score so more or less normal on some bench marks, bad on some others and not reflecting on new AAA gam titles by staying low power limit and low clock/util. any ideas why this is?
How about backing up current firmware, comparing it to default and flashing default/new one?
hey can you help, my gpu sometimes displaying to monitor sometimes it doesnt eventhough fan spinning normally. works usually after dozen times of restarting, and it works smooth as a normal gpu would. tried to stress it and no issue came up, playing games and reaching high usage for a period of time also no issue
but if i shutdown the pc, or restarting then the issue came up and to make it work I should keep to restarting until it works normally.
Dead, or unstable at current clock? (actually want to know this) I had a card that artifacted like that but a 50 mhz or 100 mhz underclock and it ran fine.
howd uk it was core and not vram issue ?
What makes you think it can't be one of the gddr6 memory modules with a solder dry joint?
Bad memory contacts can cause artifacts also. Ever tried to put liquid flux under them and do a quick reflow test?
Bad memory artifacting looks different compared to this; I've seen some and they don't look anything like what was shown in this video. It's hard to describe so I'd suggest you look it up.
@@hasnihossainsami8375 I know what your talking about, a more consistent and static pattern. However I was just saying it may be a misbehaving module also.
I did do reballing and reflows back in the day with XBOX and PS3 consoles, never on GPU's however.
I bet he won't answer, but I would like some advise on what to buy. AMD or Nvidia? Which video card is the best built? Thinking of a rx 6700 xt or Nvidia 3060 Zotac.. Great job by the way. Im not smart enough to fix video cards but enjoy watching you do it.,
I can see RX 6XXXX Card which is hard to repair due to luck of experience. I have RX6800 with Failere, everything is good, but failing under the load, all fases shows right signal. Only one test which is tserver boardtest shows EDC errors on Ch2, but how to locate bad memory or bad chnnel on card? Have you any experience how to find out memory chip or connection under the chip? 8 chips Smasung, EDC shows channels from 0 to 15.
Appreciate the content! Coming from a motorcycle technician its highly fascinating to just sit on lunch or at home and watch you work… Girlfriend thinks I’m crazy for it 😂
hmm I wonder if the artifacting goes away if under volted or under clocked, at least the card would be usable
possibly but i dont call it a fix
@@northwestrepair does it still do it with lower fps ?
@@DeepakKumar-lv4te yep
these amd reference coolers are a joke avoid them on second market go for sapphire powercolor etc
Do we know what caused the dead core though?
Hi there Tony!,
I've seen in the Internet people having a similar problem with that specific model. Some of them are having problems even with new models.
The solution or rather a workaround is to adjust core voltage, refresh rate or limit FPS.
Anyways, It seems for me rather a manufacturing flaw. Limiting the card does the trick but still you don't use a full potential of the card while the card should work on every scenario.
Btw, have you checked this card on different drivers?
Lowering core voltage is not a solution
@@northwestrepair yup true. So the solution from Internet is just a trick so that you can still use this card until it finally dies
@@northwestrepair I have a crazy theory. If the problem is quite common, can it be that maybe stock settings for this model are just wrong making this card overstressing on full load and causing the core malfunctioning?
Perhaps the voltage is incorrectly set when more computing power is needed. Maybe it is just the GPU BIOS but I don't know if GPU BIOS can control the voltage.
I'm wondering if BIOS firmware update could do anything to get rid of that artifacts.
So if my guess is true it can be just a manufacturing flaw.
@@northwestrepair I forgot to mention that perhaps this specific AMD model is just too much tweaked on stock settings making this card overstressing and dying eventually. Nowadays gamers demand more and more power, particular companies make use of this trend and push their products to the limits to attract customers.
newer cards have been literally pushing FPS until it kills itself... always try to limit FPS with their manufacturer's software (eg. Radeon Adrenalin, etc) even just to 240, 120 is better. This also helps with heat (first thing you do with any gpu = set fan control curve (eg. 70deg = 100% fan, if not set already))
when you re-ball is it leaded or unleaded?
can you get new GPU Chips to replace the old dead one?
Yes.
Yes but cost is too high.
I don't service GPU's I service laptops and PC overall plus printers and that's not economically viable. Can be done but when chip costs 2/3rds or just the price of a new one then why bother? It's exactly the same as putting brand new printhead to a Canon/Brother/Epson printer which costs almost the price of new printer - yeah there are cheap printheads/refurbs on ali but they usually don't work I tried that way..
this could be an amd driver issue. or another software problem.
While it’s unfortunate this occurred, I think it’s important to remember that there are flaws in any product mass produced. I’m not a fanboy, and I can only speak from a test group of 1, but my xfx Merc 319 AMD Radeon 6900 XT Limited Black has been running great since day 1, and sits overclocked at 2650 max 2550 min clock at all times. I do however run an undervolt as well. My old 5700 XT is still going strong in my living room PC. I upped the fan curve and power limit, and left everything else the same. My next GPU purchase decision will be based on frames per dollar, just as the last two have. I don’t care who makes it, I care about value. Just like cars, different revisions, even with the same name, can be good or bad, this is life. Don’t let yourself get drawn into the marketing. If the green team made a card with the same fps as the red for the same price, that’s the point I’d look at other competing features. Also a little research into your apps and games might reveal whether those features will ultimately be worth the cost or go under utilized. At the time ray tracing was not important or wide spread, so it wasn’t very heavy in the final decision. Best wishes and greatvideo. To all the team red and team green people: You don’t need to be loyal to a brand, however it never hurts investigating all your options before making an informed decision. While your opinion is valid or at least is a reflection of your experience/needs, others may have had different experiences and or needs. Simply chill and let others be chill.
Thanks for the feedback about this it really does help and we need a choice as without AMD Nvidia will charge us $45,000 for a mid range GPU if they had things there way.
I was thinking of getting the 7900XTX and it would have been a nice match for my 7950x but a 3060 fallen into my lap for next to nothing.
i have xfx swft rx6900xt my temps is around 70c or 69 with undervolt and overclock but my junction temp can go up to 95c dont you think high junction temp is the problem and can lead to something like this gpu that was shown in the video
"There are flaws in any product" "I'm not a fanboy" "My AMD Cards run great and don't have any flaws"
Yeah typical fanboy arguments...
But hey I could say the same thing about Nvidia, all my NV Cards run great. My 4090FE runs with 580W OC without problem or with UV and below 300W.
"If the green team made a card with the same fps as the red for the same price"
It depends on the view. Ever tried to use pathtracing? A 7900XTX runs as great as a 4060 with PT. In that case a 7900XTX is a really bad value...
Since RT/PT is future tech you could even argument, that AMD is HINDERING game developement, since developers have to consider AMD users...
And that is why I don't go down that spiral.
You are ALLOWED to buy any card you want, be it red or green OR blue. But stop defending any of them, all of them make money of you.
Bought a Sapphire Nitro+ 7900 XTX back in July that I ended up sending back (Zero problems with heat. Temps were actually STELLAR). I tried for 3 weeks to get that thing to run right. No matter what I did I had crap FPS on a massive amount of older games and just in firefox. The brand new new games were mostly fine. After trying everything.. setting minimums on the core.. setting min/max's.. messing with the power slider.. messing with different ram kits. Different power supply.. 800w Corsair and 1000w EVGA. I even stuck it in another computer to see the same problems. I know two other friends that also had their share of a TON of problems with their 7900 XTX's. One of them stuck it out and turns out its the terrible drivers. Some of the older games we played have gotten better for him but he still has a ton of problems. Not worth the headache for something that costs so much. Honestly if it doesn't just plug in and work... and you have to tinker with all this crap just to get something usable.. That seems like trash to me. I'm still waiting to buy another GPU and I'm glad I did. Will most likely pick up a 4070 Ti Super as dumb as the name is.
@@gucky4717 No, developers have to consider users having low-midrange products from multiple generations from either brand. That's why there's this thing called an options menu. It's far more likely users will have 4060's or below than flagship models and they're working on large projects for years ahead of time. Saying cards like the 7900XTX is hindering game development is absurd on so many levels. And no, I'm not a Radeon user as if that would matter anyway.
The problems with 6000 series amd refrence cards, 90% of the time are memory chips, due to poor heat dissipation or gpu sag, had one with the same issues and it was the C1 memory chip (if I recall correctly)
If i remember, that light into the game is from bad memory on the GPU, but you should work on both GPU processor and memory. Try to lower frequency of GPU processor and frequency of memory, and make fans to spin a low or high - play with it until light in game are gone. I solve that years ago with AMD GPU, but i change value directly into bios of the card (saved bios before any work). But you can try first with drivers to set the settings. I used Ati RBE (Radeon Bios Editor).
still you deserve payment for diagnosing and repair attempt sir,,
I hope so.. Nobody can work for free.
Ever heard of No cure no pay?
@@drood78 cure is to buy new card so...
It could be a faulty bios chip onboard,or the firmware on it maybe corrupted somehow.
Im curious why dont ppl return their cards/ RMA them? Isnt there buyers protection in the countries the owners of those card are located?
Often out of warranty, secondhand (warranty not transferable), and the companies will deny warranty for the slightest ding, bend, or scratch as "physical damage by owner" even if it cannot affect performance.
I'm just curious to know. Why not heat up the balls when the template is still on so that it won't get misaligned.
Hello, did you ask if the card was on a mining rig before?
Don't matter.
"It would be more interesting if it was only artifacting whilst watching TikTok"
That's called a sign from the gods...
Good little backhanded compliment to TikTok, not that its a bad thing LOL
Are those artifects not memory artifects?
Could you please explain me how you go through trouble shooting this card.
I mean you not even consider a ram at all.
And i have a recommendation i dont know if it's possible or not, but when we want to fix some dead led light bulb we just spot a dead led and just shorting it to use others that not Dead.
In this case we can't actually disable one of gpu madules by hand but what if we could tell gpu to not using that damaged mdule at all
Just telling it to ignore that part and use rest.
I mean its a lot going in core for example IO module , memory controller and ...
But this issue might be related to some of the CUs in gpu.
If this could be possible we have 6900xt minus 10% working fine.
What is considered an acceptable delta range between core and hotspot for these cards?
15 and under. 10 is a good number.
@@northwestrepair So say if the delta is 20c, that would be bad? Otherwise, the card runs flawlessly. What would be the recommendations for an issue like that?
I have heard amd say around 20c delta is fine and normal. Anything under 110c is 'fine' for the hotspot aslong as the delta isn't too great according to AMD. I have seen mine go up to 108c when overclocked in the summer and never had any issues so far. I have stopped overclocking the memory so much . I usually have around a 20c delta . Nvidia generally has lower deltas in the 10-15c range as far as I am aware@@Sandmansa
Thanks@@Iowcatalyst Mine gets up to around 70c on the core. 90c ish on the hotspot. And giving the fact that mine is a factory OC'ed and water cooled card, I was wondering if that was normal. It's a power color liquid devil 6900 XT.
I certainly wouldn't be worried about that, they seem like good temps to me, but I am no expert@@Sandmansa
I just got a 7800XT XFX Merc and it works good for the most part. But I got a Win11 Blue Screen crash while watching a youtube video at 1440p, so is it the graphics card or google chrome or youtube? But I rebooted and sent bug report to AMD. While playing a browser Tanki MMO HTML5 game with Brave Browser it crashed but with Google chrome worked good same with Opera GX. Anyone know why this is happening let me know? Also with the adrenaline OVERLAY the fps: n/a always not fps #s....
I wonder if reducing the clock speeds by 10% would "fix" this card
This helped me with my old 280x and made her last a while longer until I upgraded to a 6800xt. Had to reduce core clocks and mem by about 300mhz. One funny glitch that happened is Fallout 4 would run in 2X normal speed, it was hilarious when engaged in conversation but really wasn't fun to play that way 😂😂😂
10-90%
It might but i will not consider it a FIX.
Edit: I tried. It crashed before it helped with anything.
@@northwestrepair As an experiment you could also try measuring the ripple on Vcore under heavy load, at the back of the core. Maybe the caps have aged and lost too much capacitance. Beyond a certain point ripple can cause artefacting. To get accurate measurements on this you need to probe both Vcore and GND at the back of the core, not with a long ground lead.
I prefer to lower the texture filtering since you won't be able to notice the difference
the flashes can be removed by flashing the BIOS to a slightly lower frequency
I'm suprised you didn't try to unervolt/underclock to try and aleviate the issue. Another one could be a bad gpu bios.
Good work
Funny how the artifacts of a dead core can be way more subtle than those of a bad memory for instance, which could be fixed...
When you say bad memory can be fixed, usually that requires replacing the defective chip. The same could be done with the core but it's just much much more expensive.
@@C0smicThund3r replacing the core is pretty much equivalent to buying a new GPU. So no point in replacing at that point.
You _fix_ bad memory? Most people just replace it. And you can do the same with the GPU, naturally, if you can get a new one cheap enough to justify it.
Firts of all, sorry for my English. Second, Thank you for your job, it is really intersting and great done job.
The artifacts on certain shaders looked like when I undervolted one of the power levels (P5) of my Vega too much, but apparently the power was good here.
usually that kind of artifacts also resulting from one or more memory chips on board could have gone broken. since most memory chip/ram chip can still be working to certain extent while being broken, and resulting in errors. like how RAM stick's chips that could be broken and not reporting its frequency but still working to run a computer. but with errors. i think its similar.
I had a GTX 1070 who showed artifacts only IF I watched SPECIFIC videos on TH-cam WITH the 39X series drivers (as the 4xx series simply freezed the computer if I tried said videos), and while I was watching them then I could see graphics annoyances on some websites on the second monitor.
That was something.
To the point that I was unsure if I could return it but there weren't so much time left before being out of warranty so I sent it with a paper explaining with details how to reproduce the artifacts. I was lucky, it was accepted.
Hol' up I have the same problem with artifacting while watching youtube on my 3060... were you joking about that one being a dead core or was it actually dead? I reinstalled windows and drivers and still the same thing happens, at a point it started to appear in other 2d apps too but now its back to only youtube, no other sites, nowhere else to be found, no matter how hard I stress it. Should I be concerned about this?
i have never herd NVIDIA does that.
Try to disable hardware acceleration in your browser. It may not even be GPU related at all.
@@northwestrepairThanks! Disabling hardware acceleration actually did stop the artifacting that was appearing all over youtube interface. But I think I will live with them, I've got quite a few needed sites that don't perform as good when the browser is running on cpu only.
What did the Artifacting while watching TH-cam look like cos rdna3 has glitch that only happens under a specific 3D load that happens while game is running that has the same glitches as a glitch AMD had fixed previously in drivers, elements would teleport all over the page especially with TH-cam ambient mode, this used to be a bug in steam point shop while scrolling or resizing as well which they fixed
Awesome video
whaht is the name of usb device?
It said right on the device
I find something like a dead core on any GPU brand a bad thing I just wish AMD could make it so their GPUs (almost all the time) last 5 years running at stock settings would last that long.
Sure a bad card is going to go bad no matter what but a 3000 from nvidia or 6000 from AMD have no reason to blow up (the core anyway) for another few years.
I bought a 6600xt brand new in january this year and its also artifacting at first, how i fix it is run furmark for like 5 hours and the artifact is gone, and its still working fine today (funny thing is that i asked for rma at first and they just tell me to run the card until its hot and see if its still artifacting and its actually working)
That's what I called a feature
No, you do not "run it until it is hot" as a fix. You RMA it. That card is faulty.
could be the vram. try to reduce the frequency of the vram or gpu
@@luisff7030 its working fine now, its been 11 months since i bought it
lol
This is extremely interesting. You did everything reasonable, it was a good try.
You don't think it's a faulty/dying memory chip?, I assume it's 8 x 2gb. Can you replace those?
i can but there are no memory issues that i am aware of.
@@northwestrepair You would probably know better.. I thought artifacting was usually a memory chip problem.. When I have encountered it, sometimes underclocking the memory "fixed it" and/or also cooling it better.
A few questions as I'm saving up for a gpu. Will this be covered under warranty / rma (within a year, after an year?)
How long should I stress the gpu after purchase to identify any issues? I understand some of these are time-based, so I'm hesitant putting a lot of money into gpu if this happens, and I have no recourse :/
Most cards will have a three year warranty, and this would definitely be covered, unless you've done something to void the warranty. Even if you have overclocked it, they would probably have no way of knowing whether you did or not, and would have to honour your warranty even if you had overclocked it. A modest or minor overclock shouldn't be a big issue as long as it's well cooled, but likely would reduce the lifespan of the graphics card.
GPU's can last a long time with care
@@syncmonism overclocking is literally inside the card's software you download drivers from...
Yes it would be covered. Like others have said most come with a 3 year warranty. Personally I keep all of the original packaging for every component at least until the warranty is over.
First of all you need to learn how to install it and use it correctly. Install heavy cards using the sustaining bracket provided or if it's not provided, you need to buy one and use it from the day one. GPU sag is bad. The installation of drivers must be done after you use DDU to completely remove previous drivers or install driver normally if it's a freshly installed Windows.
Monitor maximum temperatures using GPU-Z. You can run GPU-Z, let it run in the background and play a game. After You close the game, GPU-Z can show you not only the current temperature of the GPU but you can select the maximum one to see. If it's above 75 Celsius and the horspot is above 85 you need better airflow or you may need a better paste. A repair shop should know all this and be able to do what's best fro your card. I said they should know because not always they know what to do.
The master 🙌
gtx 1060 6gb hetting 79 - 83 C full load its good ? but I repaste and rethermal pad is still the same
sell ur card and buy a ps5
@@IlIIlIllI 100? What R U? :D 60% is loud enough.
@@IlIIlIllI there is no point running gpu fans at full speed only thing it is going to do is wear out the fan bearings
It was fine. All this nonsense about "repasting" is often causing more problems than leaving the original phase change.
As for youtube artifacts that is an ongoing driver issue. It is a known current issue. And it is an Nvidia issue too. I have an rtx3080 which only has occasional issues with youtube. No other issues. Temps are good. Can play games just fine at 100 percent utilization.
my radeon showing something like that but not exactly, the game is nfs heat when i put settings to ultra, the reflection is like ants
Ants=dead core.
the thermal paste was almost solid dust....
Manufacturers stopped using consumer style thermal "paste" a long time ago. They use phase change materials in many cases now, causing people to think it's "dried up" and spreading myths about needing to "repaste", thus creating a problem with a paste that actually does need to be replaced regularly d/t thermal pumping.
that thumbnail is gold
isn't artifacting happening by a memory issue ?
Could it be one of the hundred or so caps on the back behind the GPU? They seem to be the only thing exposed when the card is completely assembled. Almost like the GPU is mooning you with its pants down at all times. Idk, seems like an easy spot for something to jab a trace or a cap and knock something loose even with the card fully assembled.
When the card is in a system it must look like the card is assss up with its pants down just mooning the CPU. 🤣🤣
Hey Tony, I have the same question as yet another of your fans in here. Isn't there any way that this GPU could remain useful while gaming? Underclocking, undervolting the main chip memory anything whatsoever? Seriously man there has to be something.
Undervolting and underclocking can help make your card last way longer, but gotta make sure to find the sweet spot and not over do it.
Most likely yes if undervolted.
I do not consider it a fix. Fix is when card works no matter what
@@northwestrepair Undervolting is only recommended if the card is too hot out of box. I agree though, cards should be produced in a better way to where the customer doesn't have to modify it to lower temps and increase performance
I think it might need underclocking and overvolting. Undervolting can stave off degradation long term but it does give you a worsened flank margin, which is something that might be the root cause of the issue already.
@@SkateClipsAndTips Interesting man. Any where I can find info about that? Like... I want to study that. You're saying that undervolting and underclocking can extend lifespan of a GPU? Can you give a figure? Like what... 2x's the normal lifespan? And... Undervolting is not the same as underclocking? I know they are not but... If you undervolt... automatically you lower the core frequency no?
love my rtx3090 suprim x, by default it has power draw of max 420watts, but it's easy to set to any wattage you want via geforce experience overlay -> performance without the use of msi afterburner (afterburner will cause some games to crash). right now I didn't play graphics intensive games and I just set max power of 270watts, the card will auto regulate everything to run at that wattage, very convenient
I wonder why he did not consider running a mem test on the card? Generally from my experience it's hard to tell a GPU vs DDR problem when it comes to artifacting.
I had those artifacts when overclocking DDR 3 mem past some limit on my old HD8670m card, but it would keep rendering, and the artifacts would get worse and worse, until a hang would happen.
cant clock speed be reduced to remove artefacts
most likely
Happened to my Radeon HD 5970 GPU, which I got in 2009. Years later I put it into the oven. Cooking it in the oven actually fixed it for awhile
good job
i think gpu which artifacting only in youtube its becasue h264 area not working
Might it be a Vram overheating issue?
maybe one of the Vram is not making proper contact with the heatsink !
The video description seems inaccurate to me. The problem is that it's a dead core, could be caused by mining/aggressive OC with inadequate cooling/not supporting the card properly. We don't know how the owner used this card.
The "revelation" at the end is really more of a theory and he doesn't seem "shocked" either by his hypothesis. Theories and hypotheses are not discoveries either. That's a little clickbaity. It also would not cause the dead core problem. One reason I was skeptical of this video description is that this GPU came out over 3 years ago. It's a little late for design flaws that actually "shock" people to be likely to actually be discovered. Not impossible, but very unlikely.
EDIT: Oh, and I just noticed you put horns on the AMD Ceo's head on the thumbnail? WTF?!
In your opinion what make this happen
Are you asking him to theorize what part of the GPU core is failing? Because he obviously thinks that's the problem.
Stuff's gonna break, not every card is going to make it. AMD reference designs tend to be pretty robust though.
Except for the bent core :D
@@draakevil don't forget the cracking GPUs