Hi Just wondering if anyone is able to comment about how boards with PLX PEX chips behave within unraid? 

For example on the Asus WS x299 sage.  

 

And if anyone is able to comment on the IOMMU groupings for that board 

 

https://www.asus.com/Motherboards/WS-X299-SAGE/overview/

 

Appreciate the help

Hi, I'm currently troubleshooting an EVGA SR-X. On my machine, slot 1 is the unRAID console (GTX650) and slots 3+5 are GTX1070's meant to be passed-through to separate VM's running simultaneously.

 

Haven't gotten it to work, but doesn't look pretty. The SR-X has a PLX Chip (PEX8747) onboard, and I think it is causing GPU passthrough issues. If my theory is correct, it really depends on your motherboard's design.

The SR-X's PCI-E Lanes are wired terribly, one reason being that all 4 bottom slots to come from the PLX chip, and only CPU1's PCI-E lanes are used (Thats right, the entire CPU0's 40 LANES go to Waste...)

There's this rant by mosie on the EVGA forum about this exact issue (not unRAID, but PCI-E Lane design), and it's where I get my reasoning from.

According to mosie, both slot 3 and 5 are connected directly to the PLX chip, and if you look at my current scenario:

They are grouped together even with ACS Override enabled.

 

I think it inherently has to do with the way PLX chips work

 

So for me, my plan is to replace the top GTX650 with a single slot card, and move slot 3's card up to slot 2 (which is controlled by the CPU), and give it a try.

 

Again, I must stress that this is coming from me, a newb at unRAID myself, so I can't guarantee anything. Just know that you have at least one case against PLX unRAIDing...

 

Cheers

Edited March 24, 20188 yr by hihihi8
Coherency

Thanks for letting me know. Let me know how the swab up goes