Frage Msi 5700 xt Gaming X völlig instabil

[Casi030]

Ist dies die Einstellung, die Sie kommentieren?
Wie verbessert diese Optimierung die GPU-Stabilität?

Ja genau.
Ist der Ram nicht 100% Stab,stürzt der Treiber schnell ab und resettet sich.
Auch Spiele Stürzen schnell ab.
Ist nur als kleiner Gegentest zum Takt und Leistungsgrenze herabsetzen in Wattman.

 

[itruji]

From GPU-Z I want to see the sensor side, but in a situation where there is a lot of power, not from idle. And I want to see the attitude from Wattman. Something like this: picture!

One question that we need to clarify is how big the difference is between the temperatures of the CPU core and hotspot under load. If the difference is very large, it can indicate poor heat conduction.
Otherwise I spontaneously do not see any error in your bios (MPT) that explains the instability.

Hello again. I reinstalled my GPU 5700 xt and took the screenshots you asked for.
The first is with the default configuration, the second drops the frequency to 2000 and the third to 1905 (the latter is what makes my games work with better temperature).

 

[ApolloX]

Ok, now I can give some more meaningful comments.
Look at image 2 and check the maximum temperatures, CPU core had 73°C while the hotspot had 111°C. I consider this as most likely the main issue. First, if there is such a big difference between both values this is a rather clear sign that heat conduction is poor. You or someone who is confident in doing this should remove the fans and heatsink and check the thermal gel. It might be insufficient, dryed, whatever... After properly fixing this issue, the card might work much better. Second, with 111°C the hotspot is far too high for normal operation. Chech the shown BIOS values at techpowerup for your card: the thermal limit fpr your card's hotspot ist 110°C, in the benchmark shown by pic2 it was already 111°C - so it's more than clear that your card throttles down, maybe even crashes.
So stop istalling drivers and thinking about RAM --> fix your heat conduction first! Then you will also see that 100% fan terror can be avoided.
Then check about your in-game/real usecase problems, I'd assume then the card runs fine.

BTW: this currency/frequency curve function is not so popolar - nobody uses this. It's sufficient toset min and max values, but the curve is - to my opinion - a useless deceptive nonsense.

 

[itruji]

Ok, now I can give some more meaningful comments.
Look at image 2 and check the maximum temperatures, CPU core had 73°C while the hotspot had 111°C. I consider this as most likely the main issue. First, if there is such a big difference between both values this is a rather clear sign that heat conduction is poor. You or someone who is confident in doing this should remove the fans and heatsink and check the thermal gel. It might be insufficient, dryed, whatever... After properly fixing this issue, the card might work much better. Second, with 111°C the hotspot is far too high for normal operation. Chech the shown BIOS values at techpowerup for your card: the thermal limit fpr your card's hotspot ist 110°C, in the benchmark shown by pic2 it was already 111°C - so it's more than clear that your card throttles down, maybe even crashes.
So stop istalling drivers and thinking about RAM --> fix your heat conduction first! Then you will also see that 100% fan terror can be avoided.
Then check about your in-game/real usecase problems, I'd assume then the card runs fine.

BTW: this currency/frequency curve function is not so popolar - nobody uses this. It's sufficient toset min and max values, but the curve is - to my opinion - a useless deceptive nonsense.

Hallo!
Ich verstehe, dass die GPU-Stresssituation mit Timespy extremer ist als mit Spielen, aber ich war überrascht, so hohe Temperaturen zu sehen. Ich habe Wärmeleitpaste bei Amazon bestellt und werde sie selbst ändern. Irgendeine Empfehlung?
Normalerweise verwende ich die dritte Einstellung: Senken Sie den GPU-Takt auf 1905 und 950 mV. Sie empfahlen mir, die Spannung zu erhöhen, und ich habe dies bis zum Erreichen von 1050 mV getan, um sicherzustellen, dass kein Strom fehlt. Mit dieser Konfiguration in Valorant steigen die Temperaturen niemals über 55/75 ° C und in Warzone auf 60/85 ° C. Aus diesem Grund habe ich immer gedacht, dass die Temperatur auf meiner GPU nicht das Problem sein sollte.
In Bezug auf die Strom- / Frequenzkurve habe ich Ihre Empfehlung leider nicht verstanden. Welche Konfiguration wäre für meine GPU geeignet? Vielen Dank!!

 

[ApolloX]

Ich würde die Spannung-Frequenz-Kurve garnicht aktivieren. Entweder im Treiber den Bereich komplett ausschalten oder halt nur die Maximal- und Minimalwerte der Frequenzen nutzen, so in etwa: Link Geht das nicht für diese Karte?

Für WLP bei Grafikkarten hab ich keine Empfehlung und keine Erfahrung. Aber bitte vorher mal recherchieren, damit du weißt, was du tust.
The temperatures 55/75 and 60/85 already show rather big differences of 20°C or more. The absolute values, especially the hotspot, are ok, but the relative difference already shows, that there might be an issue. And then yesterday's 73/111 reveals a difference of 38, which is far too much.

What is the third setting?

When looking again to the three images you posted above: You try to blow rather much power/voltage into your card. 1,167 V is a lot, 261W is beyond 'a lot'. High voltage and power result in high temperatures. So besides the assumed poor heat conduction you also should lower the power input. Read here or at hardwareluxx.de or other internet forums and sources about UV (undervolting) of Radeon cards. In most cases it works fine and you get the same performance than stock performance or even slightly more by investimg significantly less power/voltage. So UV also would help your card.

BTW: you speak Spanish and English, right? The German texts are machine translated? There's no need to translate into German here, most of us speak English and the machine translation sometimes makes the texts difficult to understand.

 

[itruji]

Ich würde die Spannung-Frequenz-Kurve garnicht aktivieren. Entweder im Treiber den Bereich komplett ausschalten oder halt nur die Maximal- und Minimalwerte der Frequenzen nutzen, so in etwa: Link Geht das nicht für diese Karte?

Für WLP bei Grafikkarten hab ich keine Empfehlung und keine Erfahrung. Aber bitte vorher mal recherchieren, damit du weißt, was du tust.
The temperatures 55/75 and 60/85 already show rather big differences of 20°C or more. The absolute values, especially the hotspot, are ok, but the relative difference already shows, that there might be an issue. And then yesterday's 73/111 reveals a difference of 38, which is far too much.

What is the third setting?

When looking again to the three images you posted above: You try to blow rather much power/voltage into your card. 1,167 V is a lot, 261W is beyond 'a lot'. High voltage and power result in high temperatures. So besides the assumed poor heat conduction you also should lower the power input. Read here or at hardwareluxx.de or other internet forums and sources about UV (undervolting) of Radeon cards. In most cases it works fine and you get the same performance than stock performance or even slightly more by investimg significantly less power/voltage. So UV also would help your card.

BTW: you speak Spanish and English, right? The German texts are machine translated? There's no need to translate into German here, most of us speak English and the machine translation sometimes makes the texts difficult to understand.

Writing in english is easy for me, better than use google translate to write in german.
My GPU doesn't has the option min and max frequency and voltage you shown in the link, this is why I use the curve.
I ordered Grizzly Kryonaut that will arrive at the end of this week.
As I am a teacher and I start lessons next week, I have reinstalled my old rx 570 (just to keep the maximum stability of my PC).
And when I have enough free time, I will try again to fix the GPU.

 

[Mini_Me]

Writing in english is easy for me, better than use google translate to write in german.
My GPU doesn't has the option min and max frequency and voltage you shown in the link, this is why I use the curve.
I ordered Grizzly Kryonaut that will arrive at the end of this week.
As I am a teacher and I start lessons next week, I have reinstalled my old rx 570 (just to keep the maximum stability of my PC).
And when I have enough free time, I will try again to fix the GPU.

Greetings,

If you like, I can optimize the vbios file for your card for free, and believe me many members were satisfied with the optimization,

You can send me the latest vbios file for your card, after I finish I will update you whenever it is possible,

If you can identify the memory type of your card that will be a bonus.

 

[itruji]

Greetings,

If you like, I can optimize the vbios file for your card for free, and believe me many members were satisfied with the optimization,

You can send me the latest vbios file for your card, after I finish I will update you whenever it is possible,

If you can identify the memory type of your card that will be a bonus.

@Mini_Me
Thank you so much for the offer, I accept it !!!
This thursday the thermal paste will arrive and I'll apply at the weekend (I will take a look to identify RAM manufacturer).

 

[Mini_Me]

[USER = 5105] @Mini_Me [/ USER]
Thank you so much for the offer, I accept it !!!
This thursday the thermal paste will arrive and I'll apply at the weekend (I will take a look to identify RAM manufacturer).

Very well, please take your time, no need to rush,

By the way, if there is a fujipoly's factory near your area then you can request a free sample of their premium thermal pad Sarcon XR-UM, 17 W/m-k, 2 mm thickness, for the memory chips.

 

[itruji]

Very well, please take your time, no need to rush,

By the way, if there is a fujipoly's factory near your area then you can request a free sample of their premium thermal pad Sarcon XR-UM, 17 W / mk, 2 mm thickness, for the memory chips.

@Mini_Me

I just open my gpu and reapply the thermal paste.

I contacted the local Fujipoly's seller but no response yet.

I attach a RAM picture (seems Micron) and the original msi gaming X bios (I attatch two other bios that I found at techpowerup with different powerlimits, TDC GFX and SoC, if it can help).

I will be waiting for you to send me the BIOS with the modifications you mentioned.

Thanks in advance !!

 

[Mini_Me]

@Mini_Me

I just open my gpu and reapply the thermal paste.

I contacted the local Fujipoly's seller but no response yet.

I attach a RAM picture (seems Micron) and the original msi gaming X bios (I attatch two other bios that I found at techpowerup with different powerlimits, TDC GFX and SoC, if it can help).

I will be waiting for you to send me the BIOS with the modifications you mentioned.

Thanks in advance !!

Great,

Actually I managed to get the latest vbios from MSI for your card a couple of hours ago, I am going to send you the optimized vbios file after I finish whenever it is possible.

 

[itruji]

Great,

Actually I managed to get the latest vbios from MSI for your card a couple of hours ago, I am going to send you the optimized vbios file after I finish whenever it is possible.

That sounds great !!!
Only one question: I see that my vram is micron 9UA77 D9WCW, but in others teardown of the same model i have seen Micron 9KA77 D9WCW.
Is this an issue to consider?

 

[Mini_Me]

That sounds great !!!
Only one question: I see that my vram is micron 9UA77 D9WCW, but in others teardown of the same model i have seen Micron 9KA77 D9WCW.
Is this an issue to consider?

Well that a very good news as I believe the U version is newer and better than K version.