I used to have an old junk GPU that began overheating and automatically shutting off, I ended up trying undervolting to keep it alive which did indeed work

>>3854
Yeah, that was my other assumption about undervoltinng. I thought it was a last ditch effort to save a card. Turns out it should be done as soon as you get the card home so you can keep it running efficiently and make it last longer.

I try to do this with all my cards. It seems like a no-brainer. Sometimes you can overclock and undervolt at the same time if you're lucky.

At most I limit my cpu usage when it goes to 70-100% since that puts my temp at 100C.
I don't play that many performance-heavy games but once palword server goes up I think this is worth a try. My hardware already struggled with the game some but if performance hit is minimal like you're saying I'm willing to take it to lower the temps.

Wonder if I can do it to my laptop's igpu chip and get similar benefits... Been hitting +100C a lot recently, wouldn't be surprised if the chip's burnt. It's a shittop but it still is something I can make good use of.
God I hate laptops.

>Misconfiguration of CPU voltage settings might result in permanently damaged hardware. You have been warned!
I don't know if I want to risk it...

>>3854
>>3872
repaste your chips
>>3875
undervolting won't result in any damage unlike overvolting

Never done either! I've always been aware of undervolting, but never thought there would be much of a reason to even look into it. I'd also be worried about encountering performance issues in "unforeseen areas" even if other things seem to work the same after undervolting, though I might still end up experimenting with this after I got some new hardware. Thanks for the idea!

Also! Would you happen to be able to help me out with this? >>3790
I'm wondering because generative AI, as well as editing, is my main reason for wanting to update after over a decade and thought perhaps you might have a few thoughts on that.

>>3876
>undervolting won't result in any damage unlike overvolting
I would be VERY dubious of this claim. If I remember right, part of the reason that 13/14th gen Intel CPUs had an issue with CPU degradation was that they were improperly thermal velocity boosting; CPU voltage would remain low, but wattage would increase, increasing resistance and long-term, leading to degradation.

I don't think it's out of the question that changing the operating voltage of a CPU may lead to damage, regardless of whether you're over- OR under-volting. In either case, you're naively changing the operating voltage to a highly complex part that not even motherboard manufacturers are adept at setting reasonable limits for... On the other hand, let's say you manually decrease the PL1 and PL2 TDP wattage limits, that should be safe because the CPU and BIOS will automatically apply values based off of the TDP wattage limits instead of changing a single value which has unforeseeable downstream effects.

>>3884
Your reasoning doesn't apply to GPUs which have more integrated control than CPU+motherboard. You can't change voltage offset at all on modern GPUs, they are totally locked down, and the only things you can change are power limit and frequencies.
"Undervolting" on GPUs is no different from power limit + overclocking.

>>3880
>Would you happen to be able to help me out with this?
Will do!

>>3875
I don't think it can cause damage, but I admit I've mostly read up on GPU stuff to start with. GPU undervolting is a lot simpler since you can do it from a UI in Windows. I do plan to mess with CPU stuff soon after I get this stuff stable.
GPUs are the main power hogs of modern computers these days as it is, although with laptop CPUs often have integrated graphics I'm not sure how it works.

It's been a decade and I finally upgraded from my i7-4770K to an i9-12900KS! It's paired with a Z790 Aorus Elite X Wifi7 and 2x 48GB CL34 RAM seemingly running stable at 6800. I've been very, very happy with this system so far!

I haven't really tinkered with CPU settings ever since I got my 4770K new, but I kept this thread in mind and have been experimenting some with undervoltig the 12900KS. Running Cinebench R23 at stock will immediately get most cores to 100+ C (D15 cooler, fans full speed), so I've steadily been going down and an offset of -0.130V seems to work the best for now, reducing the temps under load to around 90 - 93C and increasing my score from around 28750 to 29282. I guess that means I'm doing at least something right!

>>4349
Sounds good! The point difference seems negligible when it comes to performance.
After a bunch of GPU tinkering (I'm OP) I noticed that I can get a Stable Diffusion generation from 500-550 watts to 340-400 watts on a 5090 at the price of going from 40.2 seconds to 40.4 in generation time.
It really is excessive at how much power these things are set to receive by default.

>>3852
sounds like a really stupid thing that could be fixed by installing one more fan by your computer chasis