8 Jun
2013
8 Jun
'13
5:31 a.m.
On Wed, 5 Jun 2013, Cory Smelosky wrote:
On Wed, 5 Jun 2013, Mouse wrote:
The key factor is the build quality of the hardware. I very much like
Socket 370 and even Pentium-M based embedded boards, which keep running
pretty much forever. Higher end Socket 370 based servers (along the lines
of the HP DL380) tend to have a similar build quality.
The single most failure prone component (and this applies to late
generation Socket 7 boards too) are cheap, knock-off (and even
counterfeit) electrolytic capacitors. Even the better Asus boards were not
immune to these problems, and I have a Asus board sitting on the bench
right now that contains some obvious counterfeit Rubycon capacitors used
as filters in the CPU voltage regulation circuitry.
....I've run a lot of Suns, VAXen, etc. And
I've run a lot of peecees.
(More of the former than the latter, because people throw out more of
them as they scroll off their idea of the low end.) Even after
correcting for machine-hours run, the peecees exhibit much higher
failure rates. Maybe one or two of my dozen or two Suns have failed.
I'm not sure I have even _one_ working peecee made contemporarily with
those Suns.
My PII was made around the time as my Netra t1 105. Both are still 100%
functional. I don't like any PeeCee hardware that is Pentium III or
later. It's just absolutely unrealiable garbage! I just /do not/ trust
it. If I don't check it for a month I want my systems to not throw any
crazy errors. Only my Pre-PIII systems I trust to not give errors in
that time period (this is only talking about PeeCee hardware).