Du verwendest einen veralteten Browser. Es ist möglich, dass diese oder andere Websites nicht korrekt angezeigt werden. Du solltest ein Upgrade durchführen oder einen alternativen Browser verwenden.
[Sammelthread]Offizieller AMD RX 9070 (XT) Overclocking und Modding Thread
How to do it software was explained by me. I2C with soldering wires was indeed older idea. But I have explained how to do it software, I posted on 2nd of May 2025 here:
This way people don't have to solder anymore, they can temporary or permanently alter their RDNA3 and RDNA4 power levels, + other modifications (VID etc, current limit etc).
I also have no problem with others replicating my work, I encourage it. I was just expecting common sense that is all.
It is also not "my behavior", I came with good intentions and your friend bullied and ridiculed me, for not knowing as much as he does. I cannot accept such behavior, I do not care who you are or how famous you are in the overclocking world. He was wrong on my project and I didn't want to accept it, sorry.
Again, everyone can build as many tools using this idea, I was just expecting common sense, that is all.
Beitrag automatisch zusammengeführt:
Actually just so you don't think I don't like sharing, here is something you all might be interested in, the full MP2868A register list, ALL of them, with explanations for every single bit.
Datenschutzhinweis für Youtube
An dieser Stelle möchten wir Ihnen ein Youtube-Video zeigen. Ihre Daten zu schützen, liegt uns aber am Herzen: Youtube setzt durch das Einbinden und Abspielen Cookies auf ihrem Rechner, mit welchen sie eventuell getracked werden können. Wenn Sie dies zulassen möchten, klicken Sie einfach auf den Play-Button. Das Video wird anschließend geladen und danach abgespielt.
Youtube Videos ab jetzt direkt anzeigen
If you follow the description of video, you will run into a list of interesting dumps. With small modifications you can get to full register list. You should be able to figure it out.
Ja, war echt eine gute Idee von dir. In igors Forum hattest du ja auch gleichgesinnte gefunden.
Das hast du dir aber dann ganz schnell mit Beschimpfungen und fehlender Kritikfähigkeit verscherzt.
Und wenn du genau hinschaust, nimmt das RebelsTool andere Register für die Power.
Wenn Du diese Anzeige nicht sehen willst, registriere Dich und/oder logge Dich ein.
Ja, war echt eine gute Idee von dir. In igors Forum hattest du ja auch gleichgesinnte gefunden.
Das hast du dir aber dann ganz schnell mit Beschimpfungen und fehlender Kritikfähigkeit verscherzt.
Und wenn du genau hinschaust, nimmt das RebelsTool andere Register für die Power.
If I bring something to the table, you keep your criticism to yourself. That's common sense I think. I was friendly with all of you until some of you started acting weird towards me.
As for power altering, there's offset and there's also scaling, which I recently added to my script after I ran into the full register list for MP2868A, there's more ways to do it. But this is besides the point anyway, this isn't about what registers or this nonsense, I just offered the full register list (with a bit of work you get to it), it is about common sense courtesy. My work is open, else it would be a binary with no clue how it does what.
If you build based on someone else's idea, the least bit you can do is mention it, don't you think? Without the "how" you cannot make your tool anyway.
Beitrag automatisch zusammengeführt:
In the resource I shared you will also be able to get full register list with explanation for each of them, in detail, also for MP2856 and MP2857, so both RDNA3 and RDNA4, full VR register list, you can fully reprogram every single aspect of it (most likely), via I2C. RDNA4 still a mystery with saving page 0x02 mods. Must be some special command, like 0x15 is for pages 0/1. What should be EEPROM pages are 0x0000 on all registers.
Block read of 3 bytes reveals one more byte, but doesn't seem related. Per-phase current also is a bit strange, might require block read for more data, there's third byte there as well, changing with load.
Das habe ich nur am Rande mitbekommen, aber hast einigen so auf die Füße getreten, das du aus Igors Forum gelöscht und gebannt wurdest.
Das muss man erstmal schaffen.
Das habe ich nur am Rande mitbekommen, aber hast einigen so auf die Füße getreten, das du aus Igors Forum gelöscht und gebannt wurdest.
Das muss man erstmal schaffen.
🤭 Nobody banned me. I asked Mr. Wallossek to delete my user name and close my account because I didn't want to be associated with that forum, he can attest to that. I left that place because I felt unwelcome there, no reason for me to go somewhere where I am not appreciated. Not sure who told you I was banned. There was someone there who wanted to ban me, but he didn't manage to, but in the end I fulfilled his wish.
I think you got it all wrong.
🤭 Nobody banned me. I asked Mr. Wallossek to delete my user name and close my account because I didn't want to be associated with that forum, he can attest to that. I left that place because I felt unwelcome there, no reason for me to go somewhere where I am not appreciated. Not sure who told you I was banned. There was someone there who wanted to ban me, but he didn't manage to, but in the end I fulfilled his wish.
I think you got it all wrong.
Yes, took a few emails to convince Mr. Wallossek to do it, because normally they cannot, so he offered the solution to add my name to the "naughty list" of words, so it gets masked with **** 😁. That was still me, I insisted my name is not visible there, initially he didn't want to do it but he managed to come up with this solution, which I appreciated.
As for power altering, there's offset and there's also scaling, which I recently added to my script after I ran into the full register list for MP2868A, there's more ways to do it.
I made a recent update that added power limit scaling instead of an offset, so now you can have accurate readings in HWinfo if you add the same multiplier for the sensors. With power offset you had no reported power if it was under the offset value. For example with 100W offset, any power draw under 100W shows as 0W in HWinfo.
But with scaling it shows from lowest level, no gap in reporting.
You can use any tool that fits your needs, I do not insist people use mine. If anyone else has more/better options use that, I just wanted to share the idea with everyone.
As for the other discussion, that's related to tribal politics from another forum, and there is apparently more than one person trying to pass my idea as their own, without giving me any credit for it. Strange times.
Scribd is the source for 300M+ user uploaded documents and specialty resources.
www.scribd.com
From it I was able to figure out which register offsets power for RDNA4, since VR is similar to it. That was after I figured out TBP is a composite value made up from GPU Core + SoC Input Power sensors, which was a clue the data comes from VRs, so it can be altered, when everyone was saying it cannot be done for RDNA4 because it has current measuring chips at connector, it's not VR reporting anymore.
And I was able to do that after I found MP2856 Linux kernel driver, which offered more clues + monitoring. I shared all of these resources. Even if you find some new register that does something, you wouldn't have found it without the key work I did. I had to open a few doors so you can "find" another way of doing it.
For you to find "another way" you needed me to show you how to access the VR from Linux, how to permanently save mods, full RDNA3 register list with names, which you can use as clues for other register functions that can enable power limit increase.
The registers you looked at didn't come from nowhere. Same for the full list I posted few posts back. After you see their names yes, not so hard finding something that works, if you are willing to test different values in different registers (this time you have software which shows full details on every register).
I tell you PIN_IIN_SYS_OFFSET offsets the power level, if you test IIN_CONFIG register yes, most likely you will find another way of doing it. There would be no IIN_CONFIG register for you to test if I didn't show you the whole thing, isn't it? You suddenly having an idea after I show it and explain it to you is kind of sus lol.
But finding new functions/registers/etc is fine, even building more tools is fine. But this is the second person from the same group "of friends" trying to pass my ideas as their own, without a single mention of my project, like they came up with them themselves, instead of their "ideas" being fully based on the work that I did, that is my problem. Their "ideas" wouldn't be possible without the numerous key steps I spoon fed them, based on that yes, you find "different ways".
This is not only about "I told you about i2c-tools and SMU 0 on Linux", this is about numerous key points where my work enabled the next step in the process, I built upon it for months, so in the end we have a RDNA4 solution. Then you use it to simply test a few other registers from what I worked so hard to find/figure out, and pass your findings as your own work, or even worse, repackaging my information and passing it as your EVC2 XML contribution, that is insane.
With power offset you had no reported power if it was under the offset value. For example with 100W offset, any power draw under 100W shows as 0W in HWinfo.
I increased TPB to 410W via your "rdna4.sh" script (not the one wilth "old"):
(btw, the card should have a stock TBP of 330W.)
And it confirms the offset:
But HWinfo is reading values <69W at idle, too:
I confirmed via power monitoring at the wall socket that the limit is increased, so I believe its already the "new scaling" rather than the "old offset"?
As for the other discussion, that's related to tribal politics from another forum, and there is apparently more than one person trying to pass my idea as their own, without giving me any credit for it. Strange times.
Besides that I think one should respect and give credit to other peoples work (and of course I have my own subjective opinion), I try to stay neutral as I do not have the technical knowledge to join the discussion.
But I will check out OCN, too.
I increased TPB to 410W via your "rdna4.sh" script (not the one wilth "old"): Anhang anzeigen 1178861
(btw, the card should have a stock TBP of 330W.)
And it confirms the offset: Anhang anzeigen 1178862
But HWinfo is reading values <69W at idle, too: Anhang anzeigen 1178863
I confirmed via power monitoring at the wall socket that the limit is increased, so I believe its already the "new scaling" rather than the "old offset"?
Besides that I think one should respect and give credit to other peoples work (and of course I have my own subjective opinion), I try to stay neutral as I do not have the technical knowledge to join the discussion.
But I will check out OCN, too.
If you get the latest one you can add a power scaling user preset in the script file. You can calculate the factor by dividing your wished power limit to your actual one. If you have 363W limit and want ~435W you 435/363=~1.2, so you set 1.2 as scaling factor for TBP.
You can also run it with -m and get to the menu, which is interactive. Both offset and scaling methods work fine. The offset one can be permanently saved, the scaling one cannot, it goes away on power off.
As for the other discussion yeah, I felt like I should mention the fact that at least a passing mention somewhere at the bottom would have been appropriate, even if the author doesn't like me.
Well, after the fact it's kind of hard to believe. What made you think of a register anyway? Did you have this random idea all of a sudden? Just like that? Out of nowhere, people using EVC2, and you thought
What if, I use Linux, I install I2C tools, somehow SMU 0 is accessible with no reason whatsoever for this thought I'm now having, and not only that, maybe I somehow find a full list of register names for RDNA3 VRs, and based on it I get some ideas of what other registers I could modify so I get another different way of increasing the RDNA4 power limit, which would also for some reason be fully different than the method anyone else might have come up before me.
Who has an idea like that, out of nowhere? Strange idea no? After years of no ideas, you suddenly have this idea. Somehow right after I give you a lot of details about everything. The mind boggles.
Glad I could help, enjoy it!
If anyone wants more options I would consider adding them to my script. MP2868A has a new SVI3 VID offset register, seems pretty similar in behavior to the classic VID one. Another interesting one is VID trim register, seems to exhibit slightly different behavior to VID offset, has some 63-64mV max offset, +/-. Since we now have full register details for RDNA4 we can modify almost anything.
PWM frequency can be changed, it's 560KHz/500KHz set on both VRs, at least on mine. GFX/SoC are 560KHz. Will try to go a bit up/down see what happens.
Well, after the fact it's kind of hard to believe. What made you think of a register anyway? Did you have this random idea all of a sudden? Just like that? Out of nowhere, people using EVC2, and you thought
Who has an idea like that, out of nowhere? Strange idea no? After years of no ideas, you suddenly have this idea. Somehow right after I give you a lot of details about everything. The mind boggles.
I will say again, I'm not looking for that, I was just expecting a small mention somewhere at the end, that's it. Minimum common sense, nothing else. Even if we don't get along on other things, we can avoid awkward situations like this, that is why it is the mature thing to do. But also thank you. I also hope you get what I am trying to say.
