Re: first attempt ot OC

timO' wrote:


the power supply you sent happens to be on sale.
yeah i was asking about a case recommendation, but you are right.
i'll take my mind off that for a while.
i have two hdds and one of them is really not being used, but i am
clueless about linux. were you saying it would be wise to set up
another hdd in case of a crash?

what are the baby steps for me to try and OC in small steps. I'd be
fine with 3.6; at least while i'm just beginning to OC
thanks again for your thoughtful and informed responses

Download a copy of CPUZ from cpuid.com . You can
use that to monitor what is happening after an experiment.

Go to the BIOS.

1) AI Overclocking [Manual]
2) CPU Frequency will be 133MHz. That corresponds to FSB533,
the normal (quad pumped) transfer rate of the front side
bus.
3) Adjust the CPU Frequency, to 5MHz higher. That means your
first test will be at 138MHz. The test is bound to succeed,
without need for extra Vcore.
4) The manual says the DRAM frequency offered with a FSB533
processor, stops at DDR2-533. Say the "DRAM Frequency"
setting in the BIOS is DDR2-533 right now. That is a
1:1 ratio with the FSB. The DRAM increases when the CPU
speed increases, as they are tied together. The BIOS
may read DDR2-533, but the actual speed is higher.

With these settings, we changed

1) CPU input clock from 133 to 138MHz.
2) New FSB becomes 4*138 = FSB552
3) New DRAM becomes FSB552*(1/1) = DDR2-552 (even though the
BIOS display reads DDR2-533).

This is less than the DRAM rating of DDR2-800 (PC2-6400 / 8bytes)
The memory should not be stressed by running at DDR2-552.

Now that the CPU clock is bumped a little bit, save the BIOS
changes, and boot into Windows. Run CPUZ and verify the
settings. On the CPUZ DRAM window, you're supposed to multiply
the 276MHz value displayed, by 2, to get DDR2-552.

The RAM experienced a small change in this case, but
as you progress in overclocking, larger changed will be
applied. You should also be correcting the four timing
numbers. You can use the SPD table in the CPUZ program,
to get values to use at various actual speeds.

In the example below, the RAM is rated at 3-3-3-8 at DDR2-533.
And at 4-4-4-12 at DDR2-800. When we increased the CPU
clock, the BIOS RAM setting may have been left at 3-3-3-8.
The timing should be relaxed to 4-4-4-12 value of the DDR2-800
entry, as we know the RAM will be stable up to DDR2-800
with those timings. And our new proposed speed is
"DDR2-552 actual", and we should be using timings that
will work between just above DDR2-533 up to DDR2-800.

http://www.madshrimps.be/articles/ASRock4Core1333-eSATA2R50IntelP31Mother-Sidney%20Wong-19169.png

It is also possible to extrapolate those memory timings,
for operation above "DDR2-800 actual". The arithmetic
should involve ratios, as in 1066/800 * 4 = 5.33, and
rounding to the next integer gives 6. As in 6-6-6-18
or so. 6-6-6-18 would be good up to DDR2-1066, while
we can guess that 5-5-5-15 would be good for some
small range of frequencies just above DDR2-800.

In any case, the timings should be compensated, for
the degree to which the clock speed is changed to
the RAM.

I know this is a lot to absorb, and chances are, you'll
make mistakes once you apply more ambitious frequency
increases. And that takes us to the topic of crashing.

If you boot into Windows, you could crash while
doing so, and Windows could be corrupted. If you
have a second hard drive, you could use backup software,
to make a copy of the disk. Store the disk in a safe
place, away from the overclocked computer. If the
Windows boot disk is ever corrupted, return the
settings to "stock" values (CPU input clock = 133Mhz),
then do a restore using the backup image.

An alternative to that, would be to download a
700MB Linux distribution, and burn a boot CD
with it. That allows booting Linux from CD, and
a CD cannot be corrupted by overclocking. But
CPUZ doesn't run in Linux, so the available tools
to work there are different.

Anyway, those are some basic principles for an Intel
chipset board. As the CPU frequency goes up, at
first, the stock Vcore voltage for the processor
will be enough. At some point, it is going to crash.
Then, you increase Vcore by one notch, and try again.

You keep careful notes, of applied frequency and
necessary Vcore voltage. You also need to do a little
research, as to how much voltage the D 805 can take.
The Tomshardware article has a few sample voltage
values, from their overclocking adventure.

The last overclock I did, I bumped the CPU frequency
by 5MHz per time. And only increased Vcore when
stability was lacking. (I used Prime95, and if an
error was detected in less than 10 minutes, I bumped
the voltage. So in fact, the computer did not crash,
and I used a stress test like Prime95, to detect when
I was getting "close to the edge".) I plotted the
values, to get a curve, and that helps predict what voltage
will be needed to hit a higher frequency. That
could be more or less volts than the Tomshardware results.

The applied Vcore voltage, determines heat output as
well, so if you're using more Vcore voltage to get
to 3.6GHz, than Tomshardware needed, then you
can forget about 4GHz. The power dissipation of the
processor is F*C*V**2, where F is the increasing
clock speed, and V is Vcore. Note that Vcore contribution
is squared, which means power dissipation is affected
quite a bit by Vcore. C, for the purposes of this
discussion, is some constant. When you overclock, F and
V are increasing at the same time (eventually you will
need some extra V), so the power dissipated is not a
straight line relationship with frequency. it is more
of an upward pointing curve. Which makes that 6GHz
record, all the more amazing when you think about
it - the liquid nitrogen must have been boiling
like crazy.

It is always tempting to skip a step, like just read
the Tomshardware article, copy the voltage used at
3.6GHz, and dial in a CPU input clock to go right
to 3.6GHz. That will undoubtedly lead to a crash
(that is what happened to me when I tried it once).
The purpose of taking the small steps, is to become
more familiar with the controls, but also to plot
the trends, to predict how your particular processor
is working. It could be a better processor than
Tomshardware used, or it could be a worse one.

HTH,
Paul
.

Relevant Pages

  • Re: Trap exception
    ... VCORE reading (because of overclocking). ... Can you list you overclocking menu in the bios for me? ... > Here are the voltage readings: ...
    (microsoft.public.windowsxp.general)
  • Re: Trap exception
    ... not by the bios. ... > Have no overclocking menu in BIOS. ... > Can not change voltage in BIOS. ... >> VCORE reading. ...
    (microsoft.public.windowsxp.general)
  • Re: Random screen freezes with P4P800-E
    ... Speedfan now shows all voltages (except Vcore) to be at or above specs all the time, something that wasn't happening with the old PS. Vcore, which is supposed to be 1.6V, runs between 1.54 and 1.6V. ... I monitor the voltages, but they were never that far below specs in the first place. ... then the loaded voltage could be 1.40V and be good. ... As for PSU, the more test equipment you own, the more science you can apply. ...
    (alt.comp.periphs.mainboard.asus)
  • Re: ASUS P5E-VM HDMI over clocking
    ... Vcore voltage. ... as the BIOS writing out a value of VID to the regulator, ... the motherboard may use another mechanism for boosting the ...
    (alt.comp.hardware.pc-homebuilt)
  • Re: ASRock motherboard wont permanantly save custom Vcore BIOS settings change.
    ... > I've having an issue with my recently purchased ASRock 939Dual-SATA2. ... > existing the BIOS. ... the settings revert the VCore ... > back to Automatic and its default Voltage. ...
    (alt.comp.periphs.mainboard.asus)