Sapphire HD7850 1G artifacts

[tiki66]

Hello,

I have HD7850 that has artifacts. In safe mode it works ok, when trying to boot Windows in normal mode artifacts appear and after some time BSOD "BAD POOL HEADER". I unistalled drivers with DDU and now it works with Microsoft drivers.

Do you think this is a memory or a GPU chip issue? Should i try reflowing GPU/memory?

[R_J]

Re: Sapphire HD7850 1G artifacts

Have you cleaned all the dust from the video card, Are the fans even spinning? I would give the whole pc a good cleaning based on the picture.

[tiki66]

Re: Sapphire HD7850 1G artifacts

Yeah, don't worry about the dust, I will clean whole pc and replace all thermal grease.

Fans on the card are spinning. I was just wondering if there any tricks that could prolong GPU life, since new GPUs are expensive and my friend currently doesn't have that kind of money.

[taxxin]

Re: Sapphire HD7850 1G artifacts

Yeah, don't worry about the dust, I will clean whole pc and replace all thermal grease.

Fans on the card are spinning. I was just wondering if there any tricks that could prolong GPU life, since new GPUs are expensive and my friend currently doesn't have that kind of money.

In my experiences, this kind of thing happens when there is a problem with the RAM, not the ram chips necessarily but it could be the power circuits for them or the BGA connection between the chip and the PCB. I would first check to make sure that there isnt any physical damage or missing resistors on the back of the card.

if all of the chips are getting their power supply around 1.5v then it could be a possibility of broken solder joints on the BGA or faulty chips. in which case you would see if there is any kind of memory test available for that card to determine the faulty chip. if you are on a real budget you can just remove the faulty chip and edit the BIOS to run without it.

[RJARRRPCGP]

Re: Sapphire HD7850 1G artifacts

Hello,

I have HD7850 that has artifacts. In safe mode it works ok, when trying to boot Windows in normal mode artifacts appear and after some time BSOD "BAD POOL HEADER". I unistalled drivers with DDU and now it works with Microsoft drivers.

Do you think this is a memory or a GPU chip issue? Should i try reflowing GPU/memory?

Looks like a VRAM problem. Albeit, "BAD_POOL_HEADER" is usually a system RAM problem, such as a bad DIMM slot or a bad CPU IMC. Maybe a broken motherboard trace to the RAM!

Looks like you need to clean the motherboard, clean the RAM and reseat the RAM!

[momaka]

Re: Sapphire HD7850 1G artifacts

Given it's an HD7k series, most likely GPU chip went bad. 7k series has a ton of problems... or at least, now that the cards are getting old.

That said, given the horizontal artifacts, do check all of the small SMD caps behind the RAM chips for low resistance (1 Ohm or less... or whatever is the lowest your multimeter can read with shorted leads.) I did manage to save a Gigabyte Radeon HD6850 that had a single shorted cap and was exhibiting horizontal artifacts (in mobo BIOS.) Thread about that can be found here. That one was a lucky save, though, I think. Most artifacting cards I've had to deal with have all ended up being bad GPU chip. Reflow did revive a few of them back for a while (some still going OK.) But again, HD7k series is quite failure prone. If you do have more HD7k cards, make sure to run them below 55C under load if you want them to last. Even 60C is too high for them, as on these older HD cards, memory I/O area on the GPU chip tends to run 5-10C higher than the rest of the core. So at 70C core typical running temperature, mem I/O is already 75-80C in most cases... and anything above 60C tends to cook these cards fast.

[tiki66]

Re: Sapphire HD7850 1G artifacts

if all of the chips are getting their power supply around 1.5v then it could be a possibility of broken solder joints on the BGA or faulty chips. in which case you would see if there is any kind of memory test available for that card to determine the faulty chip. if you are on a real budget you can just remove the faulty chip and edit the BIOS to run without it.

Thanks, will try. That kind of "patch" I had in mind, just to make GPU work for a while so he can play older games at least.

If VRAM is bad, how do you find out which chip is bad?

Looks like a VRAM problem. Albeit, "BAD_POOL_HEADER" is usually a system RAM problem, such as a bad DIMM slot or a bad CPU IMC. Maybe a broken motherboard trace to the RAM!

Looks like you need to clean the motherboard, clean the RAM and reseat the RAM!

Sure, I will do that and run memtest, but I don't think it's a RAM problem as PC works normally without AMD GPU drivers.

Given it's an HD7k series, most likely GPU chip went bad. 7k series has a ton of problems... or at least, now that the cards are getting old.

That said, given the horizontal artifacts, do check all of the small SMD caps behind the RAM chips for low resistance (1 Ohm or less... or whatever is the lowest your multimeter can read with shorted leads.) I did manage to save a Gigabyte Radeon HD6850 that had a single shorted cap and was exhibiting horizontal artifacts (in mobo BIOS.) Thread about that can be found here. That one was a lucky save, though, I think. Most artifacting cards I've had to deal with have all ended up being bad GPU chip. Reflow did revive a few of them back for a while (some still going OK.) But again, HD7k series is quite failure prone. If you do have more HD7k cards, make sure to run them below 55C under load if you want them to last. Even 60C is too high for them, as on these older HD cards, memory I/O area on the GPU chip tends to run 5-10C higher than the rest of the core. So at 70C core typical running temperature, mem I/O is already 75-80C in most cases... and anything above 60C tends to cook these cards fast.

Yeah, I had R7 270 a while ago and it also died in middle of gaming. I think those chips in R7 270 and HD7850 are similar. On her the fuse blown and all SMD caps around GPU were in short. Removed alot of them and short was still there, so I lost my nerves and threw it away.

Will check SMD caps. If I go for a reflow, what temperature do I reflow it at and for what time?

[piernov]

Re: Sapphire HD7850 1G artifacts

https://www.badcaps.net/forum/showpo...02&postcount=2

2.1. Never reflow large BGA
It won't fix the problem since 99% of the time it's not an contact issue between the BGA and the logic board. There's a high risk of damaging the board even further.
If you want to confirm if an AMD/NVidia chip is bad, you can heat it up to 200°C max for a minute and a half, no flux. It'll come back to life for a short while and then die again, because it's bad. In this case it has to be replaced.

[RJARRRPCGP]

Re: Sapphire HD7850 1G artifacts

Sure, I will do that and run memtest, but I don't think it's a RAM problem as PC works normally without AMD GPU drivers.

That was the first RAM-related BSOD that I got while installing Windows 7, on my socket 1366 platform, after I changed the motherboard from an Asus P6T Deluxe to an Asus P6T6 WS Revolution and I never got that BSOD code before, much less before Windows 7 installation being complete. The black RAM slots are suspect with my P6T6 WS Revolution.

Now, the good news, I have yet to get a RAM-related BSOD when using Windows 7 with the 12 GB Corsair Vengeance triple-channel DDR3 SDRAM kit.

[brethin]

Re: Sapphire HD7850 1G artifacts

Its a ram issue not a gpu issue.

[tiki66]

Re: Sapphire HD7850 1G artifacts

Update:

cleaned whole PC, replaced thermal grease on GPU and CPU, installed drivers back again. Now the situation is much better and PC doesn't crash at login and you can use it normally. I downloaded OCCT as the first test that I could remember of and ran GPU error test on memory and it ran fine for about 20 minutes. Then I ran Witcher 3 for about 5 minutes and no crash either. After that I returned to OCCT and ran standard GPU test, after few minutes the whole screen went red and then I got No signal message from monitor. After rebooting, PC started normally and still works, but I didn't force GPU.

Probably a bad GPU chip...

I also ran memtest and it is ok. I have never seen a system RAM issue that would produce artifacts like these on screen. I would see these only if internal GPU was used and it used system ram as gpu ram (if ram was bad)...

[momaka]

Re: Sapphire HD7850 1G artifacts

https://www.badcaps.net/forum/showpo...02&postcount=2

2.1. Never reflow large BGA
It won't fix the problem since 99% of the time it's not an contact issue between the BGA and the logic board. There's a high risk of damaging the board even further.
If you want to confirm if an AMD/NVidia chip is bad, you can heat it up to 200°C max for a minute and a half, no flux. It'll come back to life for a short while and then die again, because it's bad. In this case it has to be replaced.

Well, I'm no expert, but IME, most GPU's I've heated up to about 200C (or at least not proper reflow temperature) did not come back to life until I heated them further. I completely understand and agree with you that the issue is not with the BGA going bad between the chip substrate and the PCB, but rather the chip die and chip substrate... which can't really be fixed, indeed. Chip just has to be replaced. However, it seems to me (or at least from my experiments) that higher reflow temperatures past 200C tend to have higher chance of "reviving" a chip, and in some cases for quite long if ample cooling is provided to the chip afterwards. I think this has to do with material expansion more than anything - at temperatures above 200C, I'm pretty sure there is a lot of thermal expansion going on with all of the materials... and the more that happens, the more likely are contacts in the GPU die to reconnect again, bringing it back to life. However, this also increases the risk of wrong contacts to form (BGA bridging both under GPU die and GPU substrate). So higher temperatures are also more likely to kill a chip, especially when going past 220C.

Of course, at the end of the day, a dead video card is a dead video card. I've fried/popcorned plenty of GPUs from trying to revive them like this... and I could care less. It's not like I could find cheap replacement GPU chips for them... or at any price, for that matter. So by popcorning the PCB/GPU, at least I know it's dead for good and only useful for salvaging parts from that point forward.

If I go for a reflow, what temperature do I reflow it at and for what time?

Depends on what you're using for the reflow. Oven? Heatgun? Stovetop burner?

For heatgun, probably at least 5 minutes on the "low" setting, so long as that's close to 200C. Just make sure to alternate between top and bottom of the PCB (place the PCB on a metal mesh or grille and suspend from a table edge, so that you have access to both sides of the card's PCB without having to flip the card by hand.)

For oven... that will depend a lot on your oven... but 200C in a pre-heated oven for about 2-3 minutes *might* do. Otherwise, if putting the card in the oven without pre-heating it, you might want to lower the temperature a bit, since some ovens can overshoot the temperature a good amount as they come up to the set heat level. On that note, electric is probably the worst, especially with heating element on the top, since the PCB will receive a lot of heating from IR, and that can end up overheating certain parts.

And as for stovetop burner... only doable if the stove uses gas (flame) or it has an exposed (curled) heating element (electric stove). With this method, though, you will need a thermocouple of some sort to probe the temperature of the PCB, as it's otherwise hard to tell when the PCB has reached a high enough temperature. Also, make sure board is suspended at least a few centimeters above the burner.

Its a ram issue not a gpu issue.

I'm guessing that's a reply to RJARRRPCGP's post above yours. Because the O/P's issue is absolutely 100% from the GPU.

Update:

cleaned whole PC, replaced thermal grease on GPU and CPU, installed drivers back again. Now the situation is much better and PC doesn't crash at login and you can use it normally. I downloaded OCCT as the first test that I could remember of and ran GPU error test on memory and it ran fine for about 20 minutes. Then I ran Witcher 3 for about 5 minutes and no crash either. After that I returned to OCCT and ran standard GPU test, after few minutes the whole screen went red and then I got No signal message from monitor. After rebooting, PC started normally and still works, but I didn't force GPU.

Probably a bad GPU chip...

Yup.

Time to reflow it... though that could also make it worse or completely kill it.
Alternative is sell it as-is for parts or repair. Maybe someone else more brave wants to try and fix it. I used to buy broken GPUs like this on eBay with intention to fix them long time ago... but I rarely do anymore. The only exception is when the price of the GPU is so low that it practically makes it worthwhile just to buy it to salvage the cooler off of it.

I have never seen a system RAM issue that would produce artifacts like these on screen. I would see these only if internal GPU was used and it used system ram as gpu ram (if ram was bad)...

Exactly.
I had this issue with an A4-5300 APU and a bad RAM module... or rather, bad contacts on the motherboard - system would freeze with all 4 RAM modules installed, but not with 2, and only if 2 particular slots were used, suggesting the other two slots were not good.

[piernov]

Re: Sapphire HD7850 1G artifacts

Well, I'm no expert, but IME, most GPU's I've heated up to about 200C (or at least not proper reflow temperature) did not come back to life until I heated them further.

See demonstration: https://youtu.be/Ty5PkmOK0Qs?t=381
I've done this dozens of times on NVidia bumpgate garbage or ATi/AMD garbage (especially Radeon HD4000-7000 and RS780/RS880 but can also work sometimes outside that range). It almost always works, I don't even remember a case where it didn't work at all. Of course in all of these cases the chip itself was bad, so they were either not repaired (most often, because good chip is not available or too expensive) or repaired by replacing the chip (with non-bumpgate for NVidia which I'd call a permanent fix, or with the same garbage for ATi/AMD which will die again a few years later anyway).

Of course, at the end of the day, a dead video card is a dead video card. I've fried/popcorned plenty of GPUs from trying to revive them like this... and I could care less. It's not like I could find cheap replacement GPU chips for them... or at any price, for that matter. So by popcorning the PCB/GPU, at least I know it's dead for good and only useful for salvaging parts from that point forward.

The whole point is that people are dumb and reflow everything for no reason, killing boards that would have been repairable otherwise. As I've already said here https://www.badcaps.net/forum/showpo...24&postcount=9 :

Well yes I'm fed up with people and their reflowing/reballing scam and I won't be kind with my words on that topic anymore. Replacing GPU with a good new one when it makes sense and they are available yes, reflowing/reballing not in hell.
And if you're not mad at people with the reflowing bullshit you probably haven't had enough boards on your desk destroyed or seen enough people killing their boards on the Internet because they took the bait of "reflow gpu bro".

The other point is that people are dumb and think that a reflowed (or even reballed) chip is reliable. So you end up with people paying for these scam repair services or buying used devices that'll die in a few weeks/months. And of course some people have insane luck and it lasts years for them (dunno how) so they spread misinformation telling others that it is indeed the solution to every problem.

And with your "advice" you take part in this insanity that's still going on 13 years after NVidia bumpgate has been figured out.

[momaka]

Re: Sapphire HD7850 1G artifacts

The whole point is that people are dumb and reflow everything for no reason, killing boards that would have been repairable otherwise. As I've already said here https://www.badcaps.net/forum/showpo...24&postcount=9 :

Oh, I completely agree.

"Reflow"/reheat should be the last option after everything else has been checked.

And with your "advice" you take part in this insanity that's still going on 13 years after NVidia bumpgate has been figured out.

I know, and I'm not too proud of it... but some of the stuff is trash-bound anyways. Might as well give the common Joe something to try out before they trash the device. And hey, if they get 2 weeks of service, that's still better than nothing, provided their time spent to "fix" the device was worthwhile to them.

I am by no way implying that a reflow/reheat is a permanent fix... or a fix at all (not anymore, at least... though in the early years of that stuff, I did believe in it, after working at a "repair shop" that did this sort of thing.) And when it specifically comes to nVidia bumpgate, I don't even attempt to save defective GPUs from that time period anymore. I've had almost zero luck (yes, luck is the right word here, as you noted) getting any older nVidia GPU to come back to life. Best case was a Compaq v6000 laptop with a GeForce 6250 chipset that is still running 4 years after I've reflowed it - and that's probably only since I have the CPU severely under-clocked to keep the temperatures low enough, and also because I just use that laptop very infrequently (few times a month, max., just for burning discs or audio signal output. - i.e. nothing heavy on the CPU or iGPU.)