Problem with RX580 8GB

[daoren123]

Bought my PC 3 years ago. Everything brand new. Immediately had crashes when playing CS:GO, but that was the only game. Found a solution by lowering MHz from base clock of 1360 to <1250 MHz. Searching for a solution to be able to set clock at its base but not experience crashes.

GPU turning off and screen going black (losing contact), restart needed, keep in mind I have to turn the PC off and on, because pressing the button once does not restart the PC. Also every form of input (audio included) is stopped, so speaking into microphone does nothing, but I still hear people and the game in the background for about 10 seconds.

I tried:
Fresh windows install (a few times)
Overvolting to from 1.150 to 1.199
Power limit set on +50% (max)
Reapplying thermal paste to the CPU (thinking it might be a CPU problem)
Reinstalling drivers
Updating BIOS(UEFI)
Checking cable connection, everything is fine
New monitor cables, as well as a new monitor completely

This sounds more like it has something to do with my motherboard, ASUS PRIME B450M-A II, but the only solution I have found while trying to fix this issue was tweaking GPU clock speed to below AMD suggested base clock.

I've added a log that GPU-Z recorded the moment my screen went black. Sorry for the resolution I hope it will help. And the base configuration of my GPU, at the time of the crash, the only thing that was tweaked was power limit set to +50%.

I can't afford testing the GPU or any component for that matter on another PC, sadly.

I have also turned Fast Boot off. GPU temperature on standby is 35-40C, while under heavy load in low 70s. Fans are just tweaked to work faster on lower temperatures because at beginning I thought that might've been the issue, temperature. I can provide more information if needed, I don't know what could help or not.

 

[izy]

What PSU do you have? Might be a power supply problem, try lowering the power limit of your GPU or/and undervolt it and see if it still crashes.
Edit: oopps i see it works with lower clocks so its probably the PSU

 

[Mussels]

Seeing how old all these RX580s are, i'd say you need new thermal paste and thermal pads (Same thickness as stock! never use thicker/thinner) - it all depends how much use its had. Overvolting and boosted power limits are going to cause it to age faster, and theres no such thing as a new RX580 any longer.




That said, the specs you mentioned for your system definitely may be the cause.
Ryzen setups prior to the 5000 series did not fully support 3200MT/s speeds, and the lower end boards had even less chance of that working problem free.

Enter your BIOS
Enable XMP for your RAM, manually set SoC voltage (vsoc) to 1.10v then lower the RAM speed to 2966MT/s
In that exact order, no other.

What happens is that SoC is the link between your CPU and PCI-E express lanes, while sharing power with the memory controller. Infinity fabric runs at the same speed as your RAM, so higher RAM speed means higher demands of the infinity fabric. Higher power needs (or errors) on the Infinity fabric can cause the PCI-E lanes to drop out causing a "GPU driver has crashed" situation or if the memory controller side crashes, a black screen lockup of the entire PC.

Despite having half a dozen AM4 systems in the house here with lots of spare parts, some combinations of motherboards and CPU's would never be fully stable at 3200 - they could boot and benchmark perfectly fine but then have issues at low loads, idle, or just powering on after being left off for a few hours. Nothing kept them stable other than lower Infinity fabric speeds (which are tied to RAM speeds)

 

[daoren123]

What PSU do you have? Might be a power supply problem, try lowering the power limit of your GPU or/and undervolt it and see if it still crashes.
Edit: oopps i see it works with lower clocks so its probably the PSU

Okay ill try this, altho after trying to lower voltage my gpu crashed, ill try doing it thru afterburner this time. Increasing power limit helped tho, because it crashed less than before.

Seeing how old all these RX580s are, i'd say you need new thermal paste and thermal pads (Same thickness as stock! never use thicker/thinner) - it all depends how much use its had. Overvolting and boosted power limits are going to cause it to age faster, and theres no such thing as a new RX580 any longer.




That said, the specs you mentioned for your system definitely may be the cause.
Ryzen setups prior to the 5000 series did not fully support 3200MT/s speeds, and the lower end boards had even less chance of that working problem free.

Enter your BIOS
Enable XMP for your RAM, manually set SoC voltage (vsoc) to 1.10v then lower the RAM speed to 2966MT/s
In that exact order, no other.

What happens is that SoC is the link between your CPU and PCI-E express lanes, while sharing power with the memory controller. Infinity fabric runs at the same speed as your RAM, so higher RAM speed means higher demands of the infinity fabric. Higher power needs (or errors) on the Infinity fabric can cause the PCI-E lanes to drop out causing a "GPU driver has crashed" situation or if the memory controller side crashes, a black screen lockup of the entire PC.

Despite having half a dozen AM4 systems in the house here with lots of spare parts, some combinations of motherboards and CPU's would never be fully stable at 3200 - they could boot and benchmark perfectly fine but then have issues at low loads, idle, or just powering on after being left off for a few hours. Nothing kept them stable other than lower Infinity fabric speeds (which are tied to RAM speeds)

It is exactly like that. low load, bunch of things open, and it crashes. While stressing it thru any benchmark or stress test, or playing a GPU intensive game, it doesn't crash ever unless I have browser open with music and discord in call with friends etc. Sounds like this might just fix it. Also I've experienced random black screens when watching youtube, or booting PC from sleep. But those last a second or two showing "No signal" and then it just turns back on. I'll report as soon as another crash happens after applying fixes you suggested with my base clock GPU, or if I experience no crashes I'll update you on this. Thanks a lot!

 

[izy]

Okay ill try this, altho after trying to lower voltage my gpu crashed, ill try doing it thru afterburner this time. Increasing power limit helped tho, because it crashed less than before.


It is exactly like that. low load, bunch of things open, and it crashes. While stressing it thru any benchmark or stress test, or playing a GPU intensive game, it doesn't crash ever unless I have browser open with music and discord in call with friends etc. Sounds like this might just fix it. Also I've experienced random black screens when watching youtube, or booting PC from sleep. But those last a second or two showing "No signal" and then it just turns back on. I'll report as soon as another crash happens after applying fixes you suggested with my base clock GPU, or if I experience no crashes I'll update you on this. Thanks a lot!

Can you post a screen shot from GPU-Z Sensor page while you are under load?

 

[daoren123]

Seeing how old all these RX580s are, i'd say you need new thermal paste and thermal pads (Same thickness as stock! never use thicker/thinner) - it all depends how much use its had. Overvolting and boosted power limits are going to cause it to age faster, and theres no such thing as a new RX580 any longer.




That said, the specs you mentioned for your system definitely may be the cause.
Ryzen setups prior to the 5000 series did not fully support 3200MT/s speeds, and the lower end boards had even less chance of that working problem free.

Enter your BIOS
Enable XMP for your RAM, manually set SoC voltage (vsoc) to 1.10v then lower the RAM speed to 2966MT/s
In that exact order, no other.

What happens is that SoC is the link between your CPU and PCI-E express lanes, while sharing power with the memory controller. Infinity fabric runs at the same speed as your RAM, so higher RAM speed means higher demands of the infinity fabric. Higher power needs (or errors) on the Infinity fabric can cause the PCI-E lanes to drop out causing a "GPU driver has crashed" situation or if the memory controller side crashes, a black screen lockup of the entire PC.

Despite having half a dozen AM4 systems in the house here with lots of spare parts, some combinations of motherboards and CPU's would never be fully stable at 3200 - they could boot and benchmark perfectly fine but then have issues at low loads, idle, or just powering on after being left off for a few hours. Nothing kept them stable other than lower Infinity fabric speeds (which are tied to RAM speeds)

Okay I've tried but was unable to figure out how to set exact voltage to 1.10V. There is no XMP, theres only DOCP which I read is the equivalent, so I tried enabling and setting RAM speed to 2966, which is the closest value to what you suggested. There is an option to select -/+, and below that its some weird options I didn't wanna change because I'm inexperienced. Any help?

Can you post a screen shot from GPU-Z Sensor page while you are under load?

Stress test or general use? Like heavy gaming.

 

[izy]

Okay I've tried but was unable to figure out how to set exact voltage to 1.10V. There is no XMP, theres only DOCP which I read is the equivalent, so I tried enabling and setting RAM speed to 2966, which is the closest value to what you suggested. There is an option to select -/+, and below that its some weird options I didn't wanna change because I'm inexperienced. Any help?


Stress test or general use? Like heavy gaming.

Use the option "Log to file" and do something that it makes it crash then post the log , i think thats better, you can also post a screenshot under stress test.
You can also check the "Event Viewer" under Windows Logs / System , see if there are any errors and what they say.

 

[daoren123]

Use the option "Log to file" and do something that it makes it crash then post the log , i think thats better, you can also post a screenshot under stress test.
You can also check the "Event Viewer" under Windows Logs / System , see if there are any errors and what they say.

its hard to replicate the crash. It's random under low load, but this the last GPU-Z log before crashing a few days ago. Yesterday it needed about 5 hours of low load usage to crash. The only game that consistently crashed was CS:GO, I'll try that and report back. Here's the last log before crash and logs during stress test, no crash during stress test or benchmarks ever btw.

 

[izy]

its hard to replicate the crash. It's random under low load, but this the last GPU-Z log before crashing a few days ago. Yesterday it needed about 5 hours of low load usage to crash. The only game that consistently crashed was CS:GO, I'll try that and report back. Here's the last log before crash and logs during stress test, no crash during stress test or benchmarks ever btw.

Just upload the .log file next time but anyway i was curios to see PCIe Slot Power [W] , PCIe Slot Voltage [V] , 8-Pin #1 Power [W] , 8-Pin #1 Voltage [V] , GPU Voltage [V] but i guess the rx 580 doesnt have those sensors.
Have you tried running your RAM at stock speeds? Usually when this happens its a PSU problem a RAM problem or a CPU having a bad OC unless ofc the GPU is showing its age, you can also try lowering the VRAM frequency and see if that fixes the problem (maybe a VRAM chip is going bad).

Edit: I forgot where is the setting on AMD but you should try to set GPU power management mode to Performance (for nvidia is in the nvidia panel) , it might help with the crashes if you say that its sometimes happening under low load, what this will do is that will not downclock the GPU under lower 3d loads to save power and also set the windows power plan to performance.

 

[daoren123]

Just upload the .log file next time but anyway i was curios to see PCIe Slot Power [W] , PCIe Slot Voltage [V] , 8-Pin #1 Power [W] , 8-Pin #1 Voltage [V] , GPU Voltage [V] but i guess the rx 580 doesnt have those sensors.
Have you tried running your RAM at stock speeds? Usually when this happens its a PSU problem a RAM problem or a CPU having a bad OC unless ofc the GPU is showing its age, you can also try lowering the VRAM frequency and see if that fixes the problem (maybe a VRAM chip is going bad).

Edit: I forgot where is the setting on AMD but you should try to set GPU power management mode to Performance (for nvidia is in the nvidia panel) , it might help with the crashes if you say that its sometimes happening under low load, what this will do is that will not downclock the GPU under lower 3d loads to save power and also set the windows power plan to performance.

Ok, so playing CS:GO 2 doesn't crash the GPU, I don't know how to intentionally crash it, I'll try setting my GPU power management to performance. I'll also try lowering VRAM frequency which I was thinking of as well. Thanks for the help so far, I'll give an update once the crash happens or if this fixes the issue.

Just upload the .log file next time but anyway i was curios to see PCIe Slot Power [W] , PCIe Slot Voltage [V] , 8-Pin #1 Power [W] , 8-Pin #1 Voltage [V] , GPU Voltage [V] but i guess the rx 580 doesnt have those sensors.
Have you tried running your RAM at stock speeds? Usually when this happens its a PSU problem a RAM problem or a CPU having a bad OC unless ofc the GPU is showing its age, you can also try lowering the VRAM frequency and see if that fixes the problem (maybe a VRAM chip is going bad).

Edit: I forgot where is the setting on AMD but you should try to set GPU power management mode to Performance (for nvidia is in the nvidia panel) , it might help with the crashes if you say that its sometimes happening under low load, what this will do is that will not downclock the GPU under lower 3d loads to save power and also set the windows power plan to performance.

nevermind, managed to crash it. It was underclocked the first time I tried to crash but I didn't notice. Here is the whole file, you can also see the 1244 cap which was me playing CS:GO 2 with underclocked settings, no crash.

 

[izy]

Did you try setting the power plan to Performance in windows / amd panel and/or underclocking the VRAM?

 

[daoren123]

Did you try setting the power plan to Performance in windows / amd panel and/or underclocking the VRAM?

no, that's default configuration for VRAM. As for performance plan, I think I'm using AMD Balanced

 

[izy]

no, that's default configuration for VRAM. As for performance plan, I think I'm using AMD Balanced

Try turning the GPU power power management to Performance and test if it doesnt work try underclocking your VRAM.

 

[daoren123]

Try turning the GPU power power management to Performance and test if it doesnt work try underclocking your VRAM.

Okay will report back after testing it.

 

[daoren123]

Try turning the GPU power power management to Performance and test if it doesnt work try underclocking your VRAM.

Okay back with an update. It managed to crash under low load with lowered MHz for my GPU XD. I also noticed VSoC is at 0.825 is that perhaps too low as Mussels said?

 

[izy]

Okay back with an update. It managed to crash under low load with lowered MHz for my GPU XD. I also noticed VSoC is at 0.825 is that perhaps too low as Mussels said?

You can try to set it at 1 - 1.1v and see how it goes.

 

[Mussels]

Okay ill try this, altho after trying to lower voltage my gpu crashed, ill try doing it thru afterburner this time

It's going to crash if you lower the voltage too far, that's just adding a second source of crashes.

Okay back with an update. It managed to crash under low load with lowered MHz for my GPU XD. I also noticed VSoC is at 0.825 is that perhaps too low as Mussels said?

That's the lowest i've seen anywhere - was that reading from Zentimings? Try 1.0v and see how that goes

 

[Jism]

Open HWInfo > Check VRM temperatures.

 

[daoren123]

It's going to crash if you lower the voltage too far, that's just adding a second source of crashes.


That's the lowest i've seen anywhere - was that reading from Zentimings? Try 1.0v and see how that goes

I'm yet to do that, discussing with people IRL about the voltage and if it should be risen to about 1-1.10V as you have said. This is a screenshot. I mean, I'm hearing that suggested voltage is 1.15-1.20 and my voltage might just be too low. I dont know how or why, but thats that

Open HWInfo > Check VRM temperatures.

If this is of any help, I think you're asking for these temps

 

[Jism]

Please put some load, while monitoring the VRM temps. Edit: it does'nt seem to be there?

 

[daoren123]

Please put some load, while monitoring the VRM temps. Edit: it does'nt seem to be there?

If I was looking at the correct window in HWinfo, then my VRM doesn't support temperature reporting. I've read about it earlier today.

 

[Jism]

I had to look for it,

But try to update HWinfo first. If it still does not show then your card does not have the proper sensors for that. In short: try to repaste the GPU and see if the VRM's are cooled sufficient. A hardware shutdown is usually related to either bad power, a too hot of a VRM or something like that. A check on bad power is apply a load (i.e Furmark) and watch the GPU VRM Voltage In.

A bad cable will 100% show that by suddenly dropping from 12.25V to 11.50V or even below. A good cable stays rock solid at 12V. I have a 6700XT now and the RX580 on the wall.

 

[daoren123]

I had to look for it,

View attachment 322236

But try to update HWinfo first. If it still does not show then your card does not have the proper sensors for that. In short: try to repaste the GPU and see if the VRM's are cooled sufficient. A hardware shutdown is usually related to either bad power, a too hot of a VRM or something like that. A check on bad power is apply a load (i.e Furmark) and watch the GPU VRM Voltage In.

A bad cable will 100% show that by suddenly dropping from 12.25V to 11.50V or even below. A good cable stays rock solid at 12V. I have a 6700XT now and the RX580 on the wall.

Yeah after updating HWinfo, I still see no VRM Voltage. I'll look into repasting the GPU. Any thoughts on aforementioned 0.825 VSoC? Am I looking at the correct voltage and if yes should I try to raise it?

Was the picture I linked showing VSoC? I think that is what I should be looking at and as you can see its at 0.825V maximum. Mussels

 

[Jism]

Idle voltage is correct; 0.8V. Load should be around 1.125V.

 

[daoren123]

Idle voltage is correct; 0.8V. Load should be around 1.125V.

Okay let me try to understand this because I'm pretty new to this. Idle voltage is when I have nothing running, which is 0.8 V. When PC is under load it should be around 1.125V. But maximum is set to 0.825V and in BIOS its also at 0.825 without offset enabled because D.O.C.P. is disabled. But you're saying under load it will automatically go to a normal ~1.1V even with it being 0.825V in BIOS and without offset? Because as far as I understood, offset in BIOS enables VSoC to go to a lower minimum or higher maximum, yet its disabled. So how does that work?