Asus Motherboard + 11700K = All Core Boost Under Advertised Intel Spec? Bad Chip?

[CyberCT]

The last PC I built was with a 9700K about 3 years ago. Overclocking with that processor on the ASROCK Z390 board was a piece of cake. 1.28V at 5.0 GHz all core. No issues.
I've been in this hobby for 10 years so I'm somewhat familiar to it but haven't kept up with the recent offerings.

I've been itching to build a powerful mini ITX build for a while and with nearly all the components on sale, I caved . I'm using the Asus ROG STRIX Z590-I GAMING WIFI ITX with the 11700K.

With everything stock in BIOS, running a stress test in AIDA64 the VCORE will jump to up to 1.41V. It will typcially settle in the high 1.3V range.
Running Cinebench, single core test, the VCORE stays in the 1.42V to 1.47V range, but running multi core it falls back down to the low 1.3V range (temp in mid 50s).
I am running the latest BIOS available from ASUS.

After lots of tinkering, it seems the Asus Multicore Enhancement option is half the equation.
Set to off, everything else stock, with the CPU under load all core AIDA64 stress test, the all core range in frequency is 4.0 - 4.2 GHz. VCORE is 1.2V range.
Set to on, voltage is high 1.3V range (with blips above 1.4V) but cores are now all at 4.6 GHz under load.
One strange thing here is that with Cinebench single core test, the VCORE is always in the 1.4V range, regardless of Asus MCE being on or off. The core frequencies are 50, 50, 49, 49, 47, 47, 46, 46 (they vary dynamically like they should)
Setting Asus MCE to off is supposed to keep things within Intel's limits … but Intel advertises this chip as 4.6GHz all core … which is only possible with Asus MCE set to ON, which bypasses Intel's limits. What gives? Is my chip defective?

With Asus MCE off, CPU temp is mid 50Cs with max core temp low 70Cs.
With Asus MCE on, CPU temp is mid 70Cs with max core temp low 90Cs with just a little bit of throttling.

Case is NZXT H1 which comes with a 140mm AIO. Having the side panels on or off doesn't matter much here.

In the past I'd never let a CPU run above 1.3V (debate on safe daily voltage). These new CPUs have a higher voltage tolerance?

I don't mind downclocking a tad to keep things cool but I at least expected to hit the advertised all core 4.6 GHz within stock voltages.

I haven't tried any of the other typical overclocking stability tests like Prime95, Realbench, etc yet.

Any help on what's going on here with my chip?

 

[AlwaysHope]

What app are you using to measure temps & Vcore?

 

[AusWolf]

What's going on is the chip's limits.

With MCE on, you're limited by the boost table (4.6-5.0 GHz). Higher frequencies need a higher voltage, that's why you're seeing 1.4 V in a single-core load, but not in multi-core.

With MCE off, you're limited by a 125 W PL2. It is in no way possible to achieve a 4.6 GHz all-core boost within this limit, that's why you're seeing much lower clocks. It's normal.

Advertised boost frequencies and power consumption aren't both guaranteed anymore. It's either/or.

What I'd recommend is leaving MCE on if your cooling can handle it. If it can't, then turn MCE off, and play with your PL values until you find a good balance.

Welcome to the world of 11th gen.

 

[CyberCT]

I'm using AIDA64 to measure temps.

Is running in the 1.4V range OK long term with these chips?

AusWolf, I looked at your post. Looks like we have roughly the same goal here. You posted your PL1s ... What are your PL2s?
What guide did you specifically use to tune your CPU to your liking?

With MCE off, I can run Cinebench, AIDA64 stress test, and God of War ... I can run them all and the temps are well within check (I also swapped the 140mm fan for the Noctua 1500RP variant). I made a fan profile with the open source"Fan Control" program and it's pretty darn quiet. The ASUS TUF 3070ti also keeps the GPU cool in the mid 60s max under a Unigine Heaven 30 minute loop, which is impressive. I added the 92mm fan on the rear inside of the H1 case, which probably helped.

 

[AusWolf]

I'm using AIDA64 to measure temps.

Is running in the 1.4V range OK long term with these chips?

AusWolf, I looked at your post. Looks like we have roughly the same goal here. You posted your PL1s ... What are your PL2s?
What guide did you specifically use to tune your CPU to your liking?

With MCE off, I can run Cinebench, AIDA64 stress test, and God of War ... I can run them all and the temps are well within check (I also swapped the 140mm fan for the Noctua 1500RP variant). I made a fan profile with the open source"Fan Control" program and it's pretty darn quiet. The ASUS TUF 3070ti also keeps the GPU cool in the mid 60s max under a Unigine Heaven 30 minute loop, which is impressive. I added the 92mm fan on the rear inside of the H1 case, which probably helped.

Sorry, I meant PL1 (the long-term one). I always confuse the two.

Right now, I have a 280 mm AIO, so I'm running my 11700 with MCE enabled.

Before I had the AIO, what I did was set PL1 and 2 to the same value starting with the default 65 W (MCE disabled), ran Prime95 for a couple minutes and looked for stability/temperature issues. When everything went fine, I went back to BIOS, increased the PLs by 5-10 W and repeated the process. When package temp reached about 90 °C, I called it a day. The sweet spot was around 125-130 W with a tiny be quiet! Shadow Rock LP. With a tower cooler or AIO, you'll be able to go well above this.

Cinebench is a good way to test too, although it uses way less power on 11th gen than Prime95 for some reason.

 

[CyberCT]

So in BIOS there's an option to limit the auto CPU Vcore. Regardless of what I put in there (1.32V, 1.38V) the VCore always hits mid 1.4V under the single core Cinebench test. Why? It should stay at the limit I give it.

 

[AusWolf]

So in BIOS there's an option to limit the auto CPU Vcore. Regardless of what I put in there (1.32V, 1.38V) the VCore always hits mid 1.4V under the single core Cinebench test. Why? It should stay at the limit I give it.

Hm... I'm guessing that the manual Vcore setting is only valid for multi-core loads and frequencies. I've never played with it myself, so maybe someone else can confirm?

Personally, I wouldn't worry about it as long as the system is stable.

 

[Deleted member 74752]

So in BIOS there's an option to limit the auto CPU Vcore. Regardless of what I put in there (1.32V, 1.38V) the VCore always hits mid 1.4V under the single core Cinebench test. Why? It should stay at the limit I give it.

My 5700G and 3700X both act the same on Asus motherboards.

 

[CyberCT]

Hmm I'm wondering if I should lower each individual core's maximum clock by 200 MHz or so to get a lower voltage. Impact would be minimal.

Everywhere I read, it's supposed to do 4.6GHz all core .... I can't believe it can't happen within a decent voltage.

 

[AlwaysHope]

Do the temps match what HWiNFO can report? I'd be cautious about setting a static Vcore value. 11th gen has Vcore voltage spikes depending on the type of code running. If your were on an MSI board, I'd recommend take advantage of the CPU lite load option & find a value that works with your particular silicon quality of that 11700k chip. Otherwise an adaptive mode with - (negative) value offset can work too.
Another thing to consider is the bios on that board up to date? sometimes this can fix unusual cpu behaviour.

 

[Mussels]

Advertised boost is for single thread.

Advertised all core for your CPU is 3.6GHz at 125W
(This would likely be under worst case AVX load conditions, which is why lighter loads can go higher)

This is the 4.9Ghz clicky: single core

The 5GHz clicky: "Uhhh so most of the cores wont achieve this ever, but one or two will. sometimes."

 

[R0H1T]

Doesn't Intel Extreme Tuning have the OC (testing) feature these days, like OC Scanner for Nvidia? You could try that if the 11700k supports it, you'll probably get better results with that. IMO manual tuning for overclocks is too time consuming these days!

 

[elghinnarisa]

Hmm I'm wondering if I should lower each individual core's maximum clock by 200 MHz or so to get a lower voltage. Impact would be minimal.

Everywhere I read, it's supposed to do 4.6GHz all core .... I can't believe it can't happen within a decent voltage.

The only spec posted is "up to 4.6Ghz" with the obligatory 48 asterisks behind that statement of course. As long as its given enough power, temperature is low enough, jupiter in the correct place relative to the moon, sun in the back, going downhill etc. etc.

 

[CyberCT]

Either late tonight or tomorrow when I'm home I'll use HWiNFO to compare the readouts in AIDA64.
I have the latest BIOS available: 1202.

With Asus MCE off (so Intel limits in place), here are my Cinebench R23 scores:
Single: 1,585
Multi: 13,548

With Asus MCE on (so Intel limits off), here are my Cinebench R23 scores:
Single: 1,585 (same)
Multi: 14,707

This 11700K review uses the same version of Cinebench and here are the scores:

Intel Core i7-11700K CPU Review

The Intel Core i7-11700K desktop processor is an attractive option given the latest round of price cuts! Join us as we explore it.

www.tweaktown.com

Single: 1,578 (just 7 points less ... essentially the same)
Multi: 15,276 (569 points more than mine w/ Asus MCE on, @ 4% more than mine)

They're using a 360mm AIO water cooler but my temps never spike to the 90s or throttle with Cinebench.

I'm going to mess with the voltage offset. I won't be doing all core locked frequencies with this processor.

I guess it's ok to hit up to 1.5X VCORE ... seeing how others online are hitting it too (probably spikes).

EDIT: TPU did a review of the 11700KF (so just IGPU is inactive) and strangely, for Cinebench R23, their single core is higher but multi core is even lower than mine (power limit on, so I assume Asus MCE off?):

Intel Core i7-11700KF Review - Almost as Fast as the 11900K

The Intel Core i7-11700KF has the same 8-core/16-thread configuration as the much more expensive Core i9-11900K, which it can almost match out of the box, and beat with a little bit of tweaking. Thanks to its very reasonable price point of only $390, the 11700KF could also be a tempting...

www.techpowerup.com

Single: 1,606
Multi: 12,874

 

[Assimilator]

Intel advertises this chip as 4.6GHz all core

No, Intel does not. As @elghinnarisa points out, the very specific language used is "UP TO 4.6GHz" - what you actually get is going to be entirely dependent on the thermal and voltage budgets available to the CPU. And with a puny 140mm radiator you are ALWAYS going to be thermally limited. This is the kind of tradeoff you need to be prepared for if you intend to go mITX.

 

[CyberCT]

So I had time to mess around with BIOS settings and don't know if I should start a new thread or just keep things here, but things get weirder. FYI I'm using AIDA64 as the monitoring metric ...

In BIOS I started messing with VCORE offset (overall, not individual), used the negative number sign from the drop down list, and reduced VCORE in increments (-0.25mv, -0.5mv, -0.75mv, -0.95mv) and ran Cinebench 2.3 and Asus Realbench 2.56 Benchmark. Got scores with no issues and much less VCORE and temps.

Then I ran the ASUS Realbench 2.56's 8 hour stress test w/ up to 8GB RAM (my system has 16GB). Any VCORE negative offsets caused failure results on multiple tests (a few dozen) even though most of the tests passed. Core temps never passed low 80s and there was no throttling). I got rid of the negative offset and changed it to "AUTO" in BIOS, ran an Asus Realbench 8 hour test again. BIOS is now set to "AUTO" for everything btw. The only failure (happened twice) during the 8 hour Asus Realbench run was related to Luxmark, which apparently is GPU related, according to a Google search. I have a brand new 3070 TI stock and temps are high 60s max so IDK why it's failing. All cores run at 4.6GHz during Realbench or Cinebench as tests.

So I now decided to just cap the max frequency for each core from 5.0GHz to 4.8GHz. I never hit 1.4 VCORE ... only max high 1.3VCORE range ... but now every core clocks up to 4.8GHz instead of the typical 4.6GHz all core frequency in the multi core tests. Strange.

I ran the ASUS Realbench 2.56's 8 hour stress test w/ up to 8GB RAM with the capped 4.8GHz (and now all cores under the test were 4.8GHz, not 4.6GHz like before) and there were multiple failed tests (a few dozen) and not only Luxmark. How can this be when all voltages are set to AUTO in BIOS? Shouldn't BIOS give enough voltage to the CPU under AUTO for the capped 4.8GHz?

I just changed the max frequency of each core to 4.6GHz and obviously all cores hit 4.6GHz in all Cinebench benchmarks (single or multi core). Max VCORE is now 1.305mv and temps are low 70s max. I hope if I ran an 8 hour Realbench test that there would be no failures (other than the bizarre Luxmark failed tests) because before, when running the 8 hour stress test in RealBench, only the two Luxmark tests failed.

Again, I have the latest BIOS available from Asus. What gives?

 

[AusWolf]

So I had time to mess around with BIOS settings and don't know if I should start a new thread or just keep things here, but things get weirder. FYI I'm using AIDA64 as the monitoring metric ...

In BIOS I started messing with VCORE offset (overall, not individual), used the negative number sign from the drop down list, and reduced VCORE in increments (-0.25mv, -0.5mv, -0.75mv, -0.95mv) and ran Cinebench 2.3 and Asus Realbench 2.56 Benchmark. Got scores with no issues and much less VCORE and temps.

Then I ran the ASUS Realbench 2.56's 8 hour stress test w/ up to 8GB RAM (my system has 16GB). Any VCORE negative offsets caused failure results on multiple tests (a few dozen) even though most of the tests passed. Core temps never passed low 80s and there was no throttling). I got rid of the negative offset and changed it to "AUTO" in BIOS, ran an Asus Realbench 8 hour test again. BIOS is now set to "AUTO" for everything btw. The only failure (happened twice) during the 8 hour Asus Realbench run was related to Luxmark, which apparently is GPU related, according to a Google search. I have a brand new 3070 TI stock and temps are high 60s max so IDK why it's failing. All cores run at 4.6GHz during Realbench or Cinebench as tests.

So I now decided to just cap the max frequency for each core from 5.0GHz to 4.8GHz. I never hit 1.4 VCORE ... only max high 1.3VCORE range ... but now every core clocks up to 4.8GHz instead of the typical 4.6GHz all core frequency in the multi core tests. Strange.

I ran the ASUS Realbench 2.56's 8 hour stress test w/ up to 8GB RAM with the capped 4.8GHz (and now all cores under the test were 4.8GHz, not 4.6GHz like before) and there were multiple failed tests (a few dozen) and not only Luxmark. How can this be when all voltages are set to AUTO in BIOS? Shouldn't BIOS give enough voltage to the CPU under AUTO for the capped 4.8GHz?

I just changed the max frequency of each core to 4.6GHz and obviously all cores hit 4.6GHz in all Cinebench benchmarks (single or multi core). Max VCORE is now 1.305mv and temps are low 70s max. I hope if I ran an 8 hour Realbench test that there would be no failures (other than the bizarre Luxmark failed tests) because before, when running the 8 hour stress test in RealBench, only the two Luxmark tests failed.

Again, I have the latest BIOS available from Asus. What gives?

Maybe your CPU can't handle 4.8 GHz all-core. There must be a reason why the 11th gen Core i7 is exactly the same chip as the Core i9 just cheaper, and I think it comes down to binning.

Or maybe the Auto voltage setting works only from a set voltage/frequency table, and the value it has for 4.8 GHz isn't enough to maintain 4.8 GHz under all circumstances. I'm just speculating again.

All in all, I don't think overclocking a modern CPU is worth the trouble. They come with pretty much the top frequency they can handle out of the factory. Silicon lottery only gives you +- 100-200 MHz max.

 

[CyberCT]

Agreed on the OC. I want the best performance for least heat/voltage possible in my NZXT H1 case. Coming from overclocking past Intel CPUs ... I don't like hitting the 1.3VCORE range. Leaving everything to auto ... the VCORE jumps up to 1.47 volts and I don't like that.

I think I'm just going to leave the max frequency at 4.6GHz with the lower volts (max 1.305V) and temps (max core low 70s C).

I just don't know why my CPU decides to bump the max all core frequency from 4.6GHz to 4.8GHz when I don't tell it to. I only reduced the max frequency for each core from 5.0GHz to 4.8GHz and here's what it does lol.

These CPUs are completely different animals from the 9700K family.

 

[AusWolf]

Agreed on the OC. I want the best performance for least heat/voltage possible in my NZXT H1 case. Coming from overclocking past Intel CPUs ... I don't like hitting the 1.3VCORE range. Leaving everything to auto ... the VCORE jumps up to 1.47 volts and I don't like that.

I think I'm just going to leave the max frequency at 4.6GHz with the lower volts (max 1.305V) and temps (max core low 70s C).

I just don't know why my CPU decides to bump the max all core frequency from 4.6GHz to 4.8GHz when I don't tell it to. I only reduced the max frequency for each core from 5.0GHz to 4.8GHz and here's what it does lol.

These CPUs are completely different animals from the 9700K family.

That they are.

What I would do is just leave everything on Auto, and play with the PL values until I find a good balance of performance and heat. I usually have a "the CPU knows how much voltage it needs" attitude.

But I like your solution too.

 

[CyberCT]

AusWolf, under your PL value modification solution, what are your core frequencies maxing out at under load and single core (CineBench 2.3 is a good example of testing both)? What max VCORE are you getting for each scenario, and typical voltage range under load?

 

[RJARRRPCGP]

Yep, looks like the 10th gen was the last to be able to OC the way you're used to OC'ing. Rocket Lake, was the start of the post-second-gen-Ryzen-like auto stuff.

 

[plastiscɧ]

dunno if it helps U :

mine runs fine. but a wise @freeagent said to me; if it does not like the overclocking it will tell u. by the way, 5,2GHz was with win 10 in parts impossible. win 11 is rocksolid and doesn't even care 5,3GHz.

ah. and the motherboard tells the following:
the deepest frequency must be two steps beyond the second highest frequency. it tells u the max range u should adjust.
the overclocking is possible as long u can cool it. [Scatter Bencher had it on 5,8GHz until it was uncoolable]

50, 50, 49, 49, 47, 47, 46, 46

50-50-49-49-48-48-47-47 then

i capped the highest possible voltage for the CPU @ 1,5V AND put the voltage he needs while working to the auto setting.

 

[CyberCT]

You have a stellar chip there. Did you run Realbench 2.56's 8 hour stress test w/ half your system RAM to confirm it's stable?

Are you worried about voltage degradation over time, seeing how high that voltage is?

I might try 4.5GHz all core and load to see what the lowest voltage is that I can achieve w/out Realbench tests failing.

I have a Kill-A-Watt meter and tested both my PCs at the outlet.
I ran benchmark tests that test both the CPU and GPU (the Division, Shadow of the Tomb Raider, Steep) with this build (4.6GHz max11700K w/ 3070ti stock) against my other build (9700K 4.8GHz max @ 1.19V and 2080ti max overclock) and the 11700K build performs within 5% of benchmark results of the other build but consumes about 60 watts more power. Somewhat disappointing.

 

[plastiscɧ]

Are you worried about voltage degradation over time, seeing how high that voltage is?

no! cus he will not get that old anyway. if it is damaged i'll buy a new one. my setup will never get older than 2 years anyway.
it is always degeneration with voltage AND heat [OVER YEARS]. my watercooling doesn't let it go over 50°C. max 80°C under hardcore stress.
i am not that cruel person stressing my CPU to look if it is stable under conditions they "never happen". My CPU is stable under conditions I use 24/7.

Did you run Realbench 2.56's 8 hour stress test w/ half your system RAM to confirm it's stable

i am playing with this GHz on 5,2. if it is under stress it goes beyond 5. and i never had issues since 1yr. i put win11 on my system

5% of benchmark results of the other build but consumes about 60 watts more power

it is a real heat-head. i regret getting an 11700k. the 10 was better in every manner

 

[CyberCT]

I'll bet your chip can do 4.8GHz all core at the same voltage mine does 4.6GHz lol. I need a good chip like yours for my build.

I'll keep tweaking. Maybe I'll figure out the better cores (are they the cores with a * next to them in BIOS?) and give just them the 4.8GHz boost while the rest are 4.5GHz. IDK.