overclocking an i7 920 - cpu VTT jumps .04v

[BTRY B 529th FA BN]

I've recently decided to turn HT on with my i7 920. I'm currently testing 201 BCLK/4Ghz with a 2:8 memory ratio/1610MHz. I've got a few things dialed in that I'm confident with except one thing. My CPU VTT reads 1.31 via CPU ID , but what concerns me is it jumps to 1.35v while running LinX and Prime 95. I stabilized my Vcore by raising:

VTT PWM Frequency, from 250 to 370 KHz
CPU PWM Frequency, from 800 to 940 KHz

where before my Vcore would jump from 1.35v to 1.38v while running LinX

Any ideas on what's going on with my CPU VTT ???
currently @ +50mV (on auto it reads 1.250v)

Thanks for your time

 

[Don Karnage]

That vtt is Alittle high but it should be fine.

 

[BTRY B 529th FA BN]

That vtt is Alittle high but it should be fine.

My concern is when it jumps .40v

EDIT: i see you're with an i7 3930K, why the change?

 

[Don Karnage]

My concern is when it jumps .40v

EDIT: i see you're with an i7 3930K, why the change?

Was 197 through intels retail edge

 

[tigersty1e]

you mean it jumps 0.04 or 0.4?

.04 is nothing.

are you using software to read or actual multimeter because software is not accurate at all.

 

[BTRY B 529th FA BN]

ok, thanks

 

[Don Karnage]

ok, thanks

Push it bro.

 

[BTRY B 529th FA BN]

the default vcore is 1.23v it will do 211 BCLK or 4.2Ghz but around 1.44v, aaaaaaaaaaaaaah, scary!

 

[coffeejunkee]

My CPU VTT reads 1.31 via CPU ID

Do I understand correctly you mean vcore instead of cpu vtt and cpu-z instead of cpu id?

If so, use loadline calibration to eliminate vdroop.

Also, if you haven't, you really should overclock Bloomfield cpu's in a 2-step fashion. First lower multiplier and find required QPI/VTT for a certain Bclk. Then raise multiplier and increase vcore as needed.

 

[BTRY B 529th FA BN]

My reference to my vcore was that I adjusted the PWM Frequency with the VTT & CPU adjustments to stabilize my vcore spiking. Before I evem did so I did have the vcore set to 'without vdroop' or aka as 'LLC enabled' and the vcore still shot up. 03v. What made it stabilized was the VTT & CPU pwm frequency adjustments. So you see adjusting something else fixed the issue with the vcore spikes so maybe something else might help stabilize the CPU VTT ... :hmm: ... and yes, I was using CPU-ID, not CPU-Z. I've ran CPU-Z & CPU-ID together and I think CPU-Z is more accurate, where-as CPU-ID will just round off the voltage reading.



EDIT: sorry, tried to edit this during work and got distracted.

 

[coffeejunkee]

Ok, I'm confused. CPUID is the company behind CPU-Z. Searching for CPU-ID I could only find a program called that for Linux. But CPU-Z is Windows only afaik.

So you either run a dual boot Windows/Linux system, found a CPU-Z version for Linux, or use a program called CPU-ID for Windows which I can't find.

 

[BTRY B 529th FA BN]

My bad. hardware monitor, by CPU id

 

[coffeejunkee]

Ah ok, I see now, CPUID Hardware Monitor is what it says in the title bar indeed. I usually say HWmon though.

But as I understand CPU VTT voltage doesn't really need to be changed for anything but the highest overclocks. Most boards don't even offer the option to change this. So I would leave it at auto and focus on QPI/VTT and Vcore. Maybe CPU PLL if all else fails.

I wouldn't even worry about the small spikes in Vcore using LLC only. It's still in very safe ranges for i7 920. But must say I have seen tighter Vdroop control, although it's debatable if that's actually healthy for your cpu.

 

[BTRY B 529th FA BN]

Apparently 1.55v is the max CPU vcore for that chip, they just didn't sell those chips with a default vcore of 1.35v or greater.

I was surprised to see the CPU vcore stable out when I adjusted those PWM Frequency settings. My vcore will now stay within .01v e.g. 1.35v - 1.36v which is better than a spike before of 1.35v - 1.38v, while running LinX

The general rule of thumb is to keep the CPU VTT within .50v of the RAM, from what I've read. e.g. if your RAM's recommended voltages are 1.65v then the lowest CPU VTT should be 1.15v. I think what EVGA did was default the CPU VTT to 1.20 to give a buffer for any drops in CPU VTT ...

if i'm wrong again, lmk

Thanks again for everyones help