View Single Post
  #7  
Old July 17th 03, 08:58 PM
John Lewis
external usenet poster
 
Posts: n/a
Default

On Sun, 13 Jul 2003 21:53:08 +1000, "MattB" matt.baranski @
bigpond.com wrote:

Hi all,

I picked up an Albatron Ti4200 128mb card today and want to overclock it a
bit. I tweaked the driver level clock frequencies to 270mhz core and 533mhz
memory and the card seems perfectly stable under D3D games with no
artifacts. My question is, will this cause the card harm in the long term
(stock heatsink/fan)? I plan to hang onto this card for at least a year,
does excess overclocking show itself as system instability or is it likely
to just die unexpectedly after being overclocked for a period of time? I
guess basically I want to know how much overclocking will reduce the life of
the card.

Cheers,
Matt


Russian roulette if you have no means of monitoring the actual GPU
chip- temperature. Estimated life is ~ divided by 2 for every 10
degreesC above 80 degreesC CHIP-temperature. Damage is irreversible.
You may or may not see artifacts before irreversible damage.
Silicon slows down significantly with increasing heat, so it may slow
enough to see artifacts before transistors or silicon-vias go POP or
it may not............

Just make sure that you have enough reserve cash in the bank
for a replacement card.

The video card memory is far less likely to go POP unless it is has
poor local cooling or is grossly overclocked. Or shares the same
physical heat-sink as the GPU plus a poor heat-sink design allowing
low thermal resistance between GPU and memory.

FYI:-

Unlike the CPU chip-temp monitoring on modern motherboards, most
video cards have no GPU chip- temperature monitors, although the
design and thermal-stress rules are exactly the same as for CPUs. The
latest video cards ATi9700, 9800, FX5800, FX5900 cost more than most
CPUs and disspate as much power as a 2.6GHz P4. Out of these only
the FX5900 has built-in user-accessible chip-temperature monitoring.

John Lewis