Change memory setting in bios

Do you have a question? Post it now! No Registration Necessary.  Now with pictures!

Threaded View
Hi, right now my 12GB is auto controlled. I want to see if I can tweak
to memory and faster FSB (I don't what its called today)
like, I was able in cmos/bios to get a better FSB and my CPU
overcooked,  Hope it can be done.You might notice the 2/4GB banks. I
was told by PC tech to do that to gain dual charnel. I never use more
than 4GB   here is what I have:

my mobo is:
 GA-880GM-UD2H (rev. 1.4)

CPU Type: DualCore AMD Athlon II X2 255, 3100 MHz (15.5 x 200)
Memory Module(s)
Memory Module : Kingston 9905471-001A00LF 2GB DIMM DDR3 PC3-10700U
DDR3-1334 (9-9-9-25 4-34-10-5)
Memory Module : Kingston 9905471-011A00LF 4GB DIMM DDR3 PC3-10700U
DDR3-1334 (9-9-9-25 4-34-10-5)
Memory Module : Kingston 9905471-001A00LF 2GB DIMM DDR3 PC3-10700U
DDR3-1334 (9-9-9-25 4-34-10-5)
Memory Module : Kingston 9905471-011A00LF 4GB DIMM DDR3 PC3-10700U
DDR3-1334 (9-9-9-25 4-34-10-5)
OS Name: Microsoft Windows 7 Professional  SP1 x64

Re: Change memory setting in bios

John Smitty wrote:
Quoted text here. Click to load it

If you don't use more than 4GB of RAM, you could unplug the 2x4GB set
and put them back in antistatic bags. Do that with all power turned
off, to avoid damaging the DIMMs.

Overclocking works best with two DIMMs, not four DIMMs. You get a
higher overclock with two DIMMs. AMD processors don't work well with
heavy memory loading, combined with heroic clock settings.

I would experiment with just changing the memory divider setting
for now. If it says 1333, set it to 1600.

Generally, there is a single clock, which is the origin of
other clocks used on the processor.

    input_clock * core_multiplier   = core_clock    (3100MHz)
    input_clock * memory_multiplier = memory_clock  ( 667MHz)

If you adjust the input_clock (a.k.a FSB clock or could use
other names), it changes both the core and the memory at
the same time. If you adjust the input_clock, it stresses
two subsystems at the same time. If you want to test just
the "core", you turn up the input_clock and turn down the
memory multiplier. That helps keep the memory_clock at
its original value.

Also, the memory timings are measured in clock cycles. You're
at CAS9 right now. Let's pretend the clock period is 1/667MHz
or 1.5 nanoseconds. The BIOS takes a setting of "9", but
the actual physics involve 9*1.5ns or a time period of 13.5ns.

Now, imagine you increase the input_clock by 20%. This
causes the memory clock speed to go up. The clock period
drops by 20%. I go from 1.5 ns per clock, to 1.2 ns.

What happens at the DIMM ? Now the CAS parameter is 9 * 1.2ns
or 10.8 ns. 10.8 is less than the factory specified 13.5ns.
We are *stressing* the memory, by expecting a certain
thing to happen in *less* time than before. Either the memory
can take this (memory is 20% better than anyone knew). Or, the
system dies with lots of memory errors.

So how do I get back to a gentle 13.5 ns setting ? I increase
the CAS value. CAS9 / 0.8 = CAS11.25 and round up to the
next nearest CAS, so CAS12 is the new setting. So if my
plan was to bump up the input_clock by 20%, I might have to
compensate by adjusting the CAS from CAS9 to CAS12. That
brings the actual CAS timing back to the desired 13.5 ns or so.

These are the basic principles.

1) The clocks are related to one another, by multiplier
    values. Changing the input_clock, changes everything.
2) Individual subsystems have their own multiplier settings.
    In the BIOS, when you select "DDR1333", that is just a
    multiplier setting with respect to the input_clock.
    Change the value downwards, to compensate for an upward
    increase for input_clock.
3) Some parameters are defined in terms of "ticks", but the
    physics work in other units like "nanoseconds". If time
    periods are involved, the setting must be changed so there
    is less stress to the memory. If the memory has a spec that
    says it needs 13.5 nanoseconds to do something, you may need
    to adjust the timing parameters to get back closer to the
    factory setting. The value of memory_clock, affects this
    timing value calculation.
4) You can test subsystems one at a time, using their multiplier
    and any associated timings. For example, you could attempt to
    overclock just the memory, increase VDimm, crank memory multiplier,
    fiddle with CAS and friends (increase CAS setting to make
    more headroom for frequency). See how far it goes. Later,
    when bumping the input_clock, you can figure out how
    far the memory_multiplier can be pushed, without seeing
    memory errors.
5) The processor also has a few bus related things. A Hypertransport
    multiplier. Cranking the input clock, may upset the bus
    running from the processor to the chipset. The HT multiplier
    should be adjusted down, to stay within Hypertransport spec.
    Another internal clock on newer AMD processors, is the uncore
    clock that handles the intermediate step between core and memory
    interface. (I haven't a clue what sets that.)

Some motherboards, have a "handy overclocker knob". You ask it
for a "30% overclock" and it adjusts stuff. But it might not
have the same objectives as you. For example, the guy in Taiwan
may have set it, so that the CPU clock shoots up, but the memory
is dialed for a more nominal setting. He may not try to stress
both CPU and memory at the same time (because it increases the
odds the system will crash, and you the customer will hate the
motherboard, and call it junk). So automated overclocks can be
both "safe" (in the sense that they don't stress things that
don't need stress). But, they can also do stupid things, like
use way more voltage on something, that doesn't need it. Like,
leave the memory at the nominal speed, but increase VDimm by 20%
for no good reason. So automated overclock settings, may seem like
a good idea at first, except they aren't designed with an
infinite amount of care. All they really want, is the customer
to see a "fat CPU clock" value, and to hell with every other
aspect of a good overclock. Parts of your system, could end
up running warmer than they need to. That's why we do manual
overclocks, for more careful adjustment. (I only use as much
VDimm as is needed to maintain no memory errors.)

I recommend finding a good web site, that walks you through the
steps for your particular processor. It's probably going to
take quite a few Google searches to find that, but don't give up.
The above is not a "recipe grade answer", it's just an attempt
to give you some idea how crude the BIOS tools provided are.
The user has to know a lot, about the lack of sophistication
of the BIOS tools. The BIOS could have been designed, such that
the user key in "objectives", and the BIOS could have computed
sample settings, plus leave room for you to "tweak" the recommended
values. Instead, we have things like memory clock settings, that
don't reflect what is really happening to the memory clock.

To learn more, you need to work in tiny steps. Get a copy of CPUZ
(the non-install version), from . Check to make sure it
has no toolbar stuff (lots of freebie tools these days, come
with attempts to do drive by toolbar installations, to make money).
CPUZ can show you the current CPU and memory clocks. Make a single
digit change in the BIOS, then use CPUZ to review what actually
happened. Some of the subsystems, probably won't display in CPUZ,
like is HyperTransport clock in there ? So maybe it doesn't cover
all the subsystems, but it's better than nothing. By working
in small steps, you can see (just like gears), how one setting,
affects more than one thing. Once you get the hang of how the
controls interact, you can be a bit more clever about your
adjustments. If you find a website with a page specific to your
family of processor, they'll take most of the pain out of the
learning process.


Re: Change memory setting in bios

John Smitty wrote:

Quoted text here. Click to load it

First, do you know what you're doing, and why you're doing it?
Overclocking the CPU or RAM can indeed "overcook" and destroy it.
Overclocking RAM, except POSSIBLY to match it to a higher CPU clock,
doesn't yield a lot of real-world performance.  You may risk shortening
its life for NO perceivable increase.

Get HWMonitor and CPUz from CPUID to monitor temperatures real-time and
find the results of your overclocking attempts, respectively.  Coretemp
from is also a nice, small utility.

Read up to find the temp limits for your system components.  Make sure
you have MORE than adequate cooling in your case, and that the airflow
is good over the RAM, CPU/cooler, and MoBo chipset.  THEN you can
follow others' guides for overclocking.

Site Timeline