Anybody planning a Ryzen build?


Recommended Posts

9 minutes ago, HellDiverUK said:

Latest update on my 1700X.

 

I can run 4GHz@1.4v, but I've settled at 3.975GHz at 1.3575v which actually benchmarks better in both 3DMark and Cinebench R15.  

 

Heat is a major issue at 4GHz - even with the H110iGT and the fans going fast, the CPU is still hitting 80C.

 

I'm quite pleased with the PRIME X370-Pro now, apart from being hit with the common problem that the RGB lighting has totally stopped working.  Seems to be an issue with this model of board.

 

Are you seeing close to bare metal 3D performance? Everyone that I know with ryzen and kvm are having big performance drops with npt. Physics scores are also very low.

Link to comment
7 minutes ago, mikeyosm said:

 

Are you seeing close to bare metal 3D performance? Everyone that I know with ryzen and kvm are having big performance drops with npt. Physics scores are also very low.

Oh, I'm not running unRAID on this machine.  I'm just updating my thoughts on Ryzen in general.  Sorry for the confusion.

 

My unRAID box is an i5-6500T that doesn't run any VMs at the moment.

Link to comment
On 27/06/2017 at 3:57 AM, Pauven said:

On my system, I notice 9-18 watts of increased idle power consumption.  Temperatures also increased 5-10 degrees C.  That said, my idle CPU temp is about 38C, and motherboard is about 34C. 

 

Hi Paul,

 

First of all thanks for the global c-state fix. I finally have a stable system! On a side note are you running Dynamix System Temp plugin to monitor temps? If so, it doesn't detect any sensors on my AMD Ryzen 5 1600 

Gigabyte - AX370-Gaming K3

 Setup. Am new to unRAID so I'm feeling my way through. Any help would be appreciated.   

 

Thanks in advance. 

 

Aaron

 

Edited by CobaltAUS
Link to comment

Well gents, just wanted to give you a update on Ryzen w/ c-states enabled locking up unRAID... or rather how we've fixed it!

 

On my Ryzen test machine, with array stopped (to keep it idle) with c-states enabled in bios, I'm approaching 7 days of uptime.  Before the changes we made to the kernel in the upcoming RC7, It would only make it a few hours and lockup.  I think you guys will find RC7 just epyc! (sorry, couldn't resist)

 

FYI, these are the two kernel changes we had to add to make it stable for Ryzen:

- CONFIG_RCU_NOCB_CPU: Offload RCU callback processing from boot-selected CPUs
- CONFIG_RCU_NOCB_CPU_ALL: All CPUs are build_forced no-CBs CPUs

 

Thanks to everyone here for testing and helping narrow it down to c-states.  That made it easier for us to test and find a solution.

 

Link to comment
1 hour ago, eschultz said:

Well gents, just wanted to give you a update on Ryzen w/ c-states enabled locking up unRAID... or rather how we've fixed it!

 

On my Ryzen test machine, with array stopped (to keep it idle) with c-states enabled in bios, I'm approaching 7 days of uptime.  Before the changes we made to the kernel in the upcoming RC7, It would only make it a few hours and lockup.  I think you guys will find RC7 just epyc! (sorry, couldn't resist)

 

FYI, these are the two kernel changes we had to add to make it stable for Ryzen:

- CONFIG_RCU_NOCB_CPU: Offload RCU callback processing from boot-selected CPUs
- CONFIG_RCU_NOCB_CPU_ALL: All CPUs are build_forced no-CBs CPUs

 

Thanks to everyone here for testing and helping narrow it down to c-states.  That made it easier for us to test and find a solution.

 

You just triggered my ryzen/unraid purchase. Thank you for the great work.
If there are other bugs I hope those get solved as well. As of now I think the only bugs that are left are related to windows virtual machines that people want to use for gaming. This is not my intention though.

Link to comment
4 hours ago, CobaltAUS said:

First of all thanks for the global c-state fix. I finally have a stable system! On a side note are you running Dynamix System Temp plugin to monitor temps? If so, it doesn't detect any sensors on my AMD Ryzen 5 1600

 

Glad it helped you too.  Hopefully the need for disabling C-states is over, but I've yet to test.

 

Yes, I am using Dynamix System Temp to monitor.  The thing to know here is that each motherboard maker gets to choose what sensor monitoring chips they use on each motherboard.  Typically you will see a vendor using the same or similar chip on most of their product line, but they can actually switch it up at any time.

 

ASRock seems to be the odd duck, from what I've seen, as their sensor chip is different than the other brands, and happens to have support in Linux kernel 4.9+.  I was one of the only Ryzen users who actually could see temps on unRAID, simply because my motherboard was using a different sensor chip.

 

Newer kernels should help, but it takes time for both kernels to add the drivers, and for Lime-Tech to move to newer kernels.  Patience is all I can recommend.

 

Paul

Link to comment
4 hours ago, eschultz said:

Well gents, just wanted to give you a update on Ryzen w/ c-states enabled locking up unRAID... or rather how we've fixed it!

 

On my Ryzen test machine, with array stopped (to keep it idle) with c-states enabled in bios, I'm approaching 7 days of uptime.  Before the changes we made to the kernel in the upcoming RC7, It would only make it a few hours and lockup.  I think you guys will find RC7 just epyc! (sorry, couldn't resist)

 

FYI, these are the two kernel changes we had to add to make it stable for Ryzen:

- CONFIG_RCU_NOCB_CPU: Offload RCU callback processing from boot-selected CPUs
- CONFIG_RCU_NOCB_CPU_ALL: All CPUs are build_forced no-CBs CPUs

 

Thanks to everyone here for testing and helping narrow it down to c-states.  That made it easier for us to test and find a solution.

 

 

Fanfreakintastic!

 

Glad there are some brainiacs figuring this stuff out.  

 

From what I've just read, sounds like these are good settings with little to no impact on performance.

 

I'm eager to test.  Any idea when RC7 lands?

 

Paul

Link to comment
9 hours ago, jonathanm said:

Were you able to measure a power usage / temperature improvement with the fix in place?

 

C-state enabled: 62 watts

C-state disabled: 80 watts

 

The numbers above remained the same before and after the kernel fix.  Will save me about $11/year (24 * 365 @ $0.07/kwh).

Link to comment
3 minutes ago, eschultz said:

Will save me about $11/year (24 * 365 @ $0.07/kwh).

Well, actual energy use in a climate controlled environment is a little more complicated. If you live in a cool climate, your saving is almost zero, because all that heat would just be offset by the extra heating cost. In a warm climate with AC running most of the time, the energy saving is more than doubled, because you no longer have to pay to pump the heat outside.

 

Computers are just tiny space heaters when you are calculating for energy cost over time.

Link to comment
1 minute ago, jonathanm said:

Well, actual energy use in a climate controlled environment is a little more complicated. If you live in a cool climate, your saving is almost zero, because all that heat would just be offset by the extra heating cost. In a warm climate with AC running most of the time, the energy saving is more than doubled, because you no longer have to pay to pump the heat outside.

 

Computers are just tiny space heaters when you are calculating for energy cost over time.

Ah yes, you're totally right.  Luckily I have just about all my computers in the basement.  Helps keep the basement warm :)

Link to comment

Hi @eschultz

 

I've been evaluating the latest stable unRaid 6.3 on my Ryzen 5 1600 / Asus Prime X370 Pro setup.

 

I haven't had much luck with it though; Kodi VM slows down and the Web Interface keeps stopping. Pretty much just been experiencing instability issues. I noticed that AM4 seems to be supported from the 4.10 Linux kernel only.

 

I'm really glad to see you're making headway on the Ryzen support.

 

Is there a way I can get a hold of the RC7 to evaluate?

Link to comment
13 hours ago, eschultz said:

 

C-state enabled: 62 watts

C-state disabled: 80 watts

 

The numbers above remained the same before and after the kernel fix.  Will save me about $11/year (24 * 365 @ $0.07/kwh).

 

The 18w delta closely aligns with what I see on my Ryzen 7 1800X.

 

Also keep in mind that disabling C-states pumps those 18w directly into the CPU.  Your 62w figure represents CPU + Memory + Motherboard/Chipset + Storage + Graphics + Fans.  The CPU is probably only about 20w-30w of that 62w, so an extra 18w pumped into the CPU is close to doubling the CPU idle power consumption and heat generation.  

 

Which is exactly why I made such a big deal about the trade-offs when using the C-state workaround for stability.

 

And also why I'm very excited to hear it is fixed! :) I keep scanning the Prerelease 6.4 Support area, but no sign of RC7 yet...  :(

 

Thanks for sharing.

 

Paul

Link to comment
On 4/24/2017 at 3:13 PM, Pauven said:

2)  The Case of the Reappearing Cache Drive:  In my previous build, I had a Samsung 840 1TB SSD installed as my cache drive.  In my upgrade to Ryzen, I installed a faster Samsung 960 1TB M.2 drive.  During this time, I left the old 840 installed and unassigned.  But on every server reboot, unRAID reassigns the 840 as the cache drive (and unassigns the 960).  I have to stop the array, change the assignment, and restart the array every time I reboot to get the 960 back into the cache drive role.  

 

Because I had so many crashes, I made a point of correcting this assignment and then rebooting the array before it crashed (thinking the change was somehow getting lost due to a crash), but this made no difference.  

 

Just a bit ago I physically removed the 840 from the server, but I haven't rebooted since then, so I don't know if this will allow the 960 to come up as cache drive.  I'm also a bit concerned that even if I fix it for now, if I ever re-install the 840 that this bad behavior will return.

 

Just wanted to report that I'm still suffering from cache drive issues.  No idea if this is a Ryzen issue, or now a 6.4.0-rc6 issue, as I've since upgraded from the 6.3.x branch.

 

It was previously pointed out to me by johnnie.black that this was a common issue to users of Chrome, and he pointed me to the FAQ.  Apparently, using another browser would address the issue.

 

I have tried both Firefox and IE, no joy.

 

Seems like the problem has actually gotten worse under 6.4.0-rc6.  At the very least I am noticing some new issues related to the problem.  Here's everything I've found.

 

A)  As before, every time I boot up, the cache drive is unassigned (unless I install the old SSD, then the old SSD boots up assigned)

B)  Looking in the disk.cfg file, I see the old SSD still being referenced (cacheId="Samsung_SSD_840_EVO_1TB_S1D9NSAF627019H")

C)  Stopping the array, assigning the cache drive, and restarting the array appears to partially fix the problem, as now caching and VM's work.

D)  Under 6.4.0-rc6, the Identification is ugly: "eui.0025385c61b08640 - 1 TB (nvme0n1)", back on 6.3.x it seems like it was pretty, something like "Samsung 960 1TB (nvme0n1)"

E)  I can browse to \\TOWER\cache and see my files, and "cache" shows up in the DISK SHARES list on the Shares tab, so both of those seem good

F)  But Mover is broken, always logging the issue "root: cache not enabled, exit"

G) Looking at the backlog of files to move, Mover not has worked in a month (June 25 is oldest file) - appears to be when I upgraded from 6.3.x to 6.4.0.-rc6

 

Paul

Edited by Pauven
tiepoes
Link to comment
6 hours ago, jonathanm said:

Have you tried booting in safe mode, assigning the cache, and restarting normally?

 

Yes I had tried that, but just did it again to make sure.  No dice.

 

I can open disk.cfg in Notepad++, which automatically notifies me if the file changes and asks if I want to reload it.  When I start the array, the file gets touched.  But not when I change the cache drive assignment.  No matter what I do, the cacheId value stays the same with the old cache drive info.

Link to comment

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Restore formatting

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.