[PolRoger]

Thought I would start one up here for the OCForums.

I picked up a 2600K batch #L039B467. I'm still trying to sort out the new BIOS o.c. and voltage settings. I have my chip running with defaults(?) HT stable at 4.4GHz which seems to have vcore go up to ~ 1.3v load. It is kind of weird I haven't been able to find my board's vtt voltage setting yet... (Biostar TP67XE). I think I may have to bump vcore to get 4.5GHz+ but it is still early on with the testing of this chip. Cooling is a TRUE Black with push/pull.

[Castiel]

I'm hoping this thread will help me decide what board to order

[MattNo5ss]

Make sure you install the BIOS with the internal PLL overvoltage fix, seems to really help in the OCing department

I've been thinking about getting the TP67XE...

[PolRoger]

Yeah, I'm sure I'll test out a few SB boards before socket 2011? comes out but I like Biostar for a lower cost board. I hope it clocks well but in the end it mostly comes down to the chip.

[Castiel]

I've kinda narrowed it down to the UD4 and the TP64XE

[retrogreq]

I'm looking at either the UD3P or P8P67. The gigabyte board has looks on it's side (for once).

[PolRoger]

Originally Posted by Castiel

I've kinda narrowed it down to the UD4 and the TP64XE

The UD4 or the UD3P may be a better choice to start with because there are going to be many more users out there with the the different Gigabyte models and it will be easier to compare and get settings on the various forums for Gigabyte boards. Biostar has a disadvantage when trying to find others to post and share BIOS settings etc. They also don't seem to update their BIOS as frequently as some of the bigger outfits.

[dejo]

got my p8p67pro and 2600k to 5100, but have to give a bump of .3 volts to get it stable. if your temps get to 90 the chip will throttle the multi to keep the temps in that range. I have a cheap cooler on mine for now as I am just testing. and for those questioning it is an l039b214 batch

[PolRoger]

Dynamic VID...

I wanted to post up some shots of my chip running "spec voltage" showing how vcore changes when increasing the overclock. Powersavings/Cstate/EIST disable in BIOS while all other voltages set to default spec. LLC disabled.

Vcore/idle rises from 1.332v to 1.356v... if you disable speedstep/turbo my chips stock 3.4GHz idle vcore is ~1.260v. I think 4.4Ghz might be this particular chip's max stable o.c at "spec" voltage.

4.0 load:


4.0 idle:


4.2 load:


4.2 idle:


4.4 load:


4.4 idle:

[deathman20]

The VID changes because of load changes in the system. It happens all the time even if you have LLC enabled on some boards. Heck my board will start low and actually up the voltage with LLC enabled.

Its nothing new with SB.

[PolRoger]

This combo (voltages) is not behaving like any of my bloomfield rigs LLC enable or disabled it is doing something different.

[deathman20]

Behaving just like my CPU does with mine LLC disabled. Enabled its the reverse since I get more juice going to the CPU under load and idle voltage is lower.

[PolRoger]

It may be something tied to this particular board's BIOS vcore settings for "spec voltage" it also has "offset" and "fixed". With my Bloomfield rigs and the powers savings features disable (C1E/EIST) once you set the vcore in BIOS to one setting idle vcore doesn't go up or down when you then change the multi or bclk. Only load vcore goes down with vdroop enabled and sometimes up with LLC enabled.

The vcore setting was the same for all three clocks but idle vcore in cpuz was different for the three SB turbo multis (40,42,44).

I'll know more with further testing and perhaps when I can get a chance to try out a different motherboard.

[wingman99]

Why would you try a different board it won't help with overclocking and you defiantly won't save any money doing that.

The CPU does it all for overclocking. If the mother board works that's all you need for sandybridge.

[burebista]

Guys, sorry to intrude by I have to ask for a little help.
Here is latest RealTemp with multi fixed (I hope) and a brand new option of showing Vcore (at last!)

Originally Posted by Kevin

I added Sandy Bridge volt monitoring to RealTemp. Yea!!!!!!!!
There should be a new option in the Settings window for this.

Some feedback/screenshots it would be nice for Kevin.

Thanks in advance guys.

[PolRoger]

Originally Posted by burebista

Guys, sorry to intrude by I have to ask for a little help.
Here is latest RealTemp with multi fixed (I hope) and a brand new option of showing Vcore (at last!)


Some feedback/screenshots it would be nice for Kevin.

Thanks in advance guys.

Thanks for the update! Dynamic VID seems to be reporting correctly. On idle the mutli in RealTemp bounces around quickly 22/24.5/41.5/27/28.5 etc.

Load:

Idle:

[deathman20]

Originally Posted by burebista

Guys, sorry to intrude by I have to ask for a little help.
Here is latest RealTemp with multi fixed (I hope) and a brand new option of showing Vcore (at last!)


Some feedback/screenshots it would be nice for Kevin.

Thanks in advance guys.

Sweet new CPU-Z!

[burebista]

Only for Sandy Bridge is that Vcore.
Thanks to Intel because at last they write Vcore in MSR so no need to hunt countless of chipsets/sensors.

PolRoger thanks A LOT for your screens.
Bouncing multi in idle is a sign about Windows Power scheme fighting against BIOS settings. How do you set up Minimum processor state in Power options? If it's default 5% put it at 100% and you should be fine.

[unclewebb]

PolRoger: Thanks for posting your tests.

MSR 0x198 used to contain the VID voltage in the Core 2 design. Intel got rid of that information in this register when they switched to Nehalem Core i7. With Sandy Bridge, this information has been returned to this register.

Edit: My best guess it that the number that RealTemp and Core Temp reports is a VID request voltage and the voltage reported by CPU-Z is actual core voltage.

Here's what the Intel docs say.



RealTemp uses the Intel approved method as outlined in their November 2008 Turbo White Paper and accurately tracks the average multiplier. With the previous processors, CPU-Z decided to report a consistent multiplier at idle for validation purposes. This didn't always accurately report what the CPU was doing internally but I can understand his reason for doing this. I'm not sure if Intel encouraged this decision or not or how he has decided to report Sandy Bridge.

Anyway, RealTemp tells it like it is. If it shows a wandering multiplier at idle it's because various C-States are being used. I'm not sure if it is possible to get a nice steady multiplier in Sandy Bridge when the CPU is idle. Turbo boost depends on individual cores being able to enter the deeper sleep states like C3/C6. If you disable all of the necessary C-States, it might prevent turbo boost from working properly. Sandy Bridge CPUs depend on turbo boost to obtain multipliers like 45.

The reason RealTemp 3.60 was showing a zero multiplier is because you likely have one or more monitoring programs on your computer that are misusing the shared system performance timers. AIDA64 / Everest was causing problems by resetting and stopping and starting the shared timers so with RealTemp 3.65, I switched to a different set of timers that hopefully are not being abused like this so the reported data should be very accurate. This is also very likely the same reason why Core Temp is not reporting the multiplier correctly. Other monitoring programs are not treating the shared timers within the CPU as a shared resource.

Does RealTemp 3.65 allow you to overclock and adjust your multiplier from within Windows? You need to turn on at least one temperature icon and then have a look at the system tray menu that pops up. You should be able to find an option for this. I'm just checking to see if it works.

[Techboy10]

Questions for PolRoger.

In your first screenshot post (#9) you said you disabled Powersavings/Cstate/EIST in the BIOS. Doesn't disabling EIST also disable Turbo, and therefor overclocking?

Also, how did you get your overclocked CPU to idle @ 4.0/4.2/4.4 GHz etc. because (unless my understanding of SB overclocking is wrong) in order to overclock you need EIST and Turbo enabled, which would throttle back the CPU (to around 1.6 GHz) when idle right?