PC Graphics

[nex1]

Hi,

I've owned the RX 5700XT for the past  2.5 years, specifically XFX RX 5700XT THICC III Ultra and ever since I've bought it I've had 3-5 daily crashes, sometimes I can go a day or two without any, but 95% of the time it crashes. I've tried multiple solutions but none of them worked so I'm hoping y'all can help me figure it out.

What I've tried so far:

-Changing PCIe 16 and PCIe to both Gen 3 and Gen 4, didn't help.

-Reseated GPU, bought GPU support, both didn't help.

-Bought new PIN connectors, didn't help.

-Installing new up-to-date Chipset drivers, didn't help.

-Overclocked and Undervolted my GPU, neither helped.

-Tried an aggressive FAN curve in case of overheating but didn't help.

PC Specs:

MOBO: ROG STRIX B550-F GAMING

GPU: XFX RX 5700XT Thicc III Ultra

CPU: Ryzen 5 5600x

PSU:  Thermaltake 750W 80+ Gold 

RAM: Crucial Ballistix 32GB 3600 MHz

OS: Windows 10 PRO

The only issue I've had is screen going black, sometimes sound loops and then my PC restarts, I also see the lights next to my RAM flashing but I think that's cause it's restarting. Didn't have any green screens of death or like that. Had 2 BSOD's in the past year and that's about it. But the black screen crashes happen 3-5 times a day and since I play competitively it causes me to get bans ranging from an hour to several days due to too many crashes leading to match leaves. 

I've noticed that 3 exes under the names "DtsApo4Service.exe", "FileOperator.exe", "ArmouryCrate.Service.exe" crash before a PC crash and the error code of the PC crash is "LiveKernelEvent ; ID:141"

[RPX100 · Accepted solution]

@nex1 hello again

I have not used Radeon Software since Jan 2020, so I can only speak from memory:
Go into performance tuning tab and enable manual performance tuning.

There you can change your cards voltage, frequency and fan speed curve.
(you might have to enable "Advanced Tuning" as well (should be right there on that tab)

Please note: AMD was trying to compete with nvidia on RX 5000 series cards.
This means: All RX 5000 series cards are overclocked by default!
This might be number one reason for your black screen crashes.
(Because your GPU clock might spike too high)

If you do a clean driver install, then your RX 5700 XT will most likely run those settings:
P3 (power state 3 or gaming mode) = max boost clock / max voltage:
- GPU core clock: 2100 MHz
- GPU core voltage: 1200 mV
- and I believe your current fan speed is limited to around 54% by default

In my experience with this card over the past 2+ years: this is far too high!
I can recommend: Core clock: 1935 MHz and voltage 1100mv or even 1050mv
Also try to adjust your fan speed to keep your hot spot below 90°C at all times.

Here are the details for your XFX RX 5700XT Thicc III Ultra (link)
"Game clock" for your card is 1935 MHz: try to stay below that value for more stability!

Please feel free to ask further questions or post screenshots of your settings.
I will gladly have a look and answer your questions.

[nex1]

Included something else in the post

[RPX100]

@nex1 hello there. Welcome to AMD community forum.

First have a look at this post: RX 5000 series crash types and reasons
This might be able to explain some very basic issues regarding the crashes.

I have RX 5700 XT myself and had to do a lot of tweaking to prevent crashing.
When I bought it in dec 2020 I had constant crashes for the first 3 days.

---
Here is what I do myself:
1) clean driver install using Display Driver Uninstaller in Windows Safe Mode
2) I do not use Radeon Software and instead just use a "driver only" install mode.
3) for the actual performance and fan tuning I use MSI Afterburner
- my settings are: 1050 mV at 1840 MHz (P3 Boost clock / max allowed clock speed)
- if I go higher than that, I notice crashes in Unreal Engine game with error "D3D device lost"

Hope this helps.
cheers

[nex1]

Hey! Thank you for replying first of all,

I've looked at your post and the only crashes I've had was black screen, system freeze and then system restart.

I've done a DDU clean plus installed the PRO drivers since people are saying they're more stable cause they're get tested more than the general gaming drivers.

I've also checked the thermal paste and it seems fine since the delta is 10° however I've noticed that when doing a 30min stress test and generally when playing my fan is sitting around 700-1.2k RPM *seems a bit to low*, sometimes spiking up to 1.5k but rarely and the temperature is generally sitting at around 85°-89° hitting 90° maybe once every 30 min. The highest I've seen it spike without crashing was up to 96°, but I'm pretty sure that was due to fans being at as low as 500 RPM.

I'm quite positive that the issue is overheating/high temperatures, but I've never done fan tweaking in the Radeon software so could you help me out on that please?

Thank you in advance.

[RPX100 · Accepted solution]

@nex1 hello again

I have not used Radeon Software since Jan 2020, so I can only speak from memory:
Go into performance tuning tab and enable manual performance tuning.

There you can change your cards voltage, frequency and fan speed curve.
(you might have to enable "Advanced Tuning" as well (should be right there on that tab)

Please note: AMD was trying to compete with nvidia on RX 5000 series cards.
This means: All RX 5000 series cards are overclocked by default!
This might be number one reason for your black screen crashes.
(Because your GPU clock might spike too high)

If you do a clean driver install, then your RX 5700 XT will most likely run those settings:
P3 (power state 3 or gaming mode) = max boost clock / max voltage:
- GPU core clock: 2100 MHz
- GPU core voltage: 1200 mV
- and I believe your current fan speed is limited to around 54% by default

In my experience with this card over the past 2+ years: this is far too high!
I can recommend: Core clock: 1935 MHz and voltage 1100mv or even 1050mv
Also try to adjust your fan speed to keep your hot spot below 90°C at all times.

Here are the details for your XFX RX 5700XT Thicc III Ultra (link)
"Game clock" for your card is 1935 MHz: try to stay below that value for more stability!

Please feel free to ask further questions or post screenshots of your settings.
I will gladly have a look and answer your questions.

[nex1]

@RPX100 Hi again!

So I've tweaked around the Radeon software as you said and got a pretty stable fan curve, GPU core sitting at around 70° no matter the scenario, except when not gaming then it's around 50°. 

Regarding the GPU core clock however I tired setting it 1950 MHz and 1100mv, but after like 5 mins of playing it already crashed, I tired setting it 1900MHz with 1050mv and it seemed fine for the time I played, no crashes and no performance loss.

However one question does come to mind, when you say GPU clock spike too high, do you mean it generally sitting at around 2100MHz as a "spike too high" or when it goes from let's say 1000MHz to 2100MHz that you consider a "spike too high" causing a crash?

Also, just to add, my current GPU tune values are:

P1: 800MHz 750mv

P2: 1350MHz 821mv

P3: 1900MHz 1043mv

That should be fine right? In case of any spikes it shouldn't be too high causing a crash I assume.

I also ran multiple stress tests before and after tuning and the result were as following:

Before:

Core was sitting at 2050-2100MHz at all times but wasn't spiking up or down so I couldn't figure out if that was the crash cause, temperatures were sitting at around 85-91° with 700-1.5k RPM. Delta was 10°.

After:
Core was sitting at 1870-1900MHz with no spikes neither up or down, temperatures were around 68-71° with 1000-2000 RPM varying depending on temperature, but at those degress it was mostly 2k RPM maintaining it. Delta was 5°, so improve.

UPDATE: For the first time ever, my PC just black screen crashed whilst just using it  normally, not even gaming or having any game running, a bit strange, but happened after the tuning.

I've added screenshots as you requested.

[RPX100]

Hello @nex1 

Thank you for providing the screenshot.
Those values are looking good to me.
In my experience, those should be more stable then the default settings.

Please note:
each GPU chip is different, so you might want to keep on tweaking in the future.
If you notice problems or crashes, try to:
1) decrease GPU core clock
2) increase GPU core voltage (try to stay below 1100 mv)
- Your GPU chip is rated for 1935 MHz game clock
- so you can focus your tuning around that value

---
Now regarding the core clock, I can explain it by using my own card as an example:

When I play Unreal Engine games with core clock above 1905 MHz,
(= max boost clock for my ASRock RX 5700 XT)
I will randomly get "D3D device lost" error and crash to desktop.

I can play some games with a custom profile (2000 MHz at 1100 mV) with 80% fan RPM.
But then I will still get random black screen crashes.
(random meaning: sometimes I can play 5 hours and sometimes 30 mins)

My card can stay at 1980 MHz for hours, but still crashes randomly.
That is why I tuned it like ASRock recommends: Game Clock around 1800 MHz
My personal daily use profile is: 1840 MHz at 943 mV 
- which is generating less heat, uses less power and still runs all my games at 60 FPS
- please note: I can not recommend these exact settings for your make and model
- you can try them - but you have to see if they work for your card

To answer your question directly:
I personally do not know why my own card randomly crashes above 1800 MHz game clock.
ASRock (manufacturer) rating for my card is 1795 MHz game clock and 1905 Boost clock.
But when I use 1905 MHz as P3 value, it becomes unstable (random crashes).

My personal guess is: it is either driver related (driver switching between power states)
or it is the GPU chip becoming unstable above its rated manufacturer specs.
(I do not understand why AMD decided to use 2100 MHz at 1200 mv as driver default!)

---
Side note: I can force my card to crash within 5 minutes (each time) when:
- I use a low fan RPM and hot spot temp goes above 90°C
- so it seems like my card can not handle the high temps
- I already opened the card and used high quality (new) thermal paste and memory pads
- the repaste and new pads improved performance and stability (and lowered temps)
- but card still crashes above 90°C within 5 minutes
- this is why I use aggressive fan curve to stay below 85°C (to be safe ^^)

I hope you can get your card to be more stable in the future.
It took me some time of trial and error to get my own card to run stable.
But it has been a long time since my last black screen / D3D device lost crash.

cheers

[nex1]

Hi @RPX100 

I've posted an update on my previous reply, as you replied to it, so I'm gonna assume you didn't read it, but basically my PC, for the first time ever, just black screen crashed whilst using it normally. By normally I mean no games running in the background and just working on some stuff, which is very odd because even before tweaking it never did that, any ideas as to why?

[RPX100 · Accepted solution]

@nex1 ah yes, I did indeed not see the updated text in your post.

First of all, please take some time and read through this topic from 2020:
ASRock RX 5700 XT Challenger Pro (triple fan)
- it has some details on my journey and all the problems with my card, when it was new
- There I also mentioned that Radeon Software Features were causing crashes for me.

quote from my old 2020 post:

Re-Installing the driver and disabling all the Radeon Software-Features solved the crashes.

things that will crash the card:
- Radeon Anti Lag, Boost, Chill, Enhanced Sync, Image Sharpening when they are enabled
- which means: only use the Standard profile and don't touch "gaming", "eSports" or "energy saver"

---
If you clicked on any of those profiles, re-install the driver
(and then enter your current performance tuning values again)

If all of that does not help, start by using 1100 mV and see if that is stable for you.
Your card might be undervolted too much - undervolting can also lead to black screens.
(remember to check the "crash types and reasons" post I made) ^^

---
side note: those problems are the main reason why I am not using Radeon Software.
I had huge issues with flickering screen while gaming.
Most of the times it would flicker for 30 seconds and then crash.
As soon as I switched to "driver only" install, the flickering was gone.

So I personally do not use / trust the Radeon Software anymore.
(I know this is quite drastic - but then again - it solved my problems)

[nex1]

@RPX100 I see, thank you for your help!

Although one  more question, what do the "P1,P2,P3" stand for in advanced tuning?

[RPX100]

@nex1 sure thing - no problem

- P1, P2 and P3 are power states.
- which power state is used, is defined via driver and GPU load %.
- P1 = light GPU usage or older games / 2D applications / CPU bound games
- P2 = medium GPU usage
- P3 = 3D games with high GPU usage

I am not sure about the exact % values, but you can think about it like this:
P1 = 0 - 25% load (not exact value, I am guessing)
P2 = 25 - 75% load (not exact value, I am guessing)
P3 = >75% load (not exact value, I am guessing)

The AMD GPU drivers decide which power state is used.
Mostly by GPU load % - so you are kinda forced to those limitations.

---
P1 will use 800 MHz core clock and 750 mV
P2 will use .... (see your screenshot)
P3 will use up to xxx MHz (the value you enter) and xxx mV (value you enter)

Your GPU also has deep sleep (DS) enabled by default.
Which means: desktop/idle usage lets the core clock drop down to a few MHz (0 - 200 MHz)
You can think of DS as power state 0 (idle/P0) - which is below the P1 that you can change.

Radeon Software does not allow you to tweak the deep sleep behavior.
There are 3rd party/community tools like "MorePowerTool" from igorslab, to disable DS.

---
You should not worry about P1 and P2 at all.
Focus on P3 and make sure to stay within manufacturer specs regarding game clock.

The only situation where an increased P1 frequency and voltage can help:
older games or CPU bound games, where you want higher GPU frequency
older games might use just a few % of your GPU, whichs leads to low frequency.
You might be able to improve performance for those game (or get rid of stuttering) 
by tweaking the P1 frequency and voltage.

[Digital_Elysium1]

OMG, thank u guys so very much for this post. I have a MSI RX 5700XT (Founders Edition). I have had mine on day 1 when they were released. So I have had thee most up & down experiences w/ this card. 1 day it's like I can tweak this thing to the fullest & it seems like it won't crash. But there's other days where I can't get the darn thing to run for more than 10 minutes. I custom water-cooled mine rig so the heat problem wasn't exactly a problem. As a matter of fact, I don't believe I have ever seen this card above 68°C. Normally though I can stress test around 48°C-54°C give or take 1 mhz or 2. N-E-way, I snapped & almost gave up when I couldn't run Witcher 2 or Halo MC edition! Seriously? I should be able to run Max settings for both of those games being older titles. Or close to it. Couldn't even run them on the lowest settings without crashing after 15 minutes. Couldn't figure out what I was doing wrong. I didn't understand why people said that Ryzen 9 3900X I was overkill for the  RX 5700 XT. From what I understand during that time that was the perfect dynamic duo. I've never had problems bottlenecking. At least not very much depending on what settings are. I tried everything with settings for my CPU tried to lower to just base 3800mhz in case it was overkill. But it didn't matter. At least not on my system it didn't. Then I read this article about such and such being too much and such and such being plenty. I would like to think that I am more in the advanced stages of PC knowledge. I wouldn't say I'm a master but I definitely say I'm more advanced than the average person. With that being said I was oblivious to knowing about settings in which they come with when you plug and play. I had no idea absolutely no idea that these cards were overclocked already out of the box. I just set my settings to almost identical to these ones as the THICC & founders numbers are almost identical. I just want to thank you guys. One for posting this article. In the other for the solution. I've been running this card stress test going on 6 hours now and it's not even breaking a sweat at these settings. I absolutely love you guys! Sorry about the long novel I just had to tell you guys how grateful and how much I appreciate you taking the time to post the problem and the tech to taking the time to giving a solution. I like to think that I still have a pretty good system after a few years. 

-Ryzen 9 3900X

-MSI RX 5700XT

-Asus ROG X570 Crosshair VIII Hero (wi-fi)

-Corsair Case (don't remember model offhand)

-EK nickel plated gpu backplate & EK water block)

-XSPC CPU water block, fittings, & PETG tubing)

-EK DC3 water pump

-G.Skill NEO RAM 3600 (4X16) 18-18-18-42

-Corsair LL fans w/ 2 NF-A14 PWM fans on rad.

-Corsair MP 600 & Sabrent Rocket m.2 SSD 1TB each & 1 Samsung EVO sata SSD 500GB

-Seasonic Focus 80+ Platinum 1200W

Again sorry about the long novel guys. Just grateful and pretty proud of my build being a proud member of the PC Master-Race & all. You guys know how it is yes save up every dime you have to get something halfway decent or beautiful. Anyway again sorry I just really wanted to put this out there.

 

[PepeJefe]

Thank you very much for this VERY usefull knowledge according to the Radeon 5700 series.

I wish I'd tried what you describe years ago.

My PC is now up and running at full speed

[chaosal]

Signed up just to say thank you to RPX100 for taking the time to comment on this with your advice and to Nex1 for taking the time to make the post!

I've suffered with my Nitro+ for years but ended up playing less GPU-intense games so hadn't been too bothered lately, until trying to play BG3 and barely making any headway before crash after crash.

A few days ago I switched to driver only rather than using Radeon Software as RPX100 suggested and scaled back to 1800MHz at 1050mV and so far so good!! Have been playing for up to a few hours at a time with zero crashes, bit absurd the hoops needed to jump through to get to a state of just being able to play a great game but again thank you for taking the time to post on here!  

[The_Woods]

Thanks for the information here already, it's good to get a solid lead on how to achieve system stability with this problematic card. 

I've tried a few different configurations to varying success, but still not stable in high demand applications.

Wondering if someone could explain the limitations of where to set the voltage and clock speed values? Might I be able to set the voltage to, max (1200), and the clock speed to 1600? Or do they need to be closer to the 'default trend'?

*edit* Is it also possible to achieve more a more stable system by playing with the power limit slider? EG, setting this to max or min?

[PepeJefe]

Hello there.

I've been using RX 5700 XT for almost 3 years by now.

I have learned that in order to play games without crashing (almost, happens once every 30-50 hours of playing by now) 

Make sure your motherboard is using the latest firmware. 

Do NOT use any of the settings using amd software (I.e. Chill or any other) only use standard profile.

I use 100% gpu fan from 65C

I use to lower Core clock: +-1900 MHz and voltage <1100mv.

Red dead redemption, days gone and the new ghost of tsushiro runs very smoothly now.

In 1440p resolution.