AB350 Pro4 incorrect manual Vcore

Message Topic Search Topic Options Post Reply Create New Topic Printable Version Translate Topic

   So i discovered some strange behavior with this board when setting manual Vcore.If i set Vcore to 1.35v, bios will report ~1.33v. But once i get to windows, hwinfo will report ~1.38v.
If i set vcore to 1.1875v (stock voltage) both bios and hwinfo will report ~1.05v.
even more perplexing, if i sleep the PC and wake it back up, 1.38v will drop down to the correct 1.35v, but 1.05v will never go up to the correct 1.1875v and remains unchanged.
this happens whether i use a fixed OC or p-states. i have tried changing every setting i can think of (CnQ, core boost, frequency, etc) to no effect.
is this normal? what is happening?

i'm using a r7 1700, Win10 x64, P3.30 bios, version 17.30 AMD chipset drivers, and a corsair 750w gold rated PSU.
CPU-Z full specs: https://valid.x86.fr/r48jzm

none of the other voltages are off, only vcore. sleeping the pc "fixes" it when the voltage is high, but it needs to be done after every boot. the fact that it drops after a sleep cycle is especially annoying since it will destabilize an otherwise stable OC.

any insight is appreciated, thanks.

Author	Message Topic Search Topic Options Post Reply Create New Topic Printable Version Translate Topic
php42 Members Profile Send Private Message Find Members Posts Add to Buddy List Newbie Joined: 14 Dec 2017 Status: Offline Points: 23	Post Options Post Reply Quote php42 Report Post Thanks(0) Quote Reply Topic: AB350 Pro4 incorrect manual Vcore Posted: 14 Dec 2017 at 8:27am
	So i discovered some strange behavior with this board when setting manual Vcore. If i set Vcore to 1.35v, bios will report ~1.33v. But once i get to windows, hwinfo will report ~1.38v. If i set vcore to 1.1875v (stock voltage) both bios and hwinfo will report ~1.05v. even more perplexing, if i sleep the PC and wake it back up, 1.38v will drop down to the correct 1.35v, but 1.05v will never go up to the correct 1.1875v and remains unchanged. this happens whether i use a fixed OC or p-states. i have tried changing every setting i can think of (CnQ, core boost, frequency, etc) to no effect. is this normal? what is happening? i'm using a r7 1700, Win10 x64, P3.30 bios, version 17.30 AMD chipset drivers, and a corsair 750w gold rated PSU. CPU-Z full specs: https://valid.x86.fr/r48jzm none of the other voltages are off, only vcore. sleeping the pc "fixes" it when the voltage is high, but it needs to be done after every boot. the fact that it drops after a sleep cycle is especially annoying since it will destabilize an otherwise stable OC. any insight is appreciated, thanks.

johndms Members Profile Send Private Message Find Members Posts Add to Buddy List Newbie Joined: 15 Sep 2017 Status: Offline Points: 24	Post Options Post Reply Quote johndms Report Post Thanks(0) Quote Reply Posted: 14 Dec 2017 at 9:21am
	As a test, set it to 1.35v then run a stress test like Aida64 stressing CPU/FPU/Cache for 10-15 minutes while watching HWiNFO64. How low does your cpu core voltage drop? Mine, with Ryzen 5 1600, has dropped to as low as 1.256v. Nearly a whole 0.1v from what is set in the bios. Not that software voltage readings are known to be terribly accurate, but it is still a concern. Welcome to the budget side of the B350 spectrum. WTB LLC. Edited by johndms - 14 Dec 2017 at 9:22am

php42 Members Profile Send Private Message Find Members Posts Add to Buddy List Newbie Joined: 14 Dec 2017 Status: Offline Points: 23	Post Options Post Reply Quote php42 Report Post Thanks(0) Quote Reply Posted: 14 Dec 2017 at 10:18am
	i ran a prime95 small FFT stress test before and after a sleep cycle vcore and core vid: before sleep: vid: 1.375 -> 1.3 vcore: 1.38 -> 1.269 after sleep: vid: 1.35 -> 1.275 vcore: 1.35 - > 1.237 vcore here is core voltage as read from the CPU, motherboard vcore reports roughly the same but seems like it's lower resolution than the CPU sensor the v-droop is less at stock settings but it will go under 1.0v i was aware of the v-droop and lack of LLC, what concerns me is this weird behavior after a sleep cycle and the high voltages needed at boot so that it doesn't crash after sleeping the voltage drop after sleeping is permanent. i had to reboot my pc to perform this test.

parsec Members Profile Send Private Message Find Members Posts Add to Buddy List Moderator Group Joined: 04 May 2015 Location: USA Status: Offline Points: 4996	Post Options Post Reply Quote parsec Report Post Thanks(0) Quote Reply Posted: 14 Dec 2017 at 11:24am
	The "VCore" voltages you are reporting, are they VIDs or the true VCore reading? HWiNFO64 shows both with my ASRock X370 board. What other programs display for VCore is random, could be VID (most likely) or VCore. What is the VCore mode set to in the UEFI? Fixed voltage? That would be the only way to maintain a constant VCore, if that is what you want to achieve. For a fixed, constant VCore, is Cool n Quiet and C6 C State set to disabled? There is another Global C6 C State option in one of the the Advanced\AMD CBS screens. Windows Power plan set to Ryzen Balanced (not really balanced) or High Performance? Lack of LLC and VDroop will cause VCore reduction under various loads, which you know. Software readings at the 1/100th of a volt level, how accurate can they be? I use Offset VCore mode, with all power saving options enabled, but the Power Plan set to High Performance. The core speeds remain at my (LOL) OC of 3.9GHz, but VCore varies from 1.360V to 0.384V according to HWiNFO. The VID readings are close to this, and the VRM output voltage is 1.35V, +0.05V with LLC set to 2. HWiNFO cannot show AM4 C State levels as it can with Intel processors, but I'm sure I'm down to C6 at idle/low loads. The changes in VCore when waking from Sleep, among many other CPU related issues (changes of core multipliers settings is the most common) are classic processor register restoration issues that have been seen on all platforms for years, Intel or AMD. They can be difficult to fix, since Windows only helps with that to a degree. Given the major changes necessary to the UEFI/BIOS code that the AGESA 1.0.7.2 update from AMD require (thanks AMD ) we will be basically starting over again when those versions are released. Once they are released (that started but was stopped, too many issues) we can see what happens with the VCore, etc, and go from there.


php42 Members Profile Send Private Message Find Members Posts Add to Buddy List Newbie Joined: 14 Dec 2017 Status: Offline Points: 23	Post Options Post Reply Quote php42 Report Post Thanks(0) Quote Reply Posted: 14 Dec 2017 at 12:12pm
	the Vcore i'm reading is listed in hwinfo as "CPU Core Voltage (SVI2 TFN)" the motherboard "Vcore" reading reflects roughly the same values, at a lower resolution i also gave the individual core VIDs for that test, but those are not what i was referencing in the OP i have the cpu frequency and voltage set to manual in the "OC tweaker" page at 3.8Ghz 1.35v, so it should be fixed voltage. i also tried modifying p0 directly with the same results. c6 and CnQ are disabled. changing power plans has made no difference, i tried all of them. i don't believe the AB350 pro4 bios has a vcore offset mode, i looked for it since p-state overclocking doesn't seem to work correctly without it. all of the power saving features should be disabled, it doesn't downvolt or downclock at idle. my concern was that that 0.03v drop after waking from sleep has actually been enough to crash the system during long rendering/BOINC sessions, and means that i either need to set the vcore higher than necessary to keep it stable, or i can't sleep my pc ever. but it sounds like i might just be waiting for a bios update then :( what i found weird was that it boots up with a +0.03v offset from the bios settings and then "fixes" itself after a sleep cycle EDIT: i went back and made sure everything was disabled including "global c-state control" and made a comparison before and after a sleep cycle, bios set to 1.35v Vcore i'm not sure what's going on with the cpu core voltage reading anymore Edited by php42 - 14 Dec 2017 at 1:49pm

delukz Members Profile Send Private Message Find Members Posts Add to Buddy List Newbie Joined: 27 Jun 2017 Status: Offline Points: 32	Post Options Post Reply Quote delukz Report Post Thanks(0) Quote Reply Posted: 14 Dec 2017 at 8:24pm
	I wouldn't worry too much about it, in all my years of PC building, i've never owned one that runs at exactly the voltage you set it. I've owned boards from Asus, ASRock, MSI, Gigabyte, Epox, ABIT, Shuttle, ECS, Jetway, DFI et etc. And from as far as I remember, they all show the same behaviour. Edited by delukz - 14 Dec 2017 at 8:28pm

zlobster Members Profile Send Private Message Find Members Posts Add to Buddy List Groupie Joined: 02 Sep 2017 Status: Offline Points: 403	Post Options Post Reply Quote zlobster Report Post Thanks(0) Quote Reply Posted: 14 Dec 2017 at 8:30pm
	This diff is way too small to bother you. It could easily be attributed to the measurement error. More important are the crashes after 'sleep'. Did you disable C6 state?
	1700X ZP-B1 (stock); X370 Taichi (UEFI 3.10); 16GB F4-3200C14-8GFX XMP; 256GB 960 EVO; RX 580 NITRO+ 8GB

php42 Members Profile Send Private Message Find Members Posts Add to Buddy List Newbie Joined: 14 Dec 2017 Status: Offline Points: 23	Post Options Post Reply Quote php42 Report Post Thanks(0) Quote Reply Posted: 15 Dec 2017 at 12:43am
	C6 is disabled. it gets worse when i enable c6 and CnQ so i'm guessing it's a problem with c-states. i originally had them enabled so i'll try leaving them disabled and see what happens

zlobster Members Profile Send Private Message Find Members Posts Add to Buddy List Groupie Joined: 02 Sep 2017 Status: Offline Points: 403	Post Options Post Reply Quote zlobster Report Post Thanks(0) Quote Reply Posted: 15 Dec 2017 at 3:24am
	I honestly have no idea how CnQ, C-states, P-states and other stuff relate to each other. More so, if you disable some sleep/power settings in UEFI but Windoze donesn't recognize they are turned off - here is another recipe for disaster. My PCs never sleep or hibernate. That's why I rely on CnQ and native power management. My belief is that manufacturers should make sure mutually exclusive options are handled internally, so users won't screw up things even if they wanted to. P.S. I nowadays recommend DOS re-flash of what's deemed to be a stable UEFI version. Although, I've never personally tried it yet, many ppl here claim it works magic for them.
	1700X ZP-B1 (stock); X370 Taichi (UEFI 3.10); 16GB F4-3200C14-8GFX XMP; 256GB 960 EVO; RX 580 NITRO+ 8GB