Volts for 2700k

[Senna]

Hello guys

Ive overclocked my i7 a little bit more to 4.2 ghz@1,296V but i noticed that the minimum core volts on 1600Mhz [speedstep] jumped quite a little bit higher [from around 0.89V to 1,05V] which is slight overvoltage for 1600mhz step. So my question is where in Asus uefi do i turn this back to normal. Its so stupidly designed - i just have few diffrent 4-5letter acronyms with voltage settings in them, without any freaking explanation whats so ever, so iam not sure which one should i turn down to make low frequency voltage lower.


Also strangely Temps jump quite rapidly when i run wPrime from ~20degrees to 50 degrees within 2 seconds then stay at max ~55 degrees Celsius. I think it is quite high for just 4.2Ghz oc and ambient temp 14C [i run my pc in my little basement server room ;P]

ps. any suggestions for the volts?
ps2 pll overvolatage disabled, loadline calibration disabled, max ECO settings.

 

[AverageNinja]

It's not that high... That 20 Degrees Celcius is your CPU on Idle. Which means the CPU is not stressed, so the temps are a lot lower.

 

[Pr3d4t0r]

sounds normal to me, prime is very cpu intensive app therefore its used for torture testing, on small fpts cpus jump up in temps by even 40c's thats normal(imho)

2700k at 4.2 running at 55c is still good.

in regards to volts, i prefer amd myself(stupid me i know :3)

i turn off all of the crap like epu, auto voltages, auto clocks/multipliers, i start with stable then lower the volts to the minimum but still stable

right now im running my cpu at 1.4(amd, 1.55 would still be fine for him) considering in running dual core, turned to quad and ocd to 3.75, not too bad v wise :3 you just have to find that sweet spot for your chip

 

[Senna]

I would like to set the volts for 1600/2400/3600/4200mhz steps myself but i cannot seem to find the option just some unknown freaking acronyms that asus didn't even give a shit about putting any explanation to the right side info panel. SO iam not sure what and where i should fiddle with. . .i do like speedstep and stuff because most of the time my pc is afk idling, or watching a movie. . i dont need 4.2ghz draining 80w of power to watch a movie or make some troll comments on youtube.

Well i know that temps should jump while in stress, the fact i dont get is how the heck it jumps from 18C now to 50c within 2 seconds. At that rate we could boil some water for tea within 5-10s.

Also found out something strange that i haven't seen before . . . 2cores are around 28 degrees while the other 2 around 16-18C. Its bothering me because thats 10 degrees difference on a very very small die size. The cooler itself touches entire cpu heat spreader so the 2 cores should warm up from the the other 2 cores. Also in wPrime 8threads one core is always ~8degrees hotter than the others. . . .


/Ambient is around 11C ;P

 

[Senna]

Okey then where do i set min cpu voltage for idle and max cpu voltage for stress at 4.4?

 

[AverageNinja]

Go to advanced tab and show us that.
P.S. 1,5volts is already a CPU-Killer. 4.4Volts would kill the chip in an instant. I think you mean Frequency?

 

[nepas]

Go to advanced tab and show us that.
P.S. 1,5volts is already a CPU-Killer. 4.4Volts would kill the chip in an instant. I think you mean Frequency?

errrr the 1.5 is for the ram.

@OP:as for the core temp difference,maybe you need to reseat the cooler again?

 

[Senna]

Go to advanced tab and show us that.
P.S. 1,5volts is already a CPU-Killer. 4.4Volts would kill the chip in an instant. I think you mean Frequency?

Well Yeah, i mean i want to change the basic volt for stock 1600mhz and for top clock 4400mhz and all steps in between will just average themselves. I dont think there even is an option to put 4V on the vcore.

Well here are the images:










And like i said i have no idea where should i play with volts. The automatic voltage seem to keep the 4,4ghz volts just slightly too high but that is not a problem. The problem is that stock 1600mhz volts went higher for over 0,7V. I want to bring down both volts.

/@AverageNinja Your rig is amazing x) I wish i still had my P3 at my place

 

[nepas]

If you think the volts are too high,you need to turn the offset away from auto and decide yourself what you want to set.

I can dig up a link that may help you later.

 

[Pr3d4t0r]

the best thing you can do about it with asus's eufi is to set the volts to offset mode, and set the voltage for the lower speed step and set the offset + to the amount that 4.4 uses.

as far as i know the offset technology is not the greatest and tends to fail from time to time but it should work.

offset is basically setting lower and upper limit, quite usefull for auto oc and trying to find lowest v to put on oc.

in regards to temps, its probably the paste is not evenly spread;

single core: spot

dual: line

quad/hex/octo: Cross/Star/even spread

 

[Senna]

See i checked this voltage offset thingy and when i set it from manual[automatic] to offset mode . . . i have only one additional option. The option is "offset voltage" and the range is 0.65V to 1.4V. . . which doesnt make sense. I expected that "offset" will allow me to add offset voltage on top of regular core voltage for both x16 and x45 state. . . but it doesnt seem like it. The min offset of 0.65v is also slightly confusing . . .if it adds 0.65v on top of regular voltage then it will fry the cpu. If it is just the max voltage for x45 state then the x16 will be servilely overvolted for no reason.

Oh btw i went for 45 multiplier this time. The x16 voltage automaticlly jumped to 1.09 and x45 to 1.38

//Oh yeah well there is second option. . . option to choose the sign of offset. i.e. i can chose "+" or "-".
I guess that if i could set negative offset of for instance -0.1 it would reduce the automatic voltage settings fby 0.1V. THe problem is minimum offset value is 0.65V or -0.65 [. . . . ?]




after around 4hour of stress testing [i.e. doing BOINC computing projects, if u have to lose some power i prefer to at least help someone with it] my pc got unstable and shut down itself without any warning. Temps were pretty normal though so not sure what was the case. I did 2 runs of 3d mark vantage and it passed on 4.5ghz. . . HWiNFO64 also reported vcore of 1.42 even though the one in CPU Z is much lower so not sure what the heck is going on anymore. . .

 

[Col55]

Set cpu to auto volts will run at 4.4 easy at around 1.256v , set all the power saving stuff to enable job done for your 1.6 no load 0.9v

 

[233Kosta]

I'd say play around with LLC. This auto volt stuff doesn't always work. I've got a Z68 and whenever it's on auto - it wants to go for 1.6V, so I say leave it at a fixed Vcore, then play with LLC settings until it's roughly the same on idle and load.

Then get Prime running and use TurboV to lower Vcore until prime crashes. Then go 1 notch up from that, save it all in BIOS and run Intel Burn In test for 10 instances at maximum memory use for the final test. It'll take a while, but if it passes - leave it be, or potentially drop more volts (doubt it'd be the case).

If it crashes - bump up the Vcore a bit and try again.

The initial IBT tests can run standard until you establish a stable Vcore. That'll save you some time (likely lots of time ).

I trust mostly CPU-Z and HWMonitor for Vcore reporting, but that is just me.

 

[Master&Puppet]

I've taken this from the ASUS ROG website. It looks like it should be a traditional offset system.

CPU Voltage:
There are two ways to control CPU core voltage; Offset Mode and Manual Mode. Manual mode assigns a static level of voltage for the processor. Offset Mode allows the processor to request voltage according to loading conditions and operating frequency. Offset mode is preferred for 24/7 systems as it allows the processor to lower its voltage during idle conditions, thus saving a small amount of power and reducing unnecessary heat.

The caveat of Offset Mode is that the full load voltage the processor will request under load is impossible to predict without loading the processor fully. The base level of voltage used will increase in accordance with the CPU multiplier ratio. It is therefore best to start with a low multiplier ratio and work upwards in 1X steps while checking for stability at each increase. Enter the OS, load the CPU and check CPU-Z to check the voltage the CPU requests from the buck controller. If the level of voltage requested is very high, then you can reduce the full load voltage by applying a negative offset in UEFI. For example, if our full load voltage at 45X CPU multiplier ratios happened to be 1.40V, we could reduce that to 1.35V by applying a 0.05V negative offset in UEFI.

Most of the information pertaining to overclocking Sandy Bridge CPUs has already been well documented on the internet. For those of you purchasing retail Ivy Bridge CPUs, we expect most samples to achieve 4.3-4.5GHz with air and water cooling. Higher overclocks are possible although full-loading of the CPUs will result in very high temperatures even though the current consumed by these processors is not excessive. We suspect this is a facet of the 22nm process.

 

[233Kosta]

Thank you! Now I understand why I get 1.6Vcore when I go on auto/offset. Much appreciated, a former AMD user.
:amdwar:

Mind you, my ROG Crosshair IV Formula never did give me such grief over offset Vcore and such, so it should be slightly more towards :intelwar: on that particular issue by the looks of it.

Thanks a lot!

 

[233Kosta]

Oh, btw I suspect the 22nm heat issues arise from the heat sensitivity of the transistors while unpowered.

Intel usually solder the IHS cap onto the chip, however they likely noticed failure or degradation caused by the excessive heat of the soldering process and went for thermal grease instead. If said thermal grease's application, quality and so on are patchy - we get heat issues after 1.3V

That's my hypothesis anyway. If someone has anything better - fire away.

 

[Master&Puppet]

Yea there has been several attempts (at least some successful) where people have removed the IHS and replaced the TIM with a better quality one which has made several degrees difference. Worth the risk? Maybe!

 

[233Kosta]

To be honest, I don't see the point of replacing that TIM if you can just run the chip bare.

I don't know about different heatsinks, but my water block doesn't have stops on it at the level of the chip's IHS, so I don't see an issue with de-capping the chip and getting a 10-20*C improvement which would translate to further overclocking headroom. I don't see anyone sending the chip back after doing this so why not go all the way? :evil:

In any case, unless there are serious thermal issues I don't see a good reason to void warranty on your chip.

Unless you got that chip specifically for the purpose of attaining the highest possible clock and benching at said clock. In that case - de-cap away! I've heard of 6+GHz on water and phase change with these little chips. I have no way of confirming this of course, but I've heard people's claims on the matter, mainly when discussing 'stupid volts'.

 

[Master&Puppet]

Honestly I still don't think there is a consensus of either of those points. I've seen people claim 10-20C better temps and others get worse results than stock. Lidless cooling has been seeing the most unreliable results of all - there are big issues with mounting the block directly to the die because of pressure. Replacing the TIM under the IHS seems more reliable but the effect, if any, is much smaller (0-5C most commonly). At some stage I will probably have a go myself but I think there is probably as much to be gained by lapping the block and IHS as there is by taking it apart and updating the TIM.

 

[233Kosta]

Don't know if lapping will have a huge effect. My WB is pretty flat and comes with a mirror shine on it, so I don't think I'd need to do that should I go IB, but I do think there's some viability to de-capping the chip, since what you're doing is taking out an element of thermal resistance. The effect however depends on how bad the TIM on the chip is, how rough and uneven the block surface is, how much and what TIM is used when this is done, any air bubbles, etc.

If your block isn't lapped but is flat, I'd recommend priming it with TIM of your choice. I still use AS5, mainly because I've got loads of it and it lasts ages. It's pretty easy to prime contact surfaces with AS5, since it's not the most viscous thing out there. All you need to do is put a tiny bit in the middle of the block and rub it in with your choice of cloth.

Far as I know, most blocks come with spring loaded screw attachments. In this case, huge contact pressure is not necessary, in fact, I'd probably suggest doing up the screws to half of what you'd normally use, but make sure it's even.

Most of the problems arise at higher load conditions (i.e. past 1.35-1.4V), where the stock TIM just doesn't keep up. I'd imagine you won't see much difference at lower load conditions, but once you start pumping the volts through, the difference will increase in favour of new TIM.

The results you achieve by taking the cap off will probably be best if you prep the contact surface and apply a line down the middle of the chip, which would then spread, ensuring no air bubbles. At this stage, I think it's probably better to have a little too much since good cover is on the priority list. Don't spread, that's just begging for air bubbles.

Have fun!