Prime95 + CoreTemp = how much voltage is drawn?

[Thermodynamic]

I've been tweaking my new system (Gigabyte GA-Z68A-D3H-B3), and applied a new cooling fan (Arctic Freezer XTREME Rev2). Corsair Vengeance (4x4GB)

I found that, with XMP enabled for RAM, the CPU vCore then goes up to 1.35v in the BIOS. With XMP disabled and the RAM frequency set to 1600 manually, the vCore (BIOS) stays at 1.23v - where I want it.

In Windows, idle, the CPU draws 0.97v and runs at 1.6GHz. Under slight load this goes up to 3.4GHz and draws 1.28~1.3v. But with Prime95 running - the voltage goes up to 1.36v (the CPU reading 3.8GHz for Turbo mode, all 4 cores).

Idle temp is 29C. Full load temp gets to 64C, which is great but the amount of voltage drawn suggests - if it can be lowered and the system remains stable - that the load temp would go down as well.

LLC is set to "auto".

Changing LLC to "disabled" using the overclocking app in Windows had no effect, but I suspect I'd have to reboot to confirm the settings...

CPU Overcurrent protection is enabled...

Real time changes in OS is disabled, so I'm definitely confused as to why it's changing the voltage requirement...

Should the voltage spike up to 1.36v under load? I know that particular voltage is needed when overclocking to 4.5GHz or higher, but for the moment I need to establish a baseline using the CPU's own defaults (1.2v, 3.4GHz normal, 3.8GHz turbo).

CPU PLL remains at "auto", but I'm not sure that's relevant.

What other settings will reduce the amount of voltage applied to the CPU under heavy load? That will help run the system cooler as well as allowing greater overclocking, under voltages I know won't damage the CPU (1.38v is the max).

Thanks much!

 

[RGE]

auto vcore always over volts when overclocking, and the newer cpus with higher bins of turbo are in essence overclocking, though its factory overclocking. It is bios (via specs) that is ramping up vcore with higher loads.

I would first disable LLC in bios and reboot and recheck prime.

If that doesnt resolve it, then I would set the vcore manually, ie get it off auto. Looks like on some bioses/mobos vcore really goes up when turbo kicks in.

 

[killerkid]

Best way would be to stop using auto vcore. You are using way to much voltage that way.

 

[ChaosSully]

I just overclocked my cpu to 4.050ghz at 1.3v and my ram at 1.6v at 1800mhz also disable the turbo boost as it does over volt the cpu my temps with turbo were 40c idle and 70c on load.the sad part was turbo boost bumped from stock clock of 3.06 to 3.2 and really didn't do much but heagt my cpu turn it off and over clock it manually and got 3.8ghz at 67c under max load

 

[killerkid]

The turbo for all the cores should be set to the same as the cpu frequency. Having turbo enabled will NOT heat your cpu up any more as long as you have the vcore set manually.

If you are using auto vcore, without turbo obviously it will be lower because the cpu frequency is lower.

By 4.050 chaos do you mean 4.5 or really 4.050 lol. Either way chaos there is a lot more left in your chip. Every SB chip can do 4.5ghz.

 

[ChaosSully]

I dnt have a sandy bridge chip I have a i7 X5675 west 6core

 

[killerkid]

Makes more sense now, a signature might help a bit considering this thread is based on a SB chip

 

[Thermodynamic]

auto vcore always over volts when overclocking, and the newer cpus with higher bins of turbo are in essence overclocking, though its factory overclocking. It is bios (via specs) that is ramping up vcore with higher loads.

I would first disable LLC in bios and reboot and recheck prime.

If that doesnt resolve it, then I would set the vcore manually, ie get it off auto. Looks like on some bioses/mobos vcore really goes up when turbo kicks in.

I did some reading last night - while the manual talks of "CPU vCore", there is no such setting. I did disable "dynamic vCore".

By disabling Turbo, CoreTemp reports 1.27v under load. Unfortunately, the CPU only goes to 3.5GHz, not the 3.8GHz (turbo). When enabling Turbo, default voltage is 1.3 and goes up to 1.38.

I also noted that even while I've set the main multiplier setting to 3.4GHz ("34" for multiplier), when saving and rebooting and going back into the BIOS, it decides to put "38" back.

Given my previous mobo (GA-Z68MX-UD2H-B3) had the same problem, I wonder if this is an issue with Gigabyte in general or if I'm overlooking another exotic setting (I've not overclocked in years, much less have had to worry about default voltage settings being too high "out of the box"...)

I'll post some pics later...

Best way would be to stop using auto vcore. You are using way to much voltage that way.

Not by choice. Dynamic vCore (DVID) is the only setting I've found that will adjust the voltage, but there is no standard "vCore" setting I've found. Worse, enabling XMP overrides everything...

I just overclocked my cpu to 4.050ghz at 1.3v and my ram at 1.6v at 1800mhz also disable the turbo boost as it does over volt the cpu my temps with turbo were 40c idle and 70c on load.the sad part was turbo boost bumped from stock clock of 3.06 to 3.2 and really didn't do much but heagt my cpu turn it off and over clock it manually and got 3.8ghz at 67c under max load

Thanks MUCH for the specs! I'm not going to OC the RAM, but those numbers will be of help as I get these settings figured out. (I'd noted this AM that Turbo mode does ramp up the voltage on its own as well, which I do not want.

The turbo for all the cores should be set to the same as the cpu frequency. Having turbo enabled will NOT heat your cpu up any more as long as you have the vcore set manually.

If you are using auto vcore, without turbo obviously it will be lower because the cpu frequency is lower.

So if I go back into the BIOS and change the four Turbo states (1-core, 2-core, etc) all to "34"?

I'll keep looking for a standard vCore setting as well. (When I get the pics posted, this will make more sense. I've found no setting that allows me to change it from 'auto' to 'normal' or '1.2v', etc... )

 

[RGE]

What your looking for is just... Vcore.

2nd EDIT: Think I found the problem, post at Anandtech, from another person with same problem of cant find vcore on your mobo.

Overclocking is achieved by setting the "stock" VCORE for idle and then using the "dynamic offset" feature to increase the vcore at load.

And another quote on this board from searching:

i currently am using this board with my 2500k overclocked to 4.7 with a vcore of 1.4. THERE IS LITERAL NO VCORE OPTION. but what we have is DYNAMIC VCORE in voltage control. you pick increments of voltage to add or subtract from your vcore.

Your on your own there, I dont care for that implementation at all. But according to manual you can change vcore to normal, so wonder if vcore is hidden setting with CTRL F1, etc. But some one with an 1155 board will have to help. Not a board I would care for, for Ocing.

 

[Thermodynamic]

Images (with notes):

(3.5GHz, 1.23v)

(this is a riot - I set the ratio to 34 and then rebooted. 38 magically reappears.)

(If I enable XMP, voltage skyrockets)

(Dynamic vCore is the only voltage setting I found. By default it read 'auto', I changed it to 'normal'.)

(when idle)

(under Prime95, note it's not going to 3.8GHz - what's weird is that, immediately when logging into Win7 and loading CoreTemp, but about 2 seconds it WILL read "3800MHz" before going down...)

 

[Thermodynamic]

What your looking for is just... Vcore. And once you hit enter on that line, you should get a list of voltages also word normal should be there. Set it manually. And you can Disable LLC (just to be sure).

EDIT: Ok I just read your mobo manual. Your bios layout sucks (typical of GB boards). You are looking for cpu Vcore. But it is not listed on bios page it should be. dynamic vcore is not what you are looking for.

What is interesting, is they state in manual you can put cpu vcore on normal, but no where did I see bios screen where vcore exists. manual with all bios screens here: http://www.manualowl.com/m/Gigabyte/GA-Z68A-D3H-B3/Manual/207322

Thanks for helping!

I'll go back and check to see if LLC has a "disable" option (I recall 'auto', 'standard', '1', and '2'...)

I'll double-check Dynamic vCore to see what the other options are, apart from 'auto' and 'normal'.

Thanks for the link as well (I still have my paper copy, but the digital version will be easier to look up...)


---

update: There is no 'disable' option, an dynamic vcore goes from -0.32 to 0 and up to +0.50...

 

[Thermodynamic]

Additional note:

When powering up, the BIOS reported 3.5GHz - so I hit 'del' to go in and try something:

I changed the dynamic vcore to -1.12. The CPU clock ratio still gets seen as "38", despite every time I try to change it to 34. (All Turbo settings and XMP are still disabled.)

I saved and rebooted. Now, in MIT status, it reads 1.212v and at 3.8GHz. Idle temp is 32C in the BIOS.

Obviously Gigabyte's brand has a few bugs in their Sandy Bridge BIOS (the Z68MX-UD2H-B3 had very similar problems). With dynamic vcore set to 'auto', 'normal', or even '0.000', there were issues. Even using the slightest negative value has shown improvement.

I may as well let this thing run at 3.8GHz, since the BIOS seems adamant in choosing the clock rate automatically (or I'll try the aforementioned forcing the 1-core, 2-core, etc, modes to"34" as well.) But with speedstep halving that, it's probably a non-issue. And I'm using the PC for 3D rendering, so 3.8 isn't going to hurt...


===========

update:

Booted into Win7. CoreTemp will read 3.8GHz now, but also reads 1.38v when Prime95 runs.

What the heck is causing Windows to override the BIOS settings?!

 

[RGE]

LLC standard should be same as disabled, ie allowing full vdroop. But doubt auto or standard will change anything.

I dont think windows is overriding your bios. I think your bios has issues. And I think GB bios writers have bigger issues.

 

[Thermodynamic]

What your looking for is just... Vcore.

2nd EDIT: Think I found the problem, post at Anandtech, from another person with same problem of cant find vcore on your mobo.



And another quote on this board from searching:


Your on your own there, I dont care for that implementation at all. But according to manual you can change vcore to normal, so wonder if vcore is hidden setting with CTRL F1, etc. But some one with an 1155 board will have to help. Not a board I would care for, for Ocing.

Just read your updated edits - thanks!

I did ctrl-f1. The advanced chipset options appears, but only one setting within exists (a video card setting).

I might be returning this card if no other options exist; my next test is to disable 1-core, 2-core, etc, modes and just run the thing flat at 3.4Hz. Maybe it'll stick to 1.20v then. ('turbo' mode going to 3.5GHz and using 1.38v is just mind-bogglingly bad... then again, my saved 'good' settings have it at 1.27v for 3.5Ghz 'turbo' when Prime95 runs. That's borderline, but if I can manage to kerjigger it to 3.8GHz on all 4 cores with 1.27v, then I'll be happy. Otherwise, 3.5GHz at 1.27 is probably acceptable, unless it is known that 1.20v should be adequate at running all 4 cores at 3.5GHz (Turbo) without the computer needing to raise the voltage?

Thx!

 

[Thermodynamic]

One last update for the moment:

Using ONE core, I can do 4.4GHz with CoreTemp showing 1.27v.

Using FOUR cores, I can do 3.8GHz with CoreTemp showing 1.361v.

Are these values typical? Or is it normal for 3.8GHz to be achieved with 1.27v? (If four cores needing 1.36v is a normal figure, then all is well... some of the info I've read hasn't gone into that level of detail between Turbo mode and number of cores enabled, and I'm still green on this as well...)

I do know that, with all 4 cores @ 3.8GHz, there was some instability at 1.23v when booting Win7, hence upping it to 1.27v. (or, rather, using -0.60v for the dynamic vcore setting since I have no actual independent "vCore" setting).

I did change LLC to "level 1", but that's easily reverted...

Also, as means for me to get up to speed on Sandy Bridge overclocking (just to know the hows and whys), I've read http://www.clunk.org.uk/forums/overclocking/39184-p67-sandy-bridge-overclocking-guide-beginners.html. I've disabled "Can change in OS" and set the Turbo modes by hand, though )1-core=44, 4-core=38, etc.)

Thanks again!


Slight edit:

I stumbled on this article:

http://hardforum.com/showthread.php?t=1592713

1.36v may not be uncommon for all 4 cores. But the Ghz rating they are aiming for is a tad higher than 3.8 for all four cores, so there is still another issue. (I've yet to disable "spread spectrum" as well...) It's also possible my 2600K isn't capable of going this far, and it's just as probable that the Gigabyte board's own issues hamper the issue.

 

[RGE]

Your 2600K potential, or vcore needed for given Ghz is an unknown, because your mobo/bios design wasnt meant for overclocking. It has poor implementation of turbo as well with overvolting.

You wont know what your 2600K needs for any given Ghz setting, unless you had a mobo that was built for even basic Ocing, ie manual vcore adjustment. The GB bios doesnt adjust vcore based on your cpu, it is going off some algorithm based on load and ghz, and aimed at allowing even the worst cpu to run stable, and in doing so gives much more vcore than necessary to majority of cpus.

 

[Thermodynamic]

Another update: While considering buying the Z68X-UD3H due to having an 8-pin power connector (the Z68A-D3H has a 4-pin connector, and the 2600k runs at 95W and that could be an issue), I found somebody who bought the board:

http://www.hardwarecanucks.com/foru...-advise-needed-gigabyte-ga-z68a-d3h-b3-2.html

He is using Gigabyte's own "TouchBIOS" utility to read the voltage rating, instead of CoreTemp.

I reset the BIOS back to defaults (3.4 regular, 3.5-3.8 for 4-core, 3-core, 2-core, 1-core Turbo). I lowered the voltage by 0.150v (so the BIOS reads 2.0v).

Now running Prime95, while CoreTemp reads 1.27v, TouchBIOS reports 1.07v.

Now, assuming TouchBIOS is correct, not only is 44C under full load FANTASTIC, but I should be able to ramp up this computer and get to run at 4.5GHz (or close to it) like the person in that post I linked to above and (in all likelihood) obtain similar results.

Why is TouchBIOS reporting 0.20v less than CoreTemp? (RealTemp also showed 1.27v). Could CoreTemp and RealTemp not be reading the voltage sensor accurately? I'm inclined to believe Gigabyte's own tools are more likely to put out accurate results than a third party... I hope! (But recalling, on my previous mobo, with temps going up to 80, the fact the PC didn't sound an alarm as I keep the BIOS alarm set to go off at 70C... hmmm...)

If this is what's going on, then all is well. Or far better. Especially as temps under Prime95 right now are at a terrific 44C (TouchBIOS - CoreTemp reports 54C).

I did notice CoreTemp showing 74.5W in use, and the 2600k processor is said to run at 95W. I don't intent to do much in the way of overclocking, but should I be concerned? (The Corsair GS600 power supply doesn't mention maximum wattage put out by the EPS cables (4+4), which doesn't help.)

 

[Thermodynamic]

Your 2600K potential, or vcore needed for given Ghz is an unknown, because your mobo/bios design wasnt meant for overclocking. It has poor implementation of turbo as well with overvolting.

You wont know what your 2600K needs for any given Ghz setting, unless you had a mobo that was built for even basic Ocing, ie manual vcore adjustment. The GB bios doesnt adjust vcore based on your cpu, it is going off some algorithm based on load and ghz, and aimed at allowing even the worst cpu to run stable, and in doing so gives much more vcore than necessary to majority of cpus.

That would make sense... a lot of sense... thanks much!!

For now, I'm not going to tinker further and let the system run as it stands (I'll keep an eye on this thread over the next few days, of course). As I learn more, do more testing, and even buy a known good O/C'ing board, then I'll know better about all of this. Especially having just read on the 4+4 EPS power cables and a different forum having a person wondering if the D3H can really handle a 95w processor.

 

[wingman99]

I have the same board about and dynamic VCore works perfect you can lower or raise you Vcore with a simple adjustment and use HWiNFO64 to read your Vcore voltage in windows.

I like not having to mess with so many settings. If you want your dynamic vcore static just turn off the power saving features.

This is a sweet overclocking board has all voltage options to play with.

You both are making this much more complicated than it needs to be leave the turbo settings stock, you don't need to touch them just overclock and adjust dynamic vcore + or -.

 

[Thermodynamic]

I have the same board about and dynamic VCore works perfect you can lower or raise you Vcore with a simple adjustment and use HWiNFO64 to read your Vcore voltage in windows.

I like not having to mess with so many settings. If you want your dynamic vcore static just turn off the power saving features.

This is a sweet overclocking board has all voltage options to play with.

You both are making this much more complicated than it needs to be leave the turbo settings stock, you don't need to touch them just overclock and adjust dynamic vcore + or -.

You piqued my interest with HWINFO64.

Sadly, it too reports the same vCore as CoreTemp and Real Temp... EasyTune and TouchBIOS (both native Gigabyte apps) show less.

I'd rather keep the energy saving features enabled, but if they are hampering a proper vCore (everything I've read basically says these CPUs will o/c to 4.4GHz with less than 1.30v... the only apps that back this up are Gigabyte's own tools... so what's different about Gigabyte's compared to every third party analyzer showing ~0.20v more for voltage and ~10C for temperature? Gigabyte's boards may not be conforming to Intel's references, which could be what the third party tools rely on - hence the differences in readouts. Assuming Gigabyte's Windows utilities are accurate, then all is well and I've been spazzing over nothing. But from what I've seen, I do agree with rge in that I think there are some issues with Gigabyte's BIOS. At least in terms of setting the frequency and voltage.

Time will tell... but while I first thought that the idle voltage was too low (0.7v ),

http://babelfish.yahoo.com/translat...nitt_undervolting&lp=de_en&btnTrUrl=Translate

The Sandy Bridge CPUs undervolt remarkably well...