Close

Register Now!

To Get More Info and Daily Reward.

Please login or register.
Pages: [1]   Go Down

Author Topic: Experiencing overclocking for the first time.need guidance(B350M Mortar+R5 1600)  (Read 4489 times)

0 Members and 1 Guest are viewing this topic.

blackr92Topic starter

  • CORPORAL
  • ****
  • Offline Offline
  • Posts: 18

Spec
Ryzen 5 1600 @stock cooler
MSI B350M Mortar
Kingston Value RAM 2400mhz
GTX 960
1TB WD HDD
SuperFlower Leadex 550W 80+ Silver

Core Performance Boost disabled/
Core C6 State disabled/
Spread Spectrum disabled/

AMD CoolnQuiet disabled

1st attempt 3.80 Ghz @ 1.3250V (also CPU NB/SOC VOLTAGE 1.0750V)

Prime95 load temp was at 86c, first 10 minutes was fine than some my threads starts to fail (CPU#2 and CPU#4 is no longer at 100% utilization compare to others it just fluctuates) just by looking at HWmonitor. Around 16 minutes mark my screen went black, also mouse/kb not working but PC still running/Fan still spinning.

Aida64  load temp was at 85c, after 7 minutes same case as Prime95 my screen went black, also mouse/kb not working but PC still running/Fan still spinning.

after the 1st attempt i just revert everything back to default in the bios cause i was a little bit scared..

2nd attempt going on right now 3.70 Ghz @ 1.3V

Prime95 load temp was at 87c, first 2-3 hours was fine than one of my threads starts to fail (CPU#3 is no longer at 100% utilization compare to others it just fluctuates). everything seems stable no black screen,mouse/kb also was working.

Aida64  load temp was at 84c, tested for 2 hours everything seems stable no black screen,mouse/kb also was working.

For the 2nd attempt i think it seems stable other than that one thread that stop utilizing at 100% in Prime95 test compare to the other thread, how should i proceed from here?
Logged

electrostingz

  • MAJOR
  • *****
  • Offline Offline
  • Posts: 1011
« Reply #1 on: 02-November-17, 21:08:49 »

Hello,

What are your default temperatures when stress testing? (3.2Ghz / no overclocking & lowest stable Vcore)
And is it stable with all cores 100%.
Logged
CPU: Intel i7-6700K | MOBO: MSI Z170A Krait Gaming | RAM: Corsair Vengeance LPX 32GB (4x8GB) 3200 DDR4 | GFX: Gigabyte GTX 970
CASE: Zalman Z11 | FC: Zalman MFC1-Plus | HS: Zalman CNPS10X | PSU: Coolermaster V850 Gold
NVMe: Samsung 960 EVO 250GB | HDD: WD Blue 1TB x2 | ODD: Sony DVDRW | OS: Win 10 x64

blackr92Topic starter

  • CORPORAL
  • ****
  • Offline Offline
  • Posts: 18
« Reply #2 on: 02-November-17, 21:50:43 »

Hello,

What are your default temperatures when stress testing? (3.2Ghz / no overclocking & lowest stable Vcore)
And is it stable with all cores 100%.
Hey i've never stress test my system on default setting on my own, but i can say my idle is around 40-46C and my highest temp so far during gaming is around 65Cish around there.

The only thing i knew when i got my pc is that the shop that build it for me done their stress test using Aida/Prime cause it was installed when i got it.
Do you want me to revert all settings to default and do it and get back to you?

i tried 3.7ghz @ Auto Cpu Voltage
CPU NB/SOC Voltage 1.100V
CPU LLC to mode 2
Enable over voltage protection and over current protection.

Prime95 stable for 3 hours but my load temp hit 86C (my country quite hot, i live in SEA)
Aida64 stable for 4 hours load temp hit 85C-86C but there will be mini froze happens randomly where it locks my mouse cursor and keyboard (do i need to change any settings to prevent this?)
Cinebench runs fine too but mini froze happens here too.
(somehow after i restarted my pc, when i run cinebench that mini froze doesnt happen anymore. Haven't try Aida yet after restart pc)

After that restart i ran cinebench/aida no froze at all. So i made a conclusion the problem could be because of my Cpu voltage is at Auto.

So now i changed from 3.7Ghz@Auto cpu voltage to 3.7Ghz@1.3V.
So far i ran 1 hour each of Aida/Prime/and few run on Cinebench looks stable so far even during gaming(PUBG/Overwatch). My idle temp around 41-47C and temp during gaming highest was close to 70C.

Whats is the next step you think i should take?
Logged

electrostingz

  • MAJOR
  • *****
  • Offline Offline
  • Posts: 1011
« Reply #3 on: 02-November-17, 22:09:29 »

Try to get your temperatures lower and yes it would be a good idea to run the stress tests without overclocking to see if it is stable.


See the yellow box, this is the VRM area.
When the CPU is under load it also gets hot and without enough airflow it can cause stability issues.
If you do not have a infrared thermometer to measure this heat sink (inside the yellow box) you must at least be able to hold your finger on it at all times when it is under load.
If not sure place a fan directly over when stress testing.

The area in the green box is the chipset  (very basic IO for the AMD chip) heat sink.
Again if this one gets hot overtime it can cause issues.

VRM heat sink and chipset heat sink should be maintained close to 60 Deg C (external temp when using a infrared thermometer).
Place fans over if you are not sure they are causing an issue with your OC.

CPU temps whilst it can run close to 90 Deg C this is the extreme limit,  you should aim for 75 Deg C MAX.
A better CPU cooler would be required.
Logged
CPU: Intel i7-6700K | MOBO: MSI Z170A Krait Gaming | RAM: Corsair Vengeance LPX 32GB (4x8GB) 3200 DDR4 | GFX: Gigabyte GTX 970
CASE: Zalman Z11 | FC: Zalman MFC1-Plus | HS: Zalman CNPS10X | PSU: Coolermaster V850 Gold
NVMe: Samsung 960 EVO 250GB | HDD: WD Blue 1TB x2 | ODD: Sony DVDRW | OS: Win 10 x64

blackr92Topic starter

  • CORPORAL
  • ****
  • Offline Offline
  • Posts: 18
« Reply #4 on: 02-November-17, 23:13:13 »

I will revert back to default and see hows my stability/temp on those settings. i will post it here when i get it done

regarding the VRM temps i did try to touch it with my finger i can say its running hot i could only hold my finger there prolly 7secs.
This are the 3.7Ghz@1.3V/ CPU NB(Soc voltage) 1.100V oc settings that im currently on.


If i were to invest on a new air cooler is tower air cooler(something like cm 212 that i have but no mount bracket) better or the low profile cooler that have fan blows down air straight down to the board that will also pass through vrms?
other than that at stock cooler what can i do to maintain 3.7Ghz and still have good temps and stability?

thanks again
Logged

electrostingz

  • MAJOR
  • *****
  • Offline Offline
  • Posts: 1011
« Reply #5 on: 03-November-17, 00:46:41 »

The tower coolers work well due to their general size and the length of the heat pipes.

If the PC is not for show then something like this can help.


For looks, a case with top mounted fans to blow air inside over the VRM area.
A rear Fan directly behind.

But if the stock cooler cannot cope with your temps it will never be stable.
To maintain 3.7Ghz you can try to get the Vcore lower.
Your screenshot shows 1.344 - 1.352.
Aim for 1.300v - 1.300v with some kind of VRM cooling.
Logged
CPU: Intel i7-6700K | MOBO: MSI Z170A Krait Gaming | RAM: Corsair Vengeance LPX 32GB (4x8GB) 3200 DDR4 | GFX: Gigabyte GTX 970
CASE: Zalman Z11 | FC: Zalman MFC1-Plus | HS: Zalman CNPS10X | PSU: Coolermaster V850 Gold
NVMe: Samsung 960 EVO 250GB | HDD: WD Blue 1TB x2 | ODD: Sony DVDRW | OS: Win 10 x64

blackr92Topic starter

  • CORPORAL
  • ****
  • Offline Offline
  • Posts: 18
« Reply #6 on: 03-November-17, 01:04:51 »

Yeah i have the tower cooler on my 2nd pc thats the Hyper 212 Evo but im waiting for the am4 bracket at the moment.

was looking at a few coolers that blow the air down to the motherboard and i found this


the guy have the same motherboard as me and he even said it that his orientation of the cooler was to also blow some air towards the VRMs.
Talking about maintaining a stable and good temp for my 3.7Ghz OC, you mention to go for lower Vcore atm im at 1.300V in bios (i dont know why its at 1.344 - 1.352V in HWmonitor)

some pics of the setting in bios





Logged

electrostingz

  • MAJOR
  • *****
  • Offline Offline
  • Posts: 1011
« Reply #7 on: 03-November-17, 14:07:42 »

BeQuiet make some very good low profile coolers, just look for the TDP rating. Higher = better.
e,g 160W: https://www.bequiet.com/en/cpucooler/1074

The Vcore you set in the BIOS is never the actual value due to the load line calibration options, motherboard variation and CPU.
Type a lower value in the BIOS to see what happens, see it says [1.344v] 1.300v.
Maybe 1.275v?
Logged
CPU: Intel i7-6700K | MOBO: MSI Z170A Krait Gaming | RAM: Corsair Vengeance LPX 32GB (4x8GB) 3200 DDR4 | GFX: Gigabyte GTX 970
CASE: Zalman Z11 | FC: Zalman MFC1-Plus | HS: Zalman CNPS10X | PSU: Coolermaster V850 Gold
NVMe: Samsung 960 EVO 250GB | HDD: WD Blue 1TB x2 | ODD: Sony DVDRW | OS: Win 10 x64

blackr92Topic starter

  • CORPORAL
  • ****
  • Offline Offline
  • Posts: 18
« Reply #8 on: 03-November-17, 20:55:49 »

BeQuiet make some very good low profile coolers, just look for the TDP rating. Higher = better.
e,g 160W: https://www.bequiet.com/en/cpucooler/1074

The Vcore you set in the BIOS is never the actual value due to the load line calibration options, motherboard variation and CPU.
Type a lower value in the BIOS to see what happens, see it says [1.344v] 1.300v.
Maybe 1.275v?
I went into bios fiddle around the voltage to get to the closest 1.3V Vcore, in the picture was the closest CPU Core Voltage used to get around 1.3Vcore.


I ran a few Cinebench with the new settings it was stable, however when i went into prime after like 10+mins one of the worker(thread?) stop working ? So i up the CPU Core Voltage to 1.275V(that put me at 1.32V) like you suggest and im testing it again on Prime95 now.
Whats the culprit could cause one of my worker(thread) stop working during the prime95 test?
Logged

electrostingz

  • MAJOR
  • *****
  • Offline Offline
  • Posts: 1011
« Reply #9 on: 03-November-17, 23:08:00 »

Usually the workers will stop if the voltage is set to low or during the test the voltage dips (Vdroop) <increase load line calibration to compensate.
Did lowering the Vcore help with the temperatures?
Logged
CPU: Intel i7-6700K | MOBO: MSI Z170A Krait Gaming | RAM: Corsair Vengeance LPX 32GB (4x8GB) 3200 DDR4 | GFX: Gigabyte GTX 970
CASE: Zalman Z11 | FC: Zalman MFC1-Plus | HS: Zalman CNPS10X | PSU: Coolermaster V850 Gold
NVMe: Samsung 960 EVO 250GB | HDD: WD Blue 1TB x2 | ODD: Sony DVDRW | OS: Win 10 x64

blackr92Topic starter

  • CORPORAL
  • ****
  • Offline Offline
  • Posts: 18
« Reply #10 on: 03-November-17, 23:27:24 »

Usually the workers will stop if the voltage is set to low or during the test the voltage dips (Vdroop) <increase load line calibration to compensate.
Did lowering the Vcore help with the temperatures?
Yeah it does lower the temp, but after the 3rd tried these CPU Core Voltages one of it threads stop working and its the same one
so far ive tried putting in those number in increments

CPU Core Voltage in bios / VCore 
-1.2625 / 1.304V
-1.275   / 1.32V
-1.2875 / 1.336V

All of this numbers i tried so far and it failed one of the worker it stop one the same thread/core.
My LLC is already on mode 2... what should i try now raise the CPU Core Voltage again? ill do it again when i get back later....
Logged

buddyw53

  • SECOND LIEUTENANT
  • **
  • Offline Offline
  • Posts: 102
« Reply #11 on: 04-November-17, 13:33:10 »

I will revert back to default and see hows my stability/temp on those settings. i will post it here when i get it done

regarding the VRM temps i did try to touch it with my finger i can say its running hot i could only hold my finger there prolly 7secs.
This are the 3.7Ghz@1.3V/ CPU NB(Soc voltage) 1.100V oc settings that im currently on.
...

If i were to invest on a new air cooler is tower air cooler(something like cm 212 that i have but no mount bracket) better or the low profile cooler that have fan blows down air straight down to the board that will also pass through vrms?
other than that at stock cooler what can i do to maintain 3.7Ghz and still have good temps and stability?

thanks again

You can see the VRM temperature as the "Mainboard" readout in the HWMonitor utility.  On yours it reads a 75C max temperature which is not very high at all for a VRM.  The FET's are usually rated at 125C or more but it's not recommended operating there in an enclosed case.

A good tower cooler usually makes it very hard to get decent airflow on the VRM.  A nice big down-flow cooler will...I think Noctua makes one.  It may seem strange but a liquid cooler, like an AIO, is nice.  Not that it cools the VRM but it leaves a lot of space to locate a fan to blow directly on the VRM heatsink and you also get un-compromised cooling for the CPU.

I modded an old blank plate to locate a 70mm CPU fan to blow directly on the VRM heatsink, then used one of the Mortar's 4 fan headers and set a profile to only kick it on if the CPU gets to 75C (meaning it's working super hard).  It never does in normal use, not even gaming.  But if I kick off an video encoding queue the fans all fire up big to keep the system cool.  I could probably tune in the profiles to reduce fans speeds but since it only does that during an encode (or stress test) I don't mind all the fans kicking up a storm then.

Oh yes: on my Mortar I have an r7-1700 clocked at 3.95G (about 1.35volts under load) using a CoolerMaster LiquidMaster 240 AIO cooler.
Logged

buddyw53

  • SECOND LIEUTENANT
  • **
  • Offline Offline
  • Posts: 102
« Reply #12 on: 04-November-17, 14:46:56 »

Yeah it does lower the temp, but after the 3rd tried these CPU Core Voltages one of it threads stop working and its the same one
so far ive tried putting in those number in increments

CPU Core Voltage in bios / VCore
-1.2625 / 1.304V
-1.275   / 1.32V
-1.2875 / 1.336V

All of this numbers i tried so far and it failed one of the worker it stop one the same thread/core.
My LLC is already on mode 2... what should i try now raise the CPU Core Voltage again? ill do it again when i get back later....
I think you're probably being a bit too easy on the voltages: you should be able to increase voltage from what you have to get stable.  But do remember that Ryzen cores are safe up to 1.425 volts for sustained use, 1.45 max. This is AMD's statement; I try to keep below 1.425 when dialing in an overclock.

But thing is that's true core voltage, after droop.  You can see that voltage reading in your HWMonitor as the 'CPU VDD (node1)' readout.  It shows that the MINIMUM voltage you drooped to is 1.265 volts.  That was probably under load during a stress test and that's kind of low.

The way I'd go about at it: at 3.800G clock speed increase BIOS CPU Core volts to something pretty high: about 1.400, leaving LLC at Mode 2 like you have it. 

Then boot and check 'CPU VDD(node0)', with no load it should be below 1.425 before you do anything else.  If it has peaks above 1.425 then reboot back to BIOS and lower VCore by a notch until it doesn't.

Once you've got 'CPU VDD(node0)' below 1.425 volts at idle start Aida64 (or whatever) stress test.  You don't have to monitor volts yet: it should be safe since you've set idle to be safe and it can only go DOWN from there.  You're interested in CPU temperature: the one to look at is Package (node 0).

For right now, let's just assume you have good enough cooling so processor temp is staying below 90C and it's holding the stress test well.  Go back into BIOS and LOWER voltage, reboot and stress test.  If it passes lower again...and again..and again until it starts to lockup, black screens, throws errors whatever.  Go back up a few notches; that's the lowest voltage you can operate at that freq.  But now you have to start stress testing for real: that means leaving it to run for an hour or more (depending on your personal preference for assurance) and monitor temperature as it soaks.   

It doesn't matter how high you actually set the core voltage in bios; it could be 1.5 volts (it shouldn't! ever! that's just an example).  What matters is what the cores are actually seeing at idle and that's read out at 'CPU VDD(node0)' in HWMonitor. Don't let that get above 1.425 and you'll be golden!

But...back to temp now.  If Package(node 0) goes over temp at 3.8G and your idle is under is 1.425v I'm pretty sure you can safely say cooling is inadequate. But the point is, that's a way to find a safe overclock once you know you've got good cooling. 

Also important is what's 'over temp': I don't think there is any consensus. Seriously: AMD has mucked up core temp readings unmercifully by dialing in arbitrary factors (on some SKU's) to force cooling to ramp up early and board mfr's have dialed in their own offsets and don't tell us what they are.  So... just some info... Ryzen won't throttle until 95C but I know I like to keep my 1700 @ 3.95G under 80C(Package(node 0)).  It only ever gets to that when stress testing but I'm also using an AIO liquid cooler with a 240mm radiator.

 If you do NOT want to go for better cooling then lower your target frequency and start again...just always remember to keep idle 'CPU VDD(node1)' under 1.425v as the start point.
Logged

blackr92Topic starter

  • CORPORAL
  • ****
  • Offline Offline
  • Posts: 18
« Reply #13 on: 04-November-17, 15:55:27 »

You can see the VRM temperature as the "Mainboard" readout in the HWMonitor utility.  On yours it reads a 75C max temperature which is not very high at all for a VRM.  The FET's are usually rated at 125C or more but it's not recommended operating there in an enclosed case.

A good tower cooler usually makes it very hard to get decent airflow on the VRM.  A nice big down-flow cooler will...I think Noctua makes one.  It may seem strange but a liquid cooler, like an AIO, is nice.  Not that it cools the VRM but it leaves a lot of space to locate a fan to blow directly on the VRM heatsink and you also get un-compromised cooling for the CPU.

I modded an old blank plate to locate a 70mm CPU fan to blow directly on the VRM heatsink, then used one of the Mortar's 4 fan headers and set a profile to only kick it on if the CPU gets to 75C (meaning it's working super hard).  It never does in normal use, not even gaming.  But if I kick off an video encoding queue the fans all fire up big to keep the system cool.  I could probably tune in the profiles to reduce fans speeds but since it only does that during an encode (or stress test) I don't mind all the fans kicking up a storm then.

Oh yes: on my Mortar I have an r7-1700 clocked at 3.95G (about 1.35volts under load) using a CoolerMaster LiquidMaster 240 AIO cooler.
Sweet i didnt know that VRM temps are also called Mainboard in HWMonitor so far trying out 3.7Ghz oc to get it to stable my temps on these are around 75C ish...

I have Hyper 212 EVO on hand atm but i dont have the mounting bracket waiting for it to arrive to my doorsteps my plan was to have these installed and have top case fans intake the air for the VRMs. Getting a down-flow cooler also crossed my mind so it can do 2 jobs which is cooling of CPU and VRMs a lil bit, AIO cooler i have it in mind.

I think you're probably being a bit too easy on the voltages: you should be able to increase voltage from what you have to get stable. But do remember that Ryzen cores are safe up to 1.425 volts for sustained use, 1.45 max. This is AMD's statement; I try to keep below 1.425 when dialing in an overclock. But thing is that's true core voltage, after droop. You can see that voltage reading in your HWMonitor as the 'CPU VDD (node1)' readout. It shows that the MINIMUM voltage you drooped to is 1.265 volts. That was probably under load during a stress test and that's kind of low. The way I'd go about at it: at 3.800G clock speed increase BIOS CPU Core volts to something pretty high: about 1.400, leaving LLC at Mode 2 like you have it. Then boot and check 'CPU VDD(node0)', with no load it should be below 1.425 before you do anything else. If it has peaks above 1.425 then reboot back to BIOS and lower VCore by a notch until it doesn't. Once you've got 'CPU VDD(node0)' below 1.425 volts at idle start Aida64 (or whatever) stress test. You don't have to monitor volts yet: it should be safe since you've set idle to be safe and it can only go DOWN from there. You're interested in CPU temperature: the one to look at is Package (node 0). For right now, let's just assume you have good enough cooling so processor temp is staying below 90C and it's holding the stress test well. Go back into BIOS and LOWER voltage, reboot and stress test. If it passes lower again...and again..and again until it starts to lockup, black screens, throws errors whatever. Go back up a few notches; that's the lowest voltage you can operate at that freq. But now you have to start stress testing for real: that means leaving it to run for an hour or more (depending on your personal preference for assurance) and monitor temperature as it soaks. It doesn't matter how high you actually set the core voltage in bios; it could be 1.5 volts (it shouldn't! ever! that's just an example). What matters is what the cores are actually seeing at idle and that's read out at 'CPU VDD(node0)' in HWMonitor. Don't let that get above 1.425 and you'll be golden! But...back to temp now. If Package(node 0) goes over temp at 3.8G and your idle is under is 1.425v I'm pretty sure you can safely say cooling is inadequate. But the point is, that's a way to find a safe overclock once you know you've got good cooling. Also important is what's 'over temp': I don't think there is any consensus. Seriously: AMD has mucked up core temp readings unmercifully by dialing in arbitrary factors (on some SKU's) to force cooling to ramp up early and board mfr's have dialed in their own offsets and don't tell us what they are. So... just some info... Ryzen won't throttle until 95C but I know I like to keep my 1700 @ 3.95G under 80C(Package(node 0)). It only ever gets to that when stress testing but I'm also using an AIO liquid cooler with a 240mm radiator. If you do NOT want to go for better cooling then lower your target frequency and start again...just always remember to keep idle 'CPU VDD(node1)' under 1.425v as the start point.
EYE opener thanks for the guidelines, i have few questions.

1st - In bios DigitALL Power, i have my LLC mode to 2 and do i also need to enable CPU Over Voltage Protection and CPU Over Current Protection?

2nd - You mention your guideline on how to start Overclock to 3.8GHZ, If i were to try and get stable 3.7GHZ or even maybe 3.75GHZ what BIOS CPU CORE i should start with and work my way down/keep trying and trying again until those clocks are stable at that GHZ?

3rd - "For right now, let's just assume you have good enough cooling so processor temp is staying below 90C and it's holding the stress test well.  Go back into BIOS and LOWER voltage, reboot and stress test.  If it passes lower again...and again..and again until it starts to lockup, black screens, throws errors whatever."   

During this try and error part what stress test you think i should run and how long? until i get comfortable and do a real stress test that are time consuming?
Logged

buddyw53

  • SECOND LIEUTENANT
  • **
  • Offline Offline
  • Posts: 102
« Reply #14 on: 04-November-17, 17:14:28 »

I think the AM4 adapter for the Hyper212 leaves the tower oriented  with the fans blowing 'north-south' instead of the normal 'east-west'.  I'm not a fan of that 'cause it won't provide any cooling assist for the VRM and complicates trying to get a fan on it.  I may be wrong though, just something to keep in mind.

I think the CPU OverVolts and OverCurrent protect are for EXTREME overclocking explorations where things are at the limits. I'm not really sure how they're beneficial since we're trying to find the LOWEST voltage and hopefully lowest current for a reasonable, 24/7, overclock.  I just leave them disabled. 

You can probably start at 1.400volts even if your target overclock is 3.700Gig.  The idea is to get a voltage you know is stable but still not exceed a 1.425v safety limit and start marching the voltage down from there.  It's very much trial and error, you'll be re-booting a lot.  I pick 3.8Gig as a start because I know the Mortar (I have one) and I think every Ryzen processor should be capable of 3.8G at least on a Mortar so it's a good starting point.

The preliminary stress tests aren't very long at all...just a few minutes.  After all, at this point you'd like it to crash fast; maybe even not boot to Windows.  That means you've found the lowest voltage and you know you have to increase it back up a couple notches. 

You've got Aida64 and it's stress test module is fine: for the intermediate tests stress using just the FPU as it provides the heaviest load on the CPU (you want it to crash fast, remember).

After you've got a stable overclock at a base frequency you can evaluate where you are.  If idle voltage is low enough, say 1.38 or lower, and processor temp at load is reasonable, say 70C, then push up the frequency to 3.85Gig and voltage a couple notches and do another full up stress test.  Just be sure to check idle voltage first: we're trying to keep it under 1.425V always.

BTW: you should save your BIOS OverClock profile settings too, just so you have them.
Logged

electrostingz

  • MAJOR
  • *****
  • Offline Offline
  • Posts: 1011
« Reply #15 on: 04-November-17, 21:06:51 »

My advice would be to keep the voltage as low as possible and work your way up.
Despite maximum operating voltages these are guidelines offered by the manufacture and the important note to take away is that a higher voltage decreases the life of the component.

VRM temperatures are also important as this will help with voltage / current stabilisation.
The higher temperatures decrease the overall efficiency of the FETs, my recommendation on this would be 60 Degrees for optimum performance.
This has a direct impact on your Vcore setting / OC and you may find that it is stable at ~1.300v once you get the VRM temperature down.

Same goes for your CPU temperature, lower is always better.
Logged
CPU: Intel i7-6700K | MOBO: MSI Z170A Krait Gaming | RAM: Corsair Vengeance LPX 32GB (4x8GB) 3200 DDR4 | GFX: Gigabyte GTX 970
CASE: Zalman Z11 | FC: Zalman MFC1-Plus | HS: Zalman CNPS10X | PSU: Coolermaster V850 Gold
NVMe: Samsung 960 EVO 250GB | HDD: WD Blue 1TB x2 | ODD: Sony DVDRW | OS: Win 10 x64

blackr92Topic starter

  • CORPORAL
  • ****
  • Offline Offline
  • Posts: 18
« Reply #16 on: 05-November-17, 00:23:37 »

I think the AM4 adapter for the Hyper212 leaves the tower oriented  with the fans blowing 'north-south' instead of the normal 'east-west'.  I'm not a fan of that 'cause it won't provide any cooling assist for the VRM and complicates trying to get a fan on it.  I may be wrong though, just something to keep in mind.

I think the CPU OverVolts and OverCurrent protect are for EXTREME overclocking explorations where things are at the limits. I'm not really sure how they're beneficial since we're trying to find the LOWEST voltage and hopefully lowest current for a reasonable, 24/7, overclock.  I just leave them disabled.  

You can probably start at 1.400volts even if your target overclock is 3.700Gig.  The idea is to get a voltage you know is stable but still not exceed a 1.425v safety limit and start marching the voltage down from there.  It's very much trial and error, you'll be re-booting a lot.  I pick 3.8Gig as a start because I know the Mortar (I have one) and I think every Ryzen processor should be capable of 3.8G at least on a Mortar so it's a good starting point.

The preliminary stress tests aren't very long at all...just a few minutes.  After all, at this point you'd like it to crash fast; maybe even not boot to Windows.  That means you've found the lowest voltage and you know you have to increase it back up a couple notches.  

You've got Aida64 and it's stress test module is fine: for the intermediate tests stress using just the FPU as it provides the heaviest load on the CPU (you want it to crash fast, remember).

After you've got a stable overclock at a base frequency you can evaluate where you are.  If idle voltage is low enough, say 1.38 or lower, and processor temp at load is reasonable, say 70C, then push up the frequency to 3.85Gig and voltage a couple notches and do another full up stress test.  Just be sure to check idle voltage first: we're trying to keep it under 1.425V always.

BTW: you should save your BIOS OverClock profile settings too, just so you have them
Am4 adapter for 212 actually have 2 variations both of them you mentioned.

atm im back on 3.7GHZ@1.3V in bios(1.344V in computer) and trying for more lower more CPU Voltage Settings and stability @3.7GHZ to lower my temps abit(my ambient temp is actually high atm).

Also when you say idle volts has to be below 1.425V "CPU VDD(node0)" at this voltage during load(during stress testing) its going to go down am i right? i tho its suppose to go up, my 3.7GHZ@1.3V(in bios, but in HWMonitor shows 1.344 CPU VCORE) when i stress test aida/prime i realise my "CPU VDD(node0)" went down to 1.269V but during idle its at 1.325V i realise that during stress testing it tends to stay lower than my idle voltage is that how it supposed to be like?

Ill try out 3.7-3.8GHZ @1.400volts later on today and work my voltage down until i hit the point, during these steps i dont really want to stress test too long right cause the temp might be super high? this the only part im worried, i just have to drop voltage until I know i have to bump it up again right?

also thanks in advance again :D
Logged

blackr92Topic starter

  • CORPORAL
  • ****
  • Offline Offline
  • Posts: 18
« Reply #17 on: 05-November-17, 00:27:38 »

My advice would be to keep the voltage as low as possible and work your way up.
Despite maximum operating voltages these are guidelines offered by the manufacture and the important note to take away is that a higher voltage decreases the life of the component.

VRM temperatures are also important as this will help with voltage / current stabilisation.
The higher temperatures decrease the overall efficiency of the FETs, my recommendation on this would be 60 Degrees for optimum performance.
This has a direct impact on your Vcore setting / OC and you may find that it is stable at ~1.300v once you get the VRM temperature down.

Same goes for your CPU temperature, lower is always better.
Yeah thats what im been practicing so far, im trying to get 3.7GHZ stable @ lowest voltage i can.
As soon my cooler bracket arrive i will find a way of cooling the VRMs i have an rough idea that i want to try.
Logged

buddyw53

  • SECOND LIEUTENANT
  • **
  • Offline Offline
  • Posts: 102
« Reply #18 on: 05-November-17, 01:33:03 »

My advice would be to keep the voltage as low as possible and work your way up.
...
The reason I don't like to 'start low and move up' to find stability is every failure ends in a hard, black screened, Windows10 crash.  Somewhere in there I never know if the crashes are still because it's not stable or Windows10 is fully corrupted and needs a repair re-install.  And if someone wants to actually use the machine between overclock tests that's a real pain: it can mean reinstalling apps and/or recovering from a backup.  I'm not an IT professional with a bunch of drives sitting around with OS installs on them so I try to minimize the crashes.  Doing it this way you're looking for the one crash that says you found the lowest voltage for a given clock frequency.

Secondly: AMD does suggest 1.425 as max safe sustained operating voltage, and 1.45 as max intermittent operating voltage and I believe it's even published in their Overclocking Guide.  But what's really strange is an 1800x processor in normal operation, when it boosts to 1.42Gig, will frequently raise core voltage to 1.475 and I've personally watched one regularly going to 1.5volts!  That's bone-stock on an Asus CHVI, no overclocking or VCore offsets at all.  So I'm perfectly comfortable using 1.425 as the safety limit to start dialing in an overclock from.  But do note I do not suggest operating there for a 24/7 overclock (unless you're the venturesome sort). 

But all that said, if you're worried about longevity and want your processor to live out it's full 20+ years service life: don't overclock. Period.
Logged

buddyw53

  • SECOND LIEUTENANT
  • **
  • Offline Offline
  • Posts: 102
« Reply #19 on: 05-November-17, 02:28:15 »

Am4 adapter for 212 actually have 2 variations both of them you mentioned.

atm im back on 3.7GHZ@1.3V in bios(1.344V in computer) and trying for more lower more CPU Voltage Settings and stability @3.7GHZ to lower my temps abit(my ambient temp is actually high atm).

Also when you say idle volts has to be below 1.425V "CPU VDD(node0)" at this voltage during load(during stress testing) its going to go down am i right? i tho its suppose to go up, my 3.7GHZ@1.3V(in bios, but in HWMonitor shows 1.344 CPU VCORE) when i stress test aida/prime i realise my "CPU VDD(node0)" went down to 1.269V but during idle its at 1.325V i realise that during stress testing it tends to stay lower than my idle voltage is that how it supposed to be like?

Ill try out 3.7-3.8GHZ @1.400volts later on today and work my voltage down until i hit the point, during these steps i dont really want to stress test too long right cause the temp might be super high? this the only part im worried, i just have to drop voltage until I know i have to bump it up again right?

also thanks in advance again :D
When the processor is loaded voltage AT THE CORE (which is the"CPU VDD(node 0)" reading) should lower from idle.   To get an idea: on my Mortar I've an 8 core, 1700, clocked at 3.95Gig.  At idle the "CPU VDD(node 0)" fluctuates 1.390-1.400V, when I run an extreme load (prime95 small FFT) it drops to 1.306V.   During and throughout this time the CPU VCore reading is 1.425-1.43 volts; that's the output of the VRM.

Doing a HandBrake video encode, my 'real-world' heavy useage, it's around 1.35V. 

(BTW: either I don't understand how it's supposed to work on Mortars or the LLC authority is not nearly enough for an 8 core.)

You're correct that you should not test too long but that's mainly just a time saver. Your goal is to find the lowest voltage first so if it boots and stays stable for a short test you know you can go lower to find the one where it won't.  Once you find that one, go back up a setting or two and then try stressing for a long time and now pay special attention to temperatures.  If it's going really hot you'll either have to fix it with better cooling or lower the clock speed. 

I can't help you too much with temperatures since I've an 8 core and it will naturally run hotter, both VRM and CPU temps, at a given frequency/load/voltage.  Guidelines are also very opinionated: just keep in mind that Tjmax for the CPU is given as 95C; I've never gotten close to that so I can't imagine a 6 core would get there either.  The same with VRM temps: mine gets 70-80C, which isn't bad, under heavy loads (I seem to remember these FET's have a 25C ambient operational rating around 140C, so de-rate that to 125C for inside an enclosure and even 100C is safe) but a 6 core should be even lower. 

And yes, lower temps are always better but you have to ask yourself if you want to overclock or maximize the service life of your processor/mobo.  It never made sense to me to be concerned about long life when I know I'll trash it in 5 years for new technology.
Logged

electrostingz

  • MAJOR
  • *****
  • Offline Offline
  • Posts: 1011
« Reply #20 on: 05-November-17, 12:41:16 »

Over volting to minimise crashes does not serve in the best interest for the CPU. The max rated voltages quoted for any CPU are there to safe guard against voltage spikes caused by the loadline calibration function as the CPU state rapidly changes. Under normal operating conditions the voltage will spike as the load changes, usually a few ms. If you run the CPU at the max rated voltage the spikes will naturally increase, peak over the range.

Whilst this is always going to occur, it is the main reason why Vdroop exists, to safeguard against spikes as the load changes. Not good for overcooking but when you start messing around with the LLC option you will remove this safeguard. Everything AMD and Intel quote are for normal operating conditions.

So my point is that someone looking for a beginner's guide should not simply dial in numbers without understanding the consequences. Furthermore the basics of overclocking require decent temperatures to begin with, you don't push a system that is already at the limits. The result will be more efficient if the VRM temps are lower and with a relatively cooler CPU.

The basic knowledge floating around is everything will be fine as long as you are in the spec quoted by X manufacturer.
This is the lazy approach to any given problem, efficiency is performance. Yes overclocking will decrease the lifespan, so does heat, voltage, spikes... It is not an excuse to use as justification. The smart overclocker will optimise all aspects to ensure maximum potential.

This is my view on overclocking and I respect your knowledge / advice however, I dislike the approach 😃
I would never want my CPU to hit 90 degrees.
Logged
CPU: Intel i7-6700K | MOBO: MSI Z170A Krait Gaming | RAM: Corsair Vengeance LPX 32GB (4x8GB) 3200 DDR4 | GFX: Gigabyte GTX 970
CASE: Zalman Z11 | FC: Zalman MFC1-Plus | HS: Zalman CNPS10X | PSU: Coolermaster V850 Gold
NVMe: Samsung 960 EVO 250GB | HDD: WD Blue 1TB x2 | ODD: Sony DVDRW | OS: Win 10 x64

buddyw53

  • SECOND LIEUTENANT
  • **
  • Offline Offline
  • Posts: 102
« Reply #21 on: 05-November-17, 14:39:11 »

Over volting to minimise crashes does not serve in the best interest for the CPU. The max rated voltages quoted for any CPU are there to safe guard against voltage spikes caused by the loadline calibration function as the CPU state rapidly changes. Under normal operating conditions the voltage will spike as the load changes, usually a few ms. If you run the CPU at the max rated voltage the spikes will naturally increase, peak over the range.

Whilst this is always going to occur, it is the main reason why Vdroop exists, to safeguard against spikes as the load changes. Not good for overcooking but when you start messing around with the LLC option you will remove this safeguard. Everything AMD and Intel quote are for normal operating conditions.

So my point is that someone looking for a beginner's guide should not simply dial in numbers without understanding the consequences. Furthermore the basics of overclocking require decent temperatures to begin with, you don't push a system that is already at the limits. The result will be more efficient if the VRM temps are lower and with a relatively cooler CPU.

The basic knowledge floating around is everything will be fine as long as you are in the spec quoted by X manufacturer.
This is the lazy approach to any given problem, efficiency is performance. Yes overclocking will decrease the lifespan, so does heat, voltage, spikes... It is not an excuse to use as justification. The smart overclocker will optimise all aspects to ensure maximum potential.

This is my view on overclocking and I respect your knowledge / advice however, I dislike the approach 😃
I would never want my CPU to hit 90 degrees.
All that is an excellent argument for not overclocking at all.  Period.  Which isn't a bad beginner's guide.

Overclocking is, by definition, exceeding mfr's specifications. It is and always has been that way so if someone is not comfortable seeing any operating parameter even close to a specified limit, e.g., core voltage, then that person should absolutely positively not be trying to 'overclock' at all.  Just set up BIOS to optimized defaults and forget you even heard the term.

Or if you really must offer up some hint of the process so a beginner can feel all tingly about it: just tell them to leave CPU Core Voltage at AUTO and push up the CPU clock frequency until it crashes, go back one and stop there.  You have them stop there since any increase you make to voltage, needed to enable the higher speed, is really over-volting.

But is it over volting? That's crazy but it does raise the question: what is over volting.   Well I believe I know what it is NOT: operating up to or at a specified voltage limit is NOT 'over volting'.  So then over volting would be going BEYOND mfr' published limits on the premise that engineering set them there, as they do most things, WITH MARGIN that may be exploited just as we do core frequency margin by over-clocking.  I don't over-volt or suggest anyone do it but the curious fact seems to be AMD (or maybe it's board mfr's) does since Ryzen 'X' CPUS are allowed to peak as high as 1.5 volts during core boost! 

Wow... based on that what even constitutes an 'over volt' situation is not very clear at all to me and would be even less clear to a beginner.   I suppose someone could consider 'over-volting' as operating with a core voltage higher than it needs to be, generating unecessary heat and wasting power.  Correctly, that is a poorly optimized system.  Any overclocker should optimize their system at the lowest stable voltage for the target clock speed.
Logged

buddyw53

  • SECOND LIEUTENANT
  • **
  • Offline Offline
  • Posts: 102
« Reply #22 on: 05-November-17, 15:56:08 »

.....

This is my view on overclocking and I respect your knowledge / advice however, I dislike the approach 😃
I would never want my CPU to hit 90 degrees.
Maybe you'd like this approach...

Start off with clock speed at AUTO and increasing Core Voltage a couple notches at a time until it is whatever CPU VDD(Node0) idle voltage is comfortable. I would use 1.400 but 1.375, 1.350 might be just as good.

Once you've got that then increase CPU clock frequency and stress test (P95, Small FFT's) for about 5 minutes.  Check CPU VDD(Node0) voltage should confirm that it goes DOWN under load, and goes back UP when at idle.  If it passes the short test then increasing clock frequency again, and again, until it crashes.  Since tests are short don't worry about temperatures yet. 

Once it crashes in a short test dial back clock speed one and check the CPU VDD(Node0) idle voltage.  If it's BELOW what it was set at above then increase core voltage a notch at a time until it's the same. 

Now, run an extended test (P95, small FFT's). Keep an eye on temp's, run it 30-40 min's now if temps allow.   If it holds and temperatures are OK then it's good, but if it crashes dial back clock speed and do a long test again.

This should result in a tolerable overclock since it established a tolerable target CPU VDD(Node0) idle voltage at a low clock speed to begin with which is used to find a target clock speed then increasing it back up to the tolerable idle voltage, thereby improving stability, once established.

If temperature just won't allow a long enough test (evaluate tolerance for temperatures) then solve those problems with 1) better cooling on CPU 2) fan on VRM heatsink or both, as appropriate.  I'd maintain there has to be sufficient cooling to handle the system when it's not optimized so do it now.

If you run an extended test, it's not crashing and temps are tolerable you can optimize: try reducing voltage a notch or two until it crashes, try fiddling with LLC until you get lowest voltage/temperature.  Even try INCREASING clock speed a notch; whatever is your goal and matches your tolerance level.   

Once you've optimized run a very long stress test:  this should run for two hours at least.  This is to prove the system is stable and can be used for 'mission essential' tasks without worry.  Then go reading around other posts on other forums to benchmark your results!
Logged

electrostingz

  • MAJOR
  • *****
  • Offline Offline
  • Posts: 1011
« Reply #23 on: 05-November-17, 19:31:00 »

All that is an excellent argument for not overclocking at all.  Period.  Which isn't a bad beginner's guide.

I stated reasons why it is better to keep the Vcore lower, nothing more.  I am not here to argue and my only intention is to provide reasonable advice and general guidance such as keeping the voltages lower and temps down. I'll say no more on this matter, blackr92 adjust your options wisely.
Logged
CPU: Intel i7-6700K | MOBO: MSI Z170A Krait Gaming | RAM: Corsair Vengeance LPX 32GB (4x8GB) 3200 DDR4 | GFX: Gigabyte GTX 970
CASE: Zalman Z11 | FC: Zalman MFC1-Plus | HS: Zalman CNPS10X | PSU: Coolermaster V850 Gold
NVMe: Samsung 960 EVO 250GB | HDD: WD Blue 1TB x2 | ODD: Sony DVDRW | OS: Win 10 x64

blackr92Topic starter

  • CORPORAL
  • ****
  • Offline Offline
  • Posts: 18
« Reply #24 on: 06-November-17, 07:14:52 »

At the moment im set at 3.7GHZ @ 1.3V in bios(Vcore reading is at 1.344V inside the computer).
- This is the only settings that during 4hours of Prime95 stress test, all the thread still working not a single thread stop working and Aida64 was stable too for 5 hours. I tried lowering the voltage in bios to try and get better temps but lower than 1.3V in bios will cause one of the threads stop working during Prime95 test but Aida64 is stable i dont get it lol....

Im kinda waiting on my am4 cooler mount for my 212 to arrive so i can stress test further atm i think my temps are at my limit and the season here in my region are quite hot.

few screenshots of temp/voltage (during stress test/idling)
Idling

During Aida64

after Prime95


temps are high i know during stress test at this settings, but when i game it almost touch 70C so far stable.
what do you think of my temps and my voltages does it look fine or is there something funny i cant detect?
also was wondering for MSI B350M Mortar the LLC features as the mode level go higher what does it do actually?(for example from mode 1 to mode 7)
Logged

buddyw53

  • SECOND LIEUTENANT
  • **
  • Offline Offline
  • Posts: 102
« Reply #25 on: 06-November-17, 12:40:28 »

At the moment im set at 3.7GHZ @ 1.3V in bios(Vcore reading is at 1.344V inside the computer).
...
temps are high i know during stress test at this settings, but when i game it almost touch 70C so far stable.
what do you think of my temps and my voltages does it look fine or is there something funny i cant detect?
also was wondering for MSI B350M Mortar the LLC features as the mode level go higher what does it do actually?(for example from mode 1 to mode 7)
I can see that CPU clock speed is boosting to almost 3.8Gig.  You might try checking that CPU Core Boost, C6 States and Cool n Quiet are all disabled in BIOS.  You can re-enable them later if you want but they affect testing and make it hard to determine if things are going like expect.

80C on the VRM for a 6 core at 3.7Gig/1.28V core just seems really high to my eye (I wonder how much it's boosting to 3.8???).  If you're using the stock cooler make sure it's maxing out RPM as it does provide air flow on the VRM heatsink.  The 212 should help with processor temp's (90C!!) but it will just hurt the VRM temp--even oriented 'east-west'.   For compare, my VRM will peak at 80C during P95-small FFT and I've an 8 core running at 3.95Gig/~1.31V core during that time.  I have a 70mm fan directed on it so that illustrates the value of providing some air movement over the heatsink.

P95, Small FFT is totally unreal and almost guarantees the highest temps.  If you pass it you are golden (even if it drops cores or crashes you probably are) since you'll not likely ever see those temps (or loading) in sustained 'real world' use-age .  I certainly haven't ever on any HandBrake video encodings I've done.

Nobody's provided any really useful advice on how the LLC modes work on Mortar's (or Tomahawk's which have the same VRM design and BIOS features).  But I've found that Mode 2 seems to work best for letting me set the lowest idle voltage and still allows stable stress test with low(est) VRM temp.

Also: always remember that what you're seeing while gaming is generally based on single or 2-core loading since games don't really load up even a 6 core very much. Even Doom using Vulkan is only using 4 cores intermittently and that combo is supposed to be 'multi-core friendly'.  I guess if you're also streaming though... maybe using a couple more cores to process the video??
Logged

blackr92Topic starter

  • CORPORAL
  • ****
  • Offline Offline
  • Posts: 18
« Reply #26 on: 06-November-17, 16:11:43 »

I can see that CPU clock speed is boosting to almost 3.8Gig.  You might try checking that CPU Core Boost, C6 States and Cool n Quiet are all disabled in BIOS.  You can re-enable them later if you want but they affect testing and make it hard to determine if things are going like expect.

80C on the VRM for a 6 core at 3.7Gig/1.28V core just seems really high to my eye (I wonder how much it's boosting to 3.8???).  If you're using the stock cooler make sure it's maxing out RPM as it does provide air flow on the VRM heatsink.  The 212 should help with processor temp's (90C!!) but it will just hurt the VRM temp--even oriented 'east-west'.   For compare, my VRM will peak at 80C during P95-small FFT and I've an 8 core running at 3.95Gig/~1.31V core during that time.  I have a 70mm fan directed on it so that illustrates the value of providing some air movement over the heatsink.

P95, Small FFT is totally unreal and almost guarantees the highest temps.  If you pass it you are golden (even if it drops cores or crashes you probably are) since you'll not likely ever see those temps (or loading) in sustained 'real world' use-age .  I certainly haven't ever on any HandBrake video encodings I've done.

Nobody's provided any really useful advice on how the LLC modes work on Mortar's (or Tomahawk's which have the same VRM design and BIOS features).  But I've found that Mode 2 seems to work best for letting me set the lowest idle voltage and still allows stable stress test with low(est) VRM temp.

Also: always remember that what you're seeing while gaming is generally based on single or 2-core loading since games don't really load up even a 6 core very much. Even Doom using Vulkan is only using 4 cores intermittently and that combo is supposed to be 'multi-core friendly'.  I guess if you're also streaming though... maybe using a couple more cores to process the video??

CPU Core Performance Boost, Core C6 State and AMD CoolnQuiet are all disabled why does it still go more than set ghz? which is more than 3.7ghz.

Ive already set custom fan curve so it will go to 100% when hit 70C around there, also how do you read that 3.7GHZ/1.28V?(something about 80C on VRAM for 6 core at 3.7Gig/1.28V core)
Logged

buddyw53

  • SECOND LIEUTENANT
  • **
  • Offline Offline
  • Posts: 102
« Reply #27 on: 07-November-17, 00:32:08 »

CPU Core Performance Boost, Core C6 State and AMD CoolnQuiet are all disabled why does it still go more than set ghz? which is more than 3.7ghz.

Ive already set custom fan curve so it will go to 100% when hit 70C around there, also how do you read that 3.7GHZ/1.28V?(something about 80C on VRAM for 6 core at 3.7Gig/1.28V core)
That's really crazy: I have no idea why it's boosting to 3.8ghz, especially since Core Boost is disabled. 

Also: All I'm seeing is VRM temp went to 80C max and core voltage is 1.28V min so I can only assuming they both occurred simultaneously while running the stress test at 3.7Ghz.  You can drag any HWInfo64 sensor reading to the desktop   and it will create a graph.  Do that for CPU Core Voltage (STI2 TFN) and core clock freq's, CPU (Tctl/Tdie) temp and  whatever interests you to see how things are tracking over time as well as when/where that crazy 'core boost' thing is happening.  It may just be a transient reading.
Logged

blackr92Topic starter

  • CORPORAL
  • ****
  • Offline Offline
  • Posts: 18
« Reply #28 on: 07-November-17, 04:03:34 »

Testing out different LLC Mode @ 3.7GHZ/1.3V

Mode 1 /1.344V-1.352V in BIOS
CPU Core Voltage(SVI2 TFN)-
1.275V(min/load) 
1.331V(max/idle)

Mode 2 /1.344V in BIOS
CPU Core Voltage(SVI2 TFN)-
1.262V(min/load) 
1.325V(max/idle)

Mode 3 /1.336V in BIOS
CPU Core Voltage(SVI2 TFN)-
1.256V(min/load) 
1.325V(max/idle)

Mode 4 /1.328V-1.336V in BIOS
CPU Core Voltage(SVI2 TFN)-
1.250V(min/load) 
1.325V(max/idle)

Mode 5 /1.320V-1.328V in BIOS
CPU Core Voltage(SVI2 TFN)-
1.237V(min/load) 
1.319V(max/idle)

Mode 6 /1.312V-1.320V-1.328V in BIOS
CPU Core Voltage(SVI2 TFN)-
1.237V(min/load) 
1.319V(max/idle)

Mode 7 /1.312V-1.320V in BIOS
CPU Core Voltage(SVI2 TFN)-
1.225V(min/load) 
1.319V(max/idle)

Mode 8 /1.304V-1.312V in BIOS
CPU Core Voltage(SVI2 TFN)-
1.219V(min/load) 
1.319V(max/idle)

Should i go higher mode other than 2 that is what im on right now? or later on maybe during Prime95 stress test not enough core voltage during Load?
I notice on higher modes temps go down but also voltage go down prolly would effect 3.7ghz clock.
Logged

buddyw53

  • SECOND LIEUTENANT
  • **
  • Offline Offline
  • Posts: 102
« Reply #29 on: 07-November-17, 12:35:07 »

Testing out different LLC Mode @ 3.7GHZ/1.3V

.....

Should i go higher mode other than 2 that is what im on right now? or later on maybe during Prime95 stress test not enough core voltage during Load?
I notice on higher modes temps go down but also voltage go down prolly would effect 3.7ghz clock.
Try a higher mode coupled with a higher core voltage setting in BIOS to find a balance that helps temperature while keeping voltage close to the same at load.  I'd surely try any setting that lowers temps with a stress test, though.  My experience with optimizing LLC, though, is finding the lowest idle voltage while still keeping it stable at high load. 

Although it may not seem so for a 1600 (max boost to 3.6, stock) but 6 Zen cores @3.7Gig is really a very mild overclock.  If your processor temp is hitting 90C you really need to get that Hyper 212 cooler in place. 

Have you tried re-seating the CPU heatsink?  What could be happening is temperature's driving the processor into instability, requiring more voltage to keep it stable, generating even more heat. A well-known vicious cycle called thermal runaway.  If so, all you have to do is get a few more joules out of the processor and you'd be able to lower voltage or catch a higher core clock strap at the same voltage even while operating at a lower temperature.

Fix the temp problem.
Logged

blackr92Topic starter

  • CORPORAL
  • ****
  • Offline Offline
  • Posts: 18
« Reply #30 on: 07-November-17, 17:57:51 »

Try a higher mode coupled with a higher core voltage setting in BIOS to find a balance that helps temperature while keeping voltage close to the same at load.  I'd surely try any setting that lowers temps with a stress test, though.  My experience with optimizing LLC, though, is finding the lowest idle voltage while still keeping it stable at high load.  

Although it may not seem so for a 1600 (max boost to 3.6, stock) but 6 Zen cores @3.7Gig is really a very mild overclock.  If your processor temp is hitting 90C you really need to get that Hyper 212 cooler in place.  

Have you tried re-seating the CPU heatsink?  What could be happening is temperature's driving the processor into instability, requiring more voltage to keep it stable, generating even more heat. A well-known vicious cycle called thermal runaway.  If so, all you have to do is get a few more joules out of the processor and you'd be able to lower voltage or catch a higher core clock strap at the same voltage even while operating at a lower temperature.

Fix the temp problem.
"finding the lowest idle voltage while still keeping it stable at high load"
lowest idle voltage to my set CPU CORE Voltage? also when you say lowest idle voltage is it the VCORE or CPU Core Voltage(SVI2 TFN)-
Logged

buddyw53

  • SECOND LIEUTENANT
  • **
  • Offline Offline
  • Posts: 102
« Reply #31 on: 07-November-17, 23:07:51 »

"finding the lowest idle voltage while still keeping it stable at high load"
lowest idle voltage to my set CPU CORE Voltage? also when you say lowest idle voltage is it the VCORE or CPU Core Voltage(SVI2 TFN)-
As reported by HWInfo64 it's CPU Core Voltage(SVI2 TFN).  That is the processor's core voltage reported by the processor through SVI2 telemetry.  It's considered the most accurate voltage that the processor cores are actually seeing (but not always).

The VCore reading I think is basically the output of the VRM, maybe the voltage sense, as reported by the VRM controller.  I can only surmise that because the voltage readings I took on the back of my motherboard agreed with it.  The reason they are frequently different values in the readings is that as the current passes through the motherboard circuitry to the processor it will drop.  It will drop a lot when the processor is working really hard under heavy load as it's drawing a heavy current.  It's all basic DC circuit analysis. 

You want the SVI2 voltage to not exceed whatever value you want to establish as your safety limit; I use 1.425V.
Logged

blackr92Topic starter

  • CORPORAL
  • ****
  • Offline Offline
  • Posts: 18
« Reply #32 on: 07-November-17, 23:44:58 »

So if my 1 hour Prime95 stress test are stable at mode 2, during load the CPU Core Voltage(SVI2 TFN) is at 1.262V(lowest) is it fair to say if im able to reach that volt at load on different mode that might give me lower Vcore or CPU Core Voltage(SVI2 TFN) at idle would work good aswell?

when you say "My experience with optimizing LLC, though, is finding the lowest idle voltage while still keeping it stable at high load."
I dont really understand this part sorry :worship:
Logged

blackr92Topic starter

  • CORPORAL
  • ****
  • Offline Offline
  • Posts: 18
« Reply #33 on: 08-November-17, 04:12:06 »

tried Mode 1-3.7GHZ@1.2875V in BIOS, got lower Vcore and CPU Core Voltage(SVI2 TFN) during Idle compare to Mode 2-3.7GHZ@1.3V in BIOS and about the same CPU Core Voltage(SVI2 TFN) during load 1.262V lowest recorded for both settings(during prime).

Pictures regarding voltage/temp on both MODE, for LOAD  1.5 hours prime95 is used for stress test.

Mode 1-3.7GHZ@1.2875V in BIOS


Mode 2-3.7GHZ@1.3V in BIOS


between the two, Mode 1 is better right in terms of lowered voltage?
Logged

buddyw53

  • SECOND LIEUTENANT
  • **
  • Offline Offline
  • Posts: 102
« Reply #34 on: 08-November-17, 11:34:15 »

So if my 1 hour Prime95 stress test are stable at mode 2, during load the CPU Core Voltage(SVI2 TFN) is at 1.262V(lowest) is it fair to say if im able to reach that volt at load on different mode that might give me lower Vcore or CPU Core Voltage(SVI2 TFN) at idle would work good aswell?

when you say "My experience with optimizing LLC, though, is finding the lowest idle voltage while still keeping it stable at high load."
I dont really understand this part sorry :worship:
Answer to first question is yes, it should work as well for stability.

Answer to second:  Since I have a 240mm AIO that's cooling my 1700 really well I'm not optimizing LLC to try and find lower temperatures, but to find a lower idle voltage that still gives me stability (by not letting loaded voltage drop too low) when stress testing.  

Your voltages really are nice and low, I can't see any reason to be concerned at all with what your running in that regard, and should be at only 3.7Ghz.  But your processor is still pulling 90C on a stress run and that's just too much.  So, if you want to try and find an LLC mode that lowers it can't hurt.
Logged

buddyw53

  • SECOND LIEUTENANT
  • **
  • Offline Offline
  • Posts: 102
« Reply #35 on: 08-November-17, 11:59:24 »

tried Mode 1-3.7GHZ@1.2875V in BIOS, got lower Vcore and CPU Core Voltage(SVI2 TFN) during Idle compare to Mode 2-3.7GHZ@1.3V in BIOS and about the same CPU Core Voltage(SVI2 TFN) during load 1.262V lowest recorded for both settings(during prime).

Pictures regarding voltage/temp on both MODE, for LOAD  1.5 hours prime95 is used for stress test.

Mode 1-3.7GHZ@1.2875V in BIOS

---
Mode 2-3.7GHZ@1.3V in BIOS

--
between the two, Mode 1 is better right in terms of lowered voltage?
Yes, it does appear so.  But in both cases, since you've demonstrated pretty good stability in each,  the voltages represent a problem that is vanishingly unimportant .  It's not doing anything for that processor temp problem.   If you're trying to optimize an overclock it's really just pushing the rubble around, so to speak.

Is it appropriate to ask what your goal is for overclocking, or have you established one?
Logged

blackr92Topic starter

  • CORPORAL
  • ****
  • Offline Offline
  • Posts: 18
« Reply #36 on: 08-November-17, 16:36:19 »

Yes, it does appear so.  But in both cases, since you've demonstrated pretty good stability in each,  the voltages represent a problem that is vanishingly unimportant .  It's not doing anything for that processor temp problem.   If you're trying to optimize an overclock it's really just pushing the rubble around, so to speak.

Is it appropriate to ask what your goal is for overclocking, or have you established one?

Basicly wanted to try out overclocking and also i heard that Ryzen 5 1600 can be easily OC(i heard its quite a waste to not OC it cause most people said its easy and little cons) in my case mildly at 3.7ghz cause i dont want to push it too far as long as its stable im good with it, but yeah temps are high maybe just because of my ambient here are too hot living in Asia thats what you get.

I might revert back to default and see how my temps everything on auto.

Also CoolerMaster support over here said that they could only send the mount bracket early next month, and atm im eyeing on these 3 coolers (AIO/down flow cooler/tower cooler)

FROSTFLOW+ 120 SNOW (RM189/USD45)


Cryorig H7 - Mid tower cooler (RM209/USD50)
heard its better than 212EVO


Be Quiet Shadow Rock LP - down flow cooler (RM210/USD50)


i got these 3 im looking atm, and tbh the AIO Frostflow looks good they even have 240 version but i dont know how big is the difference. One thing i have to worry about too if i get any of the cooler how do i cool VRMs especially if AIO/Tower Cooler.
Logged

buddyw53

  • SECOND LIEUTENANT
  • **
  • Offline Offline
  • Posts: 102
« Reply #37 on: 09-November-17, 02:09:28 »

Basicly wanted to try out overclocking and also i heard that Ryzen 5 1600 can be easily OC(i heard its quite a waste to not OC it cause most people said its easy and little cons) in my case mildly at 3.7ghz cause i dont want to push it too far as long as its stable im good with it, but yeah temps are high maybe just because of my ambient here are too hot living in Asia thats what you get.

I might revert back to default and see how my temps everything on auto.

Also CoolerMaster support over here said that they could only send the mount bracket early next month, and atm im eyeing on these 3 coolers (AIO/down flow cooler/tower cooler)

FROSTFLOW+ 120 SNOW (RM189/USD45)
...
Cryorig H7 - Mid tower cooler (RM209/USD50)
heard its better than 212EVO
...
Be Quiet Shadow Rock LP - down flow cooler (RM210/USD50)
...
i got these 3 im looking atm, and tbh the AIO Frostflow looks good they even have 240 version but i dont know how big is the difference. One thing i have to worry about too if i get any of the cooler how do i cool VRMs especially if AIO/Tower Cooler.
How I approached VRM cooling with an AIO cooler is I modded an old blank plate by drilling a couple holes in it to attach a 70mm CPU fan. I then attached the blank plate with the fan to one corner of the case fan and directed it's airflow towards the VRM.  Put it on one of the four fan headers all alone and set a profile so it only turns on when CPU temp goes above 65C.  Works perfect and is utterly quiet when not encoding.

If your routine useage doesn't include continuous, long duration intensive compute tasking with all 6 cores/12 threads heavily loaded, e.g., video encoding/image rendering/scientific & engineering modeling or simulation, then you might just consider the VRM temp problem part of the bargain. While gaming probably can benefit from overclocking, it rarely loads more than a couple threads heavily and even then rather intermittently and short in duration.  That being the case the VRM is probably never going to have any real opportunity to get nearly that hot.  During stress tests, when it WILL get that hot for several hours, just be sure to cool things off by opening the case and directing a BIG fan at the VRM to keep it cool.

However, even for gaming you must keep the CPU cool.  The reason is even a single-core, short duration extreme load can pop high temps so you need a heat sink that will absorb it quickly to keep it stable.  That needs thermal mass, proper CPU TIM contact and a way to shed the heat.  And QUIETLY too!  Nobody should have to live with a turbine engine sharing their desktop anymore!

A 240mm cooler makes a major difference with an 8 core...mine keeps my 1700 @ 3.95Gig around 60C during any encoding I've done and even stress tests at or around 70C.  I have to think it would also be the best choice for a 6 core IF your goal is to push your overclock up to the 3.8-3.9Ghz level any Ryzen processor should be capable of.
Logged

blackr92Topic starter

  • CORPORAL
  • ****
  • Offline Offline
  • Posts: 18
« Reply #38 on: 09-November-17, 14:07:25 »

How I approached VRM cooling with an AIO cooler is I modded an old blank plate by drilling a couple holes in it to attach a 70mm CPU fan. I then attached the blank plate with the fan to one corner of the case fan and directed it's airflow towards the VRM.  Put it on one of the four fan headers all alone and set a profile so it only turns on when CPU temp goes above 65C.  Works perfect and is utterly quiet when not encoding.

If your routine useage doesn't include continuous, long duration intensive compute tasking with all 6 cores/12 threads heavily loaded, e.g., video encoding/image rendering/scientific & engineering modeling or simulation, then you might just consider the VRM temp problem part of the bargain. While gaming probably can benefit from overclocking, it rarely loads more than a couple threads heavily and even then rather intermittently and short in duration.  That being the case the VRM is probably never going to have any real opportunity to get nearly that hot.  During stress tests, when it WILL get that hot for several hours, just be sure to cool things off by opening the case and directing a BIG fan at the VRM to keep it cool.

However, even for gaming you must keep the CPU cool.  The reason is even a single-core, short duration extreme load can pop high temps so you need a heat sink that will absorb it quickly to keep it stable.  That needs thermal mass, proper CPU TIM contact and a way to shed the heat.  And QUIETLY too!  Nobody should have to live with a turbine engine sharing their desktop anymore!

A 240mm cooler makes a major difference with an 8 core...mine keeps my 1700 @ 3.95Gig around 60C during any encoding I've done and even stress tests at or around 70C.  I have to think it would also be the best choice for a 6 core IF your goal is to push your overclock up to the 3.8-3.9Ghz level any Ryzen processor should be capable of.
Between all 3 coolers i cant go wrong with any of them right?
Logged

buddyw53

  • SECOND LIEUTENANT
  • **
  • Offline Offline
  • Posts: 102
« Reply #39 on: 09-November-17, 14:50:58 »

Between all 3 coolers i cant go wrong with any of them right?
I can only comment on the approach since I really couldn't say about those specific coolers...I"m not even sure they're available in the US.  Asetek has limited the marketplace offerings in US for AIO coolers through restrictive patent licensing for combination waterblock/pump (the part that fits on the CPU) designs .

Some coolers work pretty well but the fans have to be screaming to get the performance.  My advice would be to search for reviews that compare them to competition for cooling performance and fan noise.
Logged

binderasaf1

  • PRIVATE E-2
  • **
  • Offline Offline
  • Posts: 3
« Reply #40 on: 13-March-18, 12:58:25 »

That's a long one. Didn't read all...
In case you didn't manage yet, one word , soc voltage.
Culprit in Manny occasions. Put it at 1.15 and try stressing again. You might be surprised.
Logged
Pages: [1]   Go Up