AMD Confirms Its GPU Drivers Are Overclocking CPUs Without Asking

[ryao]

Summary

AMD added a feature to the Windows driver that automatically over clocks Ryzen CPUs. It was supposed to be optional, but a bug is causing it to be active even when it is not supposed to be.

 

Quotes

Quote

 AMD has confirmed to Tom's Hardware that a bug in its GPU driver is, in fact, changing Ryzen CPU settings in the BIOS without permission. This condition has been shown to auto-overclock Ryzen CPUs without the user's knowledge.

 

My thoughts

Given that the official line is that overclocking voids the warranty, this means that AMD’s software is effectively voiding warranties despite the user doing nothing wrong. Those of us using Linux do not have to worry about this.

 

Sources

 https://www.tomshardware.com/news/amd-confirms-its-gpu-drivers-are-overclocking-cpus-without-asking

[X-System]

It means the benchmarks between AMD GPU and NVIDIA GPU with AMD CPU by YouTubers are wrong because NVIDIA GPU driver doesn't overclock AMD CPU ?

[AnonymousGuy]

Insert suspicion that AMD was trying to inflate their own benchmark numbers to make AMD CPU + GPU look better.

[Mark Kaine]

This explains why my PC instantly runs like shit whenever i install an AMD GPU+driver. I guess?  (i would love this to be a joke but Im actually serious lol) 

 

 

3 hours ago, ryao said:

changing Ryzen CPU settings in the BIOS

But… how does that even work… do you need TPM or what? 

 

ps: also serious, for example MSI OC software cannot do that on MSI board, it has to load the software / profile in Windows first - maybe thats what they're saying?

[Moonzy]

Does this mean that GPU benchmarks done on AMD CPU platform is all inaccurate?

 

Or still accurate since it's how it'll run anyways

[Deadpool2onBlu-Ray]

5 hours ago, ryao said:

Summary

AMD added a feature to the Windows driver that automatically over clocks Ryzen CPUs. It was supposed to be optional, but a bug is causing it to be active even when it is not supposed to be.

 

Quotes

 

My thoughts

Given that the official line is that overclocking voids the warranty, this means that AMD’s software is effectively voiding warranties despite the user doing nothing wrong. Those of us using Linux do not have to worry about this.

 

Sources

 https://www.tomshardware.com/news/amd-confirms-its-gpu-drivers-are-overclocking-cpus-without-asking

This is INSANE 

[dizmo]

And this after a few people tried to tell me AMD drivers haven't had problems for years... 

Eh @Helpful Tech Wiard? 

[Helpful Tech Witch]

5 minutes ago, dizmo said:

And this after a few people tried to tell me AMD drivers haven't had problems for years... 

Eh @Helpful Tech Wiard? 

No need to spread an argument across multiple threads

[Mister Woof]

haha jokes on them

 

i use AMD GPUs with Intel CPUs

[Moonzy]

20 minutes ago, dizmo said:

And this after a few people tried to tell me AMD drivers haven't had problems for years... 

What? They probably never tried to use AMD's shadowplay equivalent then

 

Don't think I've had as much issue on Nvidia's driver throughout the years than me with AMD drivers in the half year I used an rx460 in my secondary PC

[spartaman64]

oh its boosting the cpu clock speed also. but if it works so well for gpus is this really an issue? and nobody is going to know you oced if you dont tell them lol

[porina]

17 minutes ago, spartaman64 said:

oh its boosting the cpu clock speed also. but if it works so well for gpus is this really an issue? and nobody is going to know you oced if you dont tell them lol

AMD CPUs running stock already boost. Question then is, how is it doing overclocking? If it is as simple as increasing power/current limits, this is relatively safe. AMD historically have enforced a PPT limit, although I hear some mobos may OC by default by enabling PBO. I've not owned one though. Intel do not consider adjusting power limits to be overclocking and thus warranty voiding, which AMD does.

[RejZoR]

Everyone making drama how AMD is gaining advantage where in fact, "overclocking" Ryzen CPU's usually just means REGRESSION in performance because system just dumps more power into CPU with raised PBO parameters and Ryzens just poop themselves when they get more power at same cooling capability as with stock, because their entire algorithm is heavily temperature biased. More power being dumped into them means more heat and more heat means lower clocks, not higher. And if anyone thinks raising PBO max clock limits does anything, it doesn't. It just shows nice clock frequency numbers in monitoring tools, but in reality, you'll again just regress in performance because clock stretching is somehow a thing (been overclocking since Celeron 333MHz days and never heard or experienced such a thing till now with Ryzen CPU's).

 

Bottom line, in 99% of cases, "overclocking" or "increasing power limits" of Ryzen will just result in REGRESSION of performance, not increase. If anything, AMD needs to fix it because it would actually make them look worse in benchmarks...

[dizmo]

1 hour ago, Moonzy said:

What? They probably never tried to use AMD's shadowplay equivalent then

 

Don't think I've had as much issue on Nvidia's driver throughout the years than me with AMD drivers in the half year I used an rx460 in my secondary PC

I was still choosing between an RX6700XT and a 2070Super, but someone's coming tomorrow to buy the 2070 Super so I guess RX it is!

Hopefully I don't encounter too many issues. I just play a few games and don't stream or anything like that, so fingers crossed...

[Mark Kaine]

On 4/8/2022 at 7:28 PM, porina said:

how is it doing overclocking

yeah… i mean i already asked, but more simplified… *can windows software change BIOS settings and how?*

 

 

If thats the case, isnt that a huge security risk and why arent there more programs doing this? Afaik all overclocking programs such as Ryzen master, Afterburner etc, load their settings on startup, not earlier (and even then Im not sure how they do it, somehow  they must access the BIOS, in a way?) 

 

 

Also, bonus? IF windows software can change BIOS settings, why arent there programs that allow the user to do just that? Would cut out the whole "spam delete button and then try changing settings in a laggy as heck BIOS at 0.5fps…" thing… lol. 

 

 

ps: I get it, this will work similar to how Ryzen master works, the question remains… how does it work and how is it *not* a huge security risk?  What stops random software to set CPU voltage to 50v or something, *if* its that easy?? o.o 

[porina]

2 hours ago, Mark Kaine said:

yeah… i mean i already asked, but more simplified… *can windows software change BIOS settings and how?*

Then my question was a little different. I wasn't so concerned about the mechanism, but the style of overclock. Power limit adjustment I not bothered with. Other methods will be higher risk of instability.

 

2 hours ago, Mark Kaine said:

Afaik all overclocking programs such as Ryzen master, Afterburner etc, load their settings on startup, not earlier (and even then Im not sure how they do it, somehow  they must access the BIOS, in a way?) 

OC software have some way to talk to the relevant parts related to control of whatever chip. It doesn't necessarily follow that every setting in bios is exposed that way.

 

In the short time I tinkered with Intel Optane Memory, that software seems able to adjust the storage mode needed for it to work. I think it needs raid so it configures both bios and windows accordingly.

 

2 hours ago, Mark Kaine said:

Would cut out the whole "spam delete button and then try changing settings in a laggy as heck BIOS at 0.5fps…" thing… lol. 

shutdown /r /fw /t 0

 

Run above in cmd as admin to reboot the system to bios. I think it needs uefi era system to work.

[porina]

1 hour ago, huilun02 said:

Would imply that GPU benchmarks using AMD processors have been showing cosistently higher CPU frequency, power draw, and temps when also using an AMD GPU

 

Any nobody noticed

Someone did notice. If not we wouldn't be discussing it now.

 

It will be a short term embarrassment for AMD. Like any error like this, they'll fix it, put in place measures to prevent a repeat, and everyone will move on.

 

Maybe more reviewers will observe for similar in future now that we are aware of this possibility.