That is correct Threadripper does not have a integrated GPU. It is looking as they are having the APU line and then a pure CPU die for high performance users and gaming enthusiasts. As the APUs now have the Zen desktop processors unless a heavy gamer they are more than powerful enough.


My Ryzen compiles slightly slower than my 8-core Xeon, 4GB KVM VPS. In a long compile like LibreOffice or Qt5 then the difference will be more noticeable in favour of the Xeon.


When more software uses OpenCL then the APU advantage will be much more noticeable. Ryzen also has CrossFire built-in so one can drop in an AMD GPU card and it will use the integrated and dedicated in concert.

1 members found this post helpful.

If this is being asked in the wrong thread (it's (Ry)zen-ish) or has already been answered, I apologize.

My 10yr update interval is upon me and my hand was forced, so I now have a threadripper system incoming (albeit in parts, scattered over the next few days/weeks), I am presently on slackware 14.2 stable, multilib (with some sbo packages for good measure). From what I have read both here and outside LQ, is that TR doesnt play well with much below 4.10 to 4.12 kernels, so I figured I should ask the experts and those with possibly better experience in this matter (which is not myself), if I should grab and compile a custom kernel or 'simply' move up to current? The last time I moved to current was slackware 3.0 (or there abouts) mainly due to it looking fun at the time, but not for the reasons now (to make a cpu function mostly without errors).

I currently have the root filesystem on a ssd, all other data lives elsewhere, so I am content with installing fresh if this were smarter. The current working plan is to change kernels (by some method), run stability tests on the present system. So when the day arrives that the last parts come, any troubleshooting that occurs then, would more likely (hopefully) be due to bad hardware rather than software (although yes, there could still be quirks in software).

Would it be best to stay on stable but compile a newer kernel?
Would it be best to revert to a clean system (as if it was recently cleanly installed) and move up to current, re-compiling/installing as needed?
Would it be best to use AlienBob's's liveslak-current to effectively install a clean current system, then update/multilib/install/compile as tests suceed?

Im leaning towards the liveslack (a clean install of current) partly due to me being lazy and partly due to at least having some method to boot if lilo and this newer fancy bios (uefi?) don't play well with each other.

Any advice would be greatly appreciated.

I'd just install (not compile) the kernel for -current in a stable system. That works, here at least.

I documented (https://alien.slackbook.org/blog/bui...all-slackware/) how I installed Slackware 14.2 on a server with Ryzen CPU and a NVMe M2 flash disk. If you install to a regular SSD or HDD disk, then you can just use the Slackware 14.2 installer, but support for NVMe was only added to slackware-current.

Before installing anything, first check for a newer BIOS than the one on your motherboard. Ryzen stability increased a lot with new versions of BIOS.
After installing Slackware 14.2 you should build yourself a 4.19.x kernel first thing.

In the beginning (With the BIOS that was factory-flashed to the mobo and a Slackware 14.2 stock kernel) I had hard lock-ups every few days and I was unable to run virtual machines for long. My server is now rock stable with a 4.19.2 kernel and the latest BIOS.

If you upgrade every ten years, I would start with Slackware 14.2 on the new hardware and upgrade to 15.0 once that gets a release. Do not try -current.

2 members found this post helpful.

What inconveniences would have using a pre-built kernel intended for -current instead?

I once had a lockup very often, but after upgrading BIOS from the motherboard vendors, it has never failed again. That, along with newer kernel (4.9+) should be good for newer hardware especially AMD products (Ryzen and Threadripper).

For Ryzen 2200U a 4.18 kernel at least is a must or you have serious problems. Mine would not even startx until 4.18. That said, since 4.19 it has been smooth sailing and the C6 bug from Carrizo (power state where the CPU is essentially off waiting for an interrupt but never leaves the power state: the infamous black screen) has stopped ailing me since 4.19.

I third the advice of Didier Spaier and Alien Bob.

Essentially that you won't be able to build any drivers/modules (like NVidia) for it as you do not have -current's toolset. The kernel as such should work OK.
This also means it isn't necessary to install that kernel's sources.

2 members found this post helpful.

My Ryzen 2400G system with B350 chipset has been running great since kernel 4.16.
Only some KMS related hangs on poweroff and reboot with 4.19.0 and 4.19.1. Things have been working again since 4.19.2.

Since the 2200U is a mobile CPU I am assuming you have a Ryzen Laptop?

1 members found this post helpful.

Yes, and HP. Which considering their firmware history my issue may have been with the firmware. I think though since this hardware is all early to mid-2018 the kernel has caught up to the hardware.

Would grabbing the -current source (mainly for the config file) of the 4.19.x kernel work well? Or would it be better to get the source package from kernel.org?

Hardware gets upgraded every ten, but yes, I understand, as I do prefer the stability.

Grabbing the source from -current is a fine option. Using Pat's config will help a lot. But grabbing the source from kernel.org and using Pat's config is fine as well. If you're building the same kernel version, the source tarballs will be the same anyway (gotta love Slackware's vanilla approach!).

1 members found this post helpful.

Ok, can someone help me here as well? I have a 14.2 system, running 4.12.2 (upgraded from 4.9.37, IIRC) with some Ryzen sensor modules compiled in. I TRIED upgrading to 4.15.2 and 4.15.13, but had no end of issues with the then-new NVIDIA 390.77 kernel. I could eventually compile the new module, but then I'd get a blank screen in Xorg. Going back to 14.2/390.42 thankfully gave me back a working system.

I have a Ryzen 1700 (replaced under RMA, it's hung up once since). Will look if there's a newer BIOS for my MSI board.

What's the best/easiest way to upgrade the kernel and nvidia driver? I used to compile kernels A LOT back in the old days (like 10-15 years ago), but am somewhat out of practice, since it's been so stable recently and have forgotten exactly what's needed. I THINK I managed to get my original kernel headers back, but I know I swapped to the 4.15.x headers at one point. Do I need upgraded Xorg/clib?

If I upgrade to 4.19.x, do I upgrade the headers or not? Will the latest nvidia drivers compile? Any other upgrades needed? I've already held off on "trying again" for quite some time. If I knew 15.0 was coming out in < 6 months, I'd just wait, but, obviously, no one knows anything :-)

Which Nvidia card are you using?

Just upgraded my GTX 950 to a GTX 960.

1 members found this post helpful.