AGESA 1.0.0.6b Might Fix The Ryzen Linux Performance Marginality Problem

[timofonic]

Because a small group of the 3% market share sometimes has an error? A disaster really. If the affected people never again bought an AMD product in their whole life, no one would even notice.

3%? Are you serious? These days mostly all iron is deployed in server infrastructures, called them Cloud or SaaS or WTF. And mostly all those run Linux. I don't see it a 3% marketshare but the opposite, maybe a 97%.

Gamers is a really marginal and often irrational (stupid brandings, buzzwords, style over substance, placebo) market compared to servers. On the other side, servers stress hardware a lot more than the average dumb desktop use.

And Linux ecosystem have been the one that really supported AMD hardware. Without Linux geeks being tech authoritative figures, AMD would have been death many years ago because of their really idiotic management.

[Azrael5]

I have a Ryzen 7 1700 that was built in week 28 ("UA 1728PGT" is on the chip) and have crashes too when running kill_ryzen.sh.
Unfortunately there isn't a bios update existing for my board yet.

Being an hardware defect you can ask for the replacement as stated by AMD.

[mat69]

Being an hardware defect you can ask for the replacement as stated by AMD.

Yeah, but I don't even know if a new CPU would fix it, since the system is not stable at all. I have regular hang-ups where only Sysrq commands help me to reboot.

I posted in the AMD monster thread though as suggested here.

[campcreekdude]

I know prior to buying my Crosshair V Formula Motherboard AM3+. Asus had an excel spreadsheet of Motherboards that were tested in Linux and what version.
Although, the Ubuntu versions were old and not the latest version that just came out yesterday.

Anyways, Ubuntu is very compatible with lots of hardware. I have that feeling with it. They drop graphics card acceleration on older models with new Ubuntu versions but I'm still able to get it to launch and do most things but gaming.

[Marc Driftmeyer]

I would recommend shying away from gigabyte motherboards for linux support. Gigabyte doesn't seem to care about linux and the linux community had to work around an gpio issue with some of their ryzen motherboards.

https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1671360

Basically all they could do was make the code more robust so that when this stuff blew up, the kernel was able to recover.

https://lkml.org/lkml/2017/5/23/1099

Never had a problem because I never buy v.1.0 models of any vendor, especially on a new chipset socket. V 1.1 or V 2.0 will always address the kinks, along with the second generation of Zen being Zen+.

[Marc Driftmeyer]

3%? Are you serious? These days mostly all iron is deployed in server infrastructures, called them Cloud or SaaS or WTF. And mostly all those run Linux. I don't see it a 3% marketshare but the opposite, maybe a 97%.

Gamers is a really marginal and often irrational (stupid brandings, buzzwords, style over substance, placebo) market compared to servers. On the other side, servers stress hardware a lot more than the average dumb desktop use.

And Linux ecosystem have been the one that really supported AMD hardware. Without Linux geeks being tech authoritative figures, AMD would have been death many years ago because of their really idiotic management.

And over 90% of that `iron' is Intel motherboards specifically curtailed for a few options.

[Azrael5]

Yeah, but I don't even know if a new CPU would fix it, since the system is not stable at all. I have regular hang-ups where only Sysrq commands help me to reboot.

I posted in the AMD monster thread though as suggested here.

You have to prove your statements and ask for replacement of the defective CPu to AMD providing the technical details of the device and its id then AMD will provide you of new working device.

[bridgman]

And over 90% of that `iron' is Intel motherboards specifically curtailed for a few options.

The split used to be more like 60/40 after AMD64 started to catch on... hopefully EPYC will get us heading back in that direction.

[soulsource]

Ran the ryzen-test and mprime, individually and together for at least 30 minutes, before and after AGESA 1.0.0.6b on ASRock Fatal1ty AB350 Gaming-ITX/ac with 1700. Never got a segment fault but the BIOS update did fix stability issues. System still ran smooth while running both tests together after the update.

Originally had an issue loading a VirtualBox export off a USB stick before the update, which took an excessive amount on time with USB 3.0. VirtualBox stated everything was successful but OS never booted. Had to first copy the export to a local a SATA drive so it would import uncorrupted.

mprime will likely not segfault at all, and ryzen-test needs to run at least several hours/days to be more or less certain that the CPU is not affected by the segfault issue.

And at those who still get segfaults during compilation: Stop waiting for magical software fixes, just write to AMD's customer support and get the faulty chip replaced. It doesn't cost you a thing (except 3 days or so without a CPU - depending on where you live obviously, as sometimes customs take longer...) and you'll get a crash-free Ryzen experience.

[mlau]

And at those who still get segfaults during compilation: Stop waiting for magical software fixes, just write to AMD's customer support and get the faulty chip replaced. It doesn't cost you a thing (except 3 days or so without a CPU - depending on where you live obviously, as sometimes customs take longer...) and you'll get a crash-free Ryzen experience.

Reports around the internet indicate that even some of the replacement CPUs still exhibit the problem.