Ryzen Overclock thread. JUST OC results, and hardware/settings

[Soulkeeper]

crazy
we used to run our 90nm chips at that

 

[imported_jjj]

crazy
we used to run our 90nm chips at that

Ryzen does best up to 1V, above that the efficiency decreases fast.

 

[Elfear]

How did you get such a low voltage at 3.8? Is it just temp based? Or is your sample golden and mine is brick?

Anything below 1.325v at 3.8Ghz on my 1700 and in Handbrake it will crash and reboot.

Also under load CPU-Z and AI Suite are showing between 1.34 and 1.375v

I think a lot of it has to do with temp as I'm on custom water. Part of it has to do with luck though I'm sure. I need to check the Asus CH6 section of HWinfo and make sure the voltage I listed is accurate. Someone over at OC.net (Elmor I think) said the most accurate CPU temp value is located in the motherboard section and I was taking my reading from the CPU section.

 

[Mr Evil]

crazy
we used to run our 90nm chips at that

Welcome to the post-Dennard scaling world!

 

[Soulkeeper]

There is a solid 10c difference going from 1.35 to 1.40v here
surprisingly overrall system power usage is about the same as my llano system

 

[Shlong]

3.9ghz @ 1.30v (set in Taichi bios) stable, so far for me in the limited amount of testing. If I set to 4.0ghz, voltage and temperatures rise a lot.

 

[malitze]

1800X @ 3.9 with +0.03125v (1.38125v for p0). Also hitting a wall when trying to get to a stable 4GHz.

 

[CraptacularOne]

Had a unfortunate mishap with my Gigabyte X370 Gaming motherboard while transplanting my system to a new case. I picked up a MSI Krait Gaming board in the mean time while I wait on an RMA but I think I'm going to keep this MSI board instead. I can run my memory at it's full 3200Mhz 16-16-16-36 1T speed and timings with this board without issue. I still have the same CPU overclock of 4Ghz with 1.35v but I can finally get all that my memory is capable of. Even with the latest BIOS on my Gigabyte X370 memory could not run past 2933Mhz without crashing.

 

[Agent-47]

I am getting a bit confused here. but vcore/vid readouts are different in different software and not what I expect. maybe someone how has been dealing with ryzen for the past month can sort me out.

in the MSI BIOS, i have manually set the volts to 1.275v. LLC is off. Freq 3.7Ghz

In HWMonitor> VID is 1.231v and VCORE is 1.280v at ideal and 1.27 at load. Why is the VID stuck at 1.231, it should report what I requested.

in HWINFO> VID is 1.231v on core 0 and 1.169v on other cores. ??? Also VCORE is 1.265 at idea and 1.2v at load. This is really strange as it does not match with HWMonitor or AIDA64.

I changing the VCORE in BIOS to 1.35 V LLC1 in BIOS, just to see what happens, as well as increase freq to 3.8Ghz. VID does not change in either HWMonitor or HWiNFO. ???

In the later case, VCORE in HWMonitor and AIDA64 is 1.36 v, but both BIOS status and HWiNFO states 1.34 during idea and 1.27v during stress.

OC is not stable beyond 45 min stress-test despite temps being only 59-60 C. system does not reboot but AIDA64 haults tests saying Hardware failure detected but does not say what failed.

I have been pulling my hairs out for the last few hours over this. Which do i believe??

Spoiler

 

[Zoffster1]

Without spending too much time on it I get the following:

CH6 1800x 4033MHz @ 1.42v
2x16GB G-Skill cl14 2988MHz @ 1.35v

Strangely I get different voltage readings in HWinfo and HW monitor

HWmonitor reads it as 1.417 going up to 1.439 under load at times based on MB sensor
HWinfo 1.375 going down to 1.369 under load (?) based on cpu sensor

I'm probably going back to stock for now until we get a few more bios updates.

 

[Agent-47]

Without spending too much time on it I get the following:

CH6 1800x 4033MHz @ 1.42v
2x16GB G-Skill cl14 2988MHz @ 1.35v

Strangely I get different voltage readings in HWinfo and HW monitor

HWmonitor reads it as 1.417 going up to 1.439 under load at times based on MB sensor
HWinfo 1.375 going down to 1.369 under load (?) based on cpu sensor

I'm probably going back to stock for now until we get a few more bios updates.

What does you bios say for ideal vcore

 

[Despoiler]

I am getting a bit confused here. but vcore/vid readouts are different in different software and not what I expect. maybe someone how has been dealing with ryzen for the past month can sort me out.

in the MSI BIOS, i have manually set the volts to 1.275v. LLC is off. Freq 3.7Ghz

In HWMonitor> VID is 1.231v and VCORE is 1.280v at ideal and 1.27 at load. Why is the VID stuck at 1.231, it should report what I requested.

in HWINFO> VID is 1.231v on core 0 and 1.169v on other cores. ??? Also VCORE is 1.265 at idea and 1.2v at load. This is really strange as it does not match with HWMonitor or AIDA64.

I changing the VCORE in BIOS to 1.35 V LLC1 in BIOS, just to see what happens, as well as increase freq to 3.8Ghz. VID does not change in either HWMonitor or HWiNFO. ???

In the later case, VCORE in HWMonitor and AIDA64 is 1.36 v, but both BIOS status and HWiNFO states 1.34 during idea and 1.27v during stress.

OC is not stable beyond 45 min stress-test despite temps being only 59-60 C. system does not reboot but AIDA64 haults tests saying Hardware failure detected but does not say what failed.

I have been pulling my hairs out for the last few hours over this. Which do i believe??

Spoiler

I noticed the same thing. Those programs do not have the same voltage values and my BIOS is different from both of them. I have nothing, but a rough estimate of how much voltage I'm using. Ahh gotta love the wild west of a new platform!

 

[IEC]

On my MSI board the VID reported in BIOS changed depending on what clockspeed I set for manual OC. As if it was referring to a voltage table for the chip.

LLC settings will change your voltage readings and spread, sometimes strangely (or opposite of what you may be used to e.g. ASRock). If you are really dropping 0.07V during stress you should try LLC level 3 and see if that helps with stability.

 

[Agent-47]

After reading about it today, and from what I understand,

VID -> is a CPU value that it provides to the Motherboard to initialization. IIRC, it should be the same as what I assigned in the BIOS. in any case this is not interesting as the VCore is assigned correctly at idle at least.

VCore (from HWMonitor)-> is what the motherboard tells the VRM to supply to the CPU using PWMs. Now it seems the MB is giving the right value for idling CPU. but why does it not increase when the CPU is under load. that is what LLC is supposed to do in order to tell the VRMs to supply more current to the CPUs while not suffering from VDdroop. The source of this value is the motherboard itself

VCORE (SVI2 TFN from HWiNFO) is directly read from the VRM outputs that goes directly to the CPU input. this is what the CPU sees. it is okay for SVI2 TFN to fall during load, but that drop seems too sharp for me. the source of this input is the CPU


=================
Can someone with 1.35v or higher OC please download HWiNFO and please confirm that they also have such sharp drop in SVI2 TFN.

I am concerned that it might be a symptom of poor VRM, so either all MSI mortars are utter crap, or I have to RMA the board.

 

[Agent-47]

On my MSI board the VID reported in BIOS changed depending on what clockspeed I set for manual OC. As if it was referring to a voltage table for the chip.

LLC settings will change your voltage readings and spread, sometimes strangely (or opposite of what you may be used to e.g. ASRock). If you are really dropping 0.07V during stress you should try LLC level 3 and see if that helps with stability.

for me LLC seems to be virtually ineffective. Went from Mode 1 to Mode 8.

Can you please check HWiNFO to confirm if your VDdroop is as large as mine. this is rather strange.

 

[Crumpet]

R7 1800x on Asus C6H
4.1ghz @1.39v

And I have the overclocking skills of a drunken chimpanzee, so I'm happy with that!

 

[IEC]

for me LLC seems to be virtually ineffective. Went from Mode 1 to Mode 8.

Can you please check HWiNFO to confirm if your VDdroop is as large as mine. this is rather strange.

My vcore doesn't seem to droop much or at all as reported by Aida64. I'm using Auto-LLC settings though.

 

[Crumpet]

Ooh.. managed to run a firestrike bench at 4.2ghz (reported as 4.192) succesfully..

Using auto voltage and LLC 2.

 

[CuriousMike]

You got the golden chip.

Ooh.. managed to run a firestrike bench at 4.2ghz (reported as 4.192) succesfully..

Using auto voltage and LLC 2.

 

[Crumpet]

You got the golden chip.

It seems so, though I can't get it stable under 1.45v so i'm "settling" for a daily driver of 4.15ghz @1.39v

THAT is what I paid the extra cash for over the 1700.

 

[guskline]

It seems so, though I can't get it stable under 1.45v so i'm "settling" for a daily driver of 4.15ghz @1.39v

THAT is what I paid the extra cash for over the 1700.

"Daily driver" at 4.15 is darn fast!

 

[malitze]

for me LLC seems to be virtually ineffective. Went from Mode 1 to Mode 8.

Can you please check HWiNFO to confirm if your VDdroop is as large as mine. this is rather strange.

I would advise using the latest beta of HWiNFO since it is very actively developed regarding Ryzen. And don't worry about the droop, mine does that as well. When trying to get 4GHz stable I had drops as low as ~1.375V from 1.4675V during load.

 

[Crumpet]

Indeed. I'm half tempted to put the power back to auto and see if it could bench at 4.3.. Don't want to risk damaging anything though as i'm an utter overclock newbie.

Unfortunately i'm held back by my early adopter tax of pre-purchasing my ram, so thats stuck at 2666 cas16, which probably balances me out vs you other guys. Guess I have a good reason now to convince the wife to let me splash on some Samsung B dies to get that bit more out of it. The benefit being she can have these sticks for her R5 1600 build that i'm slowly convincing her into....

 

[Agent-47]

I would advise using the latest beta of HWiNFO since it is very actively developed regarding Ryzen. And don't worry about the droop, mine does that as well. When trying to get 4GHz stable I had drops as low as ~1.375V from 1.4675V during load.

1.475v? you are a brave man. I went to 1.425v to get to 3.9Ghz semi-stable (45 mins of AIDA64, and then it crashed, PC was still running) and then did not push any harder.

3.7Ghz seems to be the sweet spot for me at 1.25 BIOS input with LLC1. VCORE droops to 1.2v minimum. also can be achieved via 1.265v BIOS and LLC2-- not sure which is the better option though.

What board are you on?

 

[richierich1212]

Indeed. I'm half tempted to put the power back to auto and see if it could bench at 4.3.. Don't want to risk damaging anything though as i'm an utter overclock newbie.

Unfortunately i'm held back by my early adopter tax of pre-purchasing my ram, so thats stuck at 2666 cas16, which probably balances me out vs you other guys. Guess I have a good reason now to convince the wife to let me splash on some Samsung B dies to get that bit more out of it. The benefit being she can have these sticks for her R5 1600 build that i'm slowly convincing her into....

You can save money by waiting to see what improvements the May update brings.