Opteron 165: 2.55 = stable, 2.60 not; other weirdness

[swaaye]

I've been spending the last month or so tweaking this rig. I've determined that it is Prime stable for at least 12 hours (never errored) at 1.45v (def + 110%) 2.55GHz (283*9). However, in my search to attain 2.60 GHz as a nice rounded number, I've found that it is not stable with this extra 50 MHz even with 1.50v (+113%). Core 0, the whimp of the two, will fail Prime within 15 minutes at 1.45v and within an hour at 1.50v.

 

I'm running naked with AS5 between the core and my Freezer 64 Pro. Core runs ~45C at most, but it's hard to tell cuz the mobo sensor sucks in accuracy to say the least. It sways between a difference of 7C at times. LOL. I've seen it running 38C while Primeing only to be 45C a moment later. I had to set the BIOS to keep the fan at full blast if the core is 34C or it will sit there revving up and down. I could tell the nakedness helped by this new behavior.

 

BTW, how stupid was it for DFI not to put some hysteresis in their mobo fan controller? Instant response to temperature changes is just plain dumb. I yearn for a quality temperature-based controller.

 

I've dropped the HTT multiplier to 2X, run the RAM far underclocked, loosened RAM timings, tweaked chipset voltages, etc. It just seems like it's at some sort of brick wall.

 

I also tried briefly to lower the CPU multi to 8X (8.5X won't boot) and up the bus to 325 MHz, with a corresponding reduction in memory and HTT multiplier. But this caused Windows to freeze as its loading into the GUI; vid card (PCIe) prob?

 

Which reminds me, I've been having an odd bootup screen corruption issue with my X800GTO2 even at 278 FSB. Periodically the vid card will literally show half of a Windows login screen (weird! where does it get that from after a power cycle!) during POST and it requires a Ctrl-Alt-Del to get going. Once it's in Windows it's perfectly content. But I do wonder if this is related to BIOS settings somehow, and whether its limiting my vid card core overclock (which will lock up if above 525 or so, with no screen corruption, just a freeze). I recently dropped the LDT and chipset voltages to default and, while I'm not certain at all, I think it may have helped?!

[red930]

For your Ultra-D, which Bios version are you running? These Opteron's don't need 1.6v or 1.7v as you have mentioned. They can overclock high with even a slight voltage bump. Maybe you have one that doesn't want to overclock that high, i don't know. Sounds like you've done all the things i would try with that Processor, so! I just now noticed the watts on your PSU, which could be one of the reason why you're having problems overclocking that high. Problems getting to login screen with corruption ect. Have you tried testing with a better PSU yet? What's the spec on that PSU you running anyhow? Why not, i would of done that already, because it's worth a try. Thankfully, i own a OCZ 520 Modstream which has been working flawless for my systems overclock plus video card as well. Please do verify if my suggestions work for you or not, good luck!

[swaaye]

BIOS = NF4LDB14 (official latest, supporing FX60). I am baffled by the whole custom bios thing so I haven't messed with that. Any recommendations?

 

About voltages, what I was trying to say was that 1.45v gets me 2.55GHz while 1.50 will not stabilize 2.60GHz.

 

I have been wondering about the PSU myself. But it's not very old and an Enermax 475W isn't much to sneeze at. It seems to be doing a fine job at 2.55GHz with games. http://www.silentpcreview.com/article149-page1.html.

 

BTW, video corruption doesn't occur in Windows. It's as the puter is turning on at POST, before the hard drive even kicks in. It's the strangest thing to see 1/3 of a login in screen before Windows even begins loading. I don't know where the card gets the image! LOL