Sapphire Pulse RX 580 4GB black screens when in PCIE x8 3.0 mode!

[emothxughts]

Specs in the drop down menu, I'm using a Sapphire Pulse RX 550 2GB now, for obvious reasons.

I have an old Sapphire Pulse RX 580 4GB that I bought used, that I used while waiting for my Gigabyte RX 5600 XT RMA turnaround (now my seller is saying it'll take 1 month!), and it ran my games almost perfectly fine and crash-free, were it not for the fact that it topped out at PCIE x2 3.0 mode, as checked with GPU-Z. Seeing that this is way lower than the x16 3.0 that it's supposed to be, I cleaned the GPU contacts with isopropyl alcohol and the PCIE slot (only one GPU slot on my MB) with compressed air. At first, when I checked GPU-Z, it still tops out at only x4 3.0, so I did the process again, and behold, it went all the way to x8 3.0. Not quite x16, but acceptable for me.

What's absolutely not acceptable however is when I actually try playing a game with it. On The Witcher 3, I black screened immediately after starting dialogue with the Dandelion guy, black screen as in the monitor loses signal outright, though I still can hear the game music. WinKey+Ctrl+Shift+B did nothing, so I had to cold restart my PC. Note that the fans don't spin at 100% speed, at most it'll spin at the same speed before it crashed, at times it stops outright. After it restarted, I first checked the Event Viewer, where I found only Event 41 Kernel-Powers and "The previous system shutdown at...was unexpected", nothing specifically about graphics drivers. Afterwards I opened GPU-Z and tried the render test thing. At first, it reads that my RX 580 is running at PCIE x8 1.1. No problem, that's just how GPUs work in idle. I started the render test, bus interface now says PCIE x8 3.0, then almost immediately I get a black screen. I reseated the card, unscrewed it from the back of the case, reclean the contacts and the slot, it still black screens immediately after reaching the PCIE x8 3.0 state.

After a few cold restarts, the card ended up running on the MS basic display drivers. I took this opportunity to uninstall graphics drivers with DDU and then reinstalling the drivers. Then as soon as the part of the driver installation where the display flickers comes, the card...black screens again, the display doesn't go back on. It was at this moment that I decided to replace the RX 580 with my spare RX 550 (I bought it new, somehow). The RX 550 card runs flawlessly, though of course it isn't the most powerful GPU. It tops out at PCIE x8 3.0, and it can run x8 3.0 perfectly fine with no black screens whatsoever. Proof in the picture posted.

So now the question is, what is wrong with the RX 580 that it black screens in x8 3.0? Is it the card or something else in the system? What can I do to get it working again?

Also another problem with my PC, which may or may not be related to my RX 580 situation. My PC sometimes fails to wake up properly from hibernation, instead starting up as if it was fully shutdown. Event viewer says "Windows failed to resume from hibernate with error status 0xC0000411."

 

[emothxughts]

Update: 8 Jan 2023:
The shop sent me a video of the card running Furmark for 50+ minutes, plugged into the replacement HDMI port (the one that was previously physically broken), temps at 74 degrees C. They are now shipping the card back to me, after I paid the fees of course.

Replies:

• As for why I can't use DP at home, it's because my monitor LG 24MP400 has no DP, only a D-Sub and an HDMI port. Might consider getting myself a DP to HDMI adapter, in case the HDMI port(s) act up again.
• When the card & my display black screens (cut off signals), I can still hear sounds from my game and everything else in my PC still runs, lights up and spin, except of course the card, whose fans then stop. The PC does not restart by itself, I have to long press the power button. If this is indeed a PSU problem (it's a Cooler Master MWE Bronze V2 650 W 80+ Bronze, 230V non fullrange, non-modular), could it be the PCIe cable that's at fault? Looked up my PSU on the tier list, it's tier C apparently...

 

[Mussels]

My first comments in this thread were all way off target since I thought this was issues at booting, not while the system was running/gaming and we didnt have the system specs tehn


Nothing in your last post says ANYTHING is wrong with the HDMI port at all - just because the video card turns off doesnt mean its the problem, as the video card relies on everything prior to it working fine as well
Your PC itself can cause the GPU driver to crash from many many causes, from an unstable CPU or RAM overclock, bad power/power cables, PCI-E extensions and risers, and so on.

Don't focus on the HDMI port because it's not the problem

Please post a screenshot from Zentimings
ZenTimings (v1.2.9) Download | TechPowerUp


Black screen crashes/freezes can be related to general instability issues on ryzen, especially the 3000 series - and the code in your specs for your ram shows up as DDR4 2400, but you say its at 2133

The SoC controller on ryzen controls your PCI-E express lanes. It also controls your USB ports and SATA ports, and your memory controller. Issues on one, affects the others.

 

[emothxughts]

Here's the Zentimings screenshot:

The SoC controller on ryzen controls your PCI-E express lanes. It also controls your USB ports and SATA ports, and your memory controller. Issues on one, affects the others.

Now that you mentioned it, the RX 550 card I'm currently running right now occasionally drops to PCIE x4 1.1-3.0 according to GPU-Z's Bus Interface readings, when the card itself reports supporting up to PCIE x8 3.0. (My MB's GPU slot goes up to x16 3.0, it's just the card that's limited to x8). Any relations to the SoC? As well as the RX 580 that's the subject of this topic dropping to x8 from x16?

 

[Mussels]

PCI-E should never drop the number of lanes but it will drop generation at idle to save power
So 16x 4.0 on my 3090 drops to 16x 1.1, but could never drop to 8x anything without physically changing something while the PC was off

The only normal reason it should be at 8x, is if you have ANYTHING in the second GPU slot on the motherboard, or on some rare boards using an NVME slot wired as if it was that second slot
Your motherboard has no second GPU slot or NVME slot, so those are ruled out

The 'bad' reasons that can happen is dirt or damage to the GPU slot or the contacts of the GPU, or damaged or missing pins from the CPU (such as thermal paste on the pins/in the socket)


Zeintimings is great since it shows all the relevant hardware here, including current settings and voltages - it's fantastic

Your RAM is DDR4 3200, not 2133 like your specs have listed - that's the 'safe mode' JEDEC speed for when XMP is disabled.
Your ryzen 3600 CPU does support 3200, but as an "up to" which means it can have issues at times, or with certain combinations.


Your motherboard only officially supports 2667 on the DRAM, with all higher clocks being "OC" (overclock) meaning "good luck, we wont help you get it stable but it might work"
And honestly, not a single AM4 CPU will run DDR4 4400 so they're really optimistic with that list)

I'd suggest raising your SoC voltage (Sometimes called VSOC or system on chip) in the BIOS from it's auto 1.085v to 1.10v and see if it helps - that voltage controls all the SoC devices (memory controller, USB, PCI-E, etc) and will definitely improve stability, but worst case you may need 1.15v (which is perfectly safe, but may raise CPU temperatures 2-3C at worst)





You're on motherboard BIOS 3802 and there is a new 3810 - they dont list the improvements, but there is a chance they'll help this situation (BIOS updates reset the BIOS to factory defaults, so you'd need to enable XMP and raise the SoC voltage *after* updating)

PRIME B450M-K II｜Motherboards｜ASUS Global

 

[emothxughts]

The only normal reason it should be at 8x, is if you have ANYTHING in the second GPU slot on the motherboard, or on some rare boards using an NVME slot wired as if it was that second slot

There's no 2nd GPU slot, but I do have 2 shorter PCI slots, and in one of them is a wifi adapter card, TP-Link TL-WN881ND.

 

[Mussels]

I did put something in about "A lot of edits coming, refresh soon"
real world interrupted so i posted what i had, to avoid the risk of losing it


all those edits are in now. Your motherboard doesn't support your RAM speed, but i've listed a possible solution - if SoC voltage doesnt work, lower ram speed to 2667 (After enabling XMP so voltages and timings are still applied) and see what happens.


If you need help flashing the BIOS, let me know if you're flashing within the BIOS or using bios flashback and i can offer help there

 

[emothxughts]

Thanks for the tips, I'll try some and update on this topic once the RX 580 arrives.

 

[Mussels]

Sometimes it's a common problem that just doesnt work with common logic (too high DRAM speed = video card/lan/wifi vanishing on ryzen)


Other times it's just WEIRD stuff. I've had smart TV's confuse a game for a chromecast and spam the network, an Xbox controller BSOD a PC when a fridge compressor turned on, and yesterday a GTX970 that turned on every 3rd time exactly, but only because of a somehow faulty DVI to HDMI adaptor

Don't rule out anything, think of them as very unlikely but sometimes it truly is the weird unlikely stuff that causes these issues
(And the reverse, don't assume the first answer you find is the solution - sometimes it's a coincidence)

 

[emothxughts]

My RX 580 just came back in the mail, but now I'm feeling anxious about installing it in my PC again. Do I really have to DDU my current RX 550 before I install the RX 580? Asking because they're both AMD GPUs, RX 500 series, Sapphire cards even.

Edit:
I installed my RX 580 without running DDU first, plugged into the 2nd HDMI port, the replacement port that was previously physically destroyed.. First thing I did after booting was to check GPU-Z. It runs at PCIE x16, and it didn't black screen and cut off display when I ran the render test. So far so good. Then I ran Unigine Heaven twice, once at low and once at ultra. No black screening either. So far so good too.

 

[Mussels]

No you do not have to run DDU and remove drivers

The only time that's ever neccesary is if the official driver uninstaller doesn't work, maybe a decade ago it was a problem - but it's not now

 

[emothxughts]

It ran a bit hot (and consumes quite a lot of watts) on stock settings, so I undervolted it. Tested on Heaven, Superposition and Genshin Impact (specifically the relatively intensive desert area), everything works smoothly and a bit cooler.

 

[Mussels]

It ran a bit hot (and consumes quite a lot of watts) on stock settings, so I undervolted it. Tested on Heaven, Superposition and Genshin Impact (specifically the relatively intensive desert area), everything works smoothly and a bit cooler.

Could you define "a bit hot"

The RX580's should be under 80c in extreme testing, for example

 

[emothxughts]

Could you define "a bit hot"

The RX580's should be under 80c in extreme testing, for example

75 degrees C in Heaven & Superposition (and Furmark too back when it was in the repair shop, I'd never run Furmark myself though), about 68 C in Genshin. It's below 80 C as you mentioned, but I'm quite a bit anxious about anything above 70 C, considering it's a used card, even if it's been repasted.

Barely cracks 66 C in Heaven after undervolt. Might be a tiny bit slow, but I can barely notice the slowness and can definitely notice temps and power usage decrease.

 

[Mussels]

75c is completely normal for them
Sure undervolt away - I would to, but be aware that's a completely normal stock temperature

This is from the 8GB review, but the GPU's are the same between them

 

[emothxughts]

Updates:

• The card is running perfectly fine so far, no black screens, stays at PCIE x16 after hibernation. Undervolted it a bit to cool it down, reduce power usage and save on electric bills. (That's just average life in the tropics.)
• There were driver timeouts, but only on my modded Skyrim setup, and I'm pretty sure that was because of some improperly configured mods.
• I will mark this topic as solved in the original post once 1 full week has passed without any problem from the card.

 

[emothxughts]

Update, 22 January:

After 1 week straight of powering my PC on and off (in hibernate mode) and gaming on it after I got my RX 580 back, I have received zero black screens and display signal loss, and I have only frozen in Skyrim due to bad mod setup. I have also opened GPU-Z every time I wake my PC up from hibernation, and the number of lanes the card runs on as displayed in Bus Interface never changes from x16. I am running this card with the monitor plugged in to the card's 2nd HDMI port, previously broken.

Edit: I was going to edit the original post, but there's no edit button available for it. So, next post is the summary & solution to the problem described in the OP.

Summary & solution:

The RX 580 card black screens (of the "monitor just turns off and says no signal" variety) when the card is put under load and goes to PCIe 3.0, for example right after I start the GPU-Z render test. The card also reports only running on x8 lanes in GPU-Z when it is specced for x16 lanes.

I sent it to a repair shop, repair shop messages me, "card hot the port problem" and "service port problem solved". They fixed the HDMI ports (I'm using an HDMI monitor, never used DP on the card), changed thermal paste & pads while they were at it, they tested it out in Furmark after repairs, no black screens and the tests went on for 40+ minutes on top of that. I received the card back a week ago, played games on my PC for a whole week, no black screens either.

So there it is, a graphics card borked because the HDMI ports (supposedly) went bad, fixed it and the card runs good again.

 

[Imperator]

Nice that you have a working card again.

 

[emothxughts]

Hi guys, it seems I celebrated too soon, unfortunately my RX 580 blackscreened again.

I was opening Genshin Impact, and as soon as the game loaded, the monitor loses signal, just like in my original post. I had to force restart my PC. And then it black screens a minute after Windows 10 has loaded.

So I opened my PC again and reset BIOS. Then I checked GPU-Z and now it says Bus Interface: PCIe x16 3.0 @ x8 1.1. Seems that all of a sudden it's only running on 8 lanes now.

Also, Event Viewer only ever shows Kernel Power 41 events, which just means that the system was not shut down properly (because I force restarted with the power button).

Now it looks like the ports thing was just a placebo. I'm thinking my GPU blackscreening has to do with the sudden change in lanes. What do you think?

Edit:
I also should mention, I'm running off the 2nd HDMI port, the previously broken one. And I currently have no blackscreens playing Genshin Impact at PCIe x8 for now.

 

[Imperator]

Last week i encountered the same problem as you do.
I noticed that it appeared after i changed the memory speed in BIOS (mainboard BIOS).
So i lowered the DDR4 speed in BIOS and the error is gone.
I did not reinstall the driver nor i moved the video card from slot.
All i did was to change back the memory speed to initial values.
Even i have ddr4 3600 i must use them at 3200 cause i receive the same error like you do.
Maybe it is a start...to look somewhere else not only at video card.

 

[emothxughts]

Last week i encountered the same problem as you do.
I noticed that it appeared after i changed the memory speed in BIOS (mainboard BIOS).
So i lowered the DDR4 speed in BIOS and the error is gone.
I did not reinstall the driver nor i moved the video card from slot.
All i did was to change back the memory speed to initial values.
Even i have ddr4 3600 i must use them at 3200 cause i receive the same error like you do.
Maybe it is a start...to look somewhere else not only at video card.

Before the new black screen but after I received this card back, I set my RAM speed to 2666 without XMP/DOCP on. After I reset BIOS with the touch two jumpers method, I kept my RAM at the default 2133. No blackscreens yet..."yet" being the key word here.

 

[Mussels]

Last week i encountered the same problem as you do.
I noticed that it appeared after i changed the memory speed in BIOS (mainboard BIOS).
So i lowered the DDR4 speed in BIOS and the error is gone.
I did not reinstall the driver nor i moved the video card from slot.
All i did was to change back the memory speed to initial values.
Even i have ddr4 3600 i must use them at 3200 cause i receive the same error like you do.
Maybe it is a start...to look somewhere else not only at video card.

Becuase on ryzen, the USB, SATA, PCI-E controller and DRAM controller are part of the same hardware - the SoC
If you get PCI-E errors you can either downclock the RAM (and infinity fabric, which is synced to the RAM) or raise the SoC voltage

I covered this in post 27


You'll likely swap SoC errors for DRAM errors running at 2133, since it's not running the timings and voltage the DRAM was designed for

"black screen" crashes on ryzen at the transition from load load to higher load levels is almost exclusively SoC problems, almost always triggered by the RAM setup, second most common by faulty PCI-E riser cables

 

[emothxughts]

Becuase on ryzen, the USB, SATA, PCI-E controller and DRAM controller are part of the same hardware - the SoC
If you get PCI-E errors you can either downclock the RAM (and infinity fabric, which is synced to the RAM) or raise the SoC voltage

I covered this in post 27


You'll likely swap SoC errors for DRAM errors running at 2133, since it's not running the timings and voltage the DRAM was designed for

"black screen" crashes on ryzen at the transition from load load to higher load levels is almost exclusively SoC problems, almost always triggered by the RAM setup, second most common by

No riser cable in my PC, so RAM setup it is then. Let me raise SoC voltage to 1.1v...

Edit: Forgot to mention, its default is 1.025v.

 

[Imperator]

Becuase on ryzen, the USB, SATA, PCI-E controller and DRAM controller are part of the same hardware - the SoC
If you get PCI-E errors you can either downclock the RAM (and infinity fabric, which is synced to the RAM) or raise the SoC voltage

I covered this in post 27


You'll likely swap SoC errors for DRAM errors running at 2133, since it's not running the timings and voltage the DRAM was designed for

"black screen" crashes on ryzen at the transition from load load to higher load levels is almost exclusively SoC problems, almost always triggered by the RAM setup, second most common by faulty PCI-E riser cables

Yeah, but in my case it is the intel system with Corsair Vengeance LPX 3600 Black - the one who gave me errors.

 

[Mussels]

Yeah, but in my case it is the intel system with Corsair Vengeance LPX 3600 Black - the one who gave me errors.

Gotcha, with an AMD in your specs i assumed it was the same problem

Similar issues can occur on the intel side too, but they usually have different symptoms due to the different design, combined with intel locking RAM speeds down if you dont have a K/Z combo - most people cant reach the RAM speeds that cause issues

You'd have to look up that specific CPU and see what voltages or settings people reccomend changing for memory overclocking because 3600 isn't standard on any DDR4 platform, IIRC it was VCCSA (System agent) voltage - but i cant recall the value


The best answer for simple stability is to enable XMP to get DRAM voltages and timings happy, and then lower the DRAM down to a stock value for the CPU - but if you want the higher clocks with higher density memory setups, tweaking is often required on any platform (again, this is why intel locked down the non K/Z setups)

 

[emothxughts]

The best answer for simple stability is to enable XMP to get DRAM voltages and timings happy, and then lower the DRAM down to a stock value for the CPU

Not sure if this for me or the other guy, but if it's for me, does this mean I have to turn on my XMP/DOCP, then manually set my RAM speed to the default 2133?

Btw, my ram is currently set to 2133, no XMP.