Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
2 entries for Renesas ISL69138 - which one?
#1
Exclamation 
I am trying to understand which entry to trust since the VR OUT (true core voltage instead of vCore) and VR loop 1 (VRM temps) are different.
I've attached an image which I hope helps.
All BIOS settings are stock EUFI defaults - no overclock applied so I can show baseline values.
Board is AsRock Z370 Extreme4, i7 8086K (purchased hand picked 'binned'/delidded)

I'm inclined to believe the second entry as the VR Loop matches the VCC Temp (SVID) on the CPU and the CURRENT (IOUT) are inline with what I would expect (the first entry is clearly incorrect), BUT the VR OUT is higher than the reported vCore by an entire 0.1v and if this is correct then it would mean that my current stable overclock (5.2Ghz @ 1.392 vCore), would be an insane 1.492 VR OUT! 
This would make no sense since P95 temps running custom test with FFT's min and max 1344 and 'run FFT's in place' checked are only in the 50-60C range after 4 hours on just a Corsair AIO cooler and decent case airflow.

Running Cinebench (stock BIOS settings) I see the vCore in HWInfo go up about 0.1v and the VR OUT (on the second entry) go down by about the same 0.1v
This is still concerning that if I apply my overclock with LLC1 (seems to be the recommended by AsRock) that the VR OUT would still be just under 1.5V at idle, and that's really not good I don't think?
Watching DerBauer's 8600/8700K overclocking tutorials, he states that setting a vCore in BIOS upto 1.42v is fine for every day use, but he makes no mention of VR OUT, which seems to be the more important factor?

Also, is it still generally better to use fixed vCore vs adaptive? I don't care about power usage and usually turn all C-States off. I have read that idle voltage is not harmful, just high voltage and high power usage.
So if the VR OUT goes down by say 0.1v under heavy load, then that's the most important thing to consider, and just ignore vCore?

Appreciate feedback as obviously this can make a significant difference in overclocking.
Thanks!


Attached Files Thumbnail(s)
   
Reply
#2
I cannot find a good correlation between those two PWM (VRM) values and anything else in the system. For example the 2nd one minimum VOUT value seems too high to be the Vcore rail. Also the minimum VIN looks pretty off, so it's quite possible that those Renesas controller don't provide accurate telemetry data.
As for the 1st controller, that might be the GT (graphics) part of CPU rail, but that also seems a bit off.
Reply
#3
Not sure I understand your feedback?

In bios I set fixed voltage to 1.395 and LLC1 (most aggressive on this board).
Under full load the vCore I'm HWinfo and cpu-z shows the same value of 1.392v.
The VR OUT value when under full load shows 1.368v and idle 1.42v.

Not sure if that helps?
Reply
#4
What I wanted to say is that I wouldn't trust the values provided by the Renesas controllers as they seem somehow erratic.
Reply
#5
(05-06-2019, 08:34 AM)Martin Wrote: What I wanted to say is that I wouldn't trust the values provided by the Renesas controllers as they seem somehow erratic.

Ok, so just use BIOS vCore and then assume that under load the VR OUT should be less than the vCore given all the detailed explanations on why fixed vCore with aggressive LLC will always be higher idle voltages than under load due to vdroop by 0.01 - 0.02mv it seems in my case even if just viewing the VR OUT sensors in real time. I think someone with an oscilloscope did measurements on my board, will try to find the thread. 
My board has very good power phase/Mosfets and despite it's lower price has been reviewed against boards twice the price and the main differences are features - I don't need 'dancing ninjas' on an LCD display on a motherboard! Smile

After setting my BIOS overclock at vCore (fixed LLC1) 1.395v you can see from the attached image that the vCore barely changes but the VR OUT drops by quite a bit under Realbench -although the average is still within the 0.02v range.
Is this likely due to the sensor not being read properly?

On another note, these Renesas controllers seem pretty common - any chance you will be adding better support for them?


Attached Files Thumbnail(s)
   
Reply
#6
Unfortunately I'm unable to find full documentation for those Renesas controllers. Such documents are usually quite secret.
So some information there is based on more-less accurate assumptions and partial information obtained so far.
Reply
#7
(05-06-2019, 07:28 PM)Martin Wrote: Unfortunately I'm unable to find full documentation for those Renesas controllers. Such documents are usually quite secret.
So some information there is based on more-less accurate assumptions and partial information obtained so far.

Thanks.
I'm emailing AsRock  who said they will open a work order and test with oscilloscope.

I also posted in Reddit to see if 'Elmor' might know something about this.

So you think Ada64 Extreme would show different settings than HWinfo and perhaps coded with these controllers in mind?
Thanks Martin.
Reply
#8
Martin,

Would this oscilloscope work for measuring VR VOUT?
I don't want or need a several hundred dollar lab unit, this one seems to get good reviews.
I'm attaching stats
https://www.amazon.com/dp/B07QS1W96W/ref...il_1?psc=1

Analog Bandwidth: 0 - 200KHz
Sensitivity: 5mV/Div - 20V/Div
Sensitivity error: < 5%
Resolution: 12-bit
Input Impedance: 1M ohm
Maximum Input voltage: 50Vpk
Coupling: DC, AC, GND
Trigger Modes: Auto, Normal, Single
Trigger Types: Rising/falling edge
Trigger Position: 1/2 of buffer size fixed
Display: 2.4 inch color TFT LCD with 320 x 240 resolution
Reply
#9
I don't think that AIDA64 supports this controller and that elmor has experience with this controller either, but I'll check with him directly.
Reply
#10
Thanks Martin!
Hopefully AsRock will confirm with some data from a calibrated oscilloscope.

Thoughts on that inexpensive oscilloscope?
Would it give a reading that would be useful?
Reply
#11
I really don't know, don't have such experience.
Reply
#12
I have just obtained detailed information about this chip.
I was correct in all my assumptions how to properly report information from this chip. So all information shown from this controller is correctly interpreted by HWiNFO, there's no adjustment/fixing needed.
Reply
#13
(05-07-2019, 11:52 AM)Martin Wrote: I have just obtained detailed information about this chip.
I was correct in all my assumptions how to properly report information from this chip. So all information shown from this controller is correctly interpreted by HWiNFO, there's no adjustment/fixing needed.

Thank you for the follow up Martin!

So, the VR VOUT reading is correct:

1.378v at Idle
1.355v under load

I should trust these settings more than the reported vCore of 1.392 (1.395v fixed in BIOS)?

So if I want to keep my actual core voltage under 1.40v, then I can raise the BIOS vCore and monitor the corresponding VR VOUT readings as accurate and reliable?
I can probably start by increasing BIOS vCore to 1.42v, which should translate to approx 1.39v idle VR VOUT and 1.37v VR VOUT under load.

As long as temps are good, this sound good to you?
Reply
#14
Yes, the VR VOUT value should be more accurate than the Vcore value.
Reply


Forum Jump:


Users browsing this thread: 1 Guest(s)