Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
Questions about CPU core & input voltages
#1
Hi,

First of all, thank you for this exquisite monitoring application. I've customized all information that I need in a single screen.

I've got asus sabertooth x99 and 5820K OCed to 3.8GHz. HWiNFO64 is v5.06-2640.

1) My stable core voltage is 1.076V. I get different results when I set CPU core voltage to different modes in BIOS. In manual mode, VIDs for all 6 cores in HWiNFO64 are almost the same (1.072V-1.077V), but in offset mode, VIDs for cores #0,#2,#4 are ~1.076V and for cores #1,#3,#5 are ~1.099V. Also, vcore shows 0.128V! Other monitoring applications like CPU-Z and HWmonitor report 1.076V which is the VID value for core #0 in HWiNFO64. Are these readings normal? Or my motherboard has some problems?

2) The other odd thing is that setting the CPU input voltage manually, halves the CPU Package Power (from 150W to 75W under full load and from 18W to 9W when idling), but other things like temperatures, Total DRAM Power, etc are the same.

I would be so much appreciated if you clarify these issues. Thanks in advance.
Reply
#2
1. I don't have an explanation for the voltages on odd cores, it could be a bug in the CPU logic reporting. Vcore value reporting is 'normal' for these series of CPUs (with FIVR) - when idle the measurement logic reports very low numbers. But when you put load on the CPU, it should stabilize at reasonable values.

2. That might depend on whether SVID support has also been disabled. When SVID is disabled, CPU power measurement doesn't work properly.
Reply
#3
(11-16-2015, 06:33 PM)Martin Wrote: 1. I don't have an explanation for the voltages on odd cores, it could be a bug in the CPU logic reporting. Vcore value reporting is 'normal' for these series of CPUs (with FIVR) - when idle the measurement logic reports very low numbers. But when you put load on the CPU, it should stabilize at reasonable values.

2. That might depend on whether SVID support has also been disabled. When SVID is disabled, CPU power measurement doesn't work properly.

Thank you for the reply.

Regarding the first answer, you say it could be a CPU bug and not the mobo, so is this a thing that can be fixed? Can this result in any damages to the components? Also, that reported vcore value (0.128V) was when the CPU is under full load (stress test), when system idles it is 0.016V.

Regarding the 2nd answer, SVID was on Auto and when I enabled it, that issue was solved.
Reply
#4
I don't think it's a mainboard problem - your CPU manages all voltages internally because of the Fully Integrated Voltage Regulator (FIVR).
Actually for this CPU the values for each core should be identical. So Core #0 and #1 should show the same values, because they are the HT units of core.
So it looks like a bug in VID reporting.
Did enabling SVID maybe fix the Vcore reporting as well ?
Reply
#5
(11-16-2015, 07:46 PM)Martin Wrote: I don't think it's a mainboard problem - your CPU manages all voltages internally because of the Fully Integrated Voltage Regulator (FIVR).
Actually for this CPU the values for each core should be identical. So Core #0 and #1 should show the same values, because they are the HT units of core.
So it looks like a bug in VID reporting.
Did enabling SVID maybe fix the Vcore reporting as well ?

SVID enabling only fixed CPU Package Power reporting. Still, odd and even core numbers differ in voltages by 0.02-0.025V and vcore is 0.128V under full load.
Reply
#6
Does ASUS AU Suite confirm the same Vcore value ?
EDIT: Ah, that doesn't matter - ASUS reports VID instead of Vcore in case Vcore is < 0.5V.
So I think the Vcore reporting is flawed too.
Reply
#7
(11-16-2015, 08:07 PM)Martin Wrote: Does ASUS AU Suite confirm the same Vcore value ?
EDIT: Ah, that doesn't matter - ASUS reports VID instead of Vcore in case Vcore is < 0.5V.
So I think the Vcore reporting is flawed too.

Ai suite shows exactly the same.
Reply
#8
I wouldn't worry about those values.
Reply
#9
(11-16-2015, 08:56 PM)Martin Wrote: I wouldn't worry about those values.

If those numbers are in normal ranges that everyone maybe have, I am OK with them too. My concerns are about setting something wrong in BIOS or having some bad hardware which may result in such values.
Reply
#10
The difference (~0.025V) between VIDs for odd and even core numbers, are solved by setting the CPU core voltage to [Adaptive Mode] instead of [Offset Mode]. There is still ~0.005V difference between VID values, but I think it is normal and in the [Manual Mode] it was so. However, vcore is still 0.128V under full load.

HWiNFO64 is a perfect monitoring app. Wink
Reply
#11
Thanks for the feedback.
Vcore monitoring is not working properly and that's why ASUS doesn't report it when it's < 0.5V, but uses VID instead. This is an issue on all CPUs with FIVR (Haswell, Broadwell).
Reply
#12
(11-18-2015, 05:41 PM)Martin Wrote: Thanks for the feedback.
Vcore monitoring is not working properly and that's why ASUS doesn't report it when it's < 0.5V, but uses VID instead. This is an issue on all CPUs with FIVR (Haswell, Broadwell).

I don't get this, "ASUS doesn't report it when it's < 0.5V, but uses VID instead". My AI Suite reports 0.128V. Does this mean my VIDs and/or vcore are less than 0.5V?

I know that vcore is the actual voltage CPU gets. If we don't monitor its actual value, how can we be sure it is not lower or higher than what we adjusted? Is it possible, for example, we set it to 1.1V, but its real value be 1.2V or so?
Reply
#13
Sorry for confusion, I was wrong. I checked again and indeed, ASUS reports the Vcore as HWiNFO does. It's possible that they did report VID instead in the past, but have stopped doing that.
Reply
#14
(11-18-2015, 07:12 PM)Martin Wrote: Sorry for confusion, I was wrong. I checked again and indeed, ASUS reports the Vcore as HWiNFO does. It's possible that they did report VID instead in the past, but have stopped doing that.

Thanks.
Reply


Forum Jump:


Users browsing this thread: 1 Guest(s)