22 minutes ago, John_M said:

 

Nobody had mentioned power supplies until now but since you have, the freezing issue also affects people with modern (so-called "Haswell-compatible") power supplies. But if the slight overvolt helps with that problem then it might be an even more useful way to combat the freezing.

Yes, sorry my mistake, probably PSU aren't the issue, mine is quite new, and most people is reporting the same issue with "compatible PSU".

Probably is a problem with AMD, and in the case we think is the PSU maybe Intel is masking this issue with some kind of driver or special setting in windows / linux. In fact with Ryzen/Windows this doesn't happend becase maybe windows is treating C6 in a different way, but considering that in Intel/linux it "works" adds weight to the AMD side.

https://www.eteknix.com/amd-releases-balanced-power-plan-patch-for-windows-10/

Edited March 24, 20188 yr by L0rdRaiden

7 minutes ago, pwm said:

AMD have been too optimistic and programs the PMIC to output too low voltages in the lowest power states

 

If AMD were to acknowledge this, surely they could fix it quite easily with a microcode update?

3 minutes ago, John_M said:

 

If AMD were to acknowledge this, surely they could fix it quite easily with a microcode update?

 

I have my doubts, looking at the thread there are several histories of many people that have notified them about this issue, they are aware of it, and nothing has happend yet.

The bug is open since 8/2017

Is someone of you actually running a Raven Ridge CPU in his unRAID build?

I've got some hard crashes with it and its defenitly not the C6 state problem, since it is disabled via the script and it was working with my AMD Ryzen 7 1700X..

Am I the only one with this issue?

I'm running the latest beta: 6.5.1 rc3

On 3/24/2018 at 1:15 PM, John_M said:

 

If AMD were to acknowledge this, surely they could fix it quite easily with a microcode update?

I don't know if it's part of the microcode, or the firmware of the little slave "supervisor" processor or if it is code in the Linux kernel. Or possibly code in the BIOS itself.

 

But if AMD have incorrect voltage settings for deeper power modes, then it's extremely likely that AMD can directly or indirectly release a fix. But it's an open question how much work it is to release a fix, and what spare resources AMD has for this - they don't dare make this type of changes without spending significant time with testing and we don't know if they are maybe almost burning to cinders with the workload of getting their new models out the door. Maybe AMD just doesn't have the manpower to fix what they consider "minor" problems - especially if they might have managed to get corrective Windows driver code to M$ so it's "only" Linux system that suffers.

 

AMD has a rather tough time - it isn't easy to compete with Intel since so many of the fortune-500 companies are so scared to try AMD processors.

11 minutes ago, pwm said:

AMD has a rather tough time - it isn't easy to compete with Intel since so many of the fortune-500 companies are so scared to try AMD processors.

 

Yes, it's amazing they lasted for long enough to come back from the dead. The odds really were stacked against them, not just because of their size, their lack of resources and the enormous expense of developing a new product, but because of the despicable foul play going on in the background. They are also fighting on two fronts with nVidia playing as dirty as Intel.

On 3/14/2018 at 4:03 PM, eschultz said:

we're open to including Vega iGPU drivers when we upgrade kernels

 

Hubba Hubba

On 4/2/2018 at 11:09 AM, reey said:

Is someone of you actually running a Raven Ridge CPU in his unRAID build?

I've got some hard crashes with it and its defenitly not the C6 state problem, since it is disabled via the script and it was working with my AMD Ryzen 7 1700X..

Am I the only one with this issue?

I'm running the latest beta: 6.5.1 rc3

From info tab:

 

M/B: ASRock - X370 Pro4

CPU: AMD Ryzen 5 2400G with Radeon Vega Graphics @ 3600

HVM: Enabled

IOMMU: Enabled

Cache: 384 kB, 2048 kB, 4096 kB

Memory: 16 GB (max. installable capacity 256 GB)

Network: eth0: 1000 Mb/s, full duplex, mtu 1500

Kernel: Linux 4.14.31-unRAID x86_64

 

6.5.1 rc3

 

It crashed twice within about 5-6 hours with c states enabled and the zenstates --c6-disable fix.

 

Runs rock solid with c-states disabled in the bios so far, maybe 3 weeks or so.

 

The bios does not have the option to disable only c-6.

 

17 minutes ago, FlynDice said:

From info tab:

 

M/B: ASRock - X370 Pro4

CPU: AMD Ryzen 5 2400G with Radeon Vega Graphics @ 3600

HVM: Enabled

IOMMU: Enabled

Cache: 384 kB, 2048 kB, 4096 kB

Memory: 16 GB (max. installable capacity 256 GB)

Network: eth0: 1000 Mb/s, full duplex, mtu 1500

Kernel: Linux 4.14.31-unRAID x86_64

 

6.5.1 rc3

 

It crashed twice within about 5-6 hours with c states enabled and the zenstates --c6-disable fix.

 

Runs rock solid with c-states disabled in the bios so far, maybe 3 weeks or so.

 

The bios does not have the option to disable only c-6.

 

 

Do this start with 0.1 volts just in the CPU

 

 

Edited April 8, 20188 yr by L0rdRaiden

Working on about 30 hours so far, no problems vs 2 lockups at around 6 hr point with c-states enabled before.

 

Enabled c-states in bios, commented out the zenstates --c6-disable fix for now,  & upped the cpu volts from 1.45 to 1.55

On 10/04/2018 at 6:42 PM, FlynDice said:

upped the cpu volts from 1.45 to 1.55

 

1.45 volts seems rather high, 1.55 volts dangerously so.

55 minutes ago, John_M said:

 

1.45 volts seems rather high, 1.55 volts dangerously so.

What's it supposed to be at then.  It was 1.45 before I changed anything.  I guess some research is in order....

I would not want to exceed 1.30 volts for the SOC or 1.45 volts for the cores. Left on auto (BIOS default) the core voltage varies but on my 2400Gs it never goes above 1.375 volts. I would really not be happy raising it to 1.55 volts. I'm suspicious of yours being at 1.45 volts without any adjustment. Have you tried resetting the BIOS defaults?

 

From the earlier discussion, I can see how raising the minimum voltage might help with lock-ups but this is a work round at best, not a fix, because the maximum voltage is also raised and your figures scare me. So rather than do that I would disable C-states globally if I had to.

10 minutes ago, John_M said:

I would not want to exceed 1.30 volts for the SOC or 1.45 volts for the cores. Left on auto (BIOS default) the core voltage varies but on my 2400Gs it never goes above 1.375 volts. I would really not be happy raising it to 1.55 volts. I'm suspicious of yours being at 1.45 volts without any adjustment. Have you tried resetting the BIOS defaults?

 

From the earlier discussion, I can see how raising the minimum voltage might help with lock-ups but this is a work round at best, not a fix, because the maximum voltage is also raised and your figures scare me. So rather than do that I would disable C-states globally if I had to.

Thanks for the Heads Up on this.

 

After reading up a bit I have to agree with everything you've said.  I'm trying manual 1.3 v for now on the cpu voltage to see how it works, although the hardware readings in the bios are telling me the cpu voltage is still hitting 1.37 on this setting.  It seems the ASRock bios defaults to 1.45 v for several of their boards when you enable manual vs auto.  Bad on them I would say...  To their credit the voltages over 1.44 do show as red but there's no info as to what this means.  I thought it was just alerting me that I was setting things manually.

 

15 minutes with fingers crossed so far....

Just an update on what worked & didn't work.

 

For cpu volts, manual 1.3 & 1.4 locked within 6 hours each,  zenstates --c6-disable locked at 18 hours again.

 

Back to c-states disabled for now which seems to be the safe haven for now....