HD 7850 - Powers on, fan spins, no video....

[ShortCircuited]

Bios reads fine with programmer (ch341a).

Found an area of caps showing short-circuit 4.6ohm. Normal?

Here are some pics:





Refernce: https://www.badcaps.net/forum/showth...t=35004&page=2

[ShortCircuited]

Re: HD 7850 - Powers on, fan spins, no video....

Applied 12v to 6-pin (card on bench; not in PC). Draw is 250ma. GPU is barely warm. Feeling around for hot SMD caps, etc...nothing warm except GPU (barely warm).

7805 is dropping from 12 to 5v.

[momaka]

Re: HD 7850 - Powers on, fan spins, no video....

That 4.6 Ohm resistance does look normal, as it appears to be your GPU V_core (which can sometimes be as low as 1.x Ohm on some really big GPUs.)

Check the RAM Vdd rail for short circuit too - that should be either generated by the buck circuit near the PCI-E connector (where the square inductor is) or near the fan connector on the bottom-right (again, where a square inductor is.)

If none of these are shorted (and they likely aren't if your card can still turn on without shutting down the PC), then I suspect you just have a dead GPU chip, given that the BIOS is OK too... which shouldn't be surprising. The HD7k and R9 2xx series aren't known to be very reliable - in fact, they are notoriously unreliable, based on what I've seen on the used market (mostly following them on eBay.) The number of dead/defective cards I see from the HD7k and R9 2xx series relative to how many working ones there are offered is quite high. Even the GTX 4x0 and 5x0 series don't seem to be that bad, and they were quite far from reliable too, given how hot they ran (especially the GTX 470/570 and 480/580.)

You can try giving the GPU chip a reflow or two to see if it will come back... and if it does, it most likely will be a temporary fix. However, you *may* be able to extend that a little if you make sure to keep the GPU core temperature under 53-55C at all times under load (preferable to keep it under 52C, even, if that's possible.) I use MSI Afterburner and set my own fan profiles on my cards so that GPU fan = max when card is at 57-60C... which makes the card really loud when stressed hard. So to avoid that too, I also frame-limit games to make sure the GPU isn't under 100% load, which helps lower the heat and noise. Do that to any GPU and it should last much longer. The downside is, you may not get the best game performance. But that's how these things go. Modern GPUs simply aren't built well enough to whithstand 24/7 on max load... which is why mining usually kills GPUs so quickly. The ones that do still work after being used in a mining farm were probably just the "better-made" silicon wafers. As with any manufacturing, there are always tolerances to how things are made, and some of these GPU chips just barely make it out of the factory working 100% normal, and some make it with flying colors... hence why sometimes GPUs last, and sometimes they don't. Of course, keeping a GPU cool and under a lower workload will no doubt extend its useful life.

[ShortCircuited]

Re: HD 7850 - Powers on, fan spins, no video....

Thanks for the response. I probed around in the regions mentioned and didn't find a short.

I did find pin 3 of Q621 (FDMF 6705B) is 1ohm. In this image I have marked the 2 adjacent fets laterally; Q641 bad Q601 which are both are 211ohm to pin 3. SMDs marked 1&2 are 10k and 2 ohm respectively. Those marked 3&4 corresponding to the suspect fet (Q621) read 272ohm & 209ohm respectively.


C649, C629 and c609 all 100nF.

Pulled R626 and short to ground for Q621 VDRV (pin 3) gone.


Pin 3 is VDRV-power for gate driver.

Q621 shorted maybe? VDRV is shorted to ground. FET itself or accompanying ICs?

Marked: F DC02AD FDMF 6705B.

[ShortCircuited]

Re: HD 7850 - Powers on, fan spins, no video....

Based on the block diagram I do not see that VCIN and CGND should be .9ohm (out-of-curcuit) for these DMF6705Bs. Can anyone confirm/reassure me on this? The other FDMF6705Bs are open pin 2 to 37 in-circuit.


Pulled Q621- Short from pin 2 to ground is gone on PCB. Pin 2 to CGND is .9 ohm on Q621 (FDMF6705B) out-of-circuit.

[ShortCircuited]

Re: HD 7850 - Powers on, fan spins, no video....

Installed card without Q621. No display but the fans are no longer default 100%. Now the dreaded wait for replacements....

I dont think replacing Q621 will fix this card, but I have no choice but to try.

[momaka]

Re: HD 7850 - Powers on, fan spins, no video....

Installed card without Q621. No display but the fans are no longer default 100%. Now the dreaded wait for replacements....

I dont think replacing Q621 will fix this card, but I have no choice but to try.

Wow, very interesting that you had a bad MOSFET/driver combo chip and it didn't short-out the PSU, as that is usually what happens - at least with the more classic dual MOSFET setups with older cards. But I guess these newer cards are different.

In any case, glad to hear you found the problem... or perhaps part of it anyways. Keep us posted what happens when you get the new Q621 replacement and you try it out. I'm curious to see if (and hoping for) that fixes the card.

[ShortCircuited]

Re: HD 7850 - Powers on, fan spins, no video....

Soldered in new Q621. Still no video. Fan spinning normal. Seems GPU is getting warmer than it was before. Checked with thermal cam and didn't see anything beyond 35c.

[Dan81]

Re: HD 7850 - Powers on, fan spins, no video....

If it's an VTX3D/Club3D OEM card, forget about it. I've had a 7950 that didn't show video yet it would POST. No amount of reflowing or replacing FETs ever brought it back to life again, so I saved the cooler and chucked the rest of the card.

[momaka]

Re: HD 7850 - Powers on, fan spins, no video....

It doesn't matter which OEM the card is. The HD7k and R7/R9 lines of cards are just all doomed form what I have been observing on the used market. Not sure if it's GPU chip -related or the RAM failing on these, but I think more likely the former. Just look on eBay for HD7950/7970 and R9 280/X or 290/X cards, and you will see a disproportionate amount of cards listed as not working / for parts or repair. The HD7850 and 7870 seem to be a little better, but not by much. Only the 7770 and lower seem to be a little more OK. Most likely because they have smaller cores and thus are less failure-prone.

[dicky96]

Re: HD 7850 - Powers on, fan spins, no video....

Just look on eBay for HD7950/7970 and R9 280/X or 290/X cards, and you will see a disproportionate amount of cards listed as not working / for parts or repair.

Funny thing that, Momaka. I picked a R9-290X up at the local flea market a couple weeks ago for €35 (untested the guy told me) - and mine is working

https://www.youtube.com/watch?v=EW2vuZSgbFM&t=4s

Though I know in advance exactly why you are going to well me mine still works.....

[piernov]

Re: HD 7850 - Powers on, fan spins, no video....

Well it's from the era when ATi/AMD GPUs had a pretty high failure rate, on desktop graphics card it's especially the higher end models that seem to be affected. (and after dying they often react to being heated up to 200°C and resurrect for a short time, so in my opinion I wouldn't even count on the "working" ones to be actually working properly and reliably)

[dicky96]

Re: HD 7850 - Powers on, fan spins, no video....

Hi Piernov
Did you see on the YT video what someone had already attached to modify the R9-290X card I bought..... maybe since it was new - who knows?

Interesting though it was at a flea market sold as condition 'untested' so who knows really. I'll put it in my overclocked i7-5820K Asus X99 rig shortly, as I want to use it for editing my YT videos taking advantage of the quad channel DDR4 (and playing some games also of course) so I guess I will find out eventually

[momaka]

Re: HD 7850 - Powers on, fan spins, no video....

Funny thing that, Momaka. I picked a R9-290X up at the local flea market a couple weeks ago for €35 (untested the guy told me) - and mine is working

https://www.youtube.com/watch?v=EW2vuZSgbFM&t=4s

Though I know in advance exactly why you are going to well me mine still works.....

Oh yeah, you can still find working ones too. I'm just saying that the number of bad HD7k and R9 cards I see relative to working ones is disproportionate compared to other models from both nVidia and ATI. So the working cards are probably only the "really good" ones left at this point - i.e. consider yourself a winner of the silicon lottery.

Nice score, BTW. Like you said in the video, just the cooler on that card is well worth the money you paid for it. In fact, I've bought known bad cards before for cheap, just because they had nice coolers (e.g. one of them was a R9 280 that I picked up for $15 shipped to my door, and it had a nice chunky 2-fan cooler.)

Also, that's a really good idea you brought your DMM to the carboot sale and checked for short to ground on the 12V rail and PCI-E conn. Indeed if there is one, at least on these "newer" cards, chances are the GPU could be toast. On old cards however (like AGP era and early PCI-E era), I find that a high-side FET short doesn't always result in a dead GPU. However, on some ATI cards, a high-side short (typically on RAM V_dd, which also supplies parts of the GPU too) can be caused from the GPU chip going bad and taking the MOSFETs on that rail out if it wasn't designed well (some cards don't use very robust PWM controllers that can handle a short-circuit properly.)

I also got a working HD 7950 last year for $40 shipped. It was listed as used, but "as-is". No mention if it was working or not... but it wasn't listed for parts and seller also didn't put "HD 7950" in the title or body, but did put the HIS P/N. So I was able to grab it for a little lower price than usual (at the time, these were going for about $50-60.) I haven't used it too much, but it does work. So I consider myself lucky too.

Well it's from the era when ATi/AMD GPUs had a pretty high failure rate, on desktop graphics card it's especially the higher end models that seem to be affected. (and after dying they often react to being heated up to 200°C and resurrect for a short time, so in my opinion I wouldn't even count on the "working" ones to be actually working properly and reliably)

Yup.

IIRC, aren't the earlier Xbox One and PS4 GPUs somewhat similar to the HD7770? Whichever the case, those seem to have been pretty reliable overall. Perhaps because they use smaller / less complex GPUs?

I also find it interesting, however, that cards from the HD6k era weren't nearly as failure-prone as the HD7k era, at least if kept well-cooled... or so has been my observation based on what I've been finding on the used market. That aside, I'm still putting nearly daily (ab)use on my repaired Gigabyte HD6850. Granted I do run a custom fan curve so that I've been good about keeping the GPU temps under 55C at all times, and also using frame-limiting in games to limit GPU usage to less than 70-80% usually (and typically down to 30-50% in Fornite, which is what I've been playing the most lately.)

I'll put it in my overclocked i7-5820K Asus X99 rig shortly, as I want to use it for editing my YT videos taking advantage of the quad channel DDR4 (and playing some games also of course) so I guess I will find out eventually

Cool!
Keep us posted how it goes over time.
I noticed from your Furmark score that the temps jumped quite high for a very short period of time. Granted of course, that is a 250W TDP GPU chip, so that's probably why. Even that chunky cooler can struggle. Probably need water cooling with a dual 120mm fan radiator and really high-quality block to keep temperatures more sane. Ideally, I like to keep mine under 60C... or even 52-54C, because ATI/AMD cards can have as much as 8-10C difference between where the sensor is reading the GPU temperature and "deeper" layers on the GPU chip. The HD4k, 5k, an 6k cards, for example, have a MEMIO sensor, which read the GPU chip's temperature on the memory I/O part of the GPU... and that often was the temp that was the higher by 8-10C over the GPU temperature. So that's why I've always tried to keep my ATI/AMD cards a little cooler. Ultimately, though, I don't know if and/or how much that will help down the line.

I think with the HD 7k and R9 series, there may also be something else going on that isn't related only to temperature.

[dicky96]

Re: HD 7850 - Powers on, fan spins, no video....

Hi Momaka, it could also be the generic white heat sink compound i used to test. I have some of the grey Arctic MX-4 to try before I properly test the card

I'm just waiting for a Western Digital NVMe 1TB drive to turn up before completing the build, though it is running with a cheapo sata SSD right now

I also have a used Noctua U14S cooler to fit (with dual original Noctua fans) as I struggle to clock the i7-5820K above 4.2GHz on the existing Cooler Master Hyper 212 LED as it gets too hot and starts to throttle at 4.3GHz, again with generic white thermal paste.

[momaka]

Re: HD 7850 - Powers on, fan spins, no video....

Hi Momaka, it could also be the generic white heat sink compound i used to test.

Yeah, that's a possibility too... though it would have to be really low quality TC. I have some no-name cheap gray compound from eBay, and generally it gives no more than 2-3C difference from good compounds on larger surfaces (like CPU heatsinks or GPUs with heat spreaders.) On smaller stuff, it might do worse (depends on TDP per surface contact area), but I don't recall seeing it do worse than 5C, if even that much.

That said, yes, you should probably use something a bit better - at least for the GPU. In fact, I see when you removed the cooler on the video card, it had some gray compound. If that was MX-2/MX-4, that's well worth re-using again. Yes, you read that right - thermal compound is re-usable if not dried out and/or contaminated too much with foreign objects (like dust, lint, and etc.) I've been doing it for years with both cheap and good quality compounds like MX-2/4 and AS-5... and I've never noticed a difference in thermal performance from my thermal tests. Usually I just scoop up the excess from the sides (when the cooler is removed) and re-spread it back in the middle. I've also confirmed my test results with another member here on BCN (whose name I shall not mention, in case he feels ashamed of being cheap like me for re-using TC )

I also have a used Noctua U14S cooler to fit (with dual original Noctua fans) as I struggle to clock the i7-5820K above 4.2GHz on the existing Cooler Master Hyper 212 LED as it gets too hot and starts to throttle at 4.3GHz, again with generic white thermal paste.

Well, that i7-5820k is rated for 140 Watts "Intel" TDP... which basically means the CPU will use 140 Watts under "average" load. Thus, it is actually possible for the CPU do be dissipating even more when fully stressed (Intel LinPack test is a good example, if not a bit extreme.) And then you add your OC on top, and who knows where it's going now (though Kill-A-Watt or similar meter on the wall could roughly tell you between OC and non-OC state.) So no surprise to see the Hyper 212 struggle a bit. I'd say that cooler is good for 100-120W TDP, and 140 if running stock.

Of course, if that generic white TC is really bad and allowing up to 5C higher temps... shaving off that 5C with better TC would be helpful. After all, the i7-5820k is rated for only 66.8C T_case... so be careful with the temps, as that isn't a whole lot. At least on those socket 2011 CPUs they soldered the IHS to the core, so the thermal transfer is really good. But on the regular desktop chips, it's sad how much thermal performance is lost in TC from core to heatspreader - especially on 6th, 7th, and 8th gen chips.

On a different note, you might also want to check your fan curves.
For the GPU, since the temperature is going that high, crank the fans to 100% PWM and then run the same load tests several times to see how hot it gets then. If it runs much cooler, then probably the stock fan profile was optimized for quiet computing and not performance... which is what this card is (or was back in its day) all about. I usually set my fans to hit 100% PWM when the temperature reaches 60-65C. For this I use MSI Afterburner and create a custom fan profile, then test it under load. Once I settle on a curve I like, I might edit the BIOS on some cards and re-flash them with my "optimized" fan curve.

Most modern motherboards will also let you do the same for the CPU in the BIOS of the motherboard.
Back when I worked as a restore tech in one of the computer stores around here, they had these custom PCs with retail motherboards inside. The builds were generally pretty decent. But whoever set up the CPU fan curves in BIOS (or I suspect no one did and that was just the default setting) wasn't working very well. In that case, it was the opposite - CPU fan was running way too high speed in idle, to the point that these machines would come back after a year or two (just past warranty! ) with seized up CPU fans. So whenever I had to deal with one of these (restore a returned machine for resale), I always took the extra 2 minutes it took me to fix the CPU fan curves. The result was a machine that was much more quiet when idle, but could quickly ramp up fans under really heavy load and keep CPU temperatures well within spec.

[dicky96]

Re: HD 7850 - Powers on, fan spins, no video....

Well, that i7-5820k is rated for 140 Watts "Intel" TDP... which basically means the CPU will use 140 Watts under "average" load. Thus, it is actually possible for the CPU do be dissipating even more when fully stressed (Intel LinPack test is a good example, if not a bit extreme.)

Yeah I was stress testing it when I got it to throttle at 4.3GHz. I stopped it pretty quick. So on goes the MX-4 and Noctua U14S

The Hyper 212 is not useless though - that's destined for my i7-2600K OC where it should do a very good job

[piernov]

Re: HD 7850 - Powers on, fan spins, no video....

IIRC, aren't the earlier Xbox One and PS4 GPUs somewhat similar to the HD7770? Whichever the case, those seem to have been pretty reliable overall. Perhaps because they use smaller / less complex GPUs?

Xbox One and PS4 use an APU, not a discrete CPU+GPU. From experience AMD CPU very rarely fail compared to ATi/AMD GPUs. APU may fail a bit more often.
Additionally the reliability was getting a little bit better in 2013 with the 200 series from what I've seen.
That said, there's still a non-negligible number of failing APU in Xbox and especially PS4, which I'd attribute to regularly being stressed with insufficient cooling.

I also find it interesting, however, that cards from the HD6k era weren't nearly as failure-prone as the HD7k era, at least if kept well-cooled... or so has been my observation based on what I've been finding on the used market. That aside, I'm still putting nearly daily (ab)use on my repaired Gigabyte HD6850. Granted I do run a custom fan curve so that I've been good about keeping the GPU temps under 55C at all times, and also using frame-limiting in games to limit GPU usage to less than 70-80% usually (and typically down to 30-50% in Fornite, which is what I've been playing the most lately.)

In my opinion the HD 6000 series had the highest failure rate. This is especially true with mobile chips. For example all the MacBook Pro 15"/17" 2011 had failing HD 6000M, same goes for iMac 2011. But other manufacturers were affected too of course. However it might also be biased because the HD 7000 series was less popular (in laptops at least).
Reliability had been steadily decreasing starting from around the X1000 series, but it wasn't really concerning back then, really started being a problem with HD 4000-HD 7000.
Newer GPUs seem to be much more reliable, even with the 200-300 series it's only the higher end desktop ones that I've heard failing, mobile chips seem to be mostly fine. At the same time they've been used for less time so maybe we'll see them fail more in a few years. But the failure of the HD 6000 series showed up in less than a few years.

[Dan81]

Re: HD 7850 - Powers on, fan spins, no video....

IIRC the HD3400 series (and 3200 IGPs) weren't bright either. Half of the classic DV7s (the silver ones in the 1000 range) I've seen were dead, at least on the AMD side. Intels seem to survive, but those run nVidia chips.

[piernov]

Re: HD 7850 - Powers on, fan spins, no video....

The AMD RS780 and RS880 northbridges have a pretty high failure rate yes, they are from the same era. When I see one it's typically because it's dead. They often last at least a few years before failing though.

These HP laptops using Intel CPU typically have an Intel northbridge and southbridge. They optionally have an NVidia dGPU though, but the ones after bumpgate was fixed have a low enough failure rate.
The AMD ones of the previous generation, dv6000 and dv9000, used an NVidia northbridge affected by bumpgate so they were dying a lot as well (even more than the ones with AMD northbridges).