Q9650 Overclockig help

PaulyD

PaulyD

PaulyD

Soldato
Joined
24 Mar 2008
Posts
4,654
Location
High Wycombe
Right guys need some help here!

I previously ran a 4hour small FFT test @ 1.4v and that came back fine, last night finally got round to tweaking the 4.41Ghz overclock and got this down to 1.376 under load. Small FFT's passed first self test (not a long test but a quick dirty test to just check that it was running ok as failed at lower and knew 1.4 was ok) so decided to try blend test, herein lies the problem...

Running blend test it would get to about test 6/7 on first self test and the PC would completely lock up, no mouse movement, numlock wont turn on/off, just frozen and require hard reboot. Tweaked the volts up slightly but got the same thing. I wondered if its just prime so ran a few tests of IBT and this also highlighted some more increases in volts needed.

After much playing around I got the CPU to pass a 15 run test of IBT under maximum stress, was pleased with that as its a brutal test and ran for over 20 mins. decided to run Prime blend test to see if same issue came about - which it did :confused: Thought it might just be the blend test having a problem as no error message in event viewer, no BSOD just a hard lockup.

From previous experience I have also found GTA IV a good test for stability (had numerous problems with it but the last freezes i had in it were from my overclock needing slightly more volts even though being prime stable!) Ran GTA IV and froze within a few seconds of playing. restarted PC after nudging volts up by one and after about 10-15 mins got the same kind of lockup on desktop just browsing the net..

Just a bit baffled now as CPU has been proven stable from small FFT tests (not full 8 hours obviously) and has passed 15 runs of IBT under maximum stress, but can't pass 10 mins of blend testing and will lockup randomly on desktop under very little load.

Settings from Memory:

CPU: 1.375 bios, 1.366 idle, 1.376 load with LLC enabled -> raised to 1.3875 for 1.392 load (load actually goes up with LLC )
CPU PLL: started on 1.55 risen to 1.60
FSBT: ranged from 1.36 to 1.43 when testing last night
NB: 1.57 through to 1.65
No GTL refs or clock skews set

I have 4gb of DDR3 G.Skills 1600Mhz HZ RAM, which is on 333 strap underclocked to 1569Mhz on TRD 8 and Stronger performance level.

So I don't think its RAM either as not even overclocking it at the moment, but from my testing it only sounds like a RAM issue...

Just ran Memtest from bootable USB, passed 1 test 100% at the following:
cpu, 1.375, cpu pll 1.577, nb 1.55, fsbt 1.36, nb dram 1.90

and then 6% into next test..

Any idea's guys or things to try? I'm happy at 4.275 but 4.4Ghz for a 24/7 is too tempting to not go for..
 
I found that they gave me stability at 4.275 when i set them but no combination of settings are good with 4.41ghz, will give them a try again later
 
Holy thread revival batman!

Got 4.41Ghz stable as a rock, needed to adjust CPU and NB skews to 200ps and 100ps respectively to achieve stability.

However upping the FSB to 500 and attempting to boot gives me DET DRAM on the LCD Poster, I have tried all combinations of TRD, perf level, voltages and skews but can't get it to boot at all.

Any idea's?
 
On average get 30 to 38c (depends on heating!) idle and between 58-65 load with variance on cores.

Thats with CPU, NB and 4870x2 in single loop on PA120.3 :D

Soon going to mod my case to take a double in roof for GPU :)
 
Definately, they really do aid stability when used correctly. I couldn't get 4.275ghz stable without them
 
stolen from another forum (cant remember which)
but gtl ref voltages on 45nm quads are generally stable at:

-0.032 x FSB temination voltage

i.e if your fsb voltage is 1.3V set ALL FOUR gtl refs to -0.0416V (41.6mV or as close as you can get) note this method will always give you negative gtl refs.
Let me reduce my vcore for the by about 0.06V :)
 
You must log in or register to reply here.
Back
Top Bottom