Mixing RAM speed

[philoÂ]

I build a lot of machines mostly from decommissioned P-4's and have been
re-testing machines that have been sitting for six months or so.

Three of the machines would not boot and I was getting RAM error beep
codes. (Dell and HP machines)

Upon inspection I see that in all the machines I had DDR2 of different
speeds...although all paired slots were matched.

By switching RAM around between the machines and my spare RAM box, all
machines now work fine with same-speed RAM.


My question is...why did they work in the first place but after sitting
for a while...no boot?

Each machine had a 24 hour burn-in and due to updates and such had been
on and off a quite a number of times while I was working on them.



At first I though it was just a matter of my workshop being a bit damp
in the summer and I removed the RAM and cleaned the contacts etc.

 

[Paul]

I build a lot of machines mostly from decommissioned P-4's and have been
re-testing machines that have been sitting for six months or so.

Three of the machines would not boot and I was getting RAM error beep
codes. (Dell and HP machines)

Upon inspection I see that in all the machines I had DDR2 of different
speeds...although all paired slots were matched.

By switching RAM around between the machines and my spare RAM box, all
machines now work fine with same-speed RAM.


My question is...why did they work in the first place but after sitting
for a while...no boot?

Each machine had a 24 hour burn-in and due to updates and such had been
on and off a quite a number of times while I was working on them.



At first I though it was just a matter of my workshop being a bit damp
in the summer and I removed the RAM and cleaned the contacts etc.

I've had RAM fail in storage (throws errors when tested). I had
three DIMMs in a single channel three slot machine all fail. You
could test them individually, and memtest86+ could spot errors
in them.

And three identical DIMMs from the same purchase lot (a box of
eight DIMMs), stored in their electrostatic carrier tray, when
installed in the same computer, were error free. And the only
thing I could figure, is some sort of chemical reaction must
be happening to them, while sitting in the DIMM slot in
storage. The contacts were fine, and I did not clean any contacts.
(Gold looked OK, no change there.) I think the chemistry
was happening inside the plastic memory chip package.

*******

The problem should not be a function of speed. The BIOS is
supposed to find the common speed denominator, of the lot of RAM,
and run them all at a speed they can take. A fast RAM can run at a
lower speed (like DDR400 could be run as low as DDR200). And the
CAS is adjusted for the slowest DIMM too. So if you have
DDR400 CAS3, DDR400 CAS4, DDR400 CAS5, then all the DIMMs are
run with CAS5. In other words, the BIOS ensures that no
parameter is violated.

For technologies before DDR2 (like DDR), the bus is less
than ideal, and signal quality is affected by loading. And that's
why the BIOS would turn down the memory clock if all the slots
were full.

DDR2 would be the first memory with a better termination
scheme. And it showed here, as my first DDR2 machine
was far and away better on random memory errors, than the
two DDR400 machines before it. DDR3 and DDR4 should
continue that trend (they add more calibration type
features to the design).

A board full of DDR2 or DDR3, should really be "happy",
not "grumpy".

If you reseated the RAM and it started working right,
then that could be a contact problem. If you reseated
the RAM and it was just as bad, then the chips have
probably failed. I've had unbranded RAM, where the DIMM
only lasts a year and a half, and then starts throwing
errors. And that's being run on a daily basis, so
less likely to be cold and damp.

Paul

 

[philoÂ]

And three identical DIMMs from the same purchase lot (a box of
eight DIMMs), stored in their electrostatic carrier tray, when
installed in the same computer, were error free. And the only
thing I could figure, is some sort of chemical reaction must
be happening to them, while sitting in the DIMM slot in
storage. The contacts were fine, and I did not clean any contacts.
(Gold looked OK, no change there.) I think the chemistry
was happening inside the plastic memory chip package.

*******

The problem should not be a function of speed. The BIOS is
supposed to find the common speed denominator, of the lot of RAM,
and run them all at a speed they can take. A fast RAM can run at a
lower speed (like DDR400 could be run as low as DDR200). And the
CAS is adjusted for the slowest DIMM too. So if you have
DDR400 CAS3, DDR400 CAS4, DDR400 CAS5, then all the DIMMs are
run with CAS5. In other words, the BIOS ensures that no
parameter is violated.

For technologies before DDR2 (like DDR), the bus is less
than ideal, and signal quality is affected by loading. And that's
why the BIOS would turn down the memory clock if all the slots
were full.

DDR2 would be the first memory with a better termination
scheme. And it showed here, as my first DDR2 machine
was far and away better on random memory errors, than the
two DDR400 machines before it. DDR3 and DDR4 should
continue that trend (they add more calibration type
features to the design).

A board full of DDR2 or DDR3, should really be "happy",
not "grumpy".

If you reseated the RAM and it started working right,
then that could be a contact problem. If you reseated
the RAM and it was just as bad, then the chips have
probably failed. I've had unbranded RAM, where the DIMM
only lasts a year and a half, and then starts throwing
errors. And that's being run on a daily basis, so
less likely to be cold and damp.

Paul

I have now individually tested all the RAM and found one bad stick.

Possibly it was just a poor contact due to storage in a slightly humid
area. At least I have all the machines with speed-matched RAM now.

 

[philoÂ]

X




I have now individually tested all the RAM and found one bad stick.

Possibly it was just a poor contact due to storage in a slightly humid
area. At least I have all the machines with speed-matched RAM now.

I had put all the machines away for a few weeks and tested them all
again today.

One did not boot.

Eventually I found one RAM slot that's intermittent.

I just put in a larger RAM stick elsewhere and will again set the
machine aside for a few weeks and test again. Pretty sure the machine
will be ok now