News Software Bug Lets You Overclock AMD's 5800X3D to Death In Seconds

setx

Distinguished
Dec 10, 2014
224
149
18,760
What you entered absurd multiplier into GUI without even applying means... exactly nothing! Even if you applied that it's not dangerous as multiplier can't kill the CPU. (And CPU likely would just ignore that.)

Applying voltage/power/temperature beyond practical limits is a serious issue though.
 

DingusDog

Prominent
Apr 12, 2022
23
9
515
Might want to proof read the very first sentence of your article. Maybe it was AMD's best CPU for gaming but is it still? Think about it for a minute.
 
  • Like
Reactions: helper800
What you entered absurd multiplier into GUI without even applying means... exactly nothing! Even if you applied that it's not dangerous as multiplier can't kill the CPU. (And CPU likely would just ignore that.)

Applying voltage/power/temperature beyond practical limits is a serious issue though.

Actually this is not always true. Most overclocking applications either grey out or lock out any voltage or multiplier inputs that are out of bounds from the CPU's requirements. Before you hit apply.
 

bjnmail

Honorable
Jul 20, 2018
16
15
10,515
Not at all surprising. This is why I never install any of the motherboard vendors' tuning software on my own or clients' systems, as they have all been prone to this for ages. The UEFI has limits imposed, but the software ignores them and sets the values directly, which can lead to exactly these chip-killing issues.

Obviously some daring overclockers will have fun with this, assuming raising the multiplier actually takes effect like the voltage does, but regardless, it could lead to some slight improvements in performance if you're conservative with settings (or don't touch the voltage at all).
 
  • Like
Reactions: Friesiansam

razor512

Distinguished
Jun 16, 2007
2,130
68
19,890
What you entered absurd multiplier into GUI without even applying means... exactly nothing! Even if you applied that it's not dangerous as multiplier can't kill the CPU. (And CPU likely would just ignore that.)

Applying voltage/power/temperature beyond practical limits is a serious issue though.
I think it may be fear of applying any change when the software is buggy. For example if someone's CPU died when the software was set to 1.3V even though the chip should handle 1.35V, there is a chance that what is being applied may not be what was specified in the GUI.
 
  • Like
Reactions: rluker5
I don't understand what all the fuss is about. AMD has stated over and over that the R7-5800X3D is not supposed to be overclocked.

So... someone overclocks a CPU that doesn't support overclocking and it dies.

My response to that is "Ok...and?".
 
  • Like
Reactions: PEnns
I don't understand what all the fuss is about. AMD has stated over and over that the R7-5800X3D is not supposed to be overclocked.

So... someone overclocks a CPU that doesn't support overclocking and it dies.

My response to that is "Ok...and?".
The fuss is that any normal modern CPU, for many years now, has protections build into the chip itself so that even if you put the settings to something insane the CPU will still adjust itself and not burn out.
At least not immediately as soon as you press the button.
This isn't just a software bug, it's a bad CPU.

This is how a modern CPU should react to stupidly high settings.
View: https://www.youtube.com/watch?v=jv3uZ5Vlnng
 

sonichedgehog360

Distinguished
Dec 11, 2008
23
10
18,515
"however, our testing indicates that the limitations almost certainly have something to do with the chip's heat dissipation issues."

Classic tech reviewer mistake. Rule #1: Correlation does not signify causation. Just because there is heat build-up does not mean that is the root cause of the failure. Unless you have an electron microscope and examined the die afterward, you are unqualified to make this assessment. AMD, on the other hand, has access to these essential tools and are the computer engineers who designed the chip, not Tom's Hardware. Therefore, only they are qualified to make the statement that failure would be due to the 3D cache's voltage limitations.
 
  • Like
Reactions: P1nky

peterf28

Distinguished
Apr 7, 2013
111
19
18,585
On my old P8Z68-M mainboard and i5-2500K, when I raised the multiplier in BIOS to 50x, the mainboard automatically upped the voltage to more than 1.5V under heavy load. Result: CPU destroyed in less than 5 minutes in Prime95. Because there was a setting activated by default which somehow automatically adjusted the voltage in relation to the multiplier or frequency. I don't remember which setting it was. So I bought new CPU and had to manually adjust voltage offset to negative value when overclocking, to stay under 1.4V under load.
 
And that is why the old thought of step by step still applies even on modern chips.

Being able to unlock the chip is nice, get the big cache and the big clock at once. With double checking every related setting. Auto is good when everything is on auto, but change one side of the voltage/clock balance, and the other will move as well, if left on auto.
 
  • Like
Reactions: PEnns

PEnns

Reputable
Apr 25, 2020
580
605
5,760
Just because your car allows you to drive 200 mph around a steep curve...it doesn't mean you should!!

Sometimes I question the sanity of of some allegedly smart techies. But clickbait and YT craziness sells and brings cash.

PS: RTFM, even if you're a tech "god."
 

Vanderlindemedia

Prominent
Jul 15, 2022
91
50
560
Normally on idle a high voltage shoud'nt kill a CPU. The degradation however occurs instant when high voltages are pushed with high currents.

That cache is just extremely sensitive to higher voltages then 1.35V. You will fry it. And if that happens the cpu is dead.