Norm Why wrote:
Are the RAM sticks "enthusiast" RAM or regular RAM ?

Regular RAM doesn't have heatsinks - that's one possible indicator.

Enthusiast RAM has a voltage spec. It indicates whether the
RAM is fundamentally high voltage or low voltage. An example
of high voltage RAM, was the winbond 2.5V RAM that could
run at 3.3V to 3.6V or so. Basically, it had a ton of headroom.
Other RAMs, a 2.5V stick runs at 2.7V max. The reason that
limit exists, is the RAM has an internal voltage regulator
and it starts to get hot if you raise VDimm too radically.

On enthusiast RAM, the "rated" parameters are not loaded into
the SPD. Lower values are used to ensure the operator can get
into the BIOS. When the lower values are being used, it is
hoped that no voltage boost is needed. But perhaps a
small tweak won't hurt.

There are two voltages you could adjust. There is VDimm,
which is the 1.35/1.5/1.8/2.5/3.3V value (depends on DRAM family).
There is also Vnb, a Northbridge voltage on setups where there
is still a separate Northbridge chip and it has the memory interfaces.
That's what you're using at the moment.

I bumped the Vnb on my X48 chipset by a couple of the smallest notches,
perhaps 0.050V or so total, from its nominal value. That removes
errors on my setup. The RAM, the nominal is 1.8V and the
enthusiast (Corsair) voltage is 1.8V too. It doesn't need to be
boosted to run DDR2-800 5-5-5-18, and considering the age
of the hardware, there's nothing to be gained
from abusing it. If I was having trouble with the RAM,
I could try 1.9V or if necessary 2.0V.

Some Intel systems that run 1.5V RAM, the memory controller
is on the CPU, and the CPU has a 1.65V Max spec. Intel
doesn't want warranty damage on the CPUs, so it has a spec
out there for the kids so they won't ruin stuff. And this
meant that when the enthusiast makers made RAM for those
setups, they no longer dabbled in "extreme" voltages.
There were no more "winbond style" experiments. No
senseless shooting for the moon stuff.

*******

You should be able to test the sticks, one stick at a time.
I do tests that way, so the test result ("which DIMM is bad")
is unambiguous.

The following is for "thoroughness of testing".

My favorite way to test, is two sticks in single channel mode,
so that the "upper stick" is 100% tested, while the "lower stick"
is 99% tested. You swap the two sticks and repeat the memtest
single pass, so that the stick now sitting in the high slot
gets 100% tested too. This is mainly for stuck-at faults, to make
sure any 640K reserved locations don't harbor bad RAM.

Don't forget proper etiquette when RAM testing. Power
off, wait 60 seconds (or until the green LED goes out),
before pulling the RAM sticks and pushing them in again.
That's to make sure there is no power in the slot.

*******

This is what I'm running at the moment. I used to have four
sticks of Kingston CAS6 (that failed). All I could find
at the time in a CAS5 was this stuff, so that's what is
in the machine at the moment. Four 2GB sticks. The
Northbridge is probably the flakiest part of the setup now.
I think I was pretty lucky to get this (I might have
asked at the computer store and they still had some). Usually
what happens, is after a memory generation is "Done", all
the CAS5 disappears and only some crappy CAS6 is left.
It's pretty hard to find fancy memory ten years after
the motherboard came out. By crappy, the Kingston I
bought ran abnormally hot, and the voltage wasn't
over stock. I even added a cooling fan over the
RAM just in case, for the Kingston stuff. The Corsair
right now is cooler and isn't scaring me on temps.

It doesn't look like it's still in circulation.

https://www.corsair.com/us/en/Catego...N2X4096-6400C5

Sometimes, the chipset is also a limitation. It's possible
the X48 can't run below CAS4. That's if you could find
some RAM that low. So eventually, the timing inside the
Northbridge doesn't allow "cranking CAS to zero". It has
a limit too (a minimum CAS value).

Paul
Thanks Paul. Upgraded to Q9650 CPU and renewed thermal compound on Intel
heat sink fan. As before, first memtest failed. Now running second
memtest. Samsung 4GB modules are cheap. Time to buy more from eBay for
total 16GB RAM memory.

Q9650 CPU works OK. When I set up the SSDs' SATA and power cables, I ran
into trouble. Win10 tried to boot. It failed because it did not find GTX950
HDMI display driver, even though those drivers were enabled last time Win10
booted. Win10 said the PCI VGA was not compatible with Win10. I may have a
troubleshooting CDROM somewhere. Still running Memtest.

This should not stop Windows 10.

Windows (as is traditional) has a built-in VESA driver. All video
cards have a standard legacy interface, which the Microsoft Basic
Display Adapter driver can see. The VESA style driver usually limits
the resolution to a safe value - this avoids lawsuits that happened
during the "non-multisync" era, where monitors were ruined
by OSes using too high a resolution setting. The companies have
never forgotten this, and to this day limit themselves to 1024x768
for example. And yes, a 4:3 resolution looks strange on a
16:9 monitor :-)

Is there more than one display device in the computer ? Would
the second display device be 1024x600 resolution ? There has
got to be some explanation for why the MBDA driver did not
cut in and bring up the system.

I can bring up Windows 10 with an FX5200 in the machine. And
the FX5200 doesn't have a Windows 10 driver.

Paul