View Single Post
  #6  
Old April 10th 21, 06:32 PM posted to alt.comp.hardware.pc-homebuilt
Paul[_28_]
external usenet poster
 
Posts: 1,467
Default PC occasionally freezing

Larc wrote:
On Thu, 08 Apr 2021 09:02:09 -0400, Paul wrote:

| Once upon a time, in a land far far away, there were VRMs.
|
| This is the VCore voltage regulator. It started as a module,
| but was also soldered to the motherboard. That gave it the
| name Voltage Regulator Down (VRD), when it was soldered down.
| My first one put out a measly 35W, had a single phase, and
| used a pretty good sized MOSFET.
|
| Back in the day, there was no feedback. Designers were
| worried about thermal runaway. They had to guess at what
| max power would be, and add a generous heatsink. This is
| basically open loop design.
|
| Then, more recently, power limiters were fitted. At the same
| time, you may have noticed the heatsinks are getting smaller
| and more poorly designed. The design is now closed loop, and
| it's because of feedback about how the VCore is running,
| allowed it to run burning hot.
|
| When I bought my latest board, I hadn't noticed this. I didn't
| even have it as a particular item to check on my checklist.
|
| It was the usual deal. Burned my finger on the heatsink
| when testing the board on the kitchen table. Ended up
| turning off turbo, as well as fitting a fan right next
| to the VRM cooler.
|
| In the old days, a board would not leave the factory, if
| a customer could burn their finger on it. Today, this
| seems to be OK.
|
| Some manufacturers, their "auto-overclock" is too aggressive
| and they're using the wrong VCore. The user is then forced
| to do the research, to figure out what values might better
| serve the board.
|
| There is a suggestion that the Kernel Power, is a report
| of overheat, arriving at kernel level. Adjusting BIOS conditions
| and taking it off "auto/auto" may allow reducing VCore heatsink
| below "insanely hot".

Thanks, Paul. The motherboard and CPU in question ran in my main computer for about
3 years with no problems. I didn't change BIOS settings when I moved it. Should the
VCore start giving problems now when it never did before? Besides, it's an H370
that's not set up for overclocking. I could and did disable turbo boost and will
give it some time to see if that helps. Freezing doesn't always occur.

Larc


Some brands boost VCore above the recommended value,
as part of their "stability program". You can see similar
things on memory. For example, on a server board, maybe
a memory is given "precisely 2.5V". The equivalent generation
desktop board (no overclock) might be 2.65V for memory.

Modern designs have more environment monitoring than
designs in the past. You would need to find an inspired description
for "Kernel Power", to understand what the problem is.

Since the system does not freeze (totally unresponsive, no Stop Code in
the log), then you know the system was sane when it logged the
event, so it's some kind of environmental signal.

Excessive power as recorded by a power monitor on the VCore chip,
excessive temperature (harder to get, because manufacturers are
cheapskates). But if Intel says in a VRM or VRD spec to
"do it this way", then the feature gets added quietly. No
VCore regulator company can afford to make non-compliant silicon
for a big market (millions of regulators per year as a potential market).

*******

When you move a system, there is a small chance of bumping a heatsink
or shaking something loose, and then it does not cool properly. In
years past, it was the Northbridge heatsink which was a problem. U-pins
were soldered to the motherboard, but the metal was a poor choice
and solder did not wet the pins properly. Dell was so impressed with
the Intel design, Dell

Did as they were told, used the Intel-recommended restraint.

Put an electrical continuity check in, so when the U-pins pulled
out, the computer would refuse to power up. They used a little
ohm-metering circuit, to verify the strap had fallen off.

Such are double-standards :-)

Paul