Re: Things I Don't Need Today

On Thu, 07 Mar 2024 09:46:44 -0500, Spalls Hurgenson
<spallshurgenson@gmail.com> wrote:

On Wed, 6 Mar 2024 11:06:10 -0800, Justisaur <justisaur@yahoo.com>
wrote:
>

On 3/6/2024 7:33 AM, Spalls Hurgenson wrote:

On Tue, 5 Mar 2024 15:37:08 -0800, Dimensional Traveler
<dtravel@sonic.net> wrote:

On 3/5/2024 3:31 PM, Spalls Hurgenson wrote:

 

PC's broke.

So it looks like it /is/ an overclocking issue. Which is annoying
because I don't overclock. My motherboard, on the other hand, DOES. I was aware, of course, that ASUS motherboards supported overclocking.
I was also, in fact, aware that this option was enabled by default.
However, I had - wisely, I thought - disabled this feature as soon as
I got the PC. There's an option - almost the first option in the BIOS,
actually - called "AI Overclock Tuner" that I made sure was disabled
from the start.
Unfortunately, this was not the only option restricting the BIOS from
overclocking the CPU. There were a number of other options - labeled
'Multicore Enhancement' or 'CPU Optimizations' and similar - that
would automatically ramp up CPU frequency past the specced frequency
as CPU utilization rose. These were all enabled by default, and -
lacking clear definitions on what they did in the manual - I left them
alone. These settings often ramped the CPU up to 300MHz beyond the
rated speed of the CPU. In the beginning, this didn't cause any issue
but eventually this behavior started causing memory issues (first
noticed as corrupt ZIP archives) and, finally, with the computer
BSODing whenever Windows tried to boot.
Figuring out which options it was safe to disable was a challenge;
there is no clear documentation. But even with all the 'turbo mode'
features disabled, I was still running into problems, because many of
the default CPU settings were set scarily high.
    Example: The 13900K & 14900K processors are rated by Intel     for a 'long duration power limit' (basically, 'normal mode     power usage') of 125W. The BIOS default? 253W.  The     'short duration power limit' (where the CPU ramps up briefly     for 'turbo mode', usually lasting only a few seconds) has     a specified limit of 253W. The BIOS default?     4096W.
Jesus-fucking-Christ, ASUS!
I had to manually set a large number of options to fit within Intel's
recommended standards. Once I did that, my PC started to reliably boot
again, to the point where I could start stress-testing the new
settings to ensure stability.
Normally, I don't have a problem with ASUS but now I see why other
people here have issue with them. Obviously my CPU never actually had 4096W pumped into it (as evidenced
by the fact that it's not a pile of ash). Not only can't my PSU
provide that much juice, but I am sure that the voltage was probably
throttled by hardware elsewhere on the board. The 4096W seems to me to
just be the motherboard's equivalent of 'this option isn't being
limited at all by the software'. It's still careless.
(It does speak volumes on the reliability and overhead built into
Intel's -K class processors that they survived this carelessness,
though.It's why I overpay for the -K class processors in the first
place, so that if something like this does happen, I'm provided some
modicum of protection.)
ASUS is completely to blame for this fiasco. Its 'enhancements' and
'optimizations' rode the CPU too heavily. It disguised the fact that
it was overclocking the CPU under misleading BIOS setting names. Even
when those features were disabled, its voltage settings were
carelessly high. It lacked documentation explaining what those
settings did. I might have come to a quicker solution to this problem had ASUS BIOS
not been so misleading. Overclocking issues were one of the first
things I considered, after all. But the obvious settings were disabled
in the BIOS, so it /couldn't/ be an overclocking problem. Except, as
it turns out, it was.
My PC isn't fully functional yet, but at least now I'm on track to
getting it back into working order again. The current plan is to find
a stable CPU ratio (it looks like its rated 5500 will be sufficient;
its only if it goes above that do problems start showing up), then
stress test it for a while to ensure it can maintain stability at that
speed. Only after that will I reinstall the other hardware, NVMes, and
check the status of the OS. It'll probably take a week or so but in
time, I will be back on my main PC again.
Eventually I'll have to decide if I want to replace the motherboard
and CPU. But if I can get it stable, it may be a while before that
happens, even if I'm forced to underclock it a bit. CPU/motherboards
are expensive!
But if I do replace, my next motherboard definitely won't be an ASUS.