The said:
Howdy: I'm building a new computer around a Gigabyte 880GA-UD3H. I
bought 4GB (2X2) DDR3 1333 dual channel kit for this build. I also
have 3 sticks of DDR3 1600, all of the RAM is Corsair XMS3 series. The
DDR3 1333 is 1.5v and the DDR3 1600 is 1.65v. The motherboard only
supports 1.5v.
So, I'm wondering if I install the DDR3 1600 along with the DDR3 1333,
will the DDR3 1600 work with the 1333 or will the whole thing blow up?
Will the DDR3 1600 work at the lower voltage? I'm not interested in
obtaining the most speed out of these sticks, just a normal stable
system. Any thoughts?
Thank you.
Doesn't the max speed drop, when using four sticks on an AMD board ?
This is a function of the processor design, and not the motherboard.
http://en.wikipedia.org/wiki/Phenom_II
"current Phenom II platforms limit the usage of DDR3-1333 to one
DIMM per channel; otherwise, the DIMMs are underclocked to DDR3-1066."
That means, install four DIMMs, set BIOS to "Auto", the BIOS should run
the four sticks at 1066, well below the 1600 rating of the fast sticks.
And they then won't need the 1.65V to work well. If you only install
two DIMMs, one per channel, then 1333 is the automatic setting with two DIMMs.
If so, it's likely (but not guaranteed), that 1.5V will run all of it.
I'd try to install it, as two matched pairs, and the BIOS will select
a clock and timings consistent with the slower pair of sticks. (In other
words, the BIOS would use 1066, but if you had a stick that was even slower
than that, the stick spec value would be used, like say DDR3-800.)
On AMD, the memory interface is right on the processor. That means
the memory characteristics, are not a function of the Northbridge.
The processor dictates the behavior, rather than the motherboard.
You can modify the memory settings, test and see if any
additional performance is apparent. Doing so, requires the
usage of memtest86+, before booting the OS, after a memory
settings change. That is intended to test the memory, so that
bad memory setting choices won't crash the OS or damage the registry.
Even if you use memtest86+, and it reports the memory is error free,
you can still have an OS crash. One person reported his memtest
results were clean, and yet the OS crashed on boot. So it's still
possible to have memory problems, even if you pass memtest. Using
memtest is just a suggested step, to eliminate really bad
settings choices.
*******
In terms of voltage choices, you can download the chip manufacturer
data, to see what is guaranteed to be safe.
For example, this is some "Intel 1.35V" DDR3 memory. When you look
in DRAM listings, you occasionally see RAM where they tell you it's
for an Intel system. Some of that, takes advantage of support for
lower voltage. AMD may choose not to go that low. (AMD and Intel systems
have different lower limits. The 1.35V RAM, is a recent innovation, intended
to save system power.)
http://www.micron.com/get-document/?documentId=426&file=1Gb_1_35V_DDR3.pdf
DC Electrical Characteristics and Operating Conditions – 1.35V Operation
Supply voltage 1.283 ... 1.35 ... 1.45 V
DC Electrical Characteristics and Operating Conditions – 1.5V Operation
Supply voltage 1.425 ... 1.5 ... 1.575 V
So the memory should be usable to 1.575 volts.
If I look at regular 1.5V memory, it should be the same spec.
http://www.micron.com/get-document/?documentId=425&file=1Gb_DDR3_SDRAM.pdf
DC Electrical Characteristics and Operating Conditions
Supply voltage 1.425 ... 1.5 ... 1.575 V
The RAM is likely good for more than that, as that spec is just a "tolerance"
of 5% on the nominal voltage. The "failure" voltage could be a good deal
higher.
Ah, here it is. The second spec lists an absolute max value.
Table 6: Absolute Maximum Ratings
VDD supply voltage relative to VSS –0.4 ... +1.975 V
The 1.35V RAM spec sheet is an addenda to the regular RAM. And since the
1.35V didn't have an absolute max, the same 1.975 V limit should apply.
Each brand of memory chips, can have their own value for that number.
Running all your RAM at 1.65V, isn't getting too close to that.
Another way to do this, is visit an enthusiast web forum, and see
what voltage they had their RAM "burn up overnight" with. That is
where the really good field data comes from, like what voltage a 45nm
Core2 can take (1.40V), or how much voltage DDR3 can take and so on. It's also
where we learned what killed S939 high end processors, by having the
differential between Vcore and Vdimm too high, rather than too much
voltage of either of them individually. On some architectures, when you
adjust Vdimm, you have to adjust Vcore for safety as well. The enthusiast
web forums, filled with rich guys (lots of dead RAM and dead processors),
is where the less rich guys get their data
Paul
