Galaxy 8600GE Overclocking Experience

Overclocking/Overclocking Tests by massman @ 2007-11-12

A while ago yours truly had the opportunity to review Galaxy´s improved 8600 GT video card. It featured a new design and an extra power connector for higher overclocking. Today we push this card to limit at one of Belgian´s first overclocking LANs.

  • prev
  • next

Murphy ...

Where's the cold?

Our extreme bench sessions were spread over two days; First of all the LN² day, which was together with Blind. It was not a very big success as we didn't had the voltage modifications back then. Using default voltage and some LN², we ran a couple of AquaMark3 runs, but had to stop very soon due to condensation.

http://www.hwbot.org/result.do?resultId=652090

Madshrimps (c)
Click here to enlarge picture


The second and final bench session took place at Geoffrey during the first OCTB OC LAN. Together with Blind, Jort and Geoffrey himself, we benched two different setups: one with the X2900XT and one with the inevitable Galaxy 8600GE. After we got the dry ice cooled X2900XT setup running, we mounted the cascade on the card and used a Mach2GT (modded by Jort) to cool down the E6850 (thanks Blind).

Madshrimps (c)

(picture is from the benching day at blind, at the OCTB Lan, the setup was almost the same)

The law of Murphy strikes:

Just when we were ready for some kick-ass benching, Murphy came around the corner, saying: "Whatever can go wrong will go wrong, and at the worst possible time, in the worst possible way". Running 3DMark01's nature at 1050/1050, we were enthusiast to try for more. We reduced the GPU speed to 1GHz and tried the 3DMark03 benchmark, which started good, but crashed very soon. Reboot and try at 950 ... crash. Reboot, decrease memory frequency ... crash. Reboot ... no signal!

The "no signal" sign is not good, not good at all. Without a doubt, something is wrong with either the graphics card itself or the cable to the monitor. In this case, it was the card itself.

I already knew the memory was very sensitive when it came to overclocking. My first testing led me to an overclock of 1.1GHz on stock voltages, but once I gave more than 2,15v, the memory overclock started to decrease every boot. In the end 1050-1080 was possible with 2,15v but any higher would require more juice.

Coming home from the first OCTB OC LAN, I retried the card and I found out that it was indeed the memory which had crapped out. No burned video card, but the internal structure may have been damaged by overclocking.

Madshrimps (c)
  • prev
  • next