jehoshua777 said:
Hi,
Pardon my ignorance please, but I have a new ASUS mainboard and CPU,
and seem to remember years ago, somone saying that before the computer
can be used, he had to "burn in" the CPU (or was it the mobo, I don't
know ?? ).
The details of the new parts are:
* ASUS P5GD1 (Pro) Motherboard
(
http://au.asus.com/products/mb/socket775/p5gd1/overview.htm)
* Intel LGA775 Pentium 4 CPU (3.0 Ghz)
* Kingston 512Mb RAM (DDR400)
Do I need to do anything, like leave it running for a number of hours
prior to using the new computer ? Again, pls pardon my ignorance, but
better to ask first.
Thanks,
Peter
As BigJim says, the answer is no.
But, there is a difference between the concept of "burn-in"
and the concept of testing.
You should test the computer, using any representative test, as
proof that the components of the computer are working as they should.
I like to start with memtest86+ from memtest.org. It is a
free test program, and the test program will format a floppy
diskette. You boot the computer with that diskette and a 640x480
test screen will appear. Do at least a couple of passes (of all tests),
to prove that your memory is good. This test also tests the processor
and Northbridge chip.
I don't really have a good test for disks. If you can find a test
program on the disk manufacturer's web page, you may be able to
give the disk a rudimentary test that way. Due to shipping damage,
occasionally you'll run into a disk that is dead out of the box.
Another useful test, is to test that the computer is not overheating
when it is doing some demanding computing. You can use Prime95
from mersenne.org, as that program has a lengthy computation with
a known answer. This is an excellent program to find problems with
the processor and the memory, and it uncovers memory errors that
memtest86+ won't find. Since Prime95 runs at 100% CPU, it makes
the CPU get hot. If you also run a copy of Asus Probe, that
program will display the CPU temperature. You can run Asus Probe
at the same time as Prime95, to see if the CPU gets to 70 degrees
C. If that happens, the CPU will automatically reduce its computing
rate, and a 70C processor temperature is a sign that the computer
case cooling and the CPU cooling must be improved.
You can think of these testing programs as part of a "tune up".
For both memtest86+ and Prime95, no errors are acceptable. If
you see errors reported on the screen, find out why they are
happening and fix them. Sometimes all it takes is a slight
tweak of memory voltage or a timing change, to make the machine
run properly.
Doing this kind of testing is preferable to simply screwing
together the components and trying to install WinXP. The
reason being, it takes a lot longer for trouble to show up
by using the OS as your testing tool. If you have a bad
spot in memory, it might take a very specific set of conditions
before the machine crashes. Using focused tools for testing,
means you will find problems that much sooner.
To get back to your original question, why is "burn-in" done ?
In a factory, burn-in is a separate process from actual product
testing. Factory testing consists of two parts - structural and
functional testing.
An example of structural test, is testing the blank motherboard, to
see whether all the copper tracks are present and making contact.
Doing a test at this stage (no components soldered to the board),
reduces the chance that the motherboard will have to be thrown
away later.
An example of functional test, is when a test person at the
factory, inserts RAM, processor, video card and the like, into
your new board, and they actually boot and run special factory
test programs. I think Anandtech had a video of a test person
doing that on one of their factory tours. A functional test is
not good at covering everything thoroughly on the motherboard,
but it does test that the product is doing what it is supposed
to do.
A combination of these two approaches helps to maintain a high
level of quality as the products leave the factory. The functional
test is one reason the majority of motherboards work when you
assemble them.
Now, let us consider some other products. There have been some
video cards sold by Newegg, where the video RAM chips on the
board went bad about three weeks after the purchasers started
to use them. The experience was repeatable, with many customers
experiencing the same kinds of failures. A failure early in
the life of a product is termed "infant mortality".
To battle infant mortality, depending on the cost of testing,
and the value of not pissing off your customers, some factories
use a "burn-in" process. This is also termed "accelerated life
testing", when it involves abusing the product in some way.
For example, a potentially brutal test, is alternately heating
and cooling the new products in a thermal chamber, while the
product is powered. The heating and cooling is useful for tugging
on bad solder joints, or bad motherboard copper tracks etc. The
idea is, the accelerated testing weed out the bad products, so
the product does not die on the customer's door step. A burn-in
process for those video cards described above, might have
stopped most of those cards from being shipped to customers.
Burn-in is generally considered to be an expensive process,
because of the space and time it takes. Asus produces more
than 2 million motherboards per month now, and if each board
received 48 hour burn-in, they would need a huge facility for
such a process. It is likely they rely on other test methods
to yield a quality product, without going to the trouble of
doing that kind of testing.
Now, why would an end customer do "burn-in" ? Say you were
working in the IT department of a large company. For some
strnage reason, you architect a server setup with no
redundancy - this means you need ultra-reliable computers
in the server room. A person doing something like this,
would undoubtedly run the computer for a number of days
on the bench, before putting the computer "into production".
That would be a good reason for a "burn-in" or customer
acceptance test process (i.e. a failure is simply
unacceptable).
If you feel that your application is critical, like the
one described in the paragraph above, then perhaps you
too would benefit from beating the hell out of the
computer for a few days, before doing real work with it.
My acceptance criteria, is that the computer pass a few
simple tests like memtest86+ and Prime95 - if some part
of the computer is going to die three weeks from now,
that is no different to me, than if the computer dies
three years from now. If you are building a computer,
chances are you have a backup computer available anyway,
in case the new build runs into trouble.
HTH,
Paul
