cancel
Showing results for 
Search instead for 
Did you mean: 

Drivers & Software

flyinryzen1700
Adept II

Re: gcc segmentation faults on Ryzen / Linux

Got the new processor, it did not appear to have been opened or tested.

Batch UA-1737SUS

Still segfaults, but after 35 minutes now.  That was tested at 3066Mhz memory speed on 3200Mhz chips (32GB of 16GBx2 DualRank).  Technically, this is still overclocked, so I'm retesting now at 2400Mhz.  I know they offically changed supported memory speeds with latest BIOS, but haven't seen the speed chart.

Anybody seen that?

I wonder if they are just sending these out now hoping they pass without any extra binning.  If this still segfaults at this memory speed, I'll ask for advanced shipping instead of cross-ship.

0 Likes
flyinryzen1700
Adept II

Re: gcc segmentation faults on Ryzen / Linux

New 1800X running kill_ryzen for over 4 hours without problems with memory at 2400Mhz.   Previous cpu died within minutes at 2133Mhz.

Couldn't find official AMD memory support list, https://www.amd.com/system/files/2017-06/am4-motherboard-memory-support-list-en.pdf  has been removed.  Can't find any copies anywhere.  Anyway, ASRock lists 2933Mhz offically for my memory (GSkill F4-3200C14-16GTZ)   I'll retest at that speed. 

Has anybody had luck reducing segfaults with higher ram voltage? 

0 Likes
constantinx
Adept II

Re: gcc segmentation faults on Ryzen / Linux

flyinryzen1700

The new CPU shouldn't segfault at all.

Try again at 3066Mhz, and see if the system is stable. Use a memory stress test. Maybe the new processor can't sustain the same memory speed as the old one.

0 Likes
flyinryzen1700
Adept II

Re: gcc segmentation faults on Ryzen / Linux

Ran test overnight 8+ hours without segfault at 2933Mhz.  The memory controller is in the CPU, and for my memory, the max officially supported speed is 2933Mhz.

I'll try the test again at 3066Mhz with a slight bump in voltage.  I was hoping to be able to run at the full 3200Mhz, but I'm not sure how much of a difference that will make to actual performance.  At this point, I'd be fine were I'm at with a stable CPU.

0 Likes

Re: gcc segmentation faults on Ryzen / Linux

@flyinryzen1700

What asrock mobo you have?

The RAM you have is very good - have the same (two sets) and it runs fine at 3466@14-14-14-28. But as you know memory on ryzen is tricky so it requires some manual tweeks to run stable. 3200@14 should be stable 100% with proper DRAM, SOC and core (yes core voltage too) voltages.

synthor
Adept II

Re: gcc segmentation faults on Ryzen / Linux

Fixating the voltages to their default values, instead of using the auto option, made the biggest difference. Testing higher volages didn't make any difference.

Posted my results with all the settings I used a few pages before. (Gigabyte AB350 Gaming 3, F9d, AGESA 1.0.0.6b)

flyinryzen1700
Adept II

Re: gcc segmentation faults on Ryzen / Linux

I have to SLI killer x370.  I'll try fixing the voltages.  3066 segfaulted in 2 min with fixed 1.37v.  I didn't manually set any timings.

_blackbishop_ what are the other voltages you are using with that memory? So, you have 64GB total?

0 Likes
flyinryzen1700
Adept II

Re: gcc segmentation faults on Ryzen / Linux

Does that kernel include the patch to read Ryzen CPU core temps?  That's the next item in my to-do list.

0 Likes
scorpio810
Adept III

Re: gcc segmentation faults on Ryzen / Linux

flyinryzen1700 a écrit:

Does that kernel include the patch to read Ryzen CPU core temps? That's the next item in my to-do list.

You could replace  yourself k10temp.c in your custom 4.13 kernel.

https://download.tuxfamily.org/qet/joshua/workstation/bios/k10temp.c

0 Likes

Re: gcc segmentation faults on Ryzen / Linux

flyinryzen1700

I have to SLI killer x370. I'll try fixing the voltages. 3066 segfaulted in 2 min with fixed 1.37v. I didn't manually set any timings.

_blackbishop_ what are the other voltages you are using with that memory? So, you have 64GB total?

I have 32GB (4x8GB).

1.37V for 3200 may be too low.

Start with 1.4V. For 3466 I need 1.4 for new chip - old one required 1.42 for it and 1.4 for 3200. With 1.42 I can run 3600@16-16-16.

For SOC start 1.15V and downgrade in steps if you want after validating stability.

Try different procODT - most likely it will be between 40-60 ohm for you. I have to use 53.3 (48 is unstable, other don't even boot).

Additionally CAD bus settings (in AMD CBS/UMC options menu) helped me a lot with cold boot and overall stability at high speeds (I use 40-30-40-40, some recommend 40-20-40-40).

For timings "Stilt's 3200 safe settings" https://i.imgur.com/86WUZup.jpg  should be a good start.

If still stability is not 100% try changing VPPD up and down (not sure what is your default) and CLDO_VDDP (this is needed if you even can't boot, finding this one is pain in the... and very small changes make huge difference and every time requires cold start).