I am running a Duron Applebred 1800 in a Via 266 mobo by Syntax
with 768 KB of 133 SD-RAM. CPU temp under load = 44 degrees C.
Phoenix BIOS.
My crappy mobo doesn't permit very many setting. (EG can't set cpu
multiplier.)
---------
Can I check with you guys about some basic stuff:
If I slightly increase the cpu voltage from the 1.52 V at present
(nominal voltage for my Duron 1800 is 1.50 V) then will that help
increase cpu stabilty?
If it is stable at 1.5V and you don't overclock, it should
not need changed. If you do overclock by any significant
margin, yes you're likely to need raise the voltage some BUT
because you can't change the multiplier, you will soon reach
a stability limit of the northbridge and memory- typically
that limit on Via KT266 is around 150-odd MHz FSB, at which
point your IDE (hard drives and data) are at risk, as well
as USB and network adapters starting to act flaky. Other
parts might too, I mention only the most common. Of
particular concern is the hard drive data, since it may
become corrupt entirely and you have to format and start
over from scratch. For that reason it's a VERY good idea to
make a complete backup of all HDD data if you seek to push
the FSB past (roughly) 145MHz.
I can increase by only 25 mV, 50 mV or 75
mV. I figured that 75 mV feels safe - QUESTION ONE: is this
right?
Frankly, those are too little increase to get you very much
benefit. To fully exploit your chip you'd probably need to
up the voltage to around 1.8V, a 300mV increase. That
increases heat a lot though, and your board limits you too
much to get benefit from doing so.
Generally, you might go ahead and raise it the 75mA, then
proceed to do whatever things you were to overclock, then
once you have it confirmed 100% stable, try lowering the
voltage again and retest stability- there is no need to
raise the voltage if CPU stays stable at stock voltage.
In the past with a slower processor, I increased the "CPU Host
Clock" (which sounds to me like the FSB) from a starting value of
100 to 104 (an extra 4 %) or a slightly unstable 106. I think
memory was the limiting factor for stability this time.
It it was KT133 (not KT133A) then it was the northbridge
that limited you. Generally though, such a minor increase
is hardly worth fooling with. Generally it's good to just
shoot for the moon, ie- use a board that supports the next
higher FSB range and see if it'll post at the max you hope
to use. For example, with a Duron 1.8GHz you would have
13.5X multiplier (IIRC), so you'd raise the voltage to 1.8V
or so and raise the FSB to 166MHz/DDR333.
Since your motherboard won't do that, the options you have
left that are of enough gain to bother with will revolve
around finding another way to increase the CPU multiplier.
Sometimes a bios mod will give you that feature. Sometimes
there's a hidden (undocumented) jumper on boards, or plated
holes (soldered) where there "could" be pins for a jumper
but the manufacturer didn't add the jumpers. On your
particular board I have no idea. Other times you might need
use a low voltage and needles to burn a CPU multipler bridge
on the CPU (risky unless you had another board you could use
to confirm whether the CPU was stable at the new multiplier
before permanently modifying it). last there is the
wire-wrap or wire-socket method where you insert a piece of
wire in the socket or wrap it around two pins to change the
multiplier. Google should turn up some info on this, I
don't recall the specifics for that CPU.
I have currently got all the memory settings at what I believe is
their most aggressive:
SDRAM cycle length = 2,
Trp = 2T,
Tras = 5T,
Trcd = 3T.
Have you confirmed they're stable at those settings? If you
haven't yet, run Memtest86 for several hours. Keep in mind
that IF you decide to increase the FSB any (which will
probably be useful at least up to 140MHz or a little higher)
you may easily loose stability with those memory timings, so
you might want to set them more conservatively and then
later, after you've removed that variable and have checked
stability with memtest86, THEN reduce the timings again and
retest with memtest86. That is the general procedure, but
frankly, these days it's not much return in performance for
all the time spent. A newer motherboard, CPU, memory would
run circles around it.... if you spend hours just to get
another 10% performance gain that's not much benefit
relative to the 80% gain you'd get from newer parts.
QUESTION TWO: Should I be able to increase the FSB to about the
same value as before because, presumably, the faster new processor
(Duron 1800 versus old Duro 700) should have no effect on the FSB
setting?
Your prior CPU used 100MHz FSB (DDR200). The Duron 1800
uses 133MHz. As mentioned previously, you can probably
raise that to at least 140MHz, IF your memory stays stable,
but if you go much beyond 145MHz be sure your data is backed
up. After overclocking it and testing with memtest86, it
would be good to check CPU stability with Prime95's Torture
Test. That should run for at least several hours, or until
the very first error- at which point you can stop the test
and need to take appropriate meaures to regain stability and
begin testing again. Again, a lot of work for minimal gain
these days. Some people might prefer to just leave the CPU
at stock speed, 1.5V, and see how quiet they can get it, as
a Duron 1800 can be the foundation for a silent system, as
it produces less than 1/2 the heat of a modern P4.
QUESTION THREE: Would I be able to increase the FSB bus more if I
made the above memory settings less aggressive? This would seem
to be the only way to increase the CPU speed.
yes, it seems I answered your questions before I'd ever read
them.
Finally, is "Vcore 2" (reported by Motherboard Monitor 5)
essentially only used by the CPU for I/O operations?
You cannot assume what MBM5 is reporting unless you have
compared it to a bios reading. MBM5, being generic in it's
ability to run with different boards, may report different
things for different people/systems. I wouldn't worry
about that voltage at all, you really only need be concerned
about the CPU vcore, memory voltage (if applicable, it may
not be on your board), and the power supply voltages such as
3.3V, 5V, 5VSB, 12V. I doubt yours even has chipset voltage
levels or reporting but if it does you can usually ignore
those on your board too, as the chipset itself is the
biggest limitation and it just won't go very far using
reasonable measures.