AMD RX 6800 RT cores defective or PSU problem?

[Marcus L]

He's not wrong, you are. But you've already made it clear you have no intention of replacing your power supply, which means you should sell your RX 6800 to someone who can actually use it and install something more appropriate, like an RX 6400 or a GT 1030. I'm outta here.



It literally IS the problem. Take the time to understand what a load line is and what are the effects of transient spikes, ripple, etc. and you'll quickly understand that PSU is actually hazardous to your system.

No I haven't made it clear I won't be replacing the PSU obviously you have reading comprehension as I stated I am planning to that in the op, I'm not some kid who threw together his first pc so honestly, see yourself out of the thread cause your coming off as arrogant now, thanks for the input, I'll. Disregard that and anything else you have to say, take your attitude elsewhere

Ok

You are welcome to bash your head against a brick wall when presented with an obvious solution, sure. You came with an issue seeking advice. In the very first posts you were advised sourcing a better PSU, if even temporarily, and testing with that. You decided that you don’t like this option. But you not liking it won’t change the fact that it IS the most obvious and effective solution that can be provided given the information you gave.
But suit yourself.

Thank you, off you go as well, was nice, let's do it again sometime

 

[Vya Domus]

When you say crash what does that mean, driver timeout, blue screen, black screen, restart, shutdown ?

Insufficient power/bad PSU will almost always manifest itself in the form of hard shutdowns. It's really easy to rule this out, a crappy PSU wont give you a driver timeout or bsod, it will instantly crash the system = shutdown.

 

[evernessince]

So I bought a 6800 used from eBay in November, it works fine in raster but will crash in most games when I enable ray tracing, is it possible the ray tracing cores are defective?
My PSU is not up to scratch for this GPU it is an EVGA 600 W1 80+ which I knew at the time was likely not going to be good enough to handle the transients of the 6800 though thought with an underclock and undervolt it would be fine, but it never crashes when I am gaming using just rasterization, only when RT is enabled, games that exhibit this behaviour include but are not limited to fortnite, Doom eternal, call of duty MW, and some others, ray tracing will work in the odd game without crashing though they are few and far between and I can't remember off the top of my head which ones, maybe resident evil 4 remake? ressident evil 7 also crashes when RT is enabled. I have tried different drivers to no avail.

So crashes are exhibited only when RT is enabled but not in all games where RT is enabled? If you are thinking that there's some defective cores on the GPU (in this case perhaps RT accelerators), the behavior would need to be consistent across all games with RT enabled. That it isn't suggests the issue lay elsewhere.

You seem to already be aware that your PSU is a potential culprit. Aside from rectifying that potential issue, you can try stress testing your GPU via furmark, reslotting the GPU, reseating the power connectors (and make sure you are not daisy chaining multiple connectors from the same cable), trying the GPU in a different slot, swapping the monitor cable and port (could be the esoteric pin 20 issue). I assume you did a clean driver install but if not that's a definite thing to check off the list.

Also, have you ever tested your CPU / Memory? I know in a lot of cases the only demanding applications people tend to use are games which means they often only find out about computer issues in games and automatically assume it's the GPU when in reality they've just never bothered to stress their system before to ensure stability. Ray Tracing just so happens to put more strain not only on the GPU but also the CPU and it could be simply bringing out instability in those components. Checking the CPU / memory is something I would check after ticking off the GPU troubleshooting tips above first though as it's less likely.

 

[Vya Domus]

Furmark is not a good stress test anymore, just load something like fortnite with max settings and an unlocked framerate, switch between RT on and off, we're overcomplicating this to a nonsensical degree.

If your PSU really is the issue it should crash either way and by crash I don't mean driver timeout/bsod, those things are not indicative at all of a PSU issue. I have a 7000 series AMD card, it draws the same amount of power whether or not RT is enabled. If anything RT on should put less stress on the PSU not more, lower framerate = longer time delta in between transient spikes.

 

[evernessince]

Furmark is not a good stress test anymore, just load something like fortnite with max settings and an unlocked framerate, switch between RT on and off, we're overcomplicating this to a nonsensical degree.

Both Northwestrepair and Northbridgefix's (the two most prominent repair shops that publish GPU repair videos) use furmark because it pushes the shader cores hard without running into a CPU bottleneck. You could certainly get a better assessment of system stability by testing with additional software (same as any stress testing) but to say it's not good at stress testing is categorically incorrect.

Fortnite is a CPU / Memory bound game meaning there are many scenarios in which it won't be in fact stressing the GPU. Enabling RT will utilize the RT accelerators on AMD cards but it won't suddenly drastically increase the rasterization load that would be needed to fully stress a card.

OP should use a more demanding game or stress test before fortnite, something that isn't CPU / Memory bound like fortnite is.

If your PSU really is the issue it should crash either way and by crash I don't mean driver timeout/bsod, those things are not indicative at all of a PSU issue.

For the most part yes but do note that a failing power supply can manefest in several different ways. I've repaired systems wherein a failing power supply resulted in very poor performance when entering 3D clocks but was completely fine under 2D clocks. It didn't crash the system either.

I have a 7000 series AMD card, it draws the same amount of power whether or not RT is enabled. If anything RT on should put less stress on the PSU not more, lower framerate = longer time delta in between transient spikes.

What you are describing is a bottleneck in your system or application. If a GPU is using RT cores in one instance and not in another, the additional utilization should cause an increase in power consumption assuming RT isn't introducing a bottleneck. That your system is using less indicates an obvious bottleneck. Here's TPU's 4090 chart comparing RT on vs Off power consumption:

In a non-bottlenecked scenario RT on will increase power consumption and the associated transients that come with that.

 

[Toothless]

6800XT DOES have spikes that are big enough to knock out power supplies. I had one.

An EVGA 850w Gold rated unit couldn't keep up and if that doesn't tell you anything then I don't know if you're listening. Get a quality unit (Enermax DF, Corsair RMx/HX, etc)

RDNA2/3 have horrid spikes. This is nothing new. Your current unit is not quality.

 

[Vya Domus]

Fortnite is a CPU / Memory bound game

It's a heavily GPU bound game with all the settings maxed out, just play the game and convince yourself of this if you don't believe. It's a perfect game for OP's issue because it's heavy on raster and you can turn on both software and hardware RT.

Here's TPU's 4090 chart comparing RT on vs Off power consumption:

You don't understand how this works and you also picked a terrible card to prove your point because using a 4090 in raster is going to almost certainly cause a CPU bound situation, when RT is enabled of course it will use more power, most Nvidia GPUs use less power when RT is on not more.

Notice how the 4090 is the only card that's using more power, the rest all use less and with AMD cards it's basically identical. I am 100% correct on my claim that there should be no real difference in OPs case as far as power consumption is concerned, that's why I am adamant this isn't a PSU issue, if it was the system would always crash no matter what.

6800XT DOES have spikes that are big enough to knock out power supplies. I had one.

An EVGA 850w Gold rated unit couldn't keep up and if that doesn't tell you anything then I don't know if you're listening.

This is a 6800 non XT. But what you are saying is not possible, for transient spikes to be a problem they must exceed what the PSU can supply on the 12V rail, meaning your card had 850W+ spikes ? No way, the 6900XT had 600W spikes and that was by far the worst of the bunch, OPs card probably barely hits 350W. Your unit was just bad, it wasn't a transient spike issue.

 

[Macro Device]

6800XT DOES have spikes that are big enough to knock out power supplies. I had one.

However, the OP has an RX 6800 non-XT. This one is way less cruel on PSUs.

But I agree, such a low quality unit must not be enough for this card.

Here's TPU's 4090 chart comparing RT on vs Off power consumption:

AMD GPUs operate differently. Lemme show you.

Spoiler: 1440p High FSR Quality RT disabled

Spoiler: 1440p High FSR Quality RT Reflections only

Spoiler: 1440p High FSR Quality RT Ultra

 

[Toothless]

It's a heavily GPU bound game with all the settings maxed out, just play the game and convince yourself of this if you don't believe. It's a perfect game for OP's issue because it's heavy on raster and you can turn on both software and hardware RT.


You don't understand how this works and you also picked a terrible card to prove your point because using a 4090 in raster is going to almost certainly cause a CPU bound situation, when RT is enabled of course it will use more power, most Nvidia GPUs use less power when RT is on not more.

View attachment 332227View attachment 332228

Notice how the 4090 is the only card that's using more power, the rest all use less and with AMD cards it's basically identical. I am 100% correct on my claim that there should be no real difference in OPs case as far as power consumption is concerned, that's why I am adamant this isn't a PSU issue, if it was the system would always crash no matter what.


This is a 6800 non XT. But what you are saying is not possible, for transient spikes to be a problem they must exceed what the PSU can supply on the 12V rail, meaning your card had 850W+ spikes ? No way, the 6900XT had 600W spikes and that was by far the worst of the bunch, OPs card probably barely hits 350W. Your unit was just bad, it wasn't a transient spike issue.

However, the OP has an RX 6800 non-XT. This one is way less cruel on PSUs.

But I agree, such a low quality unit must not be enough for this card.

AMD GPUs operate differently. Lemme show you.

Spoiler: 1440p High FSR Quality RT disabled

View attachment 332229

Spoiler: 1440p High FSR Quality RT Reflections only

View attachment 332231

Spoiler: 1440p High FSR Quality RT Ultra

View attachment 332233

I missed the non-XT part, but point still half stands with the 6800. (I had one of those too)

Spike does not have to go over what the power supply is rated. I've mentioned in other threads that my 1300w could not keep up with my XTX where my better built 850w can. Quality>quantity.

Don't know why people keep arguing when I've had multiple RDNA cards and literally tested them myself, where the data lines up with other people in the world.

 

[Vya Domus]

Spike does not have to go over what the power supply is rated.

They have to if that was in fact the issue. That's why this was a problem, spikes would go over what the PSU was rated on the 12V rail and it would trigger over current protection.

Don't know why people keep arguing

Because people mix things up, you seem to think that every PSU issue = spikes where the problem. But that's just not true, if you had a 1300W unit that would fail it wasn't because of the spikes, it's just not possible for that to have been the case, it probably just had an overzealous/defective OCP protection circuit. How else would a 1300W unit be unable to handle spikes that aren't even half of what it is rated for, even if it was straight up junk that would be really unlikely to happen.

 

[Ripcord]

looks like a faulty card to me

 

[Toothless]

They have to if that was in fact the issue. That's why this was a problem, spikes would go over what the PSU was rated on the 12V rail and it would trigger over current protection.


Because people mix things up, you seem to think that every PSU issue = spikes where the problem. But that's just not true, if you had a 1300W unit that would fail it wasn't because of the spikes, it's just not possible for that to have been the case, it probably just had an overzealous/defective OCP protection circuit. How else would a 1300W unit be unable to handle 600W spikes, even if it was straight up junk that would be really unlikely to happen.

Or the power supply trips on those spikes which sure, the OCP could be tripping but this isn't the case on better built units. With OP's issue it's most likely junk unit giving junk power however there is no harm in getting a quality unit in case they want to upgrade down the road.

With plenty of data out there where many low end units, and some decent ones still not able to hold up.

 

[neatfeatguy]

No I haven't made it clear I won't be replacing the PSU obviously you have reading comprehension as I stated I am planning to that in the op, I'm not some kid who threw together his first pc so honestly, see yourself out of the thread cause your coming off as arrogant now, thanks for the input, I'll. Disregard that and anything else you have to say, take your attitude elsewhere

Ok

Thank you, off you go as well, was nice, let's do it again sometime

I always love having a spare, good working PSU for troubleshooting. I know it sounds like a waste of money, but it is something I would strongly recommend to everyone to have a backup PSU. And not just some piddly piece of crap one, but one that can easily power your system.

About 8 months back there was a massive power outage in my neighborhood. Huge explosion and the power was out for a good square mile radius or more. One of my systems wouldn't power back on. Lots of troubleshooting and then I eventually tested my spare, good PSU and things turned back on. I've since picked up another spare PSU to have for backup. Before this issue, some years back I had a MB die on me. Extensive testing with multiple PSUs allowed me to confirm the MB was indeed dead.

I would suggest you don't be so condescending to those that suggest a PSU issue. You came here asking for help, they're suggesting what they strongly believe to be the issue. If you wish to disagree with them, that's one thing, but how you've worded your reply is childish.

Here's what I would do in your situation if I thought it wasn't the PSU.

1) Clean install of the GPU drivers.
2) Try the two previous WHQL drivers for your GPU if a clean install of the most recent one doesn't fix the issue.
3) Test the GPU in another system if you can - if you are unable, move on to step 4
4) If trying multiple drivers doesn't work and/or testing the GPU in another system. I'd now suggest doing a clean install of Windows (be sure to backup any important data first) or if you have a spare drive you could do a clean install on, that would work just as well. Sometimes Windows just sucks because MS breaks things with the awful updates that get pushed out.
5) If things still are not resolved at this point, then I would be left with one of two things:
a) GPU is faulty
b) PSU is faulty
- If I got this point I'd try a new, quality PSU first because if I need to buy a part a quality PSU is going to be 1-2 hundred dollars cheaper than a comparable GPU to the 6800. In the end, if the PSU isn't the problem, at least you have a spare PSU for troubleshooting/testing should the need ever arise.

 

[Macro Device]

a crappy PSU wont give you a driver timeout or bsod

Why did my junkyard PC have no instability whatsoever with a proper PSU and often crashed with driver errors in games if was fueled by a faulty PSU?

RT on should put less stress on the PSU not more

Not always. RT is also a CPU-intensive task so if you max your GPU out either way you could get more load overall with RT on thanks to more intensely loaded CPU.

 

[evernessince]

Notice how the 4090 is the only card that's using more power, the rest all use less and with AMD cards it's basically identical. I am 100% correct on my claim that there should be no real difference in OPs case as far as power consumption is concerned, that's why I am adamant this isn't a PSU issue

You are correct that most GPUs don't consume more power with RT enabled. That said I don't see how you can eliminate the PSU merely based on the fact that his card happens to not use more power with RT enabled. You don't know the settings he's using in games he plays and by extension GPU load those games are placing on his system. In addition, transients exceeding total PSU wattage don't have to be the source of any potential PSU issue if that is considered a potential root cause. It could just be poor voltage regulation or a defective unit which you recognize in your following comment.

Because people mix things up, you seem to think that every PSU issue = spikes where the problem. But that's just not true

Of course I'm not saying this is exclusively a PSU issue, there's a great lack of information provided and troubleshooting steps taken to come to any kind of determination. The OP really needs to go through some of those before anything can be eliminated as a potential variable. Ultimately our goal is the same, to help the OP. You just disgree on method. Honestly I don't care what software OP uses to stress test so long as it's properly stressing their system. That's the entire reason I recommend official tests like Furmark, Unigen, and futuremark as there's a greater degree of certainy that you are fulling stressing the component being tested. You complained that was overly complicating things but I'd argue running a test designed to stress your GPU is simpler. Less variables to consider as compared to games, especially ones like fortnites with ever changing bugs and online components. Certain games like CP2077 still have trouble setting the correct options to this day.

Or the power supply trips on those spikes which sure, the OCP could be tripping but this isn't the case on better built units. With OP's issue it's most likely junk unit giving junk power however there is no harm in getting a quality unit in case they want to upgrade down the road.

With plenty of data out there where many low end units, and some decent ones still not able to hold up.

Heck even on higher quality units you can have some odd things trigger saftey mechanisms. Good example is the Seasonic Prime units where Nvidia 3000 series GPUs were causing noise to be fed back to the PSU resulting in protections tripping (which was fixed on the 4000 series and by Seasonic providing replacement cables). This is why it's best to rule things out through testing instead of assuming based on the wattage of the unit.

 

[lawood]

When you say crash what does that mean, driver timeout, blue screen, black screen, restart, shutdown ?

Insufficient power/bad PSU will almost always manifest itself in the form of hard shutdowns. It's really easy to rule this out, a crappy PSU wont give you a driver timeout or bsod, it will instantly crash the system = shutdown.

I had a bad psu in the past that crashed nvidia driver easily under load.

Never happened again after I upgraded.

Btw, this is 1000% psu, that unit is literal dogshit.

 

[Athlonite]

it may have been used for mining so a good cleaning and repasting may be needed and new PSU that GPU can pull 225W going full noise atleast mine does just playing starfield at 1080p never bothered with RT on it as it sucks fps to badly to use it

 

[ShrimpBrime]

So I bought a 6800 used from eBay in November, it works fine in raster but will crash in most games when I enable ray tracing, is it possible the ray tracing cores are defective?
My PSU is not up to scratch for this GPU it is an EVGA 600 W1 80+ which I knew at the time was likely not going to be good enough to handle the transients of the 6800 though thought with an underclock and undervolt it would be fine, but it never crashes when I am gaming using just rasterization, only when RT is enabled, games that exhibit this behaviour include but are not limited to fortnite, Doom eternal, call of duty MW, and some others, ray tracing will work in the odd game without crashing though they are few and far between and I can't remember off the top of my head which ones, maybe resident evil 4 remake? ressident evil 7 also crashes when RT is enabled. I have tried different drivers to no avail.

Hello! I did read most of the thread, but the diagnostic techniques here are limited to information..... that said,

What are settings for your set up?
And what have you tried so we don't suggest a bunch you've done already.

Ideas, depending on some variables.

1. Try the card in a different rig
2. Set your current system to defaults.
3. Enable/disabled resize the bar
4. Limit the RT settings in the games
5. DDU completely, then try the drivers. Remove amd/nv/Intel
6. Fresh OS on a different drive
7. Monitor with HWInfo64, create logs.
8. Monitor with GPU-Z. Logs.
9. Different display port/cable/monitor
10. Freddy's coming for you.....

That's just a few things we could suggest I suppose. But I'm going to guess you've tried all but #10.

GL!!

 

[AleXXX666]

So I bought a 6800 used from eBay in November, it works fine in raster but will crash in most games when I enable ray tracing, is it possible the ray tracing cores are defective?
My PSU is not up to scratch for this GPU it is an EVGA 600 W1 80+ which I knew at the time was likely not going to be good enough to handle the transients of the 6800 though thought with an underclock and undervolt it would be fine, but it never crashes when I am gaming using just rasterization, only when RT is enabled, games that exhibit this behaviour include but are not limited to fortnite, Doom eternal, call of duty MW, and some others, ray tracing will work in the odd game without crashing though they are few and far between and I can't remember off the top of my head which ones, maybe resident evil 4 remake? ressident evil 7 also crashes when RT is enabled. I have tried different drivers to no avail.

use 7000 AMD or Nvidia for Rt....

it may have been used for mining so a good cleaning and repasting may be needed and new PSU that GPU can pull 225W going full noise atleast mine does just playing starfield at 1080p never bothered with RT on it as it sucks fps to badly to use it

225W is ok with good 550W PSU, it's not such a "big consumption"

 

[Toothless]

225W is ok with good 550W PSU, it's not such a "big consumption"

As long as the power supply is of quality, yes.

 

[Marcus L]

I'll be getting a PSU, personally I don't think it is the PSU, though we shall see, it was on my list of things to do anyway

 

[Toothless]

I'll be getting a PSU, personally I don't think it is the PSU, though we shall see, it was on my list of things to do anyway

Make sure it's a good one so we don't repeat the thread.

 

[bug]

So I bought a 6800 used from eBay in November, it works fine in raster but will crash in most games when I enable ray tracing, is it possible the ray tracing cores are defective?
My PSU is not up to scratch for this GPU it is an EVGA 600 W1 80+ which I knew at the time was likely not going to be good enough to handle the transients of the 6800 though thought with an underclock and undervolt it would be fine, but it never crashes when I am gaming using just rasterization, only when RT is enabled, games that exhibit this behaviour include but are not limited to fortnite, Doom eternal, call of duty MW, and some others, ray tracing will work in the odd game without crashing though they are few and far between and I can't remember off the top of my head which ones, maybe resident evil 4 remake? ressident evil 7 also crashes when RT is enabled. I have tried different drivers to no avail.

You can try those games at really low resolutions (720p or even lower), v-sync enabled. If they still crash (the power draw should be low at those resolutions), it's the GPU.

 

[lawood]

Power draw doesnt really matter with a bad PSU, the spikes will still be there and it will end up tripping the PSU or the driver anyway.

RDN2 and ampere demand good PSU's.

I had nvidia driver timeouts consistently when i first got the 3070 ti, I had a GOOD PSU, not the best, but it was seasonic branded, silver rating, 650 watts, solid all around but not even then was enough, even if I was using just 50% power limit. Sometimes i'd get driver timeouts, other straight up restarts..

I legit thought the card was cooked, but got a new PSU first (Seasonic GX 750) and never had a single problem since.

The moral of the story is, get a good PSU. Bad PSU's can manifest problems in many many ways, not worth the headache at all.

 

[Marcus L]

I have ran multiple GPU's on said PSU just fine, it's also curious it only happens with RT on even with much lower load, ran a 2080 super no problems also, but I am replacing it, I just don't think it is the issue, I may stand to be corrected however