Z370 Bios Asus

[BitMaster]

Thanks Bitmaster for advice.

 

Looking the OC guide I checked the different BIOS option related to Power saving settings. With my OC (game boost), the different settings were disable (EIST, turbo boost, C-state) - but CPU current limit was on auto (it's supposed to decrease clock frequ when current limit is reached).

 

Feeling not confident about changing that I found an other settings that seems to make the tweak : CPU Ratio Offset When Running AVX [Auto]. It was set to [-4] wich correspond to the decrease ratio I have frequently (4.8Ghz down to 4.4Ghz).

I set that to 0 and now the OC don't decrease from 4.8. I have to test with some sims/game but it seems to be the good "button".

 

Be happy when Current Limit only reduces the MHz. The link I gave you tells otherwise, abrupt power cut if the ampere asked for is higher than the limit. I think i have had that sometimes. Sudden OFF, black, dead. Could be different from manufacturer to manufacturer and model to model.

 

Have you solved the zigzag CPU speed ??

[toutenglisse]

Yes no more zigzag - not tested yet ingame with high load on cpu, but since some hours I use my computer there were not a single drop (usually at idle there were always down peaks to 4.4Ghz).

[toutenglisse]

I tested with a quick flight over Batumi with many ground units and a 1080p resolution to highly load the cpu (hyperthreading enabled).

 

As you can see with the afterburner screenshot, the Cpu OC is now completely stable at 4.8Ghz, instead of showing a sawtooth graph like before. Cpu temps at 70°c.

 

So setting the "CPU Ratio Offset When Running AVX [Auto]" to "0" instead of "-4" did the job. This option in Bios only shows up if I enable the "expert explorer" mode.

 

Thanks again for showing to me the right direction.

[BitMaster]

OK, good to hear you solved it.

 

I wonder if your CPU runs a bit too hot, 70°C in DCS is a bit high, iirc mine was at around 50ish °C before delid. Is that the peak temp or average temp ? Average temp would be too high imho.

 

Either Volts are too high, too high LLC, bad TIM between cooler and CPU, bad cooler, its gotta be something.

 

Can you prime or do any stresstest at 5G without the CPU going into throttling ?

 

Happy Tweaking ;)

 

..and ...what is your cooler and what are your volts and LLC and Phase ? Maybe make a screenshot in MSI GUI or such. I bet we can lower that 70°C by a good bit.

[toutenglisse]

You know it's an air cooler (pure rock) so it can't be really better with actual status.

When I turn hyperthreading off and load 4 cores at 100% the max temp stabilize at 75°c (with the actual winter low temps, in summer it's easy 85°c or more iirc).

The OC is auto from the motherboard, Vcore is 1.296V at 4.8Ghz - I can't tell for LLC/Phase ?

[BitMaster]

LodLineCalibration and Phase Control ( ampere ). usually found in VRM settings, CPU Current etc...

 

Those temps are rather high tbh.

 

What is the PEAK VOLT measured in HWinfo after 20min priming or OCCT ?? Take that value as your mark.

[toutenglisse]

Here 2 screenshots of CPU-Z stress test. First with hyperthreading off after 13 minutes stress, and second with hyperthreading on (higher temps) after 5 minutes stress.

 

Of course, temps are lower when using DCS VR (55°c max), but the "warm" (70°c) or "high" (80+°c) temps I see when using high cpu impacting games (usually 70-75°c max), I guess are not unnatural for a 7700K oc'ed at 4.8Ghz with air cooler ("cheap" !).

 

From what I understand from intel charts these are normal temps for highly used stock 7700K air cooled (non OC) - intel says (I think exaggerating) that temps below 100°c are safe with throttling at 100°c (I prefer to see Nothing above 90°c max)

[BitMaster]

Heat does not kill any CPU anymore, they shut off with a safety circuit. What ruins your CPU over time is TOO HIGH VOLTAGE.

 

Therefor, install HWinfo and check what the VID is, make a screenshot like this one:

[toutenglisse]

I made 2 screenshots : the first just before ending the 26minutes stress test (Cpu-Z) and the second just after stopping the test.

[toutenglisse]

With these VID it seems to be easy to lower my voltage in manual or adaptative and get less temps.... I have to try

[toutenglisse]

Not sure - here another screenshot of HWinfo when stress-testing with stock values (Cpu in turbo mode 4.5Ghz on all cores) : VID are higher than with game mode (OC 4.8Ghz) on.

 

Edit : setting Vcore 1.22 in bios (adaptative) - don't know yet about stability but after 4-5 minutes quick test the temps are 10°c lower at 4.8Ghz 100%load.

Edited March 4, 2018 by toutenglisse

[BitMaster]

you can lower the volts by 0.10 while you stresstest, go down until it fails, then go 1 up and you should be close to the sweetspot.

 

More LLC can help to reduce the vcore by as much as 0.10v as in my case. It must lower vcore as it adds heat itself, baseline should still be less total heat ( WATTAGE, heat = wattage, U x I ).

 

You could delid that thing. I can send you my stuff and you send it back when done ?

France = EU, no big deal

[toutenglisse]

Lol thanks Bitmaster - I've watched a video about delidding - it seems absolutely doable but I don't want to go there, at least for now and next guaranty year.

 

It's already fun and usefull to see that settings can be optimized with voltages and températures.