920 overclock problem

Naff8000 · 19 Nov 2009 at 05:17

I've stabilised at 3.8ghz, or at least it seems so after 30 min tests on small and blend. Yet now it's often resetting back to stock when I turn on the pc, it begins to post then resets itself and it's back at stock when I get in windows. When I go in the bios it gives me the warning sign that the overclock has failed.

The same happens at 3.6ghz, although it seems ok at 3.4. Anyone know what this might be?

CPU Clock ratio: 20x
QPI Link speed: x36
BLCK: 190mhz
Vcore: 1.3
QPI/Vtt Voltage: 1.355
DRAM: 1.64V

BigBruiserAl · 19 Nov 2009 at 09:02

That will be because the overclock is not stable

would say at that clock your QPI is to high try 1.28 (ish) for 3.8

Naff8000 · 19 Nov 2009 at 15:53

Well anything less than 1.355 and it doesn't last 7 minutes in stress tests

Combat Fighter · 19 Nov 2009 at 18:06

What sort of cooling are you using?

What are your load temps?

DragonQ · 19 Nov 2009 at 18:10

1.355 seems high for 3.6-3.8GHz, are you sure you tried voltages in the 1.25V range before going that high? If you started at around 1.3V then try going lower.

BigBruiserAl · 19 Nov 2009 at 18:48

whats your ram div and uncore

Nocky24 · 19 Nov 2009 at 18:54

Naff8000 said:
Well anything less than 1.355 and it doesn't last 7 minutes in stress tests

Mines just below that at 4Ghz and is entirely stable, tested in prime. Whats your uncore multi set to? Also what are your ram timings? And whats your ram multi set to? Also i find leaving the cpu multi at x21 more stable than at x20.

Naff8000 · 19 Nov 2009 at 21:27

My uncore multi is x20. I've never touched ram timings before. I spent 3 hours on this the other day, the pc isn't interested unless QPI is 1.355. At 1.25 it doesn't post, whatever the vcore (tried up to 1.425) At 1.3 QPI it didn't last 20 seconds in prime 95, 1.325 about 3 mins. Progressively on to 1.355 where it lasted the full 30 mins.

Naff8000 · 19 Nov 2009 at 21:30

Using a TRUE 120 cooler. Load temps after 30 mins we're 68c

Naff8000 · 19 Nov 2009 at 21:35

DRAM Timing selectable set to auto
mem multiplier is 8.0

BigBruiserAl · 20 Nov 2009 at 08:59

try an uncore of 17 mate and qpi at 1.285 (ish) vcore 1.3 for 3.8ghz see whats happens again.

Take the mem off auto and put the timeings as directed on your stick

Nocky24 · 20 Nov 2009 at 11:48

Try vcore and qpi volts just over 1.3 (maybe 1 or 2 steps over) then have your ram multi at x8 and your uncore multi at x16 (double your ram multi) and as i've said try the cpu multi at max x21 and adjust the bclk to around the 3.6Ghz-3.8Ghz mark. I've used the same settings for 4Ghz and i have a similar setup, but just try that for now as not all chips overclock exactly the same. Obviously make sure your cooling is good before attempting any of this and i'd recommending running prime for a while at stock settings first to see if it will cope with the overclock.

Naff8000 · 20 Nov 2009 at 14:37

Ah thanks. The auto had the timings set to 7-7-7-16. I'm doing a small fft test right now with vcore: 1.3 and QPI: 1.295 and uncore at 16x @ 3.8ghz. Everything seems fine after 8 mins so far with a high temp of 68c.

Nocky24 · 20 Nov 2009 at 14:56

Naff8000 said:
Ah thanks. The auto had the timings set to 7-7-7-16. I'm doing a small fft test right now with vcore: 1.3 and QPI: 1.295 and uncore at 16x @ 3.8ghz. Everything seems fine after 8 mins so far with a high temp of 68c.

Yeh that sounds about right, good stuff

BigBruiserAl · 20 Nov 2009 at 15:22

BigBruiserAl said:
Take the mem off auto and put the timeings as directed on your stick

Naff8000 said:
Ah thanks. The auto had the timings set to 7-7-7-16. I'm doing a small fft test right now with vcore: 1.3 and QPI: 1.295 and uncore at 16x @ 3.8ghz. Everything seems fine after 8 mins so far with a high temp of 68c.

I did the same thing mate learn from experince by the way the jump from settings 3.8ghz is big to 4.0ghz if you go for it. Let me know if you get stuck

Dont forget to run blend test