What voltage are you running your CPU at?

[jihe]

+0.1v adaptive offset.

 

[jihe]

Went through 17 chips and de-lidded them all to get that chip, which is a Costa Rica 3313B414.

As an intel stock holder I thank you.

 

[jihe]

1.44v on a X5650 @ 4.3 at 60C steady on air using Prime 95 blend.

p95 blend is not a good heat generator, at least small fft. Run IBT and watch temp explode.

 

[DrMrLordX]

Yup, AMD CPUs like their volts high. My father has an Athlon similar to yours (don't remember exactly, as it's in a soon to be retired rig) that unlocked to a quad core. Voltages were so high that I thought the chip was defective. It stopped being able to have four cores enabled after a point, but it's still running in tri-core mode. Voltage is close to 1.4V for something like 3.7GH

I can't speak for any of their newer products, but for 45nm process AMD stuff, 1.5-1.55v is usually the limit on air/water. I personally don't like going past 1.48-1.49 since I know I weakened by Sempron 140 doing silly SuperPi runs at around 1.55v. The cooling was overboard for that chip, but it just didn't matter. Electromigration is going to happen either way, and it did.

 

[tolis626]

p95 blend is not a good heat generator, at least small fft. Run IBT and watch temp explode.

Well, on my CPU, when Test 6 completes in Prime95 (Both in small and large FFTs), temperatures explode alright. It hits 100C and I stop the test out of fear. I think that after test 6 it changes the FFT size or something and produces more heat. Anyway, it's bad. Unless the H110 is crappier than I think or if I have put too much TIM on it, I'd say stay away from Prime95 if using Haswell. Watching temps soar when using 1.4V is expected. But I found out what's going on when testing 4.6GHz at 1.28V. That shouldn't overheat.

Same goes for AVX-enabled Linpack and LinX of course.

@Ed1

I tried RealBench and it's a stress test alright. It managed to make my semi-stable CPU crash within 10 minutes, and did that without unreal loads and volcano temperatures. It maxed out at 83C (4.7GHz/1.312V core and 4.4GHz/1.23V cache) when it made a 30 minute successful pass. Nice thing is, it represents a more realistic scenario and tests the system as a whole. I had it use up to 16GB and it used up about 75% of my memory. And thank you for suggesting my new stress test of choice Ed!

 

[JimmiG]

Adaptive, tops out at about 1.26V during normal use, >1.36V running AVX instructions.

 

[SlowSpyder]

One can say a lot of bad things about AMD CPUs, but they can handle some voltage alright! 1.35V isn't even close to too much for Piledriver overclocking.

I've tried as high as 1.7v on my FX 9370. More interested in undervolting lately, but winter isn't too far away... ()

 

[tolis626]

Adaptive, tops out at about 1.26V during normal use, >1.36V running AVX instructions.

0.1V difference? What the hell? On my machine running AVX (Or really anything that really pushes the CPU to its limits) will result in 0.02V being used. I've set VID to 1.312V adaptive and when running P95 it will use 1.332V.

I've tried as high as 1.7v on my FX 9370. More interested in undervolting lately, but winter isn't too far away... ()

Sweet! Water or LN2? And what frequency did you manage to hit?

 

[TeknoBug]

My i7 3770K clocked to 4.2GHz hits 1.068v under load due to -0.115v offset.

^ Lottery CPU

 

[SlowSpyder]

Sweet! Water or LN2? And what frequency did you manage to hit?

Under water. I think the setting was actually 1.65v but my LLC settings can operate a bit oddly. I topped out about 5.38GHz (not entirely stable, finished most benches, but definitely not a 24/7 overclock).

 

[tolis626]

Under water. I think the setting was actually 1.65v but my LLC settings can operate a bit oddly. I topped out about 5.38GHz (not entirely stable, finished most benches, but definitely not a 24/7 overclock).

Really? I thought I'd hear some crazy number to be honest. Not that 5.4GHz is anything to laugh at.

And I think you'd be crazy to call anything that needs 1.7V a 24/7 overclock.

 

[SlowSpyder]

Really? I thought I'd hear some crazy number to be honest. Not that 5.4GHz is anything to laugh at.

And I think you'd be crazy to call anything that needs 1.7V a 24/7 overclock.

With my old board and old power supply (both inferior to what I have now) I could get into Windows at 5.5Ghz if I disabled modules. But, with everything enabled it seems once I get anywhere above ~1.6v (and well into the 5GHz+ range) adding more voltage just adds heat faster than my cooling setup can remove it. At those settings I can get just the CPU alone to gobble up nearly 400 watts. :awe:

But for now I'm pretty happy at 4.5GHz / 1.275v. Sips power (relatively speaking), runs very cool, and is stable in everything I've thrown at it. But it is pretty cool where I live, think I'm going back to 1.5v+ soon.

 

[Server2008Sucks]

From voting choice, none of the above. I run at 0.625V at 800mhz full time. Idle or load, it's set to 0.625V, same for the on-chip graphics voltage.

Haswell i3 Core...



What does voltage offset for CPU do in BIOS? I have no idea what da heck those are for. I left it default.

 

[Ed1]

My i7 3770K clocked to 4.2GHz hits 1.068v under load due to -0.115v offset.

^ Lottery CPU

So your auto voltage would of been 1.183v , what LLC do you run , i take its not auto or lowest setting .

 

[Schmide]

Set to 1.21 in bios. Reads 1.32 during burn.

 

[Dufus]

0.1V difference? What the hell? On my machine running AVX (Or really anything that really pushes the CPU to its limits) will result in 0.02V being used.

FWIW I also only see ~0.02V difference when running Linpack AVX2 on an i7-4700MQ, 172GFlops at 20000 size. ~3.5GHz 4 cores ~1.000V adaptive. Biggest problem is cooling as Linpack pushes the power at over 1.6x TDP.

 

[TeknoBug]

So your auto voltage would of been 1.183v , what LLC do you run , i take its not auto or lowest setting .

LLC 4 IIRC (1 being highest, 5 being lowest in BIOS scale), setting to LLC 5 has weird behavior.

 

[Ed1]

LLC 4 IIRC (1 being highest, 5 being lowest in BIOS scale), setting to LLC 5 has weird behavior.

Ok, I have mine LLC one notch higher for 4.3 , but i only run -0.010 offset for load of 1.135v approx .
For 4.2 I can leave on auto .

Honestly, I have not tried to go lower .

 

[tolis626]

FWIW I also only see ~0.02V difference when running Linpack AVX2 on an i7-4700MQ, 172GFlops at 20000 size. ~3.5GHz 4 cores ~1.000V adaptive. Biggest problem is cooling as Linpack pushes the power at over 1.6x TDP.

I have no idea what the power consumption is on my system. I ordered a kill-a-watt a few days ago and it should be here soon (Maybe even tomorrow), so I will see how AVX loads affect power consumption. I hope I don't get any suprises.

With that said, the power draw with my settings must be quite high. I mean, the H110 barely manages to keep the damn thing "cool" (It WILL hit 80C+ under heavy loads no matter what). Haswell does ask for water...

 

[RU482]

120V

 

[Dufus]

Haswell does ask for water...

Maybe, but not much of on option on a laptop thankfully Linpack is not usually real world usage.

 

[groovieknave]

Just got my 5820k stable at 4.5ghz 1.275vcore, but that was running prime95. It does heat up to 80c on my h110 and I have to set it the fans on high... but it cools down from there
Never goes over 80c and thats after an hour of running prime95...

I am gonna try that realbench you guys were talking about...

 

[tolis626]

Just got my 5820k stable at 4.5ghz 1.275vcore, but that was running prime95. It does heat up to 80c on my h110 and I have to set it the fans on high... but it cools down from there
Never goes over 80c and thats after an hour of running prime95...

I am gonna try that realbench you guys were talking about...

Hmm... Then there must be something wrong with my setup. The instant Prime95 changes the test (Around the 15-20 minute mark both in small and large FFT mode) my temps will soar to 100C and stay there. Maybe I've applied too much TIM? Gonna reseat the cooler and see what happens, although I think this may be a TIM-vs-solder type of situation.

 

[SpecChum]

1.21v for 4690k at 4.6ghz.

Tried 4.8 at 1.33 but no go, didn't really want to go higher.

 

[country2]

1.366v at 4.6 ghz on 2500k