Also be careful with overclocking, because the usual advice of "just running EXPO/XMP" often results in motherboards setting voltages on very sensitive components to more than 30% over their stock values, and this is somehow considered normal.

It looks completely bonkers to me. I overclocked my system to ~95% of what it is able to do with almost default voltages, using bumps of 1-3% over stock, which (AFAIK) is within acceptable tolerances, but it requires hours and hours of tinkering and stability testing.

Most users just set automatic overclocking, have their motherboards push voltages to insane levels, and then act surprised when their CPUs start bugging out within a couple of years.

Shocking!

▲

danieldk 5 days ago | parent | next [-]

Unfortunately, some separately purchasable hardware components seem to be optimized completely for gamers these days (overclocking mainboards, RGB on GPUs, etc.).

I'd rather run everything at 90% and get very big power savings and still have pretty stellar performance. I do this with my ThinkStation with Core Ultra 265K now - I set the P-State maximum performance percentage to 90%. Under load it runs almost 20 degrees Celsius cooler. Single core is 8% slower, multicore 4.9%. Well worth the trade-off for me.

(Yes, I know that there are exceptions.)

▲

hedora 4 days ago | parent | next [-]

I’ve had multiple systems that crash or corrupt data when underclocked, so running at 90% might not be what you want.

You can always play with the CPU governor / disable high power states. That should be well-tested.

▲

anonymars 4 days ago | parent | next [-]

It sounds like you are conflating undervolting for underclocking. Undervolting runs it out of spec, while underclocking simply runs it slower

▲

danieldk 4 days ago | parent | prev [-]

Setting a P-State max percentage is completely reliable/stable, it will clock lower on average to hit the lower performance target. It’s kinda similar to setting a powersave governor, but more granular.

I think you are confusing with undervolting.

	▲	hedora 2 days ago \| parent [-]
		No. For giggles, set your DRAM frequency (MT/sec) to something crazy low, then try using the machine for a week.

▲

mschuster91 5 days ago | parent | prev [-]

> Unfortunately, some separately purchasable hardware components seem to be optimized completely for gamers these days

It turned out during the shitcoin craze and then AI craze that hardcore gamers, aka boomers with a lot of time and retirement money on their hands and early millennials working in big tech building giant-ass man caves, are a sizeable demographic with very deep pockets.

The wide masses however, they gotta live with the scraps that remain after the AI bros and hardcore gamers have had their pick.

▲

danieldk 5 days ago | parent | prev | next [-]

Also relevant in this context:

https://www.pugetsystems.com/blog/2024/08/02/puget-systems-p...

to;dr: they heavily customize BIOS settings, since many BIOSes run CPUs out-of-spec by default. With these customizations there was not much of a difference in failure rate between AMD and Intel at that point in time (even when including Intel 13th and 14th gen).

▲

ahartmetz 5 days ago | parent | prev | next [-]

Since you mention EXPO/XMP, which are about RAM overclocking: RAM has the least trouble with overvoltage. Raising some of the various CPU voltages is a problem, which RAM overclocking may also do.

	▲	eptcyka 5 days ago \| parent [-]
		The heat is starting to become an issue for DDR5 with higher voltage.

▲

electroglyph 5 days ago | parent | prev [-]

yah, the default overclocking stuff is pretty aggressive these days