Opteron 170 oc'ing brick wall

Squidge · 5 Mar 2006 at 11:58

System is: Seasonic 600W, DFI Lanparty UT NF4 SLI-DR Expert, Opteron 170, TwinMOS 3200 RAM.

Goes straight in at 250Mhz, stock volts and seems stable (Can do rounds of 32M pi no problems), everything on default apart from the RAM which I've currently got on a divider of 2/3 (making 166Mhz). I've noticed the HT bus is still on x5, so this means it's running at 1.25Ghz, but hey, it's stable, so why change it?

===

Now,

At 256Mhz, it starts producing failures in the BIOS's built in MemTest86 after about 15-20 minutes. I've not attempted to boot Windows at this frequency as I'm thinking it'll most likely only trash the registry or something daft anyway.

I've tried reducing the HT multiplier from 5 to 3 and changed the voltage slowly upwards all the way upto 1.5V (I don't want to go above default+15%), but it doesn't seem to make any difference whatsoever?

Is there something I'm missing here? Normally I'd expect a bit of extra voltage to at least stabilise a slight extra overclock, but it seems that the extra vcore makes no difference whatsoever.

I've not got the stepping as I was too eager to stick the HSF on :-/

Any ideas to get this chip to 2.6 ?

Squidge · 5 Mar 2006 at 12:52

208Mhz? How?

I assumed that 256Mhz going through the processor with the RAM divider at 2/3 would give 256/3*2 = 170Mhz, which the 3200 RAM, so DDR400, should be able to handle easily enough?

Have I assumed wrong, and the RAM speed is calculated differently than this?

RAM voltage is currently 2.68V.

Squidge · 5 Mar 2006 at 13:27

I'm sure 2/3 I stated in my original post is the 133Mhz setting, so that would be 256 * (133/200) = 170Mhz ?

Squidge · 5 Mar 2006 at 13:51

ok, ran CPU-Z and it says 170Mhz, as per expectations.

I've increased the ram voltage anyway, just in case, to 2.84V, but it still makes no difference.

So, the chip seems to be stable at stock volts @ 2.5, but unstable at 2.56, even with 1.5V through it.

Any other ideas?

Squidge · 5 Mar 2006 at 15:56

I've left it a few hours running two copies of "Prime95" at the "stable" speed (only ran memtest86 and super_pi before), and it came back with a "fatal error".

So I've now dropped *everything* back to defaults - including running the core at 2ghz etc.

Prime95 still fails!

"FATAL ERROR: Rounding was 0.5, expected less than 0.4"

erk! I take it that failing at default everything means my ram has become faulty? (I'm assuming a faulty processor would be detected a lot sooner)

For cooling the processor, I use a Artic Cooling Freezer 64 Pro. Idle temps are low 30's and and load temps seem to max out at around 45c