On 1/17/2008 1:43 PM, James G. Sack (jim) wrote:
Karl Cunningham wrote:
On 1/16/2008 9:28 PM, James G. Sack (jim) wrote:
on the subject of cheap computers, Frys has some cheap (<~ $100.)
MoBo+CPU packages, which lately seem to use ECS MoBo's.
Are these any good?
About three years ago I heard the FTTO (first time turn on) for ECS
boards was about 50%. I had 1 out of 2 work. My friend had 1 out of 3 work.
YMMV, but don't expect too much.
Wow! You mean the rate of successful FTTO is as quoted?
Just out of curiosity, does anybody care to speculate on causes that
might give such results?
Ideas I can think of are:
Physical (incl. electrostatic) damage -- perhaps(?) the most forgivable"
CPU incompatibilities -- design error? -- cause for blacklist!
Violation of unspecified requirements -- cause for major stink
PS
RAM
devices
Misconfigured old-fashioned switches/jumpers -- bad docs?
What else?
I heard the 50% FTTO rate from someone who worked at Fry's.
Definitely blacklist these. The following is all speculation but may be
somewhat real. I suspect these boards are designed with cheap labor and
as little effort ($$) as they can get by with. They put together a board
design from someone else's schematic, and by copying other board
layouts. Probably don't have test equipment to check timing margins,
etc. One thing they have is cheap prototype board production and
probably fast turnaround for changes. So they go many rounds of board
changes adjusting trace lengths, spacings, etc., until they get one
working. This is nanosecond stuff and every little thing matters. Then
it's off to production. A slight change in component timing or
capacitance can cause DOAs.
Typical users will put in 1GB of RAM and if they get windows xp to run
(out of the box using only first 150MB of RAM) they declare it a
success. No tests of memory (other than POST), temperature stress, CPU
loading, etc. So far they haven't tested very much of the functionality
of the board. Failures after this are usually chocked up to user error
or software bugs. Amazing what people will put up with, "Y'know I never
could get that to work right."
I recommend at least memtest to anyone who buys a cheap MB or computer.
A lot of them don't pass the first time.
Karl
--
[email protected]
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list
