Topic: Annoying ZOMBIE cards issue when using PCI-E powered risers on server PSU's ? (Read 477 times)

I feel like that's why the GPU's run and gun like that after you power down, if you switched the power source from server to modular for the risers i'd bet they don't get crazy when rebooting. Glad it's become more stable though..they've got minds of their own sometimes it seems ha. for my setups with 650w/1200w modular/server psu's .. If I turn off my modular psu first, all my gpu's gun up to 110%. So If I ever power down I turn off everything at once.

that was my setup as well, but i did not use the modular PSU to power the risers ,  I think it was the motherboard after swapping it out the machine seems more stable no shutdowns, and when i power it off the fans spin on some cards like  @PcChips  said

but they are still detected as normal.

I swapped out the motherboard for a Asus 270-A one and it seems to not have that issue anymore .... wierd ? It seems the other board was the normal BTC board from biostar.

Maybe an issue with the bios on the motherboard ? I neevr updated that board its like 2-3 years old too.

I run a walmart-rack tower of 3 rigs, with 6 PSU's 
 
one of my server PSU's powers cards between two different rigs 
 
I don't have the issues you describe 
 
when the motherboard reboots, yes, the cards stay on, but it still works normally

I've had similar issues with server psu's and modular psu's working together to power a rig, or powering multiple setups..

Somethings not right with server PSU's when it comes to powering risers. Not sure what exactly; I have setups with and without risers and powering them w/ a server psu sucks ha.

So, I started powering the risers, mobo, and CPU with a 650w fully modular psu..leaving the server psu to do all the heavy lifting with GPU's. Not sure if it's just me.. but switching my power source from server to modular for merely power the risers, solved my glitchy power issue.

I really feel we've had the same issue. Hope this works out for you.

Sell all your rigs and build one 12 GPU rig for the new RX Vegas to handle the 4kw PSU. Problem solved.  Rumor has its going to be a solid miner compared to FE https://videocardz.com/71591/rumor-amd-radeon-rx-vega-64-to-be-great-for-mining

As you pointed out, putting graphics cards from multiple rigs on the same PSU puts more eggs in one basket. If that PSU needs to go down, or just fails, you have downtime for all connected cards.

Unfortunately, since all of the card's riser & PCIe power connections are from your big 4kW power supply, you'd need some kind of custom switch that only affects power to cards in an individual rig.

Sometimes the rigs re-boot, this is normal , say clay-more restarts the rig or my own scripts do when one GPU dies.

problem :

Unlike a normal connected PSU even if i shut down the rig completely, sometimes the GPUs hooked on the risers connected directly to the server PSU's will stay on and running , fans spinning like madmen even.


So of course when the reset occurs that GPU is not even detected since it was never powered off , even when the entire motherboard was shut down completely.

Keep in mind i have a separate small PSU, 500 watts or so that just powers the motherboard and CPU since I can't get the PICO's I ordered from Finsky to work (another problem in itself)

only way to get all the cards back up and running is to pull out the power connectors (PCI-E) from the risers so the cards can power down and then plug them back in then turn on the rig and all the cards are detected.

I cant really power of the 4000 watt server PSU in my case since the PSU is powering risers and cards on two other rigs.

Am I missing something simple here is is they basically an ugly issue when dealing with server style PSU's for multiple rigs ?

If this can't be fixed I don't really see the huge advantage of using huge server PSU"s to power multiple rigs to save costs since this annoying issue could force you to shut down all rigs connected.