Re: [gentoo-user] Re: Epic list of total FAIL.

[Alan McKinnon]

On 22/08/2015 13:25, Mick wrote:
> On Saturday 22 Aug 2015 09:18:05 Dale wrote:
>> Fernando Rodriguez wrote:
>>> On Saturday, August 22, 2015 1:52:00 AM Alan Grimes wrote:
>>>> That said, I spent the day doing diagnostics:
>>>>
>>>>
>>>> Findings:
>>>>
>>>> 1. There were a hell of a lot more memory errors than I had seen before.
>>>> 2. There was a smudge on one of the dimm's contacts and some of the
>>>> usual dust on the CPU-facing one.
>>>> 3. The motherboard was not developed by sane engineers. In a sane world,
>>>> there are two types of variables: measured variables and controlled
>>>> variables.
>>>> The RAM voltage would appear to be a controlled variable but it is also
>>>> a measured variable. In order to achieve a close approximation of 1.5v,
>>>> I had to set it to 1.530 volts. WTF...
>>>>
>>>> 4. an AMD K10 processor cannot successfully drive 8-ranks of high
>>>> density ram at 2x800 mhz -- BUT IT WILL TRY!!! I found all dimms to be
>>>> good either individually or in pairs, but the entire ram compliment of
>>>> four dims cannot be run at full speed at once with the CPU/motherboard I
>>>> have installed.
>>>
>>> Findings 3 & 4 sound like a faulty or underrated PSU...or a bad
>>> motherboard. Start by unplugging everything that you don't need to boot
>>> from a live CD and run some tests.
>>
>> It sure does.  A weak power supply will certainly cause some issues.
>
> I also concur that the most likely cause of this problem is the PSU but first,
> I would clean the RAM contacts.
>
> Then try a replacement PSU if you have a spare one, or take your multimeter
> and measure the output, checking for lower voltage values and fluctuations.
> If you get bad measurements, then take your soldering iron out and for a few
> pence inspect and replace any domed, or all capacitors on the secondary
> (output) side.


<nitpick>
A multimeter is not really a valid test. If say the 5V rail is dodgy,
then the output will still be a solid 5V. What's happening is that the
PSU regulator circuitry can't keep up so the output averages 5V (that's
what the transformer gives out) with large amounts of high-frequency
ripple superimposed. Your multimeter average's that out and displays ...
5V! When things get really bad the output may dip momentarily when load
is drawn, but by that stage the PSU has been struggling for a long time
already.

Use an oscilloscope instead, and you see immediately what condition the
output is in.
</nitpick>

Few IT techs just happen to have an expensive oscilloscope just lying
around, so a good recommendation is to replace the PSU anyway every 2
years or so - more if the thing runs hot. I consider these as wearing
items, sorta like oil filters



-- 
Alan McKinnon
alan.mckinnon@×××××.com