- Joined
- Mar 1, 2002
- Location
- Budapest, Hungary
Okay here is my little story (again)
I upgraded my rig in the sign with another 2.4C that does 3.5-3.6 on air, a ThermalRight SP-94 (with a Zalman 92mm fan and AS5), and an Enermax Noisetaker 470W PSU.
After spending a few days with testing I got it up to an FSB of 285 with 5:4, that means 3.42GHz and 228MHz for the RAM, with timings of 2-3-3-6 and GAT set to AAADD. I noticed that my GeILs love voltage, because they needed 3.05V to do this stable. It's a 2×512Mb Kit of Ultra Platinum PC3500. Factory specifications said voltage supported up to 2.95V, so I though 3.05 won't be a problem, and nobody killed their modules even with 3.3V (searched for corresponding threads here).
I tried to go higher but at 290 FSB I got lots of memory errors, and at 3.10V it got even worse, so I figure the modules doesn't like voltage above 3.05.
At 285 I was mostly stable, but Prime failed on me after about an hour and I experienced occassional drop-back-to-desktop's in games. I realized that I occassionally left the VCore at 1.575V from my prev. chip. Set it to 1.55V (1.52-1.53 real!) and guess what? Prime ran over 12 hours, WOW:
The day before I started to play with memory timings and noticed that even at 285/228 2-2-2-5 it only does a few errors in memtest86. Raising voltage did not help as I said, but that was before I lowered my VCore that made the machine totally prime stable, so I gave it a try again.
Set it to 2-2-2-5 and started memtest86, but I noticed errors. Different errors, not the same that I saw yesterday with the same settings. Only 1-2 location fails but the count is increasing, and the error-bits is always the same. It was suspicious.
Then the game started: swapping modules, DIMM 2-4 -> 1-3, relaxing timings, GAT settings, etc. The most funny thing is that it ran dual primes overnight, and now I get memory errors even at lower FSBs. Of course, now Prime95 fails immediately at the settings it was stable for 12 hours before.
Tried only one stick at a time and realized that one of the modules making the errors, even at stock speed (DDR433).
Here are some memtest pics, the failing adress changes as I played with the slots and swapped the modules, but notice the error bits:
So, latest investigations: module1 does errors, module2 does not. Did I kill one of them? The overnight Prime at 3.05V stressed them so much and trying to tighten timings was the last step to make one module fail?
I haven't heard about anybody who killed his/her modules even with higher voltages... Why is this happening to me?
The modules under full load was quite hot when running at 285/228 3.05V, but I think it was still OK, about 40-45°C on the heat spreaders. And it was stable.
Any opinions would be greatly appreciated, now I'm going to test the modules in another board.... and then (I think) start the RMA process....
I upgraded my rig in the sign with another 2.4C that does 3.5-3.6 on air, a ThermalRight SP-94 (with a Zalman 92mm fan and AS5), and an Enermax Noisetaker 470W PSU.
After spending a few days with testing I got it up to an FSB of 285 with 5:4, that means 3.42GHz and 228MHz for the RAM, with timings of 2-3-3-6 and GAT set to AAADD. I noticed that my GeILs love voltage, because they needed 3.05V to do this stable. It's a 2×512Mb Kit of Ultra Platinum PC3500. Factory specifications said voltage supported up to 2.95V, so I though 3.05 won't be a problem, and nobody killed their modules even with 3.3V (searched for corresponding threads here).
I tried to go higher but at 290 FSB I got lots of memory errors, and at 3.10V it got even worse, so I figure the modules doesn't like voltage above 3.05.
At 285 I was mostly stable, but Prime failed on me after about an hour and I experienced occassional drop-back-to-desktop's in games. I realized that I occassionally left the VCore at 1.575V from my prev. chip. Set it to 1.55V (1.52-1.53 real!) and guess what? Prime ran over 12 hours, WOW:
The day before I started to play with memory timings and noticed that even at 285/228 2-2-2-5 it only does a few errors in memtest86. Raising voltage did not help as I said, but that was before I lowered my VCore that made the machine totally prime stable, so I gave it a try again.
Set it to 2-2-2-5 and started memtest86, but I noticed errors. Different errors, not the same that I saw yesterday with the same settings. Only 1-2 location fails but the count is increasing, and the error-bits is always the same. It was suspicious.
Then the game started: swapping modules, DIMM 2-4 -> 1-3, relaxing timings, GAT settings, etc. The most funny thing is that it ran dual primes overnight, and now I get memory errors even at lower FSBs. Of course, now Prime95 fails immediately at the settings it was stable for 12 hours before.
Tried only one stick at a time and realized that one of the modules making the errors, even at stock speed (DDR433).
Here are some memtest pics, the failing adress changes as I played with the slots and swapped the modules, but notice the error bits:
So, latest investigations: module1 does errors, module2 does not. Did I kill one of them? The overnight Prime at 3.05V stressed them so much and trying to tighten timings was the last step to make one module fail?
I haven't heard about anybody who killed his/her modules even with higher voltages... Why is this happening to me?
The modules under full load was quite hot when running at 285/228 3.05V, but I think it was still OK, about 40-45°C on the heat spreaders. And it was stable.
Any opinions would be greatly appreciated, now I'm going to test the modules in another board.... and then (I think) start the RMA process....
Last edited: