*** Official AMD Phenom II X6 1055T & 1090T Overclocking Thread ***

[HeX]

eeerrrrrrr ok what the blooming heck .. lol

a NB and a motherboard NB ???? thats got me confused but thats not hard .. so is the motherboard NB a real sensor and what exactly is it measuring ? .. the stuff under the heatpipe (SB to power circuits) or something in or around the cpu socket

is it something that needs to be kept under 62C or has it got a higher threshold ... cant say I have noticed any performance issues when it gets toasty but my cpu temp is always under 50C even when maxing out doing some encoding

adding extra case fans or to be more precise more intake fans did nothing to change the TMPIN2 temp so dont think air flow influences it

CPU-NB is the IMC on the actual CPU.

NB is the chip on the mobo near the CPU socket with the big heatsink on it.

 

[gareth170]

I'm surprised you have to use so much CPU-NB voltage to get 1600 stable on your CPU.

I've got 8GB stable at 1600 C9, CPU-NB @ 2600MHz - 1.125V.

That's with the CPU at 3.9Ghz.

but your ram is CL9 where mine are CL7 so i guess thats why...

i can get 2.6ghz stable with 1.225-1.25v, but tbh theres no difference in temp from stock to 2.8ghz 1.325v

also my stock CPU-NB voltage is 1.15v..

 

[HeX]

Maybe it's 1.175V then I'm using, one or the other

Eitherway it's less than 1.2V

 

[gareth170]

eeerrrrrrr ok what the blooming heck .. lol

a NB and a motherboard NB ???? thats got me confused but thats not hard .. so is the motherboard NB a real sensor and what exactly is it measuring ? .. the stuff under the heatpipe (SB to power circuits) or something in or around the cpu socket

is it something that needs to be kept under 62C or has it got a higher threshold ... cant say I have noticed any performance issues when it gets toasty but my cpu temp is always under 50C even when maxing out doing some encoding

adding extra case fans or to be more precise more intake fans did nothing to change the TMPIN2 temp so dont think air flow influences it

im 99% sure it as a real temp sensor.

 

[Cozman]

if thats the TMPIN2 then I wonder what a safe operating temp is

I basically have a side case fan almost blowing directly at that area but it didnt impact on temps apart from lowering TMPIN1 by a good 5C or 6C

 

[gareth170]

TMPIN1 is motherboard...

 

[Cozman]

TMPIN1 is motherboard...

which I assume has a sensor under the oblong heatspreader by the power circuitry .. now your gonna say ... nope the small square heatsink over the SB .. lol

still debating if a fractual case with its 140mm fans top & bottom will give me better air flow and better dust control

 

[alexp999]

which I assume has a sensor under the oblong heatspreader by the power circuitry .. now your gonna say ... nope the small square heatsink over the SB .. lol

still debating if a fractual case with its 140mm fans top & bottom will give me better air flow and better dust control

I think the 890fx might have its own sensor actually. Not 100% tho.

What board do you have?

You know the Leo platform effectively has three chipsets? The CPU Northbridge, the 890FX (Northbridge) and the SB850 (Southbridge).

Yes just to confuse things, our systems have two northbridges XD

 

[dazz]

My first tests at 20x200 for 4GHz (stock NB & RAM yet) where a bit dissapointing. Prime95's blend failed almost imediately until I set 1,53V (being CPU-Z readings between 1.424 and 1.47).

I'm not sure I can get those volts any higher anyway, as my TMPIN0 was already in the high 50s, so I guess I should aim for something like 3.9/3.8.
What concerns me is the fact that it was always core #5 the first to fail the tests with a variety of FATAL ERRORS. Could it be a damaged core?

 

[alexp999]

Afaik, most if not all CPUs have a weak core (one in particular will tend to fail first), you just have to find the right voltage to keep them all happy.

Also that is some serious vdroop you've got. Your CPU-Z readings are in the region of what most of us are getting 4 GHz stable at.

For example I need 1.4v with Load Line Calibration on, which gives about 1.44v in CPU-Z.

 

[gareth170]

My first tests at 20x200 for 4GHz (stock NB & RAM yet) where a bit dissapointing. Prime95's blend failed almost imediately until I set 1,53V (being CPU-Z readings between 1.424 and 1.47).

I'm not sure I can get those volts any higher anyway, as my TMPIN0 was already in the high 50s, so I guess I should aim for something like 3.9/3.8.
What concerns me is the fact that it was always core #5 the first to fail the tests with a variety of FATAL ERRORS. Could it be a damaged core?

a FATAL ERRORS could be ram / cpu-NB issues,, BSOD uselly is core clock/vcore issues.

you've got 8gb of ram so u may need to up the cpu-nb voltage alittle

 

[dazz]

Makes complete sense

Afaik, most if not all CPUs have a weak core (one in particular will tend to fail first

Thanks for the tip, I'll do exactly that

a FATAL ERRORS could be ram / cpu-NB issues,, BSOD uselly is core clock/vcore issues.

you've got 8gb of ram so u may need to up the cpu-nb voltage alittle

...I wonder if anyone has ever tried OCing say, 2, 3 or four cores? so maybe keep 3 cores @3GHz and the rest @5GHz. I guess it's not that easy since, AFAIK there's no Vcpu per core, hence you'd need to set it to the max needed

 

[HeX]

...I wonder if anyone has ever tried OCing say, 2, 3 or four cores? so maybe keep 3 cores @3GHz and the rest @5GHz. I guess it's not that easy since, AFAIK there's no Vcpu per core, hence you'd need to set it to the max needed

You can kinda do that using the AMD Overdrive software.

You can set up your own Turbo profile.

So for example you could have all 6 cores running at say 3.6GHz on low volts, then setup a Turbo profile for 4 good cores running 4GHz with a small Vcore bump.

That way in highly multi threaded stuff you'll get all 6 cores churning through, while in games that tend to only really use 2-4 threads your CPU will Turbo those threads up to 4GHz.

 

[dazz]

Nice. I'll give that Turbo Core thing a try when I get a stable balanced clock. Finding out which cores are the best candidates for faster settings shouldn't be a big deal, and given I only play FSX and it can be set for core affinity, this might help me get some extra FPS.

 

[vandiesel]

My first tests at 20x200 for 4GHz (stock NB & RAM yet) where a bit dissapointing. Prime95's blend failed almost imediately until I set 1,53V (being CPU-Z readings between 1.424 and 1.47).

I'm not sure I can get those volts any higher anyway, as my TMPIN0 was already in the high 50s, so I guess I should aim for something like 3.9/3.8.
What concerns me is the fact that it was always core #5 the first to fail the tests with a variety of FATAL ERRORS. Could it be a damaged core?

If it is the asus in your sig, then try adjusting the LLC maybe try 9 or 16%

 

[dazz]

A quick google search shed some light on Vdroop/LLC.
Guess your are implying my mobo is well known for high Vdroop values, will try compensating with LLC then. Cheers, very much appreciated

If it is the asus in your sig, then try adjusting the LLC maybe try 9 or 16%

 

[Sayso]

just ordered a 1090T yet another cpu i dont need

 

[dazz]

Ok, so I'm currently at:
4080MHz / 2640MHz
1600 CL9

Disabled cores #5 & #6 and been blend priming for hours flawlessly.

So I guess it confirms I have a memory bottleneck here, right? because looks like six full cores blending clogged my RAM bandwidth or something, hence prime95 failed.

Might as well leave it like this till I get me a WC kit and try again, maybe 4.1GHz.

And sorry to be such a pest, but can I please also ask where can I get a reading of the actual Vnb (it's nowhere to be seen in HWmonitor) and if LLC also applies to it?

Thanks

 

[gareth170]

Ok, so I'm currently at:
4080MHz / 2640MHz
1600 CL9

Disabled cores #5 & #6 and been blend priming for hours flawlessly.

So I guess it confirms I have a memory bottleneck here, right? because looks like six full cores blending clogged my RAM bandwidth or something, hence prime95 failed.

Might as well leave it like this till I get me a WC kit and try again, maybe 4.1GHz.

And sorry to be such a pest, but can I please also ask where can I get a reading of the actual Vnb (it's nowhere to be seen in HWmonitor) and if LLC also applies to it?

Thanks

mate try upping the cpu-nb voltage to 1.25-1.3v with all cores enabled.

are u upping the FSB or the multipliers?

 

[dazz]

Both Gareth: 225x18

The reason I disabled those cores is that I was getting what I think could be too high temps (Vcores ~55 / Vcpu~65). I've gone through every single post in this thread and I know some of you are ok with Vcore below 62, while some others rather keep Vcpu as a reference.

I'm just about to test all six cores after upping the Vcpu-nb to 1.3