pinmods

  • Thread starter Thread starter GT
  • Start date Start date
I agree - an 85% drop will only gain me a few degrees of temperature, so the
noise output isn't going to change drastically. I have a new heatsink and I
was just investigating the possibility of running it entirely passively, or
with just a few undervolted fans to give carefully designed airflow. I have
abandonned the idea of undervolting the CPU.
Click to expand...

Ironically I've just come across a scenario where I'll end
up doing something similar (in power reduction, not a
pinmod) to decrease CPU heat production.

I had a spare A64 I knew could o'c to 3Ghz easily (at stock
voltage) but the system it was in had since been upgraded to
a X2 processor. Having the spare processor and leftover DDR
memory (was a skt 939 chip) I got an Asus A8N5X board to run
these leftover parts.

Oddly, the A8N5X board seems to do a terrible job. Perhaps
it's just a poor VRM design but it can't run the same CPU
with same heatsink it previously used ('sink pulled and
reinstalled as well, is definitely making good contact), and
same vCore, without overheating even at lower than stock
voltage and a 200MHz lower speed. At first I just thought
it was an inaccurate temp reading as some Asus boards do
seem to report higher CPU temp than other brands, but sure
enough runnning a Prime95 torture test it crashes when the
temp reported rises above about 62C.

Granted it could probably run w/o overheating if I put a
different fan on the 'sink but the whole point was to
duplicate the result seen on the other board with it running
inaudible.

Since it just wouldn't run cool enough I ended up reducing
vcore from 1.35V default to 1.2V and speed from 3GHz to
2.7Ghz to keep it stable @ 1.2V. Not entirely satisfied
with this I would've normally investigated why and perhaps
upgraded the capacitors in the VRM subsection on the board
but unfortunately Asus chose to use 8mm dia. caps so that's
a real limitation. They're Panasonic FJ caps so not
entirely junk quality but it does seem as though the board
wasn't engineered to provide enough clean power for an A64 @
1.35V/3GHz, their higher tiered boards using very similar
design have more (I think it was) UCC caps of higher
capacitance.

I'd wondered if it was a memory, bus speed, etc issue but
changing all these paramaters made no difference, it just
won't run the processor as cool at full load as the prior
(Biostar TForce4 U) board did. So much for Asus' lower-end
boards. If it were all new parts I'd just send the board
back but for it's purpose, price, and the time to redo it
all, this will have to do.
 
Oddly, the A8N5X board seems to do a terrible job. Perhaps
it's just a poor VRM design but it can't run the same CPU
with same heatsink it previously used ('sink pulled and
reinstalled as well, is definitely making good contact), and
same vCore, without overheating even at lower than stock
voltage and a 200MHz lower speed. At first I just thought
it was an inaccurate temp reading as some Asus boards do
seem to report higher CPU temp than other brands, but sure
enough runnning a Prime95 torture test it crashes when the
temp reported rises above about 62C.
Click to expand...


After writing this I decided to get to the bottom of the
problem...

The biostar board which had better results uses a different
brand of socket, 'sink mounting frame which seems to put
more pressure on the heatsink-CPU interface. While the same
heatsink had a poor result on the Asus board, substituting
another 'sink which has roughly equivalent performance
(otherwise) resulting in the temp dropping down to an
expected range. This is after I'd already remounted the
original 'sink 3 times, the same as I always did and had
good results with on the other board.

Further examination also showed that when the AN85X was set
to 1.35V vcore, the processor was actually getting about
1.4V idle and 1.42V under full load. Unfortunately when
the bios vcore setting was reduced until actual voltage
remained at the target 1.35V, the processor was no longer
stable at as high a(n overclocked) speed. These two factors
seem to account for the differences in heat but now I wonder
if Asus' design to overvolt the processor slightly was just
bad design or if they'd intentionally done so to save a buck
on their VRM circuit - at the users' expense of heat and
power.

Regardless the processor was stable at stock voltage when it
actually received the stock spec'd 1.35V so ultimately it is
just noteworthy as a board that has atypically worse
overclocking performance than most of Asus' offerings. It
reminds me of another Asus board I had in past years which
was also built to a price-point (A7V600) but had overclocked
more in line with their upper-end models. Ultimately I
think they should have sold this one as an Asrock instead of
Asus brand.
 
Back
Top