Follow along with the video below to see how to install our site as a web app on your home screen.
Note: This feature may not be available in some browsers.
"It is getting very obvious the Vmem (or probably more specifically the memory Vddq) is to high
for the Qimonda chips even though 2.00 nominal is what Qimonda specs for both memory Vdd
and Vddq with the .80 and 1.0ns speed bin chips and is what the GT's run out of the box.
Qimonda Specs 1.80 nominal for Vdd/Vddq on the 1.1, 1.2 and 1.4ns chips so they just jacked
the voltage up to get a .80 and 1.0ns speed bin on the higher speed rated chips (they all come
off the same FAB wafers). Note the practice increasing the voltage to get higher rated speed
bin chips is normal and both Samsung and Hynix do it as well.
It is known that running the I/O buffer voltage (Vddq) can hurt a memory chip over time and
that would be exacerbated and could be rapidly accelerated by running at higher frequancies
at the same time. Normally 2.0 Vmem wouldn't be considered all that high...at least with
Samsung or Hynix DDR3 chips anyway.
It is within the realm of possibility the combination of the two at the same time (higher
frequency and higher Vmem) could be hurting the Qimonda chips rapidly (although there is
no proof of that in the public domain) and that is what is being referred to on the web about
running an 8800GT's memory over 1ghz killing them. If that is so then in this case the root
cause isn't higher frequency, it is the higher Vmem.
It kinda looks like Qimonda may possibly have missed something during certification/
speed bin testing of the .80 and 1.0ns chips and nVidia in memory chip qualification testing
(like running an old benchmark with a test that point it out rapidly)."