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...
I'm glad after 30+ years of repairing electronics to component level I can finally look forward to retiring after next year. Things are so much smaller these days and my eyes are worn out along with many other things.
As an asic repair tech (who emailed you about a gpu repair recently) I feel your pain! It sucks when you spend countless hours working on a repair only to have it be a no fix. Such is our life! Despite this repair not being a success, just know you are a highly skilled technician! Kudos for what you do!
@@heatelectronicsstart by practicing on something cheap like S9’s. They are actually harder to work on the s19’s so once you are successful with them you will have a very good footing on s19 repairs. Also there isn’t much difference between asics… their architecture are all similar.
@@teddptrue but I do not consider my repairs a success unless they work as designed. It’s a high standard that differentiates a repair man from a technician. Making it work and making it work right are 2 different things!
Thanks for taking us through a technical walkthrough instead of sharing a video with an easy repair and telling us about how great your tweezers are for 20 minutes.
@@northwestrepair I got into tech-repair-tube beyond the louis rossman channel in the last 8 months or so. I found mr tweezers and liked his content, then I found you. You are so much less annoying to listen to. Mr sell-you-something had the content I was interested in, but the delivery was shit, and has gone downhill. Happy subscriber here.
As I've commented before run mods and loop tests 94 and 95 gradually increase the memory clock until it crashes shouldn't need to exceed stock clocks. Also you can exceed stock clocks to find any degraded modules.. In my experience this works every time and will lead you to the problem which is most likely a module not being able to run at stock clocks at higher temperatures.. Anyways gl keep up the good work and God Bless..
I ran this command using mods 455.127: mods gputest.js -skip_rm_state_init -oqa -test 94 -dramclk_percent 120 Mods report: Global.CheckForSpecificTestSkipped A specific test was requested to run With that said, it does not matter what test i run, error is the same so test never takes place. I dont know what to do about that. Maybe wrong MODS version ? Lowering the clock for memory does make it stable.
Sometimes we avoid the possibility that the BIOS could be a problem so much that we end up forgetting about it being a factor. The customer knew a respectable lot about what he wanted done to the card. Did you consider that, perhaps, he might have flashed an alternative BIOS trying to squeeze some power out of the card? It happened to me more times than fingers I have in a hand.
If I was the customer, I would ask you to return the memory back to hynix and still pay you anyway. Your dedication is enough for me to pay. It is my show of respect for your works.
Hynix went to the trash. If I solder it back, it may work for a week or a day or a year. Lucky me, I still have the card and am trying to figure this out.
This card looks like it has a problem getting enough supplied voltage. The likely problem is it is missing a good connection on one of the 12v+ wires in the main connector which overloads the other connectors increasing their resistance as higher amp flow goes through the connector wires and connector resulting in a voltage drop and a heavy reliance on the 12V+ board back-plane to supply 12V across the card. I've had this happen on lots of cards and the next symptoms you'll see , under heavy loads if it's not addressed, are burnt pins or connections at the PSU connector followed by overworked, overheated, and burned power phase chips that get hot and melt the solder, stop functioning or weld to the board.
I'm thinking this is probably the issue, it's not getting enough power considering the PMIC was having an issue to begin with. would've been nice for him to check the card before putting micron on it.
@@ChrisFaulkner it wasnt a pmic issue. It was a voltage monitor issue. If you remove pmic and Power on card say goodbye to gpu, since all mosfets will let 12v straight to vcore
There is something absolutely therapeutic about watching you work your magic on a core, the re-ball too, just watching those little solder balls fit into place as they melt is so relaxing :) #solderasmr
You are Amazing as usual! Sometimes it can be something simple and it can also be you changed that is not agreeing with your decision. When they started using controllers on engines and transmissions that was sometimes a real pain to troubleshoot. Transmission controllers were the most interesting to work with.
15:12 voltage drops on core is 11,2 should be 12v input voltage. thats why it crashes, nothing wrong with card?!? psu? i had corair what did the same and it was broken
Your patience is endless, your nerves are made of steel! Every time when I see you making another reball of gpu or memory replacement it's like watching a horror movie for me! Keep fighting!
Lot of work on that card to get kicked in the arse at the end. I really enjoy your vid's and this hit all the points. I sent you some candy money : ) Keep up the great work.
Before the crash, the GPU jumps to 1890 MHz. Try to lock the speed to 1800 Mhz, and maybe lower the mem clock as it goes down to 5000 MHz before crashing. Maybe one of the capacitors is failing?
Funny for my fiancee used to have this exact card! Scary that it had a huge red flag in terms of memory choices. Thank you for bringing this to everyone's attention! Am also trying to join your Discord but keep getting an error :
As the card has undergone a hefty impact there may be a cracked joint, track or via hiding somewhere. The card heats up and goes open circuit. If you can't monitor signals accurately it could be near impossible to find.
You were more than reasonable with your customer, not a full win on your part, but still the card is usable and you saved it from the landfill. I, on the other end, am a winner, because I learned a fair deal from your video.
GPU GURU strikes again, i dont know how much you charge for this... But i can tell you now... Its not enough for sure! Love your videos Greetings from Portugal 🇵🇹
Hello boss! Nice work even if you couldn't fix it at the end 😢 I have a question: the configuration straps for the memory chips is specific to DDR6 GPUs (or RTX cards) or you need to do it for every GPUs if you change the type of Ram?
@@computersrepaircotesaint-l7407I was wondering about that one, too. But as I understand, the customer requested it, right? Well, I guess you could still advise the customer to not worry about the memory for now, and only start replacing the chips when they start going bad.
Have you seen if the phases are not shutting down during load? Or perhaps the vcore ripple is too high because of a failed tantalum right on the bottom of the GPU due too all the heat. Try to increase the capacitance by soldering more capacitors in parallel, just for testing. Use your scope to see the ripple during load
Hi, According to your sheet the Hynix configuration is 00101 and micron is 00001(throw-out-issue) and 00100(line on screen issue), but is it possible 00101 (after the line being fixed) also works for micron too and solve the throw-out-of-game issue?
Sorry if it's out of this video's topic. But I got some interesting find on my XFX 319 SWFT 6800, I could push it just fine & stable up to 2500 MHz core @935mV, and 2100 MHz on the memory clock (default timing). But it's unstable af as soon as I turned on the Fast Timing on the memory no matter what I do. Even when all the other values are at stock, it will still be crashing and reboots my pc along with a broken BSOD view, like a broken monitor (even though my LG GL850 monitor is working just fine). Is this a sign of a faulty VRAM? Or am I just unlucky? Or is there something else?
Hi.. Really loved your videos.. I'm having a hard time with Zotac 1050ti OC. I'm a normal consumer with minor knowledge of electronics. I know how to use a multimeter. The case is where ever I go with this GPU for repair I get this strange feeling that people wanna buy my card for less price and not repair. Keeps telling me this card has a dead core. What dead core means and can I check the GPU main chip directly ? like directly from the chip that it is getting power !?. Please shed some light.
strange that all that work and in the end crap. have you test the card before you change the memory ? was working 100% ? , if yes might be another part that got damage or "misplace" like that resistance
I know there is, for Maxwell anyway, another BIOS version for Samsung, than one for Hynix. Maybe some double -density Samsung chips, and dif strap settings could fix it? Edit: But make sure card supports address lines for new RAM first.
Hi there Tony! Thanks for another adventure within GPU micro world :) Btw, @northwestrepair are you planning to give this card another chance to get rid of the crash? Another question, do you know the whole story of this card, except that it was dropped on the ground? See you
Err, doesn't the Black Screen mean you're supposed to lower the GPU core speed or sumpin'? Sumptin' about mis-timings between GPU core and the GPU RAM on old stuff?
5:45, but why?? I know it is a bad line of memories, but if it aint broke why fix it? can it be that the memory replacement "blew up" something? I hope you find something and do a follow up video
This is how you serve customers properly, very well!😉 There's no point in giving the customer back a defective card, so everything was done right. The only question that arises is, is it even worth investing so much work and material into a 3060Ti that is worth a maximum of $350 used?
Tony the customer might not run into the problem you describe with heat. Undervolting? That may or may not be the case. It is possible the customer will play it long term with no problem. Possible. Possible. Likely? Dunno after all the work you did for this customer my guess is a resister somewhere is not what it claims to be. Just a guess. Tony could possibly have put on a regulating diode maybe where it is not supposed to be.
I don't believe, I know that you did the best that it is humanly possible to fix that card. Even more, if you ask me, you went above and beyond the call of duty. On the other hand, you made an error but even you Mr Wizard, you're only human. After you first fixed the card with its stock memory chips you should have made it run the gauntlet (stress test it) and then and only then you should move forward to the optional change of the memory chips, for which thing alone you should be charging double because it's not a fix anymore it is something extra that you do after the request of the customer. So bear that in mind Tony, next time anyone asks you something extra you should also charge extra for it and also make certain that things work at the default configuration before you go the extra mile. Stay strong 😉
have you ever converted a 3060 ti or 3070 to a 16 gigabyte gpu with samsung M die vram before on a 3000 series gpu since 8 gigabytes isn't enough for a lot of games now
My Boardviewer is completely empty. It has no schematics. Where do I get the schematics from ? And why it didn't work the repair ? You probably have to set the memory timing in the bios. If there is such a thing. A bios update maybe ?
Are there are differences in height of the memory chips you replaced? If the Micron chips are slightly higher than the Hynix chips, then the heatsink/cooler would not sit correctly.
Got the quote back via email. $300-$500 to fix my RTX 3080...kinda not worth it when a new one is 480ish. Oh well. worth a shot at least. Guess I'll sell it for $200 and let it be someone else's problem.
I had a card like this. After time the vmem just shut down with no other signs. After crash, measure vcore and vmem. One wont be present. Or will disapear for a short time and come back. And thats the controllers fault.
I think EVGA saw a lot of things that were no longer in their control and bowed out. I can't blame them. Every company is going to have some bad stuff. They were a good company.