Ryzen-Test & Stress-Run Make It Easy To Cause Segmentation Faults On Zen CPUs

[macchky]

This is anecdotal evidence but I underclocked my A10-7870K to 0.2V below normal and while everything seemed fine in Windows, when in Linux the system had L1 parity errors that were corrected most of the time. At -0.1V the problem did not appear, so I'm not surprised a problem with voltages could lead to what people see with Ryzen.

Also in another thread on Phoronix I explained how to underclock the RX 470/480 by patching the amdgpu-pro kernel source. After a few months it appears the voltage I chose (820mV) is too low as it sometimes leads to page faults in the GPU. When that happens only a reboot helps.

This thread is about RYZEN CPU not APU mate

[bridgman]

(John there's threads you promised to drop an update in and haven't please do so)

Is "John" in this case me ? If so, I don't think I promised (or even "said") that I would provide an update, just that I would make sure the info and concerns were getting to the right people. I did that, and posted back to confirm it.

If that isn't how you read things could you try to point me to the threads ? Thanks...

[c2h5oh]

My motherboard's BIOS keeps my RAM voltage at 1.2v, even though it's supposed to be 1.3v. This alone makes the RAM very unstable.

Most DDR4 faster than 2333 requires 1.35V

[Silverthorn]

I got curious and tried ryzen-test/kill-ryzen.sh on my system and it took almost 40 minutes until I got the segmentation fault. Hopefully this behaviour gets fixed and while doing so also fixed other more rare crashes and stability issues.

[caligula]

I can also confirm I get random segfaults on a 1700X system which is mostly used for compiling packages with gcc. Let's see how it will go with this new test.

[schmidtbag]

Most DDR4 faster than 2333 requires 1.35V

Exactly, but for whatever reason my motherboard wants to set it to 1.2v, even though the default speed is 3GHz. That's actually pretty stable, but if I OC to 3.2, then I need to bump up the voltage.

[chuckula]

the ryzen GCC compiling segfault bug in the closed source microcode is the best example why open-source from top to bottom matters.
and on an opensource system from top to botton this bug would not happen...
it just monopoles the bug-fix to one company otherwise if it was open-source many companies can compete to fix the bug
in the end it is just magic security features so in the end we all can be "save" ,... and "save" in this meaning means that we are save to say that we can not and we should not fix bugs instead we need to obey our masters and hope for an update.

Yeah that's a load of crap from Quaridiot who assumes that magically open sourcing CPU microcode will mean that bugs can't happen.

Just like the open source OpenSSL has literally never had a single bug ever.

Yeah -- just like that -- with the one major exception being that OpenSSL code is at least auditable using normal tools while CPU firmware by its very nature is going to be much more esoteric be it "open source" or not.

I'd rather have one undergrad intern at AMD with a modicum of training trying to actually fix the microcode vs. a million Quaridiots drooling over themselves while they chant about open source and do nothing of value.

[Beherit]

and on an opensource system from top to botton this bug would not happen...

It doesn't happen when using Microsoft Windows. Which is the diametrical opposite of open source.

[SaucyJack]

Pretty sure that small number of users thing is BS. It happens on my ryzen constantly using gcc.

[gurv]

Yeah first fail happens on my 1700 in 36s and 5 fails within 132s
Not too happy about it

I knew I should never have bought a new architecture at its release.
Next time, I'll wait for the second or third stepping

It doesn't happen on windows so I wonder if a workaround could be implemented in the kernel.