6900xt degrades below stock speeds within 4 Weeks?

[Franzen4Real]

I helped a friend diagnose a similar issue recently. It came down to the CableMods sleeved extensions being used on the 8pin cables to the graphics card. It worked perfectly fine for a while and slowly it began to black screen and crash, increasing in frequency over a few months. Not sure if when you said that two different power supplies were used, you had kept the same modular cables in the case and just swapped the psu only? Or if it was a full psu and cable swap. Just a thought if you're running out of ideas.

 

[Falkentyne]

it happens unfortunately, a small defect or a bad solder point, that passed QA because it was working and then gives out, or a cap or the cooler not making proper contact.

It's just a tragedy when stock is this bad. Nvidia's 2080ti's had a solid % that baked their Vram chips to death at launch.

The space invaders problem? Wasn't that caused by the RAM chip closest to the PCIE slot getting too hot and the solder BGA balls coming apart due to bad soldering in combination with temps?

 

[EsaT]

i must have missed that in the reviews, damn!

okay thats a furmark only thing

Gaming has 450W transients as well:

Is AMD getting the crown back? Radeon RX 6900 XT 16 GB review with benchmarks and a deeper technical analysis | Page 15 | igor´sLAB

AMD still has one left, and that’s what we have today. They want to take the competitive Radeon RX 6800 XT 16 GB one step further and so Dr. Lisa Su pulled a last ace out of her sleeve some time ago…

www.igorslab.de

And with non-reference cards those are propably higher.
Those big sudden load variations are stressfull to PSUs and loose connectors/daisy chained power cables etc can also cause problems with them.
(RTX 3080/3090 even worser)

 

[Aquinus]

All of my Intel and Nvidia products are 100% functional.

i always liked AMD for being the "Underdog" but since AMD sells it's products in a very very early pre alpha state and does ignore almost all issues that are out there... i can certainly say.. i am done with AMD.

No wonder the OP is pissed. He has an unrealistic expectation for how often hardware can fail. I've lost a lot of hardware over the years. Just because you're just experiencing this now doesn't mean it never happens or that it only happens with AMD chips. You don't see me complaining every time I see a nVidia card having issues because one time I had an XFX 8600 GTS which relied on a Molex to PCI-e adapter to provide +5v on one of the pins and will fry DDC on your display if you used a normal PCIe power cable, or how I had a GeForce 7900 GT that had a VRAM failure much like what you're describing.

Don't be so close minded. Shit happens.

 

[Nater]

^ no kidding. My first Ti 4200 was DOA. I swore off nVidia since then.

*edit*
Guys this was an absolute rollseyes extreme sarcasm statement, I thought I had an emoji at the end. I buy the best bang for the buck for my use case, I don't care what brand it is typically. Shit happens.

 

[Mussels]

I helped a friend diagnose a similar issue recently. It came down to the CableMods sleeved extensions being used on the 8pin cables to the graphics card. It worked perfectly fine for a while and slowly it began to black screen and crash, increasing in frequency over a few months. Not sure if when you said that two different power supplies were used, you had kept the same modular cables in the case and just swapped the psu only? Or if it was a full psu and cable swap. Just a thought if you're running out of ideas.

not a bad idea tbh, some external cables could totally cause some voltage drop

 

[delshay]

I've never known anyone with so many computer(s) or computer parts that have so many issues. May have to question the environment & does the OP where a anti-static strap. Gee so many questions.

 

[Redwoodz]

Simple fact is the OP moved the goal post several times.
Fact is you can damage your gpu running Furmark. OP tried to take it back but we all saw him state he ran it overnight. Also running Furmark for an hour on a 750w psu probabably subjected the card to some very large ripple. Most of his other "AMD" issues are solved in the latest Agesa from AMD. One of the concessions of using the same socket for several generations is it takes a lot more work to do updates. Add in the Corona effect and you have these relatively small delays. Most of the other AMD issues are from people overclocking their RAM improperly, though they don't think they're overclocking they just want to run their sticks rated XMP speed.

 

[jjnissanpatfan]

Here are (some) of my Cards

ALL of them are working 100% perfectly fine.

ALL of my AMD Hardware has some kind of issues except ONE single 5700XT and a 3800X.

i have a Ryzen 5 3600 on a B450 gaming plus Board that has massive problems with NVME Drives.
my 5700XT Nitro + SE stuttered and black screened so hard that i RMAd within 24 Hours.
my RMA Unit was a non SE and a 50€ Refund and it died after around 14 days of usage (VGA LED over night.. completely toast.)
my X570 Aorus Elite has massive problems with USB Ports.
my 5800X has non stop cache hierarchy errors.
my 5900X is not stable at stock speeds in P95 (spits out errors after 3-4 minutes)
my b550 Strix F has a dead SATA Port.
a RX 580 Armor that displayed nothing but some weird acid trip colors after 5 minutes of load.
and my second RX 580 8GB Pulse that still works fine (forgot to mention this card)

All of my Intel and Nvidia products are 100% functional.
two Z97, two Z370, one Z390, two Z490 and one Z590 board
4690, 4690k, 4770k, 4790,7700k,8600k, 8086k, 9600k, 9900k,10600k,10600KF,10700K,10850K and a 10900f
980 Ti, 660, 660 Ti, 760, 1050 Ti, two 1060s, two 1070s, two 1080 Tis, 1660ti, 2060, 2060 Super, 2070 Super, 2080 Super, 2080 Ti,two 3070s, two 3080s, a 3090.
even my old GeForce 4ti 4200 still works.

i always liked AMD for being the "Underdog" but since AMD sells it's products in a very very early pre alpha state and does ignore almost all issues that are out there... i can certainly say.. i am done with AMD.

View attachment 191063View attachment 191064View attachment 191062

And your not biased to AMD...Hmmmmm. I think you should get your RMA and buy NVidia. The world is happy again. I read through all the posts and think this is an overreaction and biased. Sorry for adding my opinion but just you got a bad card...it happenens.

 

[Mussels]

Doing a thread clean up, help the guy or keep quiet

edit: issues with cleanup, behave anyway

 

[Aquinus]

So let's be candid here. So just for a comparison, my Vega 64 when power limit is raised to 330w with my 3930k at 4.5Ghz could pull over 800w off the wall with a 1000w Seasonic 80 PLUS Platinum PSU. The OP is using an arguably more power hungry GPU with a power hungry CPU, but with a 750w PSU. It doesn't take a rocket scientist to put two and two together. I think the GPU was exposed to a lot of ripple off the PSU and the VRMs couldn't completely suppress it and it damaged the VRAM. The OP probably would have been fine if everything was left stock, but it turned into a ticking time bomb once you start pushing it.

 

[Hachi_Roku256563]

e

Here are (some) of my Cards

ALL of them are working 100% perfectly fine.

ALL of my AMD Hardware has some kind of issues except ONE single 5700XT and a 3800X.

i have a Ryzen 5 3600 on a B450 gaming plus Board that has massive problems with NVME Drives.
my 5700XT Nitro + SE stuttered and black screened so hard that i RMAd within 24 Hours.
my RMA Unit was a non SE and a 50€ Refund and it died after around 14 days of usage (VGA LED over night.. completely toast.)
my X570 Aorus Elite has massive problems with USB Ports.
my 5800X has non stop cache hierarchy errors.
my 5900X is not stable at stock speeds in P95 (spits out errors after 3-4 minutes)
my b550 Strix F has a dead SATA Port.
a RX 580 Armor that displayed nothing but some weird acid trip colors after 5 minutes of load.
and my second RX 580 8GB Pulse that still works fine (forgot to mention this card)

All of my Intel and Nvidia products are 100% functional.
two Z97, two Z370, one Z390, two Z490 and one Z590 board
4690, 4690k, 4770k, 4790,7700k,8600k, 8086k, 9600k, 9900k,10600k,10600KF,10700K,10850K and a 10900f
980 Ti, 660, 660 Ti, 760, 1050 Ti, two 1060s, two 1070s, two 1080 Tis, 1660ti, 2060, 2060 Super, 2070 Super, 2080 Super, 2080 Ti,two 3070s, two 3080s, a 3090.
even my old GeForce 4ti 4200 still works.

i always liked AMD for being the "Underdog" but since AMD sells it's products in a very very early pre alpha state and does ignore almost all issues that are out there... i can certainly say.. i am done with AMD.

View attachment 191063View attachment 191064View attachment 191062

either bluffs or bad luck
ive been using amd for a long time and have recived one dud part from them
odly enough ive had 4 dud intel cpus but i still use intel
Northwood pentium 4
Core 2 quad q6700
I5 4700
I3 8130
2 dead nvidea cards i stil use nvidea
a gts 450
and a 1060 6g
and 1 3200g that killed itself right before its warrenty was up and i still use amd
very good timing
My point is dont hate a company cause you got some bad parts
they all have problems and like someone said all your amd problems are recently patched

Edit no 7billion 5 thousand and 42
You said you ran furmark overnight
thats amost certainly what killed the card
im almost certain you did that cause you said
"it could run furmark overnight"
no one says it can run x benchmark WIThOUt RuNnInG tHE BeNChmArK
My advice get a new card and dont run furmark i promise it will last longer

 

[delshay]

I think my last posting in this thread say's it all, so here's my question to the OP.

Did you check ambient room temperature before & after? This plays a big role in the card behaviour,, ie, throttling at high temperature.

It seems you did an overnight test in the evening or at night. Isn't it not colder where you are at this time?

 

[Mussels]

So let's be candid here. So just for a comparison, my Vega 64 when power limit is raised to 330w with my 3930k at 4.5Ghz could pull over 800w off the wall with a 1000w Seasonic 80 PLUS Platinum PSU. The OP is using an arguably more power hungry GPU with a power hungry CPU, but with a 750w PSU. It doesn't take a rocket scientist to put two and two together. I think the GPU was exposed to a lot of ripple off the PSU and the VRMs couldn't completely suppress it and it damaged the VRAM. The OP probably would have been fine if everything was left stock, but it turned into a ticking time bomb once you start pushing it.

if you push over the limits, OCP kicks in and the PC shuts off. that almost never results in any damage.

 

[Space Lynx]

I just ran 3dmark tests last night to compare to when i first got this card (my rx 6800) on launch day. and no noticeable changes in scores. not sure what your problem is, prob best to use warranty and RMA it

 

[watzupken]

Yup, whenever something is wrong with an AMD product (which is quite often considering all the threads on forums), there's always a legion of defenders suggesting (if not outright claiming) it's something (someone) else's fault; motherboard, bios, OC settings, ram, PSU...you name it. And while I certainly agree that all these things should be accounted for and ruled out, the deflectors often stubbornly stick with their claims, even in the face of overwhelming evidence to the contrary.

Oh and regarding what to do - I honestly suggest RMA-ing the card then either handling its replacement like a spoiled baby (low clocks, low voltage) or selling it on while they are still going for good money and putting the earnings towards a 3090.

I don't get this. The card in this case was heavily overclocked and way out of its specs. Sure there is a limit set in the BIOS, but for a RX 6900XT, that's 3Ghz. Even at 2.7 Ghz, is that a realistic clock with air cooling? I've been using my RX 6800 XT for 3 months now clocked at 2.5 Ghz and it still works fine in any games. So before we blame the product, I think it makes sense to understand who is at fault here. If this was faulty even when using at stock speed all the while, then its clear that its a faulty product. By the way when Nvidia's RTX 3090/ 3080 were crashing hard at launch, didn't people start pointing out the PSU as the issue as well?

 

[Aquinus]

if you push over the limits, OCP kicks in and the PC shuts off. that almost never results in any damage.

OCP isn't going to protect you from voltage spikes, only excessive current. A lot of times when a PSU is pushed to the max, it's the voltage you need to be careful of because typically the voltage spike is accompanied by a drop in current. There is also lingering questions as to how "dirty" the signal has to become before over voltage protection kicks in. Transients might not be detected well by the voltage and current sensors on the card.

Either way, the biggest issue I see is a PSU too small for the load being demanded of it. I don't think there should be any dispute there.

 

[LFaWolf]

he oced it and proceded to push it really hard
maybe even furmark
he does not dserve a rma

e

either bluffs or bad luck
ive been using amd for a long time and have recived one dud part from them
odly enough ive had 4 dud intel cpus but i still use intel
Northwood pentium 4
Core 2 quad q6700
I5 4700
I3 8130
2 dead nvidea cards i stil use nvidea
a gts 450
and a 1060 6g
and 1 3200g that killed itself right before its warrenty was up and i still use amd
very good timing
My point is dont hate a company cause you got some bad parts
they all have problems and like someone said all your amd problems are recently patched

Edit no 7billion 5 thousand and 42
You said you ran furmark overnight
thats amost certainly what killed the card
im almost certain you did that cause you said
"it could run furmark overnight"
no one says it can run x benchmark WIThOUt RuNnInG tHE BeNChmArK
My advice get a new card and dont run furmark i promise it will last longer

Ah, now that makes sense why you think overclocking would kill hardware components. If you pumped enough voltage, you can kill any component. Judging by your list of dead hardware, I can see that is probably what you did. You can overclock, but you still gotta do it within the confine of the hardware limits. How do you even kill locked CPUs???

I have had only 1 DOA CPU and it was an Intel 7740X. I had 1 DOA GPU, an EVGA GTX 1080. I have killed a GTS 450 when I re-pasted and inadvertently used a conductive paste. I volt-modded an AMD 6950 and that ended up catching fire Otherwise, I have overclocked and have not killed any CPU so far. And I have probably built 50+ machines.

Don't pump too much voltage. Live and learn.

 

[Hachi_Roku256563]

Ah, now that makes sense why you think overclocking would kill hardware components. If you pumped enough voltage, you can kill any component. Judging by your list of dead hardware, I can see that is probably what you did. You can overclock, but you still gotta do it within the confine of the hardware limits. How do you even kill locked CPUs???

I have had only 1 DOA CPU and it was an Intel 7740X. I had 1 DOA GPU, an EVGA GTX 1080. I have killed a GTS 450 when I re-pasted and inadvertently used a conductive paste. I volt-modded an AMD 6950 and that ended up catching fire Otherwise, I have overclocked and have not killed any CPU so far. And I have probably built 50+ machines.

Don't pump too much voltage. Live and learn.

None of this had to do with oc I don't overclock these were either DOA or had something faulty

 

[LFaWolf]

None of this had to do with oc I don't overclock these were either DOA or had something faulty

Sure, I believe you You must be one really unlucky builder.

 

[Hachi_Roku256563]

Sure, I believe you You must be one really unlucky builder.

Pentium 4 DOA
Core 2 duo refused to run at any sort of decent speed
I5 dead men controller
I3 DOA
GTS 450 would not go into 3d mode
1060 vramm died in a month
And 3200g had some broken controller that was effecting USB and mem

 

[Aquinus]

Pentium 4 DOA
Core 2 duo refused to run at any sort of decent speed
I5 dead men controller
I3 DOA
GTS 450 would not go into 3d mode
1060 vramm died in a month
And 3200g had some broken controller that was effecting USB and mem

You're incredibly unlucky then. I've never had a CPU fail on me from a P4 Northwood Celeron to today. GPUs are a totally different story though. To each their own I guess.

 

[Hachi_Roku256563]

You're incredibly unlucky then. I've never had a CPU fail on me from a P4 Northwood Celeron to today. GPUs are a totally different story though. To each their own I guess.

But the point still stands i dont hate intel because of the doa or failed parts

 

[Chrispy_]

Does this thread of salty tears exist because someone's 450MHz overclock is now unstable and he's only getting 440MHz overclock? 2700MHz is a 20% overclock of what is considered by most reviewers to already be uncomfortably-clocked silicon operating well beyond it's sweet spot.

Perhaps it's time someone read about electromigration; I have no idea what voltages were required for that behemoth overclock but I doubt they're good for infinite-lifespan on the silicon. Degredation rates and instability increase exponentially with voltage.

 

[Deleted member 205776]

Fanboyism is just so toxic. No manufacturer is perfect. There's bound to be faulty units.

And there's bound to be degradation, ESPECIALLY when you do a gigantic OC like this.