H115i PRO Sensors

[DevBiker]

Well, keep in mind that 4.6Ghz is the single core "boost". The max core at full load is 3.2Ghz. With MCE disabled, you run at the stock frequencies - so you'd have to enable MCE to see 4.6-4.7 across all cores. I run with MCE enabled but keep my vCore in mind.

[TheTiesThatBind]

Yeah vCore got some killer voltages. I'm seriously thinking of tweaking CPU Freq to about 4.5 All core with manual voltages to protect the longetivity mostly. Since I'm covered on my daily duties. With MCE Enabled, I need All Cores sync to see 4.7 across all cores. On Auto it stays at 4.4 idle.

 

But why would it underclocked under load or crash the test?

 

I mean when I activate XMP Profile, I choose No for the enchanced performance and Ai OC changes to XMP on Extreme Tweaker and CPU RATIO changes from Per Core (Default) to AUTO.

[c-attack]

The thing is that with MCE disabled and SVID Support on AUTO, after a few seconds into the Stress test (AIDA64, Prime95, OCCT) the CPU underclocks itself even below the stock speeds (!).

 

I spent a lot of time today playing today and temps didn't go beyond 60 (MAX), Average 45~53-55) and freq was 4300-4400 Mhz. I mean even when I try to stress the CPU it won't reach anywhere close to 4.7, with or without CME, which seems odd. I remember being able to run a stress test with steady frequenices on my AORUS Gaming 7 which had hot VRMs compared to Maximus X. I don't think it's a MOBO issue, I just can't figure out what's keeping the CPU 'back"

 

Go into the Advanced BIOS (F7) and the Extreme Tweaker column. Scroll down until you find the Internal CPU Management Sub-menu:

 

Long Duration Package Power Limit = set 4095

x

x

Short Duration Package Power Limit = set 4095

 

Go back to the Extreme Tweaker Menu. Scroll down just a bit more to...

 

CPU Core /Cache Current Limit = set 255.50

 

 

Surprisingly, you can hit the auto limits very easily on 8700K. I went through a week thinking AIDA was broken, before realizing I failed to reset these values after a BIOS update. All the board safety features are still in play. It just won't throttle so easily. Essential on my 5.0/1.30v settings.

[TheTiesThatBind]

5.0GHz/1.3v? Wow! This is on delid right?

Thanks for the info, i will re-configure it as soon as possible and check how it goes.

By the way do I have to change SVID to Best Scenario or Typical Scenario instead of having it run on AUTO (which is the worst scenario)? Also SVID runs on Auto as well, do I change it to enabled?

Also what about CPU Ratio, MCE? Do I leave CPU Ration on AUTO and disable MCE?

 

Also my BIOS version is 1602 (the latest) which was already installed on the mobo (guess newer models come with the latest BIOS). I'm running ASUS Maximus X Hero as I said, I changed from Gaming 7, F7 bricked my motherboard through AORUS APP Center (but that's another topic :D).

 

As I said I don't care for manual overclocking, I just want my CPU to run at Intel's stock settings without any issues (including the Turbo Boost) .

[TheTiesThatBind]

I ran OCCT but after almost 11 minutes it stopped as Core #3 got over maximum value max temp was 89 on Core 3, the rest 85, 85, 82 and 87.

 

BIOS Configuration:

XMP -> On

Enhanced performance -> No

ASUS CME -> Disabled

SVID -> Best Scenario & AUTO

CORE RATIO: AUTO

 

PLUS:

Long Duration Package Power Limit = set 4095

Short Duration Package Power Limit = set 4095

CPU Core /Cache Current Limit = set 255.50

 

Meanwhile, I tried a full load optimize settings, I kept everything at stock and the stress test appeared to work fine. Even PUMP's Temp remained between 31.6-31.7C.

 

http://i.imgur.com/EdLZe9u.jpg

[c-attack]

OK, good... no more throttling. Now you just have the normal 8700K problem - HOT. Short term fix is 1.26-1.27 vcore is probably more than ample for 4.3 GHz. You can either bump up the multiplier or reduce it to cut temperature. Long term solution is the delid and repaste process. It typically takes 20-30C off the top of your CPU temps. My peaks are 25C lower than you while at the higher vcore and frequency.

[TheTiesThatBind]

Hello again, thanks for your advices c-attack. Well I tried to approach the issue differently (almost). I did a Load Optimized Settings and followed der8auer steps for OCing as I'm new to it and I did the following steps.

 

-Activated XMP of Vengeance Led 3000/15 (Select No)

-Disabled CME

-Disabled SVID

-AVX to 0 (Should I put this to 1 or 2 offset?)

-Sync All Cores x47 Multiplier

-CPU Load Calibration to Level 6

-Set the values that you also suggested to the in the Internal CPU Power Management

-Same for Core/Cache

-Set Minimum/Max CPU Ratio Cache to 42

-Set CPU Core Voltage on Manual at 1.26V

-VCCIO to 1.1.

-System Agent to 1.1.

 

Here are my results from Cinebench R15 on those settings as well as a the results from Prime 95 after 1 hour and 9 minutes of stress test on "real-world scenario" according to the video as it didn't allow AVX to kick in.

 

http://i.imgur.com/lpVo9GE.jpg

http://i.imgur.com/DWWi2fJ.png

 

Are my results/thermals good now? Also is my system stable? My PUMP Sensor's Temp peaked at 33.1, about 80% fan speed.

 

Few last questions. Shall I change now from Level 6 to Level 5 and re-run the test to check for stability and keep it that way or even aim for a 1.25V and re-run the tests before trying to change level from 6 to 5 or greater? Also should I set an offset value to AVX at about 4.5GHz if it kicks in?

 

Looking forward for a reply

 

Kind regards!

[c-attack]

I don't see a reason to use the negative AVX offset unless you regularly use programs with AVX instructions that make the cores run too hot, but still want the higher frequency for other tasks. Most people do not utilize AVX types of programs. It will kick in when running most of these stress test programs, so that kind of defeats the purpose for the multiplier you are attempting to test.

 

I like LLC 6 and this what I use for 5.0/1.30v. It keeps a tight leash on the Vcore with my normal full load rise at 1.316. I am not as stable at LLC 5. You could drop down and retest, but this is one of the last variables I would lower until you have everything else where you want it. 5 or 6 likely has little effect on your day to day stuff, except of course when the system crashes.

 

All your other settings look good to me and are similar to mine or appropriately lower for the related frequency. If I am reading that right and your new peak core range is 67-73C at 4.7GHz, then it looks like you are in good shape. As for stability, let it ride and go do your normal stuff. If you start to get regular BSODs, then we'll need to go back and look at a few things.

[TheTiesThatBind]

Well after playing games all day long the OC so far nothing out of the ordinary occurred so I guess things are stable. I won’t touch anything for now and see how it goes. System appears stable. I will re-run all night the Prime95 test with AVX off, since I don’t use any avx apps

 

Thank you all of you for all of your help, means a lot! Hopefully I can get a stable oc now and keep a peace of mind

[TheTiesThatBind]

Quick question, sometimes when I shut down my PC the MOBO's leds stay on, is this a hint of possible instability?

 

I have everything set to off through BIOS about the RGB leds in soft-off etc modes.

[DevBiker]

Quick question, sometimes when I shut down my PC the MOBO's leds stay on, is this a hint of possible instability?

 

I have everything set to off through BIOS about the RGB leds in soft-off etc modes.

 

Not at all.

[c-attack]

Quick question, sometimes when I shut down my PC the MOBO's leds stay on, is this a hint of possible instability?

 

I have everything set to off through BIOS about the RGB leds in soft-off etc modes.

 

Absolutely proves instability --- of the people who write the Aura software. The newest release borked the off controls. Mine is turned off at the motherboard and software level and it still turns red on its own every couple of sleep/shutdowns. The newest one also broke the ability to set a basic lighting pattern when Aura is not running (solid, breathing, or rainbow). It now does rainbow all the time, unless that is what was meant by better synchronization with 3rd party applications. It now matches iCUE. You can roll back to Aura 106.17 if the issues are persistent.

[TheTiesThatBind]

Thank you for your replies! Yes C-Attack, it happens every couple shutdowns (I don't use sleep mode). Thankfully, it is static red as I set it.

 

Also I figured out what caused the underclocking from the XMP profile (ASUS likes to make things complicated on some matters I guess).

 

So I all had to do was to set the timings of RAM manually and keep MCE on. I set SVID Support to Best Case Scenario and I also changed VCCIO to 1.1. But the thing is during a stress test VRMs temp on iCUE appear at 127C! While at HWInfo its 19 ? On idle, in iCUE sits at 0. But when I use my OC profile which includes XMP 2.0 profile activated and not manual ram timings and everything else is also manually set, the Temp #5 sensor works fine and displays 45-46 Celcius idle and under load, while temp #2 reaches about 60 degrees

 

Here's a picture:

 

Any ideas?

 

Also, despite the RAM Freq. being about 1500-1507, in SPD table from CPU-Z it says "2998 MHz" is anything wrong with my RAM or what? One more thing that I noticed is that when XMP is set on by MOBO the Ram Freq is between 1499.3-1501.1 while on manual timing set is above that (maybe its because I have Maximus Tweak on Auto on the manual and not set at Mode 1?)

[c-attack]

"DDR"(3/4) is Double Data Rate and how we commonly refer to specific memory kits (2666, 3200, 4000 MHz, etc). Some programs still give the single data rate or half the box value. Same thing either way. 1500 is the single data rate. 3000 MHz is the double data rate.

 

One thing iCUE and Link have never been good at is reading the motherboard data from Asus boards (and others). You pretty much need to ignore everything in the motherboard temp box. They don't even hold the same places. What was temp 2 last time might be #3 on the next boot and vice versa. Every time I think I have one pinned down, in comes back with an impossible value the next time. Temp 5 is notorious on Asus boards for being either 0C or 127C. Neither is possible. Permanent 17C is another classic. Regardless, don't use this for monitoring critical motherboard information. On the bright side, your VRM temp should not be on the critical watch lists. VRM temp is extremely low on the Asus Z370 boards even when running hard. I considered getting the Formula for quite a while so I could water cool it. That would have been a waste. My Code never breaks 50C VRM at 5.0/1.30, LLC 6, no matter how long I run. If you want to see accurate values for PCH, VRM, etc., try HWInfo64 in conjunction with iCUE. Use the "portable" version that recalls the driver when you exit and on first launch it will give you two warnings. One will be about the Asus EC sensor. Ignore this and let it read the values (this is what you are after). The second will be about the Corsair sensors. This is you need to mind and select the option not to track Corsair data. All of that will be in iCUE, but HWINFO will give quite a bit of detail not available there. A lot of it is superfluous, but its fun to watch it all for a while on a new system.

 

XMP on or off should have no bearing on the reported temp values. It is most likely you are seeing the usual variance, but who knows. Mode 1 and Mode 2 in the manual DRAM timing will provide a tighter set of secondary and tertiary timings. You can probably use Mode 2 at 3000, but it will depend on all the usual factors regarding module, voltage, primary timings, etc.

[TheTiesThatBind]

Well my question is why on SPD page it says 2998 instead of 3000, I mean that’s my concern, but thank you for your detailed description! So all I have to do is to re-install the HWInfo64 and allow it to read ASUS sensors but not Corsair’s?

[c-attack]

Well my question is why on SPD page it says 2998 instead of 3000, I mean that’s my concern, but thank you for your detailed description! So all I have to do is to re-install the HWInfo64 and allow it to read ASUS sensors but not Corsair’s?

 

I find HWINFO to be useful for the extra details not shown in iCUE. There will be a box with all your board info, plus all kinds of other things you may or may not be interested in.

 

DRAM frequency varies a bit. Whether this is natural jitter, sensor accuracy or something else, I don't know. Mine is set 3466 and the single data rate fluctuates between 1732.2 and 1734.3. That is consistent with the last 3-4 DDR4 systems I have owned. I am afraid I can't offer a more detailed explanation other than "it's normal".

[TheTiesThatBind]

It's normal is enough for me :D Thanks once I again, yeah I've seen the section that you mention since I also use HWInfo but I had both ASUS/Corsair disabled, so I will perform a re-install and set it on

[DevBiker]

You can enable the Asus. If you are running iCue, keep the Corsair disabled.

[TheTiesThatBind]

Thank you Dev, I got it :D

[TheTiesThatBind]

"DDR"(3/4) is Double Data Rate and how we commonly refer to specific memory kits (2666, 3200, 4000 MHz, etc). Some programs still give the single data rate or half the box value. Same thing either way. 1500 is the single data rate. 3000 MHz is the double data rate.

 

One thing iCUE and Link have never been good at is reading the motherboard data from Asus boards (and others). You pretty much need to ignore everything in the motherboard temp box. They don't even hold the same places. What was temp 2 last time might be #3 on the next boot and vice versa. Every time I think I have one pinned down, in comes back with an impossible value the next time. Temp 5 is notorious on Asus boards for being either 0C or 127C. Neither is possible. Permanent 17C is another classic. Regardless, don't use this for monitoring critical motherboard information. On the bright side, your VRM temp should not be on the critical watch lists. VRM temp is extremely low on the Asus Z370 boards even when running hard. I considered getting the Formula for quite a while so I could water cool it. That would have been a waste. My Code never breaks 50C VRM at 5.0/1.30, LLC 6, no matter how long I run. If you want to see accurate values for PCH, VRM, etc., try HWInfo64 in conjunction with iCUE. Use the "portable" version that recalls the driver when you exit and on first launch it will give you two warnings. One will be about the Asus EC sensor. Ignore this and let it read the values (this is what you are after). The second will be about the Corsair sensors. This is you need to mind and select the option not to track Corsair data. All of that will be in iCUE, but HWINFO will give quite a bit of detail not available there. A lot of it is superfluous, but its fun to watch it all for a while on a new system.

 

XMP on or off should have no bearing on the reported temp values. It is most likely you are seeing the usual variance, but who knows. Mode 1 and Mode 2 in the manual DRAM timing will provide a tighter set of secondary and tertiary timings. You can probably use Mode 2 at 3000, but it will depend on all the usual factors regarding module, voltage, primary timings, etc.

 

I tried to use MODE 2 but I got a BSOD after a stress test, code was about Memory related, so I reverted back to Mode 1

[c-attack]

Then I would leave it on Mode 1 for the time being. You aren't going to get any noticeable real world performance from the bump up. So many other things to work on and measure.

[TheTiesThatBind]

Changed my PUMP from Extreme to Balanced cause noise started to annoy me while on idle/browsing :D

[c-attack]

Unless you have a lot of background noise, 3000 rpm is going to be audible. My $190 D5 is also easily heard at that speed. There should be no performance loss at the balanced 2100 rpm. Just leave it there.