2500K vcore/vid anomaly MB or advanced power management?

[Kenmitch]

I was playing around with my 2500K overclocking this AM as I'm off for the weekend. Figured I see how high she'd go and was up to 5.1ghz doing some super pi runs. Pretty quick and vcore was showing in the 1.45v(cpu-z) or so range temp was laughable in the 50's as windows power management was doing the load shifting as it's a single thread benchmark

I was thinking that's not bad untill I fired up Intel Burn Test 2.50 and hit start the vcore jumped up to around 1.50v(cpu-z) and temps spiked at 70*C I hit stop as soon as I saw the vcore. I fired up Intel Burn Test 2.51 and did the same thing again. The vcore jumped up to around 1.55v and the temps spiked just below 80*C

Went back to my conservative 4.6ghz overclock to do some more in depth testing to see if it was just my bios tweaking to get to 5.1ghz that created this anomaly. Well lets just say it wasn't for now as it was repeatable 100% as far as voltages and temps go at least.

What I'm thinking is it's the inner workings of Sandy Bridge's power management features doing what it does best. Seems like the VID is not set to a fixed amount based on speed or load only. To me it looks like it has some kinda AI and takes feature's being used to set the vcore at the same time.

Here are some examples of my testing done at 4.6ghz the vcore will be the average and the temps will be the hottest core.

Super PI 1mb vcore 1.30v temp 46*C

Intel Burn Test 2.50 vcore 1.36v temp 62*C

Intel Burn Test 2.51 vcore 1.38v temp 68*C

I figure Super PI being so old it uses old school parts of cpu's. Intel Burn Test is pretty old also and version 2.51 seems to maybe use a newer feature set than version 2.50 as the gflops are close to double the previouse version.

Kinda makes me wonder about peoples stated vcore/temps at certain overclocks.

Anybody bored and wanna see if it's repeatable feel free to post results

 

[smakme7757]

If you have Load Line Calibration enabled that will cause your motherboard to provide extra voltage under extreme load, such as Intel Burn Test. Super PI is a single threaded application so you would not be causing enough load for the motherboard to react and apply more voltage.

I would bet that Intel Burn Test 2.51 uses the extended AVX instructionset in its Linpack library causing it to 'burn' your CPU a lot more than the previous version without support for AVX.

 

[Kenmitch]

If you have Load Line Calibration enabled that will cause your motherboard to provide extra voltage under extreme load, such as Intel Burn Test. Super PI is a single threaded application so you would not be causing enough load for the motherboard to react and apply more voltage.

I would bet that Intel Burn Test 2.51 uses the extended AVX instructionset in its Linpack library causing it to 'burn' your CPU a lot more than the previous version without support for AVX.

Could be load line calibration as it was on during the quick 5.1ghz runs. Think it's set to normal at 4.6ghz settings tho. I'll have to check and see.

Maybe I'll set the 2500K to stock settings later today and disable load line calibration in bios and see what it does....Never ran it at stock speeds yet anyways for a base comparison for performance gains anyways