Re: Overclocking Failed! message...But I'm not!!

In article <1122553074.606095.178860@xxxxxxxxxxxxxxxxxxxxxxxxxxxx>,
"Desyfa" <adam@xxxxxxxxxxxxxxxx> wrote:

> Hi, (sorry for the very detailed post)
>
> Hopefully someone can help me, I'm pulling my hair out trying to figure
> this one out!!
>
> I've just recently built a PC using the following components:
>
> 640 Intel Pentium=AE 4 LGA 775 CPU (3.20 Ghz 800FSB) HT 2MB Cache
> Asus P5AD2-E-Premium
> 256Mb Connect 3D PCI-E X800 Tv/DVI ViVo
> Arctic Cooling ATI 4 (X800 Series) VGA Silencer
> 1Gb Twin Pack (2x 512Mb) Value Select DDR2, 533 MH PC4200
> 200Gb Seagate Barracuda (7200.7rpm, 8Mb) - SATA
> Samsung Combo Black 52x32X52 CD-RW + 16x DVD-ROM
> 1=2E44Mb Sony Silver Floppy Disk Drive
> Arctic Cooling T2 Silentium Silent Midi Tower Case - 350W Seasonic
> Silent PSU
> Windows XP Professional SP2
>
> When I press the power button on the front of the case, the pc seems to
> start with a bit of disk drive whirring, no beeps though, and the
> monitor does not turn on. I then have to turn off the power off at the
> wall, turn it back on, press the power button on the front of the case,
> and then the PC starts and shows the following message:
>
> Overclocking Failed! Please enter Setup to re-configure your system.
> Press F1 to Run SETUP
> Press F2 to load default values and continue
>
> As I haven't changed any settings in the BIOS, if I press F1, the BIOS
> shows exactly how it was before. The PC then starts fine and I have no
> trouble whist in Windows. If I press F2, the PC then starts fine and I
> have no trouble whist in Windows.
>
> If I shut the PC down completely, and then press the power button on
> the front of the case immediately afterwards, the PC starts up fine.
> If however, I wait few minutes to turn the pc back on, the same issue
> arises, ie the pc doesnt start up and the monitor doesnt turn on.
>
> Obviously, the easiest and cheapest thing is to do the "power on, turn
> on, power off, power on, turn on" thing every time. However, I don't
> know if anything is being damaged in the PC when I do this....so this
> could be costly in the future if for example something melts or blows.
>
> The power supply in the Arctic Cooling T2 case is a Seasonic 350W
> Continous Silent PSU. This has a 20-pin connector on it whereas the
> motherboard has a 24 pin connector. One of the things I'm wondering
> is, is whether I need a different PSU with a 24 pin connector, or get a
> 20-pin to 24-pin adapter that I have read about on some newsgroups.
>
> The Connect 3D Radeon X800 PCI-E graphics card does not have a power
> connector like the later models.
>
> I haven't flashed the BIOS, I don't know if its worth attempting an
> update. The details of the BIOS are:
>
> AMIBIOS
> Version: 0404
> Build Date: 02/03/2005
>
> To me, this seems like an issue with power, either going to the
> motherboard or to the graphics card. Maybe some settings in the BIOS
> should be changed with regard to power, I don't know.
>
> If any one can help, it'd be very much appreciated!!!!
>
> Thanks,
>
> -adam

Your computer case is reviewed here:
http://www.ap0calypse.com/showthread.php?t=1642

The review says the built-in PSU has these limits:
+3.3V@28A, +5V@30A, +12V@17A

If I try a few numbers, like 8-9 amps for the processor,
3+ amps for the video card, 2 amps to spin up the hard drive,
1 amp for case fans, I cannot quite exceed the 12V rail of
the PSU. (CD drives shouldn't be drawing any 12V unless there
is media present.) Unless there is something wrong with the
Seasonic, it looks like it should work.

Testing with a slightly more powerful PSU would be a good
test (mainly as I cannot think of any other tests to run).
I suppose the odds of having one just sitting around
are pretty remote, but a cardboard test of all your
components sitting on a table top, and using a different
PSU would be educational. You would want a PSU with a bit
more +12V. (If you use an ATX 2.0 style PSU, where the 12V
is split into a couple of separate rails, you need even more
current, as the two rails don't share. 12V2 powers the processor
and 12V1 powers everything else. Maybe 12 amps on 12V2 would
give a bit of breathing room. The 12V1 wouldn't need nearly
as much, and should be sized according to the number of
12V peripherals. Due to the lack of sharing between 12V1 and
12V2, you end up buying more PSU capacity.)

Which is 12V1 and which is 12V2, is documented on page 37 here:
http://www.formfactors.org/developer/specs/ATX12V_PSDG_2_2_public_br2.pdf

As your video card only draws slightly more than 3 amps on
+12V, and that current is coming from the single pin on the
ATX power connector, there is no reason to panic, about
the 20 pin versus 24 pin thing. If the motherboard needed
more +12V, then buying a proper 24 pin power supply would
be a good idea, but I don't see the 6 amp limit of the single
+12V pin on the 20 pin ATX connector being an issue for you.
You might be using 4 amps of the 6 amps of allowed current
flow through the pin.

I don't see a particular damage mechanism waiting for you,
if this problem is not fixed. You have about the same odds,
as if you were using a cheap power supply. Seasonic is not
a bad brand, and their forward converter designs only seem a
bit strange (their S12, for example, seems to have a weak
3.3V, and the voltage drops on 3.3V more than it should).
Other than that, their forward converter design seems to be
pretty efficient, which is a good thing.

So, the evidence so far, doesn't point to a "smoking gun".
Your config should work, but one of the two parties (mobo
and PSU) is not happy. This could be a cold boot problem
(of which there are many possible root causes), or it could
be the motherboard Vcore regulator shutting down because
it thinks there is a problem.

One of the potential problems with Vcore designs these days,
is the use of a switching regulator with "latchoff" on
overload. Some of the older converter designs (from P3 days)
operated in "hickup" mode. What that means, is if a problem
occurs, the Vcore converter will try to start again, at a
low repetition rate. Say, once a second, it tries to power
the circuit, and if the current is still too high, it might
shut off the current after 50 milliseconds. The advantage
of the hickup mode, is if there is poor transient response
of your ATX PSU during the first second or two of operation,
a couple of "hickups" and you are "away to the races". With
the current generation of "latchoff" designs, you have to
press the power switch on the front of the case, and turn
the power off, for the Vcore fault detector to clear. This
means, if there is some weird behavior at startup on the PSU,
the Vcore regulator circuit can be intolerant of the behavior,
and you'll never get anywhere. At least one Asus motherboard
design had trouble with a bunch of different models of
Antec power supplies, due to a problem like that (at least
the symptom descriptions suggested that mechanism). Only the
initial production run was affected, and I think a component
change on later motherboards fixed it.

I could be wrong, but for the "overclocking failed", the
processor has to detect that initialization did not succeed
on the last attempt. I don't think the algorithm is documented
anywhere - it could be something like setting a bit in CMOS
eariy in the initialization sequence at POST, and clearing the
bit just as boot is completing, or perhaps even during the
computer shutdown sequence. If the computer wakes up, and
finds the bit is set, the implication is the computer crashed
during POST or later, on the last powerup. But, this algorithm
also implies that the CPU got to execute some instructions,
which would not happen if Vcore shut down when the power was
first applied. Using those lines of thinking, I cannot see
a nice neat theory to explain it all - it is less likely that
the board would run for a fraction of a second, only to have
a power problem at that instant. It would be really handy
to have a precise definition of how "overclocking failed"
is detected, as that would make it easier to figure out
whether the CPU actually got to run for a fraction of
a second or not.

So, the only thing I can suggest, is a cardboard test using
a different PSU.

One problem I have with these cases with the bundled PSUs,
is you don't get a say in what size of PSU is used. Your
current PSU is adequate for the job, without a lot of spare
margin. If your PSU brand was say, Leadman, I'd suggest
replacing it immediately. With PSU specs these days, the
tough part is figuring out how much margin is needed, to get
the actual power you need (i.e. correcting for power number
fudging).

HTH,
Paul
.

Quantcast