LinuxQuestions.org

LinuxQuestions.org (/questions/)
-   Linux - Hardware (https://www.linuxquestions.org/questions/linux-hardware-18/)
-   -   3-way GTX 970 system won't boot - works fine with 2-way (https://www.linuxquestions.org/questions/linux-hardware-18/3-way-gtx-970-system-wont-boot-works-fine-with-2-way-4175600909/)

VolumetricSteve 03-02-2017 07:56 AM

3-way GTX 970 system won't boot - works fine with 2-way
 
I have an EVGA X58 Classified motherboard, which, as far as I can tell was designed to do 3-way SLI (while SLI doesn't matter in my particular case, I think it's worth noting) I'm running Crux 3.2 which is a from-source distro. With Kernel 4.10 and an nvidia driver that's no more than a month old, I'm running into a boot issue only when I have all 3 gpus in. Though all of it's regular boot output, it'll halt on "Booting the Kernel." and it just never does.

I've discovered that any two cards can work in any two slots....so it seems there's some issue specifically with having 3 cards in, I'm thinking something along the lines of an insufficient number of PCI-E lanes, but I'm not sure and I wouldn't know how to check anyway.

To anyone doing multi-gpu in linux, how'd you do it? What kernel are you using? How could I troubleshoot something like this?

business_kid 03-05-2017 05:07 AM

Nobody has had a guess at this, so allow me.

It sounds like a kernel or hardware problem. One thing you could try is to download Slackware's latest 'huge' kernel. Packages are in tar.xz format. My local mirror is
ftp://ftp.heanet.ie/mirrors/slackwar...2/slackware64/

You need the kernel-huge & kernel-modules txz packages which are tar.xz and located in the 'a' directory. Open them in a tmpdir, and copy it all into place & set it as an extra kernel. That has everything you mightn't have in your kernel, and will eliminate the kernel. Then you're down to hardware issues. I imagine you need to find someone who has done 3 GPUs running.

It may also be power supply limitations. GPUs suck power. How big is your power supply? There's a 'power good' line from the psu which monitors voltage. 5V or 3.3V could be low if overloaded. Or the tracking on the motherboard could be dropping too much voltage.

VolumetricSteve 03-06-2017 08:34 AM

Thanks for the reply. I'll give slackware a try, it's been a while.

I have a pretty serious PSU in there, 1200 watts I think, and these are the mini Asus GTX 970s, I can't imagine they're pulling enough power to bring that PSU down.

What's interesting is with 3 gpus, the system will POST just fine. If I remove bootable media, it'll happily sit with a "no bootable media found" type message. It feels like a software issue, so I'll try your slackware suggestion. Maybe some other OS could play nice too? I guess if I want to be really sure I should try windows to rule out hardware.


All times are GMT -5. The time now is 05:15 AM.