[Linux] Freezes on Asrock X370 Taichi + C6 enabled
Printed From: ASRock.com
Category: Technical Support
Forum Name: AMD Motherboards
Forum Description: Question about ASRock AMD motherboards
URL: https://forum.asrock.com/forum_posts.asp?TID=6397
Printed Date: 21 Dec 2024 at 9:22pm
Software Version: Web Wiz Forums 12.04 - http://www.webwizforums.com
Topic: [Linux] Freezes on Asrock X370 Taichi + C6 enabled
Posted By: shmerl
Subject: [Linux] Freezes on Asrock X370 Taichi + C6 enabled
Date Posted: 23 Oct 2017 at 6:57am
|
" rel="nofollow -
I have a problem with AMD based build. It periodically freezes the
system (rather randomly, and not very frequently). Strangely, sometimes
the cursor continues working, but everything else is frozen. Configuration CPU: AMD Ryzen 7 1700X Motherboard: Asrock X370 Taichi RAM: G.SKILL Flare X (2x8 GB), DDR4 3200 (14-14-14-34) OS: Debian testing Linux x86_64 RAM is installed in second and fourth slots (A2, B2) as recommended by the motherboard manual. Currently in the EFI I have these settings: XMP 2.0 profile: DDR4-3200 14-14-14-34 1.35 V DRAM Frequency: DDR4-3200 AM4 Advance boot training: auto Voltage mode: stable DRAM voltage: 1.350 V VTT-DDR: 0.680 V all other settings on auto. In this build I already RMA'd the CPU (because of known AMD segfaults bug) and the motherboard (coil whine), and also tested it with a different power supply. The freezing problem persists. I tried a few tweaks before. Such as increasing DRAM voltage to 1.375 V, increasing ProcODT to 60 Ohm (separately from the DRAM voltage though). That didn't help. Can you recommend any way to fix this by adjusting any motherboard settings? Or the only way to fix is to RMA the memory, besides running it at lower frequency? |
Replies:
Posted By: Kevin A
Date Posted: 24 Oct 2017 at 3:31am
|
" rel="nofollow - Hi there, I read your situation and am curious regarding a couple of things. Have you used just 1 stick @3200mhz in slot B2? Does it freeze with the one stick as well? If not then swap in the second stick and try running it alone @3200mhz in slot B2. Also you may truly *need* higher CPU voltage and/or SOC voltage. Such as manually setting your CPU voltage to 1.35(or higher - yes...CPU) volts and manually setting your SOC voltage to 1.1 or higher in addition to your already set DRAM Voltage of 1.35 volts(mentioned in your post). Finally you may have to attempt SOC LLC settings for level 4,3,2 or even 1. Whatever is needed. I run 64GB and *need* LLC 2, although I have run it at SOC LLC level 1 as well when I ran 64GB @3200mhz. I currently run with slightly tighter timings @3066mhz and SOC LLC 2. Hope your build finds stability as you have already invested a lot of time and effort with you system. Best success! |
Posted By: shmerl
Date Posted: 24 Oct 2017 at 8:27am
" rel="nofollow -
Thanks. If that happens, it will demonstrate that one stick is defective? |
Kevin A wrote:Posted By: Kevin A
Date Posted: 24 Oct 2017 at 11:22pm
|
Not necessarily a defective DDR4 stick, could be however. Could be a bad BIOS flash, a weaker SOC(that operates some PCIE functionality, Integrated Memory Controller, etc.) on your 1700X. It could even be something in your Linux install/configuration. It could be a bad 'DDR stick pairing', which is usually indicated by serial numbers on the DDR4 sticks that are out of 'sequence' - but not always. Thus some 'basic' steps to figure out what is 'going on' needed to be suggested. At least some steps that I would (and have over the years with various builds) try. Again, hope you have success with your build and figure out the freezing issue. As for me I am thrilled with my Ryzen 1800X, and pretty much all of my components in my build... - cept still not thrilled with the BIOS from ASRock. I complement ASRock on a very 'high' quality component level design and build on the Fatality X370 Gaming Pro which I own. However I have fought with the ASRock BIOS revisions more than any MSI/Tyan/Asus/ECS or SuperMicro system I have ever owned/configured/supported/used. But nonetheless it 'is what it is' and 'soldier on' and stuff... |
Posted By: Peter13779
Date Posted: 25 Oct 2017 at 12:27am
|
I own the same config but little differend RAM. I had to downgrade my G.SKILL F4-3200C15D-32GVK (2x16GB CL15-15-15-35 1,35V) to 2933@CL18 1.45V to use any of the latest BIOS versions. Hopefully there is an update soooooon ... |
Posted By: shmerl
Date Posted: 25 Oct 2017 at 8:29am
" rel="nofollow -
Did it work for you before the update? Did you contact Asrock support for any recommendation on MB settings? I wonder if they can help with this. |
Posted By: parsec
Date Posted: 25 Oct 2017 at 12:17pm
" rel="nofollow -
In your first post, you asked about adjusting mother board settings to get your memory stable at 3200. You seem certain that it is a memory problem, but it is not clear to me why. It was suggested you increase the SOC voltage, which is a standard configuration for a Ryzen memory over clock. The default SOC voltage is 0.900V, increase it to 1.100V, Fixed voltage mode, LLC at 1 or 2. Did you try that? Another suggestion is as said above, reduce the memory speed to 2933 or 3066, at least for a test to see if the freezing stops. Have you ever used your memory at it default speed of 2400, without any freezing? Memory compatibility and over clocking speed are the worst issues Ryzen has had since its release. Any memory speed over 2667 is a memory OC, and not guaranteed. Ryzen processors were released to the public before AMD had its CPU microcode developed to the point it should have been. The first several AGESA versions were poor, only a few memory speeds were supported and only five memory parameters were available for configuration. Ryzen memory compatibility was, and still is to some degree, rather random and unpredictable. We have no idea which UEFI version you have installed. AMD has provided several AGESA CPU microcode updates that are part of several UEFI updates. The new AGESA versions are meant to provide enhanced memory compatibility, speed, and stability. In actual use most people saw improvements, but others did not.AMD is planning yet another AGESA update (1.0.0.7) in the near future, that is so different it will require a major rework of the current UEFI code. This demonstrates how immature the CPU microcode was upon Ryzen's release, and still is currently. I have the same G.SKILL F4-3200C14D-16GFX memory, used with an ASRock X370 Killer SLI/ac board and 1700X. Only the latest UEFI/AGESA versions allowed the memory to operate at 3200, and that is still setting the secondary and tertiary memory settings manually, the XMP profile is not optimal. That is both AMD's and G.SKILL's issue, but mainly AMD's. Since you are using Linux, what do you use for the AMD chipset drivers? ------------- http://valid.x86.fr/48rujh" rel="nofollow"> 
|
Posted By: Kevin A
Date Posted: 26 Oct 2017 at 9:56am
That is a significant downgrade for your memory. I would think that @2933 speed you could get C14 timings(possibly). Your memory is a decent kit so perhaps some additional attempts may net better performance at a lower voltage as well. Then again, if you are happy @2933Mhz speed for the here and now - then 'all is good'! |
Posted By: shmerl
Date Posted: 26 Oct 2017 at 3:01pm
I suspect it's RAM related, since I already RMA'd the motherboard and CPU both (each for different reasons though as I explained in my first post), but RAM is the same I had from the time I started building this system. Plus, it works without freezes at lower RAM frequency, see below.
Since I'm using XMP profile, the SOC voltage is already set to 1.104 V (EFI does that automatically when XMP profile is selected).
It's still freezing at 3066 MHz, but works well at 2933 MHz for me.
I'm using latest available from Asrock (3.20), which ships with AGESA 1.0.0.6b and microcode 0x8001129.
Thanks, that's good to know. Hopefully they can fix some memory related issues in it.
How exactly did you set these options?
Pretty much default out of the box kernel support (kernel 4.13.4 as it ships in Debian testing). I know for sensors for example, it's using https://github.com/groeck/nct6775" rel="nofollow - nct6779 . |
Posted By: shmerl
Date Posted: 27 Oct 2017 at 8:34am
|
Just got one freeze at 2933 MHz. |
Posted By: shmerl
Date Posted: 29 Oct 2017 at 11:36pm
|
" rel="nofollow - @parsec: so can you share your timings please? |
Posted By: shmerl
Date Posted: 17 Dec 2017 at 10:28pm
| " rel="nofollow - I think I noticed the pattern that causes the freeze. After normal boot everything works fine. But if I ever use suspend and then resume the computer, that's when the freeze happens after a while. I suppose the motherboard firmware does something wrong in case of resume from sleep (may be memory training doesn't work, or something of the sort). I suppose I should open a bug with Asrock? |
Posted By: zlobster
Date Posted: 17 Dec 2017 at 10:35pm
|
" rel="nofollow - Did you disable the C6 state? Some (older) PSU can't handle the C6 state correctly.
------------- 1700X ZP-B1 (stock); X370 Taichi (UEFI 3.10); 16GB F4-3200C14-8GFX XMP; 256GB 960 EVO; RX 580 NITRO+ 8GB |
Posted By: shmerl
Date Posted: 18 Dec 2017 at 12:17am
|
I doubt that's related here, since this happens both with disabled and enabled C6 and Cool'n'Quiet. And my PSU is certified to handle C6 fine. Also, the freeze doesn't happen right away after resume, but randomly, at some point after that. So it's more likely, that the motherboard doesn't initialize something properly on resume from suspend, and it eventually glitches out. |
Posted By: shmerl
Date Posted: 18 Dec 2017 at 1:36am
|
Well, scrap that. I just got a random freeze after a normal boot. But may be sleep / resume just makes it more likely. |
Posted By: zlobster
Date Posted: 18 Dec 2017 at 3:52am
|
" rel="nofollow - Things that come to mind - try another PSU; run some memtest from a live Linux USB; some UEFI leftovers from a flash; Does this happen with all left @ auto? ------------- 1700X ZP-B1 (stock); X370 Taichi (UEFI 3.10); 16GB F4-3200C14-8GFX XMP; 256GB 960 EVO; RX 580 NITRO+ 8GB |
Posted By: shmerl
Date Posted: 18 Dec 2017 at 9:24am
|
It did happen with another PSU. I already RMA'd CPU, motherboard and RAM (first two for different reasons, but still). It still happens :) I suspect it's a bug in Asrock UEFI when it works with this particular RAM. And it happens with auto too. I'd wish someone from Asrock could shed some light on what it can be. |
Posted By: datonyb
Date Posted: 18 Dec 2017 at 6:03pm
|
" rel="nofollow - maybe some further testing to help us -help you have you tested with the ram set at 2133 safe ? if not please do so ,it may help us determine if the ram needs tweaking ------------- [url=https://valid.x86.fr/jpg250]  [/url]3800X, powercolor reddevil vega64, gskill tridentz3866, taichix370, evga750watt gold |
Posted By: shmerl
Date Posted: 19 Dec 2017 at 10:45am
|
I'm running it now with auto auto (XMP and frequency). I think my previous tests with auto didn't set XMP profile to auto as well (but just the frequency), so I just want to make sure. Such auto setting sets RAM to 2400 MHz. If I'll get the freeze this way, I'll try 2133 (I'll post the result here either way). |
Posted By: datonyb
Date Posted: 20 Dec 2017 at 2:27am
|
ok its a start to try and work out the issue i feel i can help with ram settings if required (am4 and xmp makes a pigs ear of setting them stable) ------------- [url=https://valid.x86.fr/jpg250] [/url]3800X, powercolor reddevil vega64, gskill tridentz3866, taichix370, evga750watt gold |
Posted By: cannavaro
Date Posted: 20 Dec 2017 at 11:24am
|
" rel="nofollow - Do you happen to use a RAID configuration? If yes, this may help: https://community.amd.com/thread/214245 My Ryzen 1600 with Asrock Fatal1ty AB350 Gaming K4 is having a similar freezing issue to yours but it has been greatly reduced after trying many things. Now the issue is solely related to 'rcraid' I'm using RAID0 3x WD BLUE 1TB boot drive is klev neo 240gb m.2 ssd Everytime the system freezes I will see this in event viewer: quote: The description for Event ID 129 from source rcraid cannot be found. Either the component that raises this event is not installed on your local computer or the installation is corrupted. You can install or repair the component on the local computer. If the event originated on another computer, the display information had to be saved with the event. The following information was included with the event: \Device\RaidPort0 /end quote |
Posted By: shmerl
Date Posted: 20 Dec 2017 at 11:44am
|
Nope, I'm not using a raid, so it's not the same issue looks like. Also, I'm using Linux. |
Posted By: shmerl
Date Posted: 20 Dec 2017 at 11:46am
|
And it froze with auto RAM setting. I'm now running with XMP profile 2.0 with 2133 MHz for a test. |
Posted By: shmerl
Date Posted: 21 Dec 2017 at 10:19pm
|
" rel="nofollow - So far it's running smoothly with 2133 MHz. No freezes (or if they can happen still, they must be way less frequent than with above configurations). Any idea what to do next, to make higher frequencies more stable? |
Posted By: datonyb
Date Posted: 22 Dec 2017 at 1:50am
|
yes run the ryzen ram calculator takes about 20 mins please note follow the settings exactly and use all settings also manually turn off dram power down mode while in ram bios settings and make sure amd advanced boot training is set to on/auto http://www.overclock.net/t/1640919/ryzen-dram-calculator-overclocking-dram" rel="nofollow - http://www.overclock.net/t/1640919/ryzen-dram-calculator-overclocking-dram any settings marked 'alt' are to be used if first setting dosnt take properly select settings /results for safe first ,than if stable try settings for fast ------------- [url=https://valid.x86.fr/jpg250] [/url]3800X, powercolor reddevil vega64, gskill tridentz3866, taichix370, evga750watt gold |
Posted By: shmerl
Date Posted: 22 Dec 2017 at 4:18am
" rel="nofollow -
Seems like the instruction there says to use some Thaiphoon burner tool to read input values. I looked it up, it's a commercial tool, but the problem is, it's Windows only, so it won't work on Linux. Is there some reliable way to get those readings without relying on Windows only tools? I suppose I can get that info from G.Skill support sources. And I tried running the calculator itself in Wine, and got this:
I wish such tools would be open source, so they could be ported to Linux properly. Better even, they can simply be made as Web applications, making them usable from any browser. If I understand correctly, it simply uses a bunch of formulas to calculate memory timings based on set hardware parameters. So shouldn't require any Windows specific tooling. I can try commenting there, may be author could publish the code for the tool, or give some ideas how to calculate all those values on Linux. |
Posted By: datonyb
Date Posted: 22 Dec 2017 at 4:41am
|
option b find an old hard drive install windows ,run programs, get results ------------- [url=https://valid.x86.fr/jpg250] [/url]3800X, powercolor reddevil vega64, gskill tridentz3866, taichix370, evga750watt gold |
Posted By: shmerl
Date Posted: 22 Dec 2017 at 4:45am
|
I'm not that desperate yet as to buy or let alone pirate Windows to do that (didn't use Windows for a very long time already :) ). Also, I think tools like that calculator are pretty useful for Linux users, so may be the author can do something to make it properly cross platform. I can help with that even, as long as he publishes the source. Thanks for pointing out that forum by the way. May be some folks there can have ideas how to do it all on Linux. |
Posted By: shmerl
Date Posted: 22 Dec 2017 at 4:56am
|
" rel="nofollow - I have a question though. Why would values given by that calculator be different from what Asrock firmware is setting automatically when frequency is selected for XMP profile? Asrock didn't do a good job with automatic detection or timing formulas? And is the data that Thaiphoon tool generates, dynamically detected too and can end up different in different builds from same parts, or it just provides static info based on the motherboard / RAM? |
Posted By: PetrolHead
Date Posted: 22 Dec 2017 at 5:40am
|
The XMP profile contains values determined by the manufacturer and the BIOS reads the values from the profile. Even if it needs to calculate something, every manufacturer will rather use safe values than tight values in order to not cause instabilities. Furthermore, XMP is more accurately Intel XMP, which means that the overclocked profiles are not really specified with Ryzen in mind (with maybe the exception of the few DDR4 modules that are meant for Ryzen). Then there's the fact that manufacturers rarely push their hardware to the absolute limit. Think of factory overclocked GPUs; some of them still have noticeable headroom for manual overclocking. ------------- Ryzen 5 1500X, ASRock AB350M Pro4, 2x8 GB G.Skill Trident Z 3466CL16, Sapphire Pulse RX Vega56 8G HBM2, Corsair RM550x, Samsung 960 EVO SSD (NVMe) 250GB, Samsung 850 EVO SSD 500 GB, Windows 10 64-bit |
Posted By: shmerl
Date Posted: 22 Dec 2017 at 6:19am
" rel="nofollow -
Right, but that doesn't really answer the question, how values provided by the above calculator produce more stable result, than values set by the firmware for the profile.
That can be the difference, but at least in my case the RAM is specifically supposed to be for Ryzen (by G.Skill), so whatever profile is provided should be supposedly good already?
That's as above. Supposedly that calculator can produce values that are more stable for the same frequency. My point is not to push the RAM to the limit, but to make it work stably to begin with at the frequency it was made for (3200 MHz). |
Posted By: PetrolHead
Date Posted: 22 Dec 2017 at 7:05am
" rel="nofollow - I think I misunderstood you a bit, sorry about that.
That's a good question. In theory the manufacturer should have the resources and know-how to do a better job than a seemingly simple software made by one overclocking enthusiast(?). I haven't gotten around to using the software myself, but my guess - and it's only a guess at this point - is that with the values from Typhoon the program can calculate the optimal values for your exact system, whereas the values provided by the manufacturer are somewhat generic and meant to be a "one size fits most"-type of solution.
Good point, I hadn't paid attention to the exact RAM modules you have. Have you tried running MemTest86 to test your RAM kit? Or have you considered switching to Debian stable, to see if that helps? Or maybe try some other distro, like Fedora 27 or CentOS 7.1711 (depending on whether you think bleeding edge might have better chances of working or not)? You could even leave your current installation intact and just make a live USB for testing purposes. ------------- Ryzen 5 1500X, ASRock AB350M Pro4, 2x8 GB G.Skill Trident Z 3466CL16, Sapphire Pulse RX Vega56 8G HBM2, Corsair RM550x, Samsung 960 EVO SSD (NVMe) 250GB, Samsung 850 EVO SSD 500 GB, Windows 10 64-bit |
Posted By: datonyb
Date Posted: 22 Dec 2017 at 7:28am
|
ok let me clear a few points up the pre programmed xmp is meant for intel , and am4 really dosnt do well at using these next point the flarex designed for ryzen STILL dosnt work as well as advertised it was picked timings from a very early stage in am4 launch next point the calculator dosnt read different generated timings depending on system if you read the instructions it tells you to scroll down and read the xmp profile ,entering these into calculator basically tells the p[rogram the quaility of the ram chips, this then works off tables to produce the ryzen tested and tweaked timings example petrolheads tridents have highly binned samsung b dies which would have better xmp timings than my 3200 cl15 samsung b dies tridentz so some of his suggested settings would be a click or two lower than the results given for my sticks in particular the flarex has some timings that really struggle and get loosened up a little ,and quite a lot of timings way to loose so get tightened for better performance i know it sounds strange that gskill cant update the flarex with better knowledge now ,but hey they just want your money,like the new flarex for threadripper.............seems it has identical timings to the ryzen flarex haha but costs you more as its for threadripper ------------- [url=https://valid.x86.fr/jpg250] [/url]3800X, powercolor reddevil vega64, gskill tridentz3866, taichix370, evga750watt gold |
Posted By: datonyb
Date Posted: 22 Dec 2017 at 7:30am
|
p.s. you dont need pirated windows its free install from microsoft google mediacreationtool you get a week or so to evaluate it ,then it disables some useless stuff but still works with a tiny annoying icon bottom right of screen ------------- [url=https://valid.x86.fr/jpg250] [/url]3800X, powercolor reddevil vega64, gskill tridentz3866, taichix370, evga750watt gold |
Posted By: shmerl
Date Posted: 22 Dec 2017 at 7:36am
" rel="nofollow -
I didn't think the calc reads anything, but assumed may be Thaiphoon tool does. From what you explain, Thaiphoon step can simply be skipped, if needed values can be obtained otherwise (they should be fixed for each model). This video: http://%20www.youtube.com/watch%3Fv=6-XkaF_Rp0c" rel="nofollow - https://www.youtube.com/watch?v=6-XkaF_Rp0c uses timing input and "memory type". I suppose I need to determine that for my flare x. And then figure out how to run actual calculator. |
Posted By: datonyb
Date Posted: 22 Dec 2017 at 7:40am
|
i asked for someone to post a screenshot of your ram ,someone should have it saved to desktop and no i didnt say calculator reads it ,as per the instructions thaiphoon reads it scroll to bottom xmp change to show results in nanoseconds and enter them into calc. ------------- [url=https://valid.x86.fr/jpg250] [/url]3800X, powercolor reddevil vega64, gskill tridentz3866, taichix370, evga750watt gold |
Posted By: shmerl
Date Posted: 22 Dec 2017 at 8:50am
|
I saw, thanks! Trying one suggestion from there now. |
Posted By: shmerl
Date Posted: 24 Dec 2017 at 8:48am
Here are suggested timings from overclock.net: They worked out very well! No freezes now. What I'm surprised about is, that support staff from G.Skill forum didn't offer anything of the sort. |
Posted By: shmerl
Date Posted: 24 Dec 2017 at 9:54am
|
Nope, soon after writing the comment I got another freeze %) |
Posted By: shmerl
Date Posted: 25 Dec 2017 at 2:51am
|
I'm now testing one setting I never tried before. Asrock has "voltage mode" between stable and overlock. I set it to overclock (apparently it uses wider ranges of voltages). So far so good. |
Posted By: shmerl
Date Posted: 26 Feb 2018 at 2:35am
|
" rel="nofollow - Just for the reference, most likely the fault is with the CPU, not with motherboard or RAM. See: https://bugzilla.kernel.org/show_bug.cgi?id=196683 So far AMD didn't fix it with any firmware updates. |
Posted By: shmerl
Date Posted: 01 Mar 2018 at 2:46am
|
" rel="nofollow - So, these freezes aren't related to RAM after all. It's apparently something to do with how C6 states are handled. See: https://bugzilla.kernel.org/show_bug.cgi?id=196683 https://community.amd.com/message/2848556#comment-2848556 For now, I built a custom kernel as a workaround (with CONFIG_RCU_NOCB_CPU), and set rcu_nocbs=0-15 parameter, but it really means that CPU is defective or at least doesn't handle power states properly. AMD answered this to similar support request:
Latest ASRock firmware update doesn't seem to contain this fix. Can anyone from ASRock please confirm, if this fix is eventually coming to your firmware too? |
Posted By: MisterJ
Date Posted: 01 Mar 2018 at 4:22am
|
shmerl,this is a user forum and seldom do ASRock people post here. I would suggest you open an ASRock support ticket and ask them. I also suggest you ask AMD (AMD support ticket) and ask them to specify a SPECIFIC AGESA/UEFI/MICRODE that corrects this problem. Latest tells me nothing especially with the way AMD and ASRock release code. Please let us hear how you make out. Thanks and enjoy, John. ------------- Fat1 X399 Pro Gaming, TR 1950X, RAID0 3xSamsung SSD 960 EVO, G.SKILL FlareX F4-3200C14Q-32GFX, Win 10 x64 Pro, Enermx Platimax 850, Enermx Liqtech TR4 CPU Cooler, Radeon RX580, BIOS 2.00, 2xHDDs WD |
Posted By: NeoTheFox
Date Posted: 01 Mar 2018 at 6:16am
|
AFAIK enabling Cool'n'Quiet and/or C6 would make your Linux freeze. Same happens to me, but it looks like ASRock has nothing to do with it - it's a problem somewhere between the Ryzen CPU and Linux kernel. There are multiple bug reports about this issue and even software to mitigate it - http://github.com/qrwteyrutiyoup/ryzen-stabilizator" rel="nofollow - http://github.com/qrwteyrutiyoup/ryzen-stabilizator . Also some Ryzen CPUs segfault when you are compiling heavy software, and these can be RMA'd, as it is a recognized issue. |
Posted By: shmerl
Date Posted: 01 Mar 2018 at 6:19am
|
I had the segfault issue and already RMA'd the CPU. Many report, that replacements are actually more likely to get the C6 freeze bug. Disabling C6 though is a very crude workaround for these freezes. Using CONFIG_RCU_NOCB_CPU and rcu_nocbs=0-15 is a better one. Personally, I consider buying Ryzen 2 which hopefully won't have this problem anymore (but can have new ones, haha). And thanks for the suggestions, I'll contact ASRock support directly about it. |
Posted By: shmerl
Date Posted: 04 Mar 2018 at 8:29am
|
Browsing around my firmware settings, I found this one: Advanced > AMD CBS > Zen Common Options > Power Supply Idle Control. I changed it from auto to low, let's see if it will help with stock kernel. |
Posted By: shmerl
Date Posted: 04 Mar 2018 at 12:14pm
|
Still freezing with "low current idle". Testing now with "common current idle". |
Posted By: shmerl
Date Posted: 05 Mar 2018 at 2:30am
I'm now also seeing this in dmesg:
And so on for all 16 virtual cores. |
Posted By: shmerl
Date Posted: 07 Mar 2018 at 9:07am
|
That C-state 0x0 not supported by HW happens now always, so it's not related to my test above. With Advanced > AMD CBS > Zen Common Options > Power Supply Idle Control set to "Common current idle" (instead of auto), I didn't get any freezes in a while, so I assume it's a valid workaround. I noticed what changes after it's set in the firmware, using https://github.com/r4m0n/ZenStates-Linux/blob/master/zenstates.py" rel="nofollow - zenstates.py: When set to auto (freezes happen):
when set to Common current idle (no freezes so far):
So apparently it disables package C6 state (while keeping core C6 state enabled)! Hopefully it can shed some light on what the problem is. I wonder if Ryzen 2 will be free of this issue. What exactly is "package" in this context? Is it still part of CPU, or it's something on the motherboard? |
Posted By: Prodif
Date Posted: 07 Mar 2018 at 12:34pm
|
" rel="nofollow - shmerl On a firmware of v2.36 package C6 it is possible to disable, on the version is higher this option is deleted. And you have what version of a firmware? ------------- ASRock x370 Taichi (bios v2.36) / Ryzen-1800X / Crucial CT16G4DFD824A x2 / GPU: AMD R9 NANO / SSD NVMe Samsung 960 PRO MZ-V6P1T0BW (M.2) / HDD SATA 3TB x6 / PSU: Corsair AX860i ps. RedDragon, by AG |
Posted By: shmerl
Date Posted: 07 Mar 2018 at 12:44pm
" rel="nofollow -
I'm using latest firmware 4.60. There is no explicit package C6 option. I only commented what I observe with zenstates.py after changing Power Supply Idle Control option. Apparently it affects package C6. So what exactly is package C6? Does it refer to CPU package or something on the motherboard? UPDATE: I found some info here, though it's talking about Intel: https://software.intel.com/en-us/articles/power-management-states-p-states-c-states-and-package-c-states#_Toc383778911 I suppose AMD uses similar terminology. So package C states still refers to the processor. If that's correct, there is a chance Ryzen 2 will fix this mess |
Posted By: Prodif
Date Posted: 07 Mar 2018 at 2:00pm
|
shmerl https://www.amd.com/system/files/2017-03/AMD-Ryzen-Processor-and-AMD-Ryzen-Master-Overclocking-Users-Guide.pdf" rel="nofollow - https://www.amd.com/system/files/2017-03/AMD-Ryzen-Processor-and-AMD-Ryzen-Master-Overclocking-Users-Guide.pdf CPU low power c-states (CC1, CC6, and PC6) .. p.s. Ryzen 2 in 2019, in april 2018 Zen+
------------- ASRock x370 Taichi (bios v2.36) / Ryzen-1800X / Crucial CT16G4DFD824A x2 / GPU: AMD R9 NANO / SSD NVMe Samsung 960 PRO MZ-V6P1T0BW (M.2) / HDD SATA 3TB x6 / PSU: Corsair AX860i ps. RedDragon, by AG |
Posted By: shmerl
Date Posted: 07 Mar 2018 at 2:03pm
Is PC6 - package C6? That document doesn't give any details about what it is. |
Posted By: Prodif
Date Posted: 07 Mar 2018 at 2:09pm
|
shmerl http://www.uefi.org/specifications https://support.amd.com/TechDocs/43170_14h_Mod_00h-0Fh_BKDG.pdf" rel="nofollow - https://support.amd.com/TechDocs/43170_14h_Mod_00h-0Fh_BKDG.pdf https://support.amd.com/TechDocs/54945_PPR_Family_17h_Models_00h-0Fh.pdf
------------- ASRock x370 Taichi (bios v2.36) / Ryzen-1800X / Crucial CT16G4DFD824A x2 / GPU: AMD R9 NANO / SSD NVMe Samsung 960 PRO MZ-V6P1T0BW (M.2) / HDD SATA 3TB x6 / PSU: Corsair AX860i ps. RedDragon, by AG |
Posted By: shmerl
Date Posted: 07 Mar 2018 at 2:15pm
|
Well, that explains the sequence, but not what it is. I suppose package refers to the whole set of cores in this case. Anyway, as I said, if it's all in the CPU, getting new one like Ryzen 2 can potentially provide one that works with PC6. |
Posted By: Prodif
Date Posted: 07 Mar 2018 at 11:01pm
|
shmerl http://i3.imageban.ru/out/2018/03/07/339330dabf44efa4159b67c2ee4ebb73.jpg" rel="nofollow">  http://i2.imageban.ru/out/2018/03/07/f16442f0ea74a1b8db10bf287e651ee3.jpg" rel="nofollow">  ------------- ASRock x370 Taichi (bios v2.36) / Ryzen-1800X / Crucial CT16G4DFD824A x2 / GPU: AMD R9 NANO / SSD NVMe Samsung 960 PRO MZ-V6P1T0BW (M.2) / HDD SATA 3TB x6 / PSU: Corsair AX860i ps. RedDragon, by AG |
Posted By: shmerl
Date Posted: 08 Mar 2018 at 12:55am
|
I don't disable the whole C6, that's too much and stresses the CPU. Setting it like above with power supply idle mode is already enough to prevent the freeze, and as I said, it only disables package C6, while leaving core C6 enabled. Still not ideal, but will do until I'll get Ryzen 2. |
Posted By: JohnM
Date Posted: 08 Mar 2018 at 8:49am
Ryzen 2 is Zen+ in April 2018. Zen 2 in 2019. ------------- ASRock Fatal1ty AB350 ITX P4.90, AMD Ryzen 5 2400G, 2x8GB Corsair CMK16GX4M2A2666C16, 250GB Samsung 960EVO, 500GB Samsung 850EVO, 4TB WD Blue, Windows 10 Pro 64, Corsair SF450, Cooler Master Elite 110 |