Ryzen 1700 system woes

Sojoe

n00b
Joined
Nov 24, 2014
Messages
34
Hi All,

So a year and a half ago I build myself a system with the following key components.

AMD Ryzen 7 1700 Processor (replaced last summer due to segfaults).
MSI B450I AC Gaming ( downsize from an MSI X370 Gaming Pro Carbon)
Corsair Vengeance LPX DDR4 2666MHz 32GB @ 2133
Corsair AX 850 (from 2011)
Crucial m4 256GB SSD (from 2012, newest firmware)

I run Linux (Manjaro) exclusively on it.

The system, was bought for development and simulation projects, but has never been in good enough shape to do any real work.

First it was 1 year of random segfaults and simulation crashes (faulty 1700). Then after the cpu replacement the segfaults have stopped, but the system has become unstable during daily use, it stalls 3-4 times a week.

I am this close to throwing the system to the wolves and just build a new one... I am all for AMD, but this situation kind of reminds me of the woes I had on my shitty bamboo PC in the mid 90's.

If you have any suggestions as to what the problem could be I would much appreciate it.

Cheers,
 

drescherjm

[H]F Junkie
Joined
Nov 19, 2008
Messages
14,650
Update your BIOS. I had my infrequent random hard lockups on my Ryzen 7 2700 ( under linux ) fixed by a BIOS update.
 
Last edited:

Sojoe

n00b
Joined
Nov 24, 2014
Messages
34
Thanks, I have the newest BIOS, and will be moving the system into an Ncase M1, so I will have to replace the PSU, so I will try that first :)
 

drescherjm

[H]F Junkie
Joined
Nov 19, 2008
Messages
14,650
BTW, I am using linux on the 2700. However it's gentoo. When I was having the random hard freezes I ended up making an Arduino based hardware watch dog to have it detect and reboot the system. However just after I had the watchdog running a BIOS update fixed the issue and the system has run fine for months (24/7).

Here is a link to the info on the watchdog:

https://hackaday.com/2018/02/23/arduino-watchdog-has-bite-and-doesnt-need-treats/

I replaced the relay with this one:

https://www.amazon.com/gp/product/B079FJSYGY/ref=oh_aui_search_asin_title?ie=UTF8&psc=1

I set it to high level triggered.

And made some changes in the software here:
https://github.com/drescherjm/hardware-watchdog-arduino
 
Last edited:
  • Like
Reactions: Sojoe
like this
Joined
Jul 26, 2018
Messages
39
Different people seem to have different fixes for the idle hangs (and they may not even all be the same problem), but what did it for me was disabling global C-state control (called that on at least both ASRock and Asus boards, and it disables data fabric C-states). The extra power use is negligible, and it's much better in every way than bumping Vsoc up to the moon as AMD suggested when I talked to them about it.
 

Sojoe

n00b
Joined
Nov 24, 2014
Messages
34
BTW, I am using linux on the 2700. However it's gentoo. When I was having the random hard freezes I ended up making an Arduino based hardware watch dog to have it detect and reboot the system. However just after I had the watchdog running a BIOS update fixed the issue and the system has run fine for months (24/7).
<SNIP>
Thanks alot I will definitely be trying that out.

Different people seem to have different fixes for the idle hangs (and they may not even all be the same problem), but what did it for me was disabling global C-state control (called that on at least both ASRock and Asus boards, and it disables data fabric C-states). The extra power use is negligible, and it's much better in every way than bumping Vsoc up to the moon as AMD suggested when I talked to them about it.
And this too.
 

pitingres

Weaksauce
Joined
Jul 25, 2018
Messages
95
I've had crazy hangs with 5.0.x and Asrock Taichi BIOS 2.0. The older BIOS had a "disable c6 state" which isn't in the newer one. I was blaming the hangs on my RX 550, but disabling global C-state control seems to work. (also kernel command line processor.max_cstates=1, but I had trouble waking out of sleep with the latter. early days yet to see which is better. Disabling global C-state control costs 10w at idle, 120w draw at the UPS vs 110+.)
 
  • Like
Reactions: Sojoe
like this
Joined
Jul 26, 2018
Messages
39
Disabling global C-state control costs 10w at idle, 120w draw at the UPS vs 110+.
Huh, that's interesting, the gap was <2W for me. Is this the DDR4-3200 in your sig? That means Vsoc will be a lot higher if you haven't changed it (I'm at 1.00 Vsoc on DDR4-3000), but usually high Vsoc (as in the default at high RAM / DF clocks) gets rid of the hangs itself. I haven't heard many reports of it on DDR4-2800 or higher (high Vsoc threshold on OG Zen) and don't think I've heard any on Zen+; if global C-state control fixes it, that's great, but it may be a different underlying cause.
 

pitingres

Weaksauce
Joined
Jul 25, 2018
Messages
95
Yes, it's the memory in the sig. I haven't touched Vsoc, might take a look next time I reboot. It appears to be stable again, although I won't be sure for another week or so; but I was getting a freeze every few hours before, so something it definitely better.
 

Burticus

Supreme [H]ardness
Joined
Nov 7, 2005
Messages
4,163
Yeah I'd lean towards different ram. I had a set of Patriot DDR4 3000 ram that my Ryzen 1700/asrock ab350 just hated. I could post and be functional for most part at 2133 (but who wants that) but random lockups all the time. Wierd bizarro stuff. Changed the ram and all the problems went away (and the ram worked just fine in an intel z170 system).

win10 tho
 

Sojoe

n00b
Joined
Nov 24, 2014
Messages
34
I have ordered a Corsair SF750 (near same price as 600 so..:)), and a Samsung 970Evo+ Nvme.

Will probably try a set of memory modules too.

Cheers
 

sirmonkey1985

[H]ard|DCer of the Month - July 2010
Joined
Sep 13, 2008
Messages
21,769
Yes, it's the memory in the sig. I haven't touched Vsoc, might take a look next time I reboot. It appears to be stable again, although I won't be sure for another week or so; but I was getting a freeze every few hours before, so something it definitely better.
definitely try a higher voltage on the memory as well and if your board has LLC support for memory try setting that to 3 and see if it helps. on my 1600 system in sig i had to run the ram at 1.38v and on my other system i have with a 2600x on a b450 board i had to run the memory at 1.37v. neither would run XMP at their rated 1.35v and both showed the same random freezing symptoms without actually causing the system to crash. if that doesn't help then yeah i'd look at different memory but would be hard to believe it's a compatibility issue since the 3200CL14's are samsung B-die. if they were CL16 hynix then i'd lean more toward it being a memory compatibility issue.
 

pitingres

Weaksauce
Joined
Jul 25, 2018
Messages
95
This memory and setup was rock solid for months before the mobo BIOS update in January; I was running the memory at 3133 and everything else auto. With the BIOS update I was finally able to (apparently) stabilize at 3200 with 1.36v on the memory, but I'd get a hard freeze every couple days and I dropped back to the old config. The freezes continued and with newer kernel versions, it got much worse to where it was freezing in various ways several times a day. After disabling global C-state control, I'm back on the 3200 MT/s config and it's been solid for a few days now. Fingers crossed. (FWIW I also have Idle Current Typical set.) I think what happened is that the 2.0 bios deleted the separate C6 state control, which I had had off, and it would lock up hard when linux tried to put it into a lower power state.

If I still get rare freezes at 3200 I'll try LLC 3 and take a look at the soc voltage, maybe tip the ram up another notch. I'm not too worried about getting 3200 vs 3133, but I'll take the higher number if I don't have to work too hard for it!
 

Zareek

Limp Gawd
Joined
Sep 5, 2011
Messages
191
definitely try a higher voltage on the memory as well and if your board has LLC support for memory try setting that to 3 and see if it helps. on my 1600 system in sig i had to run the ram at 1.38v and on my other system i have with a 2600x on a b450 board i had to run the memory at 1.37v. neither would run XMP at their rated 1.35v and both showed the same random freezing symptoms without actually causing the system to crash. if that doesn't help then yeah i'd look at different memory but would be hard to believe it's a compatibility issue since the 3200CL14's are samsung B-die. if they were CL16 hynix then i'd lean more toward it being a memory compatibility issue.
I agree, Ryzen is RAM picky. The default Corsair mem voltages are too low bump that up a bit and bump soc voltage a bit and the ram will be more stable. You can also just get a Ryzen approved RAM kit. I got the G.Skill kit after I had issues getting the Corsair LPX 3200 kit to run at 3200 regardless of settings. With the G.Skill kit I installed and set it to XMP 3200 profile and never looked back.
 

pendragon1

[H]F Junkie
Joined
Oct 7, 2000
Messages
15,997
i just worked on a 2600x system this weekend were the ram was defaulting to 1.2v at 3200 and it would cause all sorts of issues/bsod. id set it to 1.35v and no more problems, until i figured out that a full power disconnect would reset it to 1.2v until i rebooted once then the voltage would jump back up. apparently its a known issue on GB x470 aurous ultimates. from what i remember from around here is that anything over 3000 has to be 1.3v+.

this still sounds like a psu issue though
 

Sojoe

n00b
Joined
Nov 24, 2014
Messages
34
All good information, I finally have the time to rebuild the system this weekend, will leave it at idle and check :) Really lookning forward to building in the Ncase M1.
 

Sojoe

n00b
Joined
Nov 24, 2014
Messages
34
Have run the new system for approx 8 hours now, with a new SSD, CPU cooler, case and PSU (SF-600 i pulled from another system, I have other plans for the 750). I re-enabled my XMP memory profile.

I think I may have found the root cause. System stalled twice when I mounted and subsequently accessed my (ancient) Crucial SSD (to retrieve some settings and 'home' data). So It was probably a powerstate issue on my Crucial.

The system has been in and out of idle without issue since.

Fingers crossed.
 

elite.mafia

Broke Back [H]
Joined
Aug 23, 2004
Messages
12,150
If it's not that, I'd suspect a ram configuration issue as that is probably the cause of 99% of ryzen instability issues. Ryzen is super picky on RAM settings. ESPECIALLY 1st gen.
 

Sojoe

n00b
Joined
Nov 24, 2014
Messages
34
If it's not that, I'd suspect a ram configuration issue as that is probably the cause of 99% of ryzen instability issues. Ryzen is super picky on RAM settings. ESPECIALLY 1st gen.
Yes, that is still on the table... Despite me picking a set of ram that was on the mobo manufactures compatibility list.
 

tangoseal

[H]F Junkie
Joined
Dec 18, 2010
Messages
8,272
I agree, Ryzen is RAM picky. The default Corsair mem voltages are too low bump that up a bit and bump soc voltage a bit and the ram will be more stable. You can also just get a Ryzen approved RAM kit. I got the G.Skill kit after I had issues getting the Corsair LPX 3200 kit to run at 3200 regardless of settings. With the G.Skill kit I installed and set it to XMP 3200 profile and never looked back.
Ryzen 1 is ram picky
 

TheSlySyl

Limp Gawd
Joined
May 30, 2018
Messages
252
I'd also say its likely the ram. I'm using a launch 1700 with a launch AsRock X370 Gaming K4 and every single BIOS update required me redoing the memory timings from scratch, even if they were perfectly stable before. Something had to be changed to get it back into stability. Bumping up the voltage was an absolute must, after that there was just weirdness. Never actually got it stable at a full 3000mhz (but 2800 was good enough after the amount of time I spent getting it stable at that speed. ) After that I'd look at the SSD, honestly you're probably due for a new one anyway.

Getting my currently unbalanced 48GB of RAM to work required a ridiculous amount of reboots and weird juggling of sticks. It's stable right now though!
I should mention that my CPU is overclocked to 3.9ghz at 1.248v at the moment.
 

Below Ambient

[H]ard|Gawd
Joined
Dec 5, 2001
Messages
1,441
ram board and cpu are all mismatched... b450 really doesnt have the VRMS for an 8 core, not to mention MSI is a bit shady... that ram is godawful and more than likely the key source of your issues... a psu from 2011 may not be deliver the correct power required for ryzen, def try a different one... my bet is you change the ram out for something a bit better and your woes might go away..
 

tangoseal

[H]F Junkie
Joined
Dec 18, 2010
Messages
8,272
Okay, this thread is about a Ryzen 1700...
You said Ryzen is ram picky
But in fact Ryzen 2 is not ram picky.

Ryzen has two vastly different generations in as far as ram is concerned. It's just a semantic thing nothing more. Not being snarky.
 

Stanley Pain

2[H]4U
Joined
Apr 5, 2001
Messages
2,478
ram board and cpu are all mismatched... b450 really doesnt have the VRMS for an 8 core, not to mention MSI is a bit shady... that ram is godawful and more than likely the key source of your issues... a psu from 2011 may not be deliver the correct power required for ryzen, def try a different one... my bet is you change the ram out for something a bit better and your woes might go away..
The VRMs are more than enough to power that CPU and the PSU is a solid one and should be fine (put testing it would be a good idea none the less).

Sojoe Try the C6 disable trick I posted above. It solved all my Ryzen woes under Linux.
 

Zareek

Limp Gawd
Joined
Sep 5, 2011
Messages
191
You said Ryzen is ram picky
But in fact Ryzen 2 is not ram picky.

Ryzen has two vastly different generations in as far as ram is concerned. It's just a semantic thing nothing more. Not being snarky.
Good to know, I'm holding out for Ryzen 3xxx(Zen 2). Hopefully, it is as RAM friendly as Ryzen 2xxx(Zen+) apparently is.
 

Sojoe

n00b
Joined
Nov 24, 2014
Messages
34
System still seemingly running stable. I have run several stress tests and its my main desktop so I use it daily.

If I run into another stall, I still have a nice checklist to try off (c-state and memory), thanks to your help. If it comes down to memory I might just try a Zen+ instead.
 

Sojoe

n00b
Joined
Nov 24, 2014
Messages
34
or you could just get proper ram
Proper RAM?

Don't know what you are talking about there!
The modules are listed as compatible at their rated speed on the motherboard manufactures website! They could be defective (unlikely) but their brand and model is very proper.
 

bobzdar

[H]ard|Gawd
Joined
Jun 6, 2003
Messages
1,736
Sounds like the SSD was the issue, especially if the ram is running at xmp without issue now and was having issues at 2133 before.
 

pooloffire

n00b
Joined
Jun 22, 2004
Messages
63
Proper RAM?

Don't know what you are talking about there!
The modules are listed as compatible at their rated speed on the motherboard manufactures website! They could be defective (unlikely) but their brand and model is very proper.
They never read the rest of the post. They only read your first post and didn't see that you got new ram, and power supply.
 
Top