Re: A7V333 questions

Paul wrote:

Paul wrote:

crustybutthole wrote:

I have a machine with an A7V333 mb installed. Originally this had an
Athlon XP 1800+ installed with (2) sticks of 256K PC 2100 memory.
Board is an early version 1.01 with bios updated to 1017.

About 2 weeks ago I was given a used - but known to be in good working
condition - Athlon XP 2600+. I think it is a Barton because SiSoft
Sandra is reporting L2 cache = 512kB. As I understand, the Barton core
runs at 333mhz while the A7V333 only runs at 266mhz. Is this correct
so far?

I also recently found a good deal on Crucial 1 gig PC 2700 memory, so
I picked up two and decided to try to upgrade this older machine a
little. I installed the Athlon XP 2600+ and the 2 gig of PC 2700, and
tried to see if I could get it to work. MB is in jumper free mode, so
I went into bios and selected highest processor speed available (1917)
and all else to auto, which put the speed at 166/33 with a multiplier
of 11.5x. Machine posted, booted, and windows XP ran fine. However,
over the next week, the machine randomly froze up on about 4 seperate
occasions so I decided to do some web research and see if I could
tweek it a little and get this to stop happening.

A couple notes here: Asus probe was reporting CPU temp of 55 C which I
thought was a little high, but not dangerous. But SiSoft Sandra was
reporting CPU temp of 32-33 C. This mb does not support on die
temperature sensing, but uses a thermal sensor which I may not have
situated in the perfect location.

Also, I read about a possible misprint in the A7V333 manual that lists
the wrong settings for the ROMSIP jumper and suggested switching this
jumper from 1-2 to 2-3. I did this and it didn't seem to have any
effect at all.

Currently, I have the bios set to a speed of 165/33 with a multiplier
of 12x. Sandra confirms a FSB of 2x165 (330) and processor at 1.9GHz.
This has been runing for about 48 hours straight now with no issues.

So on to my questions. I'm kind of confused as to how this mb is
running at 330 when I thought that board version 1.01 could only run
at 266 - even with the bios upgrade. And if it can run at 333, then
why is it sometimes freezing up when set at 166/33 and 11.5x. What
would be the recommended settings for this board with XP2600+ and 2
gig of PC 2700 memory.

Thanks for reading and any advise that you might have to offer.

G


Chipsets can be overclocked. If a chipset is rated at FSB266,
it is possible it could run faster. To run faster, you need
a clock generator chip that allows going faster, and in the
case of the old motherboards, you also need the divider change
to happen at the right point.

According to this, the PCI clock divider switches to 1/5th at
160MHz. So both 165MHz and 166MHz, will yield PCI clocks
very close to 33MHz. And as a rule, AGP is just double the
PCI clock, and will be a safe value as well. PCI is good up to
37.5MHz, before there is a danger of corruption due to the
clock being too high. Some systems have run the PCI bus up to
40MHz, but in that case, no storage devices were getting their
timing from the PCI clock.

http://www.lostcircuits.com/motherboard/asus_a7v333/6.shtml

What the divider issue means, is there are "bad zones" for
clock values. For example, if you selected 159MHz, and the
above article claims the divider would be 4 in that case, then
PCI would be 39.75MHz. And that is too fast to be safe. But
selecting 160MHz for the processor, where the divider is 5,
gives a PCI of 32MHz, which is safe.

Since the CPU and memory clocks scale at the same time,
you should be careful you aren't overclocking the
memory. Say, for example, you select FSB266 and set the
memory to DDR266 (that is a 1:1 CPU:mem relationship).
Now, dial the CPU clock, from 133MHz to 166MHz. Not only
does the FSB change to FSB333 by doing that, but the memory
also changes to DDR333.

If you selected FSB266:DDR333 and then raised the FSB, the
memory would be at DDR333*(333/266)= DDR417. So you want
to dial down the memory, so that when the CPU clock is
raised, the memory stays in spec. The 1:1 setting
should do that for your PC2700 memory.

The lostcircuits article also mentions jumpers for
DDR voltage, but the default looks to be plenty to
make the memory work.

http://www.lostcircuits.com/motherboard/asus_a7v333/5.shtml



And to finish my post, on this page, a guy ran 187 x 9 on
his board. Which indicates there might be a bit of headroom
there to work with. 187 * 2 = FSB374. The post doesn't say
what revision he had, but I suspect either chipset revision
would have room for overclocking.

http://www.hardforum.com/showthread.php?t=811432

The trick, now, is figuring out why it is freezing.

1) Fix the RAM speed, so the RAM is no longer being overclocked.
I don't know for a fact that you're overclocking the RAM,
but if you selected FSB266:DDR333 (i.e. the non 1:1 ratio),
then it could be happening. Use CPUZ from cpuid.com to
verify what is happening. CPUZ runs under Windows.

2) You can certainly try both sticks at the same time if you
want. But for debugging a stability problem, I might try
just one stick. That is less stressful for the memory
controller, and if the problem doesn't improve, then that
might point a bit more at the CPU conditions. For one stick,
put the memory in slot 3 (furthest from the processor). For
two sticks, use slot 1 and slot 3. The more load on a DDR
bus, eventually, the lower the stable clock that will work
with it.

3) Use Prime95 as a test.

http://www.mersenne.org/gimps/p95v255a.zip

Prime95 is a way to accelerate stability testing.
If you can run Prime95 torture test for four hours,
that is equivalent to passing hundreds of hours of
idle desktop usage. Prime95 does a calculation with a
known answer, and detects math errors as it runs. One
error will stop the program running. When I overclock,
Prime95 will stop in ten seconds if I've overclocked
too far. (Prime95 will make the CPU hot, so make sure
your cooling is adequate.)

That version of Prime95 is multicore capable, but that is not
important to your current processor. When prompted to
"Join gimps", you want to select "Torture Test", as that
is the test you want from the program. The default
settings are appropriate if you aren't doing anything
else with the machine, but if you want to mess around
while it runs, then dial down the memory allocation
a bit. If you don't adjust the memory setting, there
won;t be any memory left to start new programs with.

Another possible reason for freezing, could be the
value of Vcore. (Asus boards traditionally overshoot
at idle, so if the voltage readout is 0.060V higher
than expected, don't panic. If nominal is 1.65V, then
a measured value of 1.71V is perfect.) For the
processor I've got, I discovered it didn't like too
much extra voltage, so cranking the voltage won't
necessarily make a positive difference. YMMV.

http://web.archive.org/web/20041213174558/http://www.qdi.nl/support/CPUQDISocketA.htm


As for detecting Barton, the one parameter that is
"sealed in silicon", is the model number. Barton is
model 10 (hexidecimal "a"). The model number may
show when you use a utility like CPUZ, so see if the
model number is mentioned. I have a Barton (not the
machine I'm typing on) and it reads out as model 10.
And that is the only number on the processor, that
isn't controlled by bridges on top of the processor
package.

Paul

I also have the A7V333 with an XP-2600; the BIOS version is 1016 with a
release date of 01/28/2003 from ASUS SEPT/2007; it's surprising that there
is an even later revision. The CPU is unlocked & runs at 12.5x167= 2100 MHZ;
the standard speed is 11.5x166= 1917.
The original BIOS had no thermal trip point settings but the newer one does
under the BIOS Power menu. It runs quite hot with the MB/CPU/Thermal-Diode
= 35/52/72C requiring a trip point of 85C since it can rise to
81C with heavy FP calculations. The most useful monitoring tool is in
Linux where
you can write a small script to monitor the output of the sensors
command showing
the current Temp & maximum seen while the script is running.


.

Quantcast