F5F Stay Refreshed Hardware Desktop Yes, undervolts tend to lose stability as they age.

Yes, undervolts tend to lose stability as they age.

Yes, undervolts tend to lose stability as they age.

Pages (2): 1 2 Next
X
xFqtal_
Senior Member
670
08-05-2018, 02:03 PM
#1
I own five Ryzen 7950X processors running in 45W ECO mode with a -0.105 V undervolt setting. They’re connected together for 3D rendering tasks, often staying active for days or weeks at a time, effectively running continuous Cinebench sessions. Initially I set the offset to -0.110 V just to test reboot stability during Cinebench; however, I soon realized that surviving short bursts doesn’t guarantee endurance over long render periods. After each restart I lower the voltage further, aiming for a stable setting that won’t trigger issues later. Recently, several machines have rebooted more often than expected, even though they’d functioned well for months. I’m wondering if recent rendering tasks—perhaps involving different CPU instructions or higher loads—might be affecting stability. It’s also possible the chips are gradually degrading, though that seems unlikely given their high operating temperatures remain low (around 61°C in summer). The main concern is whether undervolt stability can truly be maintained over time.
X
xFqtal_
08-05-2018, 02:03 PM #1

I own five Ryzen 7950X processors running in 45W ECO mode with a -0.105 V undervolt setting. They’re connected together for 3D rendering tasks, often staying active for days or weeks at a time, effectively running continuous Cinebench sessions. Initially I set the offset to -0.110 V just to test reboot stability during Cinebench; however, I soon realized that surviving short bursts doesn’t guarantee endurance over long render periods. After each restart I lower the voltage further, aiming for a stable setting that won’t trigger issues later. Recently, several machines have rebooted more often than expected, even though they’d functioned well for months. I’m wondering if recent rendering tasks—perhaps involving different CPU instructions or higher loads—might be affecting stability. It’s also possible the chips are gradually degrading, though that seems unlikely given their high operating temperatures remain low (around 61°C in summer). The main concern is whether undervolt stability can truly be maintained over time.

E
EndoHash
Member
196
08-05-2018, 02:03 PM
#2
They remain steady once started or they don't.
E
EndoHash
08-05-2018, 02:03 PM #2

They remain steady once started or they don't.

M
mousse2006
Member
157
08-05-2018, 02:03 PM
#3
It's entirely feasible. Many Overlocks and Undevolts seem to become stable under certain conditions, even though they're known for being inconsistent. I've tested hours of Prime95 before it crashes after just 10 minutes of gameplay, similar to how I expected an overclock to work in Windows but failed when trying to boot a Linux ISO. Each workload reacts differently. I've also noticed this with temperatures—after prolonged use, stability only emerges during warm-up, not at startup. Give it some time, adjust your setup, and keep testing until you find consistency.
M
mousse2006
08-05-2018, 02:03 PM #3

It's entirely feasible. Many Overlocks and Undevolts seem to become stable under certain conditions, even though they're known for being inconsistent. I've tested hours of Prime95 before it crashes after just 10 minutes of gameplay, similar to how I expected an overclock to work in Windows but failed when trying to boot a Linux ISO. Each workload reacts differently. I've also noticed this with temperatures—after prolonged use, stability only emerges during warm-up, not at startup. Give it some time, adjust your setup, and keep testing until you find consistency.

_
_Arctos_
Member
63
08-05-2018, 02:03 PM
#4
When stress tested correctly, components shouldn't show any significant changes until the system is very old. At that stage, the capacitors on the motherboard become a more important cause of instability. Software elements like BIOS, chipset drivers, and the operating system can also contribute to instability in well-tuned systems.
_
_Arctos_
08-05-2018, 02:03 PM #4

When stress tested correctly, components shouldn't show any significant changes until the system is very old. At that stage, the capacitors on the motherboard become a more important cause of instability. Software elements like BIOS, chipset drivers, and the operating system can also contribute to instability in well-tuned systems.

D
demox777
Junior Member
13
08-05-2018, 02:03 PM
#5
What is this supposed to mean? These units seem extremely low, almost like they’re barely functional. It’s like swapping them for cheaper models without realizing it.
D
demox777
08-05-2018, 02:03 PM #5

What is this supposed to mean? These units seem extremely low, almost like they’re barely functional. It’s like swapping them for cheaper models without realizing it.

C
cookiedough909
Posting Freak
782
08-05-2018, 02:03 PM
#6
Values above -0.10 may already be unstable from the start
C
cookiedough909
08-05-2018, 02:03 PM #6

Values above -0.10 may already be unstable from the start

X
xDeviantWolfe
Member
158
08-05-2018, 02:03 PM
#7
It usually takes a long time for an undervoltage to lose stability because of the natural aging of silicon. You might have been close to being stable when you first set it up, but after some months it failed. Raising the voltage should quickly fix the issue for many months ahead. More likely, the device wasn’t stable from the start, and a newer version revealed the problem. Confirming stability is difficult, especially if these renders are essential for your income. I’d recommend reducing the undervoltage now or removing it completely. If a render requires 120 hours, it’s better to skip the undervolting and have it take 130 hours instead of risking losing everything.
X
xDeviantWolfe
08-05-2018, 02:03 PM #7

It usually takes a long time for an undervoltage to lose stability because of the natural aging of silicon. You might have been close to being stable when you first set it up, but after some months it failed. Raising the voltage should quickly fix the issue for many months ahead. More likely, the device wasn’t stable from the start, and a newer version revealed the problem. Confirming stability is difficult, especially if these renders are essential for your income. I’d recommend reducing the undervoltage now or removing it completely. If a render requires 120 hours, it’s better to skip the undervolting and have it take 130 hours instead of risking losing everything.

I
iWonderTiger
Member
137
08-05-2018, 02:03 PM
#8
You can't expect stability if it's already unstable. Heavy loads work for MC, but what about SC loads? Those are the ones that will bring your system down. Also, running 45w Eco on a 7950X is unrealistic—just upgrade your cooler, okay?
I
iWonderTiger
08-05-2018, 02:03 PM #8

You can't expect stability if it's already unstable. Heavy loads work for MC, but what about SC loads? Those are the ones that will bring your system down. Also, running 45w Eco on a 7950X is unrealistic—just upgrade your cooler, okay?

F
flowerjam
Junior Member
13
08-05-2018, 02:03 PM
#9
AMD markets them this way. Dragon range chips fit that description. You'll definitely see it. They're not dealing with temperature problems as OP mentioned—it's about efficiency and power consumption. It's all about watts and energy use. The total watt-hours needed to finish the task is what matters. @OP, lowering voltage won't hurt the CPU, but if you're pushing for maximum stability at a specific frequency, it could cause issues.
F
flowerjam
08-05-2018, 02:03 PM #9

AMD markets them this way. Dragon range chips fit that description. You'll definitely see it. They're not dealing with temperature problems as OP mentioned—it's about efficiency and power consumption. It's all about watts and energy use. The total watt-hours needed to finish the task is what matters. @OP, lowering voltage won't hurt the CPU, but if you're pushing for maximum stability at a specific frequency, it could cause issues.

X
xXMrLuckyXx
Junior Member
33
08-05-2018, 02:03 PM
#10
This reduction in power isn't justified by any gains in efficiency. It actually diminishes the CPU's performance significantly. I don't mind how it runs; it feels like a huge loss of value.
X
xXMrLuckyXx
08-05-2018, 02:03 PM #10

This reduction in power isn't justified by any gains in efficiency. It actually diminishes the CPU's performance significantly. I don't mind how it runs; it feels like a huge loss of value.

Pages (2): 1 2 Next