Do you do a burn-in procedure?

  • Thread starter Thread starter TVeblen
  • Start date Start date
T

TVeblen

When you do a new build, is there a burn-in or testing procedure you do
before you install OS and software, or do you just assemble, load it up, and
hope for the best?
 
When you do a new build, is there a burn-in or testing procedure you do
before you install OS and software, or do you just assemble, load it up,
and hope for the best?
Click to expand...

I think a full pass with memtest would be a good minimum.
 
TVeblen said:
When you do a new build, is there a burn-in or testing procedure you do
before you install OS and software, or do you just assemble, load it up, and
hope for the best?
Click to expand...

I install just enough stuff to let me see if the system works, then I
run a memory test (MemTest86 ver. 3.xx, Gold Memory 5.07) for a few
hours. Next, I install the HD(s) and run the factory diagnostic. If
that passes, I install the OS. If you add and test one piece of
hardware at a time, it's easier to pinpoint bad hardware or OS
problems.
 
TVeblen said:
When you do a new build, is there a burn-in or testing procedure you do
before you install OS and software, or do you just assemble, load it up, and
hope for the best?
Click to expand...

Depends on whether the build is for yourself or you are a jobber with a
business and want to reduce returns and support calls.
 
When you do a new build, is there a burn-in or testing procedure you do
before you install OS and software, or do you just assemble, load it up, and
hope for the best?
Click to expand...

I don't bother basically because I don't buy cheap shit anymore.
Usually if it goes titsup it's the cheap ass generic RAM or that $30
motherboard you've bought so I avoid both and have no problems.
 
Memory is the component I find that fails most often. Mayby one
stick out of every couple of dozen installs is flakey after it heats
up from running for a while.
Click to expand...
Stop buying cheap generic shite then.
 
When you do a new build, is there a burn-in or testing procedure you do
before you install OS and software, or do you just assemble, load it up, and
hope for the best?
Click to expand...

If there's no problems on the build - the OS install is a pretty good
test. Pretty much have to test if it's going to be overclocked for
limits, although that can lead to quirky behavior or problems on an
overclock as a stability issue down the line. With speeds today
doesn't grab most, me either, though others will want it on principle.
 
TVeblen said:
When you do a new build, is there a burn-in or testing procedure you do
before you install OS and software, or do you just assemble, load it up,
and hope for the best?
Click to expand...

I build on a piece of plyboard on the kitchen table, with CPU/HSF, GPU, RAM,
PSU, and an optical drive. Then I run memtest86 overnight.

In the morning, if it's still running fine, I'll install windows, update
windows, install drivers, and play around with it for awhile.

Eventually it all gets stuffed in the case, at which point it is good to go.

Jon
 
Conor said:
I don't bother basically because I don't buy cheap shit anymore.
Usually if it goes titsup it's the cheap ass generic RAM or that $30
motherboard you've bought so I avoid both and have no problems.
Click to expand...

What are some good retail brands of RAM?
 
What are some good retail brands of RAM?
Click to expand...

Crucial. Guaranteed compatibility and lifetime warranty. Never had a
stick go bad in the decade I've been using them. OTOH, the times I've
stupidly bought other stuff, it's ended up causing nothing but grief -
the worst being a machine refusing to boot the very first time it was
put in.
 
Conor said:
I don't bother basically because I don't buy cheap shit anymore.
Usually if it goes titsup it's the cheap ass generic RAM or that $30
motherboard you've bought so I avoid both and have no problems.


Crucial. Guaranteed compatibility and lifetime warranty. Never had a
stick go bad in the decade I've been using them. OTOH, the times I've
stupidly bought other stuff, it's ended up causing nothing but grief -
the worst being a machine refusing to boot the very first time it was
put in.
Click to expand...

That's been my experience as well, probably because Crucial is the
only company that always uses branded chips (Micron, Samsung, Elpida)
while everybody else builds modules mostly from UTT or home-sliced
chips. Kingston was especially bad for me, with an overall 30%
failure rate.
Click to expand...
 
Why? Don't you trust the memory you've bought?
Click to expand...

I don't see any good reason to trust anything without confirmation.

Besides, memtest will stress other aspects of the system as well.
 
larry said:
That's been my experience as well, probably because Crucial is the
only company that always uses branded chips (Micron, Samsung, Elpida)
while everybody else builds modules mostly from UTT or home-sliced
chips. Kingston was especially bad for me, with an overall 30%
failure rate.
Click to expand...

Crucial have had some bad batches. With the latest
generations of RAM, Crucial has even shipped DIMMs which had
other than Micron memory chips on it. Not everything
Crucial makes, is golden. They've had their problems.
(And them changing chips, was part of an attempt to fix it.)

Some basic tests of any computer you build, is a wise idea.
A couple passes of Memtest86+ (or one of several others), plus
four hours of Prime95, will give you some idea whether the
computer works. The Prime95 is pretty important, at accelerating
the test and finding issues faster. On my current machine, I can
even see a difference with certain motherboard settings,
between using Prime95 by itself, and using Prime95 while
playing a 3D game. So running a 3D game can also uncover
problems. But bare minimum, you should be doing a Prime95
test, because it is so good at uncovering issues. It is
much better than memtest86+ at that. The reason I run
memtest86+ at all, is because it covers most of the
memory area in the computer (it cannot test the "BIOS
reserved" areas, which is a small amount of RAM). The area
of coverage of Prime95, might only be 80% or less. But it
is still a good test.

http://majorgeeks.com/Prime95_d4363.html

That is not "burn-in", it is a test, and it happens to take
a while to run. Running the computer for 48 hours, as a means
of finding infant mortality issues, would be a burn in. At
work, the factory also liked the idea of including temperature
ramping while doing burn in (being careful that the degrees C
change in temperature per hour, did not exceed the capabilities
of any of the hardware used - the burn-in chamber can be programmed
to ruin equipment if you're not careful). But the Prime95 test
I'm talking about, is just a simple test case to run, to prove the
computer actually computes correctly. It isn't a burn-in
strategy as such.

Paul
Click to expand...
 
Paul said:
Crucial have had some bad batches. With the latest
generations of RAM, Crucial has even shipped DIMMs which had
other than Micron memory chips on it. Not everything
Crucial makes, is golden. They've had their problems.
(And them changing chips, was part of an attempt to fix it.)
Click to expand...

I know they had a disasterous run with their 2.2 volt Ballistix
modules, which they claimed were damaged by being run at below 2.0V.
I've never read an explanation of that, nor have I ever seen a DDR2
chip data sheet specify a nominal voltage rating other than 1.8V.
Some basic tests of any computer you build, is a wise idea.
A couple passes of Memtest86+ (or one of several others), plus
four hours of Prime95, will give you some idea whether the
computer works.
Click to expand...

MemTest86+ vers. 1.7x and 2.xx have never reported errors for any
memory I've tested with them, but MemTest86 vers. 3.2 and 3.3 have.
Similarly, Gold Memory 6.xx always passed memory, while ver. 5.07 has
reported errors.
 
Since you have no way of knowing what I buy for memory, I must
conclude that, one more time, you have posted just to show everybody
how big an ass you are.
Click to expand...
I'm not the one who buys memory that's such low quality that my builds
have a 1 in 24 failure rate.
 
larry moe 'n curly said:
I know they had a disasterous run with their 2.2 volt Ballistix
modules, which they claimed were damaged by being run at below 2.0V.
I've never read an explanation of that, nor have I ever seen a DDR2
chip data sheet specify a nominal voltage rating other than 1.8V.


MemTest86+ vers. 1.7x and 2.xx have never reported errors for any
memory I've tested with them, but MemTest86 vers. 3.2 and 3.3 have.
Similarly, Gold Memory 6.xx always passed memory, while ver. 5.07 has
reported errors.
Click to expand...

Sorry if this is a dumb question, but if you run Memtest and it finds errors
does it just fix them, or do you have a bad RAM module that needs RMA? Do
the manufacturers ever question your reason for return?
 
Paul said:
larry said:
Crucial have had some bad batches. With the latest
generations of RAM, Crucial has even shipped DIMMs which had
other than Micron memory chips on it. Not everything
Crucial makes, is golden. They've had their problems.
(And them changing chips, was part of an attempt to fix it.)

Some basic tests of any computer you build, is a wise idea.
A couple passes of Memtest86+ (or one of several others), plus
four hours of Prime95, will give you some idea whether the
computer works. The Prime95 is pretty important, at accelerating
the test and finding issues faster. On my current machine, I can
even see a difference with certain motherboard settings,
between using Prime95 by itself, and using Prime95 while
playing a 3D game. So running a 3D game can also uncover
problems. But bare minimum, you should be doing a Prime95
test, because it is so good at uncovering issues. It is
much better than memtest86+ at that. The reason I run
memtest86+ at all, is because it covers most of the
memory area in the computer (it cannot test the "BIOS
reserved" areas, which is a small amount of RAM). The area
of coverage of Prime95, might only be 80% or less. But it
is still a good test.

http://majorgeeks.com/Prime95_d4363.html

That is not "burn-in", it is a test, and it happens to take
a while to run. Running the computer for 48 hours, as a means
of finding infant mortality issues, would be a burn in. At
work, the factory also liked the idea of including temperature
ramping while doing burn in (being careful that the degrees C
change in temperature per hour, did not exceed the capabilities
of any of the hardware used - the burn-in chamber can be programmed
to ruin equipment if you're not careful). But the Prime95 test
I'm talking about, is just a simple test case to run, to prove the
computer actually computes correctly. It isn't a burn-in
strategy as such.

Paul
Click to expand...

I have used Kingston valueram on 4 builds and never had a problem. If I
mention it in group the flames get high. What is your take on Kingston for a
non OC system on a cost/benefit basis?
Click to expand...
 
TVeblen said:
Sorry if this is a dumb question, but if you run Memtest and it finds errors
does it just fix them, or do you have a bad RAM module that needs RMA? Do
the manufacturers ever question your reason for return?
Click to expand...

I've had memtest86+ detect a stuck-at bit (same error in same location
every time). It doesn't seem to be good for much more than that.
Maybe test 5 might pick up a situation where the RAM needs more
voltage.

Memory test programs are intended to tell you whether the RAM is performing
correctly. If the program writes FF hex to a location, and reads back
7F hex, then you know that one bit is in error. That tells you the
RAM, in its current state, cannot be relied upon to work correctly.
If you were working out your income taxes, this could be a
catastrophe.

The corrective actions are:

1) Attempt to adjust Vdimm voltage, timing values like CAS, memory
frequency setting, and retest. If you're happy with the need to
use custom settings with the memory, then you're done. If the
system is being shipped to someone else, you may choose a higher
level of operability, meaning you won't ship the system unless
it works with default settings. That means you won't be receiving
a phone call, after the customer replaces a CMOS battery or
flashes the BIOS some day.

2) Return the memory and exercise the warranty. At retail, you might
have seven days to get easy satisfaction, which is why you
want to run your tests immediately. If you have to rely
on the manufacturer warranty later, things might not go as
smoothly, and you may end up buying more RAM in the interim
period, until the manufacturer replaces the memory. On some
"lifetime warranties", there may be a sliding scale applied
over time, and if the item is out of production, you may
receive a small fraction of the original purchase price.

Memory cannot be "repaired" by a program. Either a memory
cell is stuck in a particular state, or the error could be
a transient fault caused by the memory not meeting its
stated timing specs. (This happens a lot, because they
take untested memory from the chip maker, and "pick
out the good ones", to make a stick of RAM. So shoddy
testing leads to substandard memory performance. The
only cost item in the manufacturing process they can
cut back on, is the amount of testing they do.)

Paul
 
TVeblen said:
I have used Kingston valueram on 4 builds and never had a problem. If I
mention it in group the flames get high. What is your take on Kingston for a
non OC system on a cost/benefit basis?
Click to expand...

I'm using some right now. The price was right.

You can use the Newegg reviews, to spot the bad stuff.
I don't buy from Newegg, but before I buy locally, I
check the reviews to see how many DOAs there are. I'm
especially interested in reports of brands of RAM that
causes the DIMM slot to burn. That may mean the DIMM
never got tested at the factory at all.

Paul
 
Back
Top