Threadripper

[wintpe]

No one has started this topic, yet so I thought id put it out there.

Im starting to pull together a parts list to support threadripper, although ill probably leave the important bit to the end, so thats the Board and threadripper chip will be the last thing I buy, just in case theres a bug in the hardware thats not yet surfaced, like it did for Ryzen 7 and linux.

But from a Mageia 6 perspective, anyone see any issues with threadripper and or the following parts.

Im going with the Asus prime X399-a MB unless someone already knows a good reason not to, and an MZ-V6E500BW NVMe M2 drive.

this is equipped with
amd X399 chipset
Intel® I211-AT
Realtek® ALC S1220A
ASMedia® USB 3.1 Gen 2

im sticking with my existing gtx560 as it already has enough throughput for what i want.
existing plexstore dvd writers

if necessary ill start of with mageia couldron (MGA7) if that's what it requires

Im not doing this because I need it, my existing 1100T system is more than enough, but it is getting old, and it does have a buggy bios,
and it does not support many of the AES features of newer hardware. and Finally someones got to do it for mageia.

this is intended to help anyone else considering this platform make an informed decision.
this platform is cutting edge, and vulnerable to early adopters syndrome.


regards peter

[wintpe]

OK Sat all the bits arrived, and it took me a good four hours to put the thing together and install mageia 6.

by the number of replies i got, either no ones interested , or no one has an opinion.

if anyone is interested , or if the devs want logs, lspci output etc, then happy to help.

As is well known in wider Linux communities none of the temp sensors work , and im going to need a later kernel for that, so hopefully the kernel maintainer
will be able to help with that at some point, there are also a couple of minor performance stability patches, that would be nice to have.

otherwise mageia 6 does install, run, update and perform very stable on asus prime 399-a, with 1950X.

nic works, sound works, no issues with usb that i have seen, the old iommu issue is only there as by default the bios does not enable visualization.

I believe one of you has already got a Ryzen 7, and the 1950X is simply two 1800X in one package.

do the sensors work on that yet?

regards peter

[wintpe]

I notice thers the following in backports testing

kernel-desktop-4.12.14-1.mga6-1-1.mga6.x86_64.rpm 27-Sep-2017 13:23 52949574
kernel-desktop-devel-4.12.14-1.mga6-1-1.mga6.x8..> 27-Sep-2017 13:23 10087834
kernel-desktop-devel-latest-4.12.14-1.mga6.x86_..> 27-Sep-2017 13:23 70894
kernel-desktop-latest-4.12.14-1.mga6.x86_64.rpm 27-Sep-2017 13:23 70842
kernel-doc-4.12.14-1.mga6.noarch.rpm 27-Sep-2017 13:23 6331118


ill be giving that a try tonight.

regards peter

[pete910]

Not here often now, Went ryzen 7 myself on release. Due to the lack of support I ended up jumping ship (Antergos)

Regards temp sensors, I have the gigabyte k7 which I use the modded it87 driver with. Works a treat.

https://github.com/groeck/it87

No clue if the prime can use that though

Just dropped lucky a bagged a vega 64 for under 300 so back to oss drivers for me. NV drivers on pascal have been shocking this past year.

[wintpe]

thanks pete for that hint about pascal, ive got one in the post, so we will see how that goes.

yes ive been reading the thread on the sensor chips which is why i tried the above backported kernels just now.

wont be doing that again, and now i know how those users we help feel when faced with a blank black
screen that goes no where and an inability to use the drakconf cli to correct it because it keeps crashing when you select
the nvidia driver.

i guess ill find out about pascal when 1070ti turns up.

anyway if anyone is interested how to get out of the nvidia driver back after a failed kernel upgrade, when all you have is a black screen its as simple as this, as ive just done it.
boot into a prior kernel rescue mode

rpm -qa |grep kernel|grep 4.14 (if 4.14 is the version you need to back out of, but dont remove the ones you want to keep.

urpme kernel-desktop-.......... all the bits to the end

get rid of all the offending updated kernel.

you might have to clean up /boot/grub2/grub.conf as i did as the kernel i just tested did not do that either on removal.

then move /etc/X11/xorg.conf /etc/X11/xorg.conf.old

now reboot to the installation CD and perform an upgrade.

it wont do anything to your system if you do an upgrade , only if you poorly choose install when given the option.

it will then ask for kerboard region etc. just select the right ones.

it will then skip to the end, and almost immediately say your graphic card has a proprietary driver do you want to use it.

select yes, adjust the graphics to what you want

then finish, it will say installing boot loader , and then reboot.

and that recovers it far easier than some of the attempts ive seen some people go through

as for jumping ship for lack of support, i cant say i agree, i have far too much time and love invested in mageia and the people who use it to walk away.

regards peter

[wintpe]

OK not 100% sure that 1070 ti is supported by x11-driver-video-nvidia-current-384.90-2.mga6.nonfree
and i dont see any new ones in the back ports.
ill send an email to the testers and see if the kernel and driver team will be willing to help.

regards peter

[wintpe]

found a 384.98 driver in the cauldron mirror, grabbed the srpm and rpmbuilt it and that works
as a driver for my current gtx560, ill let you know what happens with the 1070ti in case anyone is thinking of getting one.
regards peter

[pete910]

You need the 387.22 for the 1070 ti

[wintpe]

this is what im actually running with the 1070 ti which came today

dkms-nvidia-current-384.98-1.mga6.x86_64.rpm
nvidia-current-cuda-opencl-384.98-1.mga6.x86_64.rpm
nvidia-current-devel-384.98-1.mga6.x86_64.rpm
nvidia-current-doc-html-384.98-1.mga6.x86_64.rpm
x11-driver-video-nvidia-current-384.98-1.mga6.x86_64.rpm

I know there seems to be info out there that supports what you say, but when i went to the nvidia site specified 1070 ti,
it also provided NVIDIA-Linux-x86_64-384.98.run

and its 100% stable, as far as i can see, although some of the mageia games bomb at the end of a scene, but im not sure how stable they are anyway, so ill have to test on both this and my old system to be 100% sure.

what issues did you find with Pascal?

=======================================================
glmark2 2014.03
=======================================================
OpenGL Information
GL_VENDOR: NVIDIA Corporation
GL_RENDERER: GeForce GTX 1070 Ti/PCIe/SSE2
GL_VERSION: 4.5.0 NVIDIA 384.98
=======================================================
=======================================================
glmark2 Score: 13694
=======================================================

regards peter

[pete910]

I see you figured it out

Don't know what games you play but do you have steam ect?

Got my rx64 in today, Nice to be back on OSS driver. 2d desktop so much smother.

bests my 1080 in the few things I've tried too!

[wintpe]

what games i play, hardly any to be honest, anything first person shooter
so quake, battlefield, you get the idea.

but more often i spend my time, trying out systems services, so not why i bought this monster.

I justify this, purchase not on what im going to do with it, but by buying something that gives me a significant multiplier of my
old systems performance, otherwise its not worth the outlay.

hopefully this will last as long as the last one and ill still be using it in 10 years or more.

regards peter

[wintpe]

another threadripper related issue and fix to collect together in this thread.

Code: Select all

Nov 26 16:19:38 localhost kernel: [ 2563.913157] dpc 0000:00:01.1:pcie010: DPC containment event, status:0x1f00 source:0x0000
Nov 26 16:19:38 localhost kernel: [ 2563.913169] pcieport 0000:00:01.1: AER: Corrected error received: id=0000
Nov 26 16:19:38 localhost kernel: [ 2563.913173] pcieport 0000:00:01.1: PCIe Bus Error: severity=Corrected, type=Data Link Layer, id=0009(Receiver ID)
Nov 26 16:19:38 localhost kernel: [ 2563.913176] pcieport 0000:00:01.1:   device [1022:1453] error status/mask=00000080/00006000
Nov 26 16:19:38 localhost kernel: [ 2563.913177] pcieport 0000:00:01.1:    [ 7] Bad DLLP             
Nov 26 16:34:00 localhost kernel: [ 3425.719225] dpc 0000:00:01.1:pcie010: DPC containment event, status:0x1f00 source:0x0000
Nov 26 16:34:00 localhost kernel: [ 3425.719236] pcieport 0000:00:01.1: AER: Corrected error received: id=0000
Nov 26 16:34:00 localhost kernel: [ 3425.719239] pcieport 0000:00:01.1: PCIe Bus Error: severity=Corrected, type=Data Link Layer, id=0009(Receiver ID)
Nov 26 16:34:00 localhost kernel: [ 3425.719241] pcieport 0000:00:01.1:   device [1022:1453] error status/mask=00000040/00006000
Nov 26 16:34:00 localhost kernel: [ 3425.719242] pcieport 0000:00:01.1:    [ 6] Bad TLP               
Nov 26 16:34:55 localhost kernel: [ 3480.398271] dpc 0000:00:01.1:pcie010: DPC containment event, status:0x1f00 source:0x0000
Nov 26 16:34:55 localhost kernel: [ 3480.398281] pcieport 0000:00:01.1: AER: Corrected error received: id=0000
Nov 26 16:34:55 localhost kernel: [ 3480.398284] pcieport 0000:00:01.1: PCIe Bus Error: severity=Corrected, type=Data Link Layer, id=0009(Transmitter ID)
Nov 26 16:34:55 localhost kernel: [ 3480.398286] pcieport 0000:00:01.1:   device [1022:1453] error status/mask=00001000/00006000
Nov 26 16:34:55 localhost kernel: [ 3480.398287] pcieport 0000:00:01.1:    [12] Replay Timer Timeou


this issue is being reported across many forums, from threadripper X399 to intel X99 chipset
so while threadripper people think its another threadripper issue, it seems its also appearing on other chipsets.

the fix for now is

[root@localhost log]# cat /etc/default/grub
GRUB_CMDLINE_LINUX_DEFAULT=" splash quiet noiswmd nokmsboot pcie_aspm=off resume=/dev/nvme0n1p3 audit=0"

ie add pcie_aspm=off

then do update_grub and reboot.

regards peter