Confusion on 4930k overclocking and turbo mode in bios

[Kaynar]

Hello,

I've been testing the overclocking capacity of my 4930k today and there is one thing that seems odd to me:

The turbo mode in the bios MUST be enabled in order for the clock multiplier to work. If I disable it, I can only do 3.4Ghz no matter what the multiplier in the BIOS says (in windows 7 its stuck to 34x100mhz).

My previous platform was X58 and there disabling turbo mode was the standard procedure for overclocking. Has this changed now? Is it safe to have turbo mode on?

 

[FreedomEclipse]

Its safe, thats how ive done it since i jumped to a 2500k than to my 3930k. Theres no harm in keeping it on so long as the OC is stable.

 

[LiveOrDie]

I'm sure turbo boost on x58 only applied to the 1st core when now its apples to all cores.

Increasing BLCK and/or Frequency: You have to use Turbo Mode to OC SBe with every board. You need to leave Turbo Mode enabled, and set each multiplier, or just set one for all cores. On some boards EIST or C1E needs to be left enabled to use Turbo multipliers. On the UD7 it isn’t necessary to leave either one on. If you want to use BLCK straps or change BLCK, please do that first. Set the strap and BLCK you want and restart, if that doesn’t work go down to the BLCK Strap Section for more help. Next after you find BLCK, whether stock or not, then start increasing the multiplier. Increase it two at a time. Please realize that you won’t be doing over 4.8 GHz 24/7, at least not many will because of heat restraints. If you fail to POST, or fail stability tests in Windows, then you need to increase your VCore. At this point you should disable power saving settings, which are EIST (Speed step), C1E, and C3/C6 states. On some boards you need to leave EIST or C1E on to maintain the OC at all.

 

[Kaynar]

Ok thanks for the explanation and the confirmation!

For now it seems I can do 4.4ghz at 1.25v at 65-70c in prime95 (haven't tried less voltage yet) for at least 30mins nonstop and I got BSOD at 4.6ghz 1.4v after 10 minutes of prime95. Anybody think 4.5ghz will be easy or that I can get 4.6ghz stable with some patience and tweaks? I've already changed basic options in bios for better oc and I just updated to new bios today.

Also, Do you guys think its worth having 1T on my ram? default is 2133 9-11-11-31-2T, so I forced that and 1T instead of 2T and allowed 110% voltage so its now 1.67v. My thoughts is that some other stuff there is on auto mode and I am pretty sure that if I set 1T then some other automatic settings become worse, bringing no improvement overall. (CPUZ reports Row Refresh Cycle Time tRFC of 300 clocks at 1T is this ok?)

 

[FreedomEclipse]

up voltage and check your loop. people are hitting 4.9Ghz@1.34v at around 70'c prime95 with watercooling.

 

[Ralfies]

Ok thanks for the explanation and the confirmation!

For now it seems I can do 4.4ghz at 1.25v at 65-70c in prime95 (haven't tried less voltage yet) for at least 30mins nonstop and I got BSOD at 4.6ghz 1.4v after 10 minutes of prime95. Anybody think 4.5ghz will be easy or that I can get 4.6ghz stable with some patience and tweaks? I've already changed basic options in bios for better oc and I just updated to new bios today.

Also, Do you guys think its worth having 1T on my ram? default is 2133 9-11-11-31-2T, so I forced that and 1T instead of 2T and allowed 110% voltage so its now 1.67v. My thoughts is that some other stuff there is on auto mode and I am pretty sure that if I set 1T then some other automatic settings become worse, bringing no improvement overall. (CPUZ reports Row Refresh Cycle Time tRFC of 300 clocks at 1T is this ok?)

The performance impact of 2T vs 1T is very small(~5%), so if 2T allows you to achieve a higher CPU clock it is definitely worth it. Your ram OC can affect your CPU OC and vice-versa. Give it a try.

 

[Kaynar]

My loop is silent on priority. I'm not gonna hit 2000rpm with my fans. 1200rpm and low fpi rads (HWlabs) is what I got. So I don't expect lower temps. Also, while some people hit 4.9ghz there are others who can't seem to get past 4.5ghz no matter what voltage and options... and im not an overclocking expect. I'll see how low voltage I can get stable on 4.4ghz and I'll report back.

And yeah tbh I know I should be trying to oc with ram at stock settings, I'm just being impatient lol.

EDIT: for as long as I had 4.6ghz at 1.4v, cpu temp was up to 76c.

 

[cantoncobaltsssc]

I'm running 4.4 auto bios everything max voltage is 1.28 turbo is left on and I'm rock stable for a year now. I got 4.5 manually putting the voltage to 1.35 but I could never get my chip to do 4.6 no matter what I did. 4.4 is a really good spot far as I'm concerned if you guys are getting i7's higher than that you are some guru's.

 

[Kaynar]

yeah I got stable 4.4ghz for a whole day of gaming at 1.25v and I intend to try to lower that voltage till if find the least possible. Only way to get 4.5ghz is 1.35v (1.3v crashes in under 5 minutes of load) so I don't think its worth the extra voltage and temps of anything higher.

 

[cantoncobaltsssc]

That's what I did I gave up trying to get 4.5 or 4.6 just requires to much extra & temps start running over 80c. 4.4 is very respectable.

 

[FreedomEclipse]

That's what I did I gave up trying to get 4.5 or 4.6 just requires to much extra & temps start running over 80c. 4.4 is very respectable.

imo 80'c is too hot

 

[cadaveca]

imo 80'c is too hot

Yap. try to stay under 80c, but 80c on cores, not on CPU sensor! That's the one thing that makes clocking these chips a bit difficult...cooling!

 

[Kaynar]

imo 80'c is too hot

The same is happening to me at 1.4v I get into the 80c area even though I got a good custom wc loop which could keep my previous i7 930 at 4.2ghz without trouble

 

[cadaveca]

The same is happening to me at 1.4v I get into the 80c area even though I got a good custom wc loop which could keep my previous i7 930 at 4.2ghz without trouble

The 930 only had 8 MB of cache. That's a big part of the heat generation, the cache, plus X58 didn't have PCie integrated. The inclusion of these other parts into the CPU makes cooling modern CPUs a completely different thing than it was in the past. Add in a VGA...CPU temp goes up.

 

[Kaynar]

@cadaveca can you please answer this:

-Why does CPU-Z show me 1.224v or 1.232v when under the "max heat and power consumption" prime95 test (got all 12 workers active) but in idle it reports higher 1.248v or 1.256v (it randomly keeps changing between the values both in load and idle)? Does this essentially mean I can easily reduce voltage in bios (currently at 1.25v)?
-Why Core temp has gone a bit crazy since my overclock and lags quite a lot in updating cpu conditions? (I am using the windows gadget additionally) Two times it totally disappeared and I could not launch the windows gadgets at all till system restart.
-Why does Core temp report 44.5x multiplier (lol?) sometimes and it even goes up to 46x, effectively reporting 4600mhz.
-Since I overclocked, my CPU seems unable to down clock below 4.4ghz. CPU-z, coretemp and realtemp all report permanent 4.4ghz even though CPU-Z still says "12-44" in the multiplier bracket.
-Do you think its possible for my cpu to reach 4.6ghz stable by tweaking other options, since I can boot at 1.35v and 46x multi with medium LLC but I crash after 2-3 mins of load and still crashes at 1.4v.

I got a quick briefing from singularitycomputers youtube channel from this video

about x79 overclocking, certainly he used 3930k since its from 2012!

My current tweaks are:
-1.25v (haven't tried less) with "medium" (50%) LLC (can still go up to high, very high, extreme if needed). 44x multi on all cores.
-cpu current capacity 140%
-VCCSA current capacity 100% (default)
-VCCSA LLC "extreme" (was like this from bios default), running 2133mhz dominator platinum.
-CPU VTT switch 1.3x (cleaner power)
-VVT over-current protection "off"
-Disabled two other options for PCI-e clocking and something else, I don't remember the name now but the video advised to do so. In bios it is also written that disabling those help oc.

 

[Aquinus]

Could we see more of the voltages and such you have setup? Where is your VTT and VCCSA voltages sitting at? I don't know if this is the case with IVB-E, but with my 3820, I try to keep my VCCSA within 0.2v of my core voltage. For example, if my core voltage is 1.4v, my vccsa is ~1.2v, maybe a tad higher considering LLC.

With this said, I don't think 1.4v is a smart voltage to leave an IVB-E chip at but that depends on your temperatures and how much they fluctuate. 80 degrees tells me to kick it back a notch. I aim for 70 as a relatively safe max.

Personally, I have turbo turned off, the bclk strap set to 125Mhz, and my multiplier to 36 and it gives me a pretty healthy 4.5Ghz. I had to adjust my timings a tiny bit, but my memory and IMC could handle the overclock fairly well.

 

[Kaynar]

@Aquinus here is what Asus AI Suite reports:

(IDLE / LOAD)

Vcore: 1.254-1.258 / 1.230-1.232 (yes, lower voltage under load)
VTT CPU: 1.084-1.112 / 1.100-1.103
CPU VCCSA: 1.131-1.137 / 1.146-1.156
CPU PPL: 1.790-1796 / 1.787-1.800
DRAM AB and CD: 1.671-1.681/ 1.665-1.681
2nd VTTCPU: 1.053-1.062/ 1.056-1.059

temperatures are 31-35c idle 64-67c under load.

 

[Kaynar]

Managed 4.4ghz at 1.230v for an hour of prime95... I still don't understand why CPU-Z reports 1.204v under load but 1.232v when idle...

 

[Aquinus]

It's called v-droop. It prevents your VRMs from overshooting the output voltage when load current changes by a large amount.

Advances in silicon technology have led to significantly faster processors, memory and switching logic. These high-performance semiconductor devices can demand very fast changes in load current, posing considerable challenges for the voltage regulator supplying their core voltage. When switching out of “sleep” mode, for example, processors often activate a large number of hitherto dormant circuits, resulting in a large step-change in load current within a few hundred nanoseconds. Minimizing peak-to-peak voltage deviation in the face of such large dynamic changes in load current involves innovative power conversion techniques, especially if large amounts of output capacitance are to be avoided.

Several techniques have been proposed to address these challenging regulation requirements. One method that has gained wide industry acceptance is called “droop” compensation.
...
Droop compensation intentionally increases the dc output impedance of a converter, lowering its output voltage, as load current increases. The technique was first employed with dc-dc converters designed for parallel operation to help balance current sharing. It is now used by a number of the leading multi-phase pulse width modulated (PWM) controller ICs and voltage regulator modules (VRMs) designed to power latest-generation processors. The effect of droop compensation is to reduce overshoot of the core voltage during large step changes in load current.

Source

 

[cadaveca]

It's called v-droop. It prevents your VRMs from overshooting the output voltage when load current changes by a large amount.

More importantly, it's caused by the "MEDIUM" LLC setting he's got. "High" or "very high" should "fix" that. Often got to wonder when users are asking for actual cause, or what setting is causing the behavior, and now, between you and me, we have both angles covered.

 

[Kaynar]

Alright your answers got me covered thanks! But I guess there is nothing bad in leaving the LLC to medium since its actually working without crashing...

 

[Aquinus]

More importantly, it's caused by the "MEDIUM" LLC setting he's got. "High" or "very high" should "fix" that. Often got to wonder when users are asking for actual cause, or what setting is causing the behavior, and now, between you and me, we have both angles covered.

It's also worth mentioning that you're circumventing what vdroop is designed to do when you start jacking up LLC. Under load-changes you very well could damage the CPU. CPU-Z doesn't show it, but the spikes are actually much higher voltage than what you see and they last for several nanoseconds. They're short spikes that occur and vdroop (lower LLC settings,) are designed to minimize such spike.

I would try to make a reasonable decision about how much less vdroop you need as opposed to just eliminating it. It is good for your CPU to keep it, even if it's not optimal for your overclock.

 

[eidairaman1]

Ensure your psu is from a reputable maker-seasonic,enermax,corsair,pcpowerandcooling