First overclocking attempts

[-]

I've just tried overclocking (AMD Athlon T-Bred 1800+ XP CPU) for the first
time. Mobo: ASRock K7VT2.

I've tried overclocking to 134 and 135 FSB speeds. At 134, the machine works
perfectly fine, at 135, the machine doesn't boot to the POST. I reckon its
my PSU (Q-Tec 400w Gold Dual fan, which I've heard is a bit crap). My temps
are fine case: 21c CPU: 32c. Is the extra power needed by upping the FSB by
1MHz enough to prevent the machine from having enough power to boot?

The machine has 3 HDs, a CDRW/DVD combo drive, TV card and 64MB PCI graphics
card, 2 case fans a USB powered device and the usual other stuff. Am I right
in thinking that the power supply is the issue? I'm not bothered if it is,
I'll stick the FSB back to what it should be and relax that my system won't
come so near to being underpowered.

Thanks for any suggestions.

 

[kony]

I've just tried overclocking (AMD Athlon T-Bred 1800+ XP CPU) for the first
time. Mobo: ASRock K7VT2.

I've tried overclocking to 134 and 135 FSB speeds. At 134, the machine works
perfectly fine, at 135, the machine doesn't boot to the POST. I reckon its
my PSU (Q-Tec 400w Gold Dual fan, which I've heard is a bit crap). My temps
are fine case: 21c CPU: 32c. Is the extra power needed by upping the FSB by
1MHz enough to prevent the machine from having enough power to boot?

The machine has 3 HDs, a CDRW/DVD combo drive, TV card and 64MB PCI graphics
card, 2 case fans a USB powered device and the usual other stuff. Am I right
in thinking that the power supply is the issue? I'm not bothered if it is,
I'll stick the FSB back to what it should be and relax that my system won't
come so near to being underpowered.

Thanks for any suggestions.

Not being able to raise the FSB even 2 MHz is a problem, IMHO... too
little margin for stability even if it was overclocked. However it
seems doubtful that a mere 2 MHz is causing this, I'd more likely
suspect a BIOS bug.. sometimes what the BIOS calls "2 MHz" isn't even
accurate, the BIOS is allowing a finer adjustment than the clockgen
actually supports for marketing purposes.

You might use Google searches to find a software overclocking app that
supports your board's PLLIC clock generator (I don't know what it
uses, maybe someone in a forum (somewhere) does or you could get a
strong flashlight or pull the board out and take the numbers off the
chip itself. Something like "CPUCool" or it's little brother "CPUFSB"
might work, don't know.


Dave

 

[-]

Thanks for the info : )

 

[-]

sometimes what the BIOS calls "2 MHz" isn't even

accurate, the BIOS is allowing a finer adjustment than the clockgen
actually supports for marketing purposes.

I've been thinking about this - are you saying that when the BIOS 2 Mhz
increase is implemented, it may actually only be being upped by a much
smaller amount so that the clockgen doesn't really complain because it is
within the clockgen's +- error figures or something like that?

When you said about using CPUCool, do you mean that when the machine has
past the BIOS that I could then tweak the FSB and the BIOS would no longer
complain? I'm a little unclear as to how that would make a difference
because I thought that the BIOS still has to deal with FSB when the OS is
running, so it would be like altering the same value just after its booted
instead of before and so would probably cause an immediate crash.

Thanks for the much valued input.

 

[kony]

I've been thinking about this - are you saying that when the BIOS 2 Mhz
increase is implemented, it may actually only be being upped by a much
smaller amount so that the clockgen doesn't really complain because it is
within the clockgen's +- error figures or something like that?

I mean that you many have more choices than the clockgen actually
supports, or those are rounded off improperly, or any number of
bizarre BIOS quirks, or it may be dead-on correct on everything.

I don't mean that it's only "a much smaller amount", but that it may
not be accurate. Also, it is possible that not all choices will work
properly due to bugs or lack of support when different revisions of
boards use different clockgens, but the manufacturer keeps releasing
the same bios to cover multiple boards.

In other words, that 2MHz increase could be impossible to use with any
CPU, power supply, memory, etc, but a 10MHz increase (or some other
number) might work. We can't conclude much from a single test of a
minor speed increae, except that something is wrong... that something
may not be anything to worry about if it's simply that the board can't
support that frequency choice but on the other hand, if you really are
that close to instability, you ought to find out why. Even on my
highly overclocked systems there's over 2% margin away from that
instable upper limit.

When you said about using CPUCool, do you mean that when the machine has
past the BIOS that I could then tweak the FSB and the BIOS would no longer
complain? I'm a little unclear as to how that would make a difference
because I thought that the BIOS still has to deal with FSB when the OS is
running, so it would be like altering the same value just after its booted
instead of before and so would probably cause an immediate crash.

Thanks for the much valued input.

Some boards, for whatever reason, can't POST and boot at certain
speeds, it's a bios glitch. It a perfect world those glitches would
be fixed, but in this world it may never be fixed. Using a software
is simple in that entering your clockgen allows choices of all
possible frequencies (IIRC), so it's relatively trivial to up it that
2MHz. It might be instable or it might not. Just keep in mind that
huge increases in speed may incur problems, you should limit testing
to reasonable speeds for the technology you're using. I'm not
necessarily advocating using a software to up the speed every time you
boot, but for testing. The other test I would do is increasing the
FSB even more than 2 MHz, trying the next few choices.

Also it might be helpful to research your motherboard in some forums.
Popular board have more end-user support than you might expect, if a
fair number of those are running you many find a lot of examples of
what speed it can run with particular revisions or CPUs, etc.


Dave

 

[-]

I really appreciate and enjoy reading your posts Dave - I think as you said
I'll do a minor increase/s of the FSB and if then if it doesn't POST, then
I'll do some more thorough research in more relevant forums on what may be
the likely cause. Thanks for clarifying the BIOS issues - I understand what
you mean now.

 

[-]

I'll do a minor increase/s of the FSB

I successfully changed the FSB to 136MHz and all now seems to be well - I
guess you were right the value I used before wasn't probably supported
properly in the BIOS - I now have a 1900xp+ whereas before I had a 1800xp+!

I can't believe that about 32Mhz constitutes a brand name difference like
that - its hardly any different at all whereas in marketing terms it may
seem like a better thing to get a faster 1900+ xp processor, which really I
don't think would be worth the extra over the 1800+ xp(don't know how much
extra exactly though).

With the extra increase in FSB, the RAM changes to 136Mhz too. Is the memory
more likely to give out before the processor when overclocking? I'm not sure
I want to push the RAM too much whereas I've heard quite a few people
overclock my CPU type to well over 2Ghz, which I would quite like to stretch
slightly further if I can.

 

[kony]

I successfully changed the FSB to 136MHz and all now seems to be well - I
guess you were right the value I used before wasn't probably supported
properly in the BIOS - I now have a 1900xp+ whereas before I had a 1800xp+!

I can't believe that about 32Mhz constitutes a brand name difference like
that - its hardly any different at all whereas in marketing terms it may
seem like a better thing to get a faster 1900+ xp processor, which really I
don't think would be worth the extra over the 1800+ xp(don't know how much
extra exactly though).

It's not just 32MHz, but being rounded off. Still I see you point, it
isn't much difference, but it's always that way for same family of
CPU, small increments to tide us over till their next gen of CPU is
ready.

With the extra increase in FSB, the RAM changes to 136Mhz too. Is the memory
more likely to give out before the processor when overclocking? I'm not sure
I want to push the RAM too much whereas I've heard quite a few people
overclock my CPU type to well over 2Ghz, which I would quite like to stretch
slightly further if I can.

Insufficient data. If all you're doing is raising the FSB, starting
with 133MHz, it depends a bit on the motherboard chipset and Athlon
family, but if it's a T'Bred "B" and you can up the voltage you may
find the first problems with the PCI bus, getting up around 40MHz
towards 160MHz FSB, and USB. It depends on what you memory can do,
some PC2100 isn't too good over 140MHz but some can do 180MHz and
more, especially if you relax the bios timings and only use 1 or 2
modules. If your motherboard supports 1/5 PCI divider it may be
easier to hit 166MHz than 158Mhz, for example... can't say, it helps
to have the CPU's multiplier LOWER so you're removing it from the
equation, are then testing the motherboard and memory, and vice-versa,
leaving the FSB at spec'd speed while raising the multiplier...
finding the max of each component so you can better make a judgement
of what your options are.

Whatever you decide, be sure to thoroughly test the memory with
http://www.memtest86.com before booting windows... HDD and/or data
corruption is the worst kind of failure since you might lose data and
not realize it till later. Whatever your max stable speed is, back it
down a bit, give yourself a margin for stablity.

It is quite conceivable that at some point the power supply fitness
will come into play, you'll have to keep an eye on that. The other
concern with the power supply isn't just short-term stability though,
it's the additional ripple, made even more significant when the system
is using more power as with overclocking (moreso the CPU than anything
else, especially with a Vcore increase).


Dave