Overclock.net banner
1 - 13 of 13 Posts

·
Registered
Joined
·
89 Posts
Discussion Starter · #1 ·
I've been fighting with this very weird issue for a while now. I have a 3090 FE and a 3080 FE. Both cards are overclocked, and completely stable standalone. However, the moment I start mining on both I get 0x116 TDR black screen reboots. I'm using a B550 GB Aourus Pro V2 board, latest beta bios. PSU is a Corsair AX1200i.

One thing I've noticed is that when both cards are active I see a 0.1V drop on the PCI-E 12V line, which is now my main suspect. 8pin power seems stable regardless of how many cards are running.

3090 alone:
2514610

3080 + 3090:
2514611

Any thoughts? I don't think I can do anything to stabilize PCI-E 12V power from BIOS can I? In which case I guess my only option is to reduce the TDP on one of them... Thanks.
 

·
Registered
Joined
·
207 Posts
I've been fighting with this very weird issue for a while now. I have a 3090 FE and a 3080 FE. Both cards are overclocked, and completely stable standalone. However, the moment I start mining on both I get 0x116 TDR black screen reboots. I'm using a B550 GB Aourus Pro V2 board, latest beta bios. PSU is a Corsair AX1200i.

One thing I've noticed is that when both cards are active I see a 0.1V drop on the PCI-E 12V line, which is now my main suspect. 8pin power seems stable regardless of how many cards are running.

Any thoughts? I don't think I can do anything to stabilize PCI-E 12V power from BIOS can I? In which case I guess my only option is to reduce the TDP on one of them... Thanks.
Under load a 5% variance in the 12 volt rail is considered acceptable. A good power supply will yield half that or less while under load. Unfortunately you did not mention how much load was on the GPU/s when you made your above screen shots, so I cannot comment on your PSU being okay or bad. Please let us know what kind of load the PSU was under when you took the screen shots.

FE cards have a problem with build quality. The cooler is poor in my opinion, Nvidia often put pads in the wrong place and have been known not to put a pad on the bottom VRAM chip, among other things. Just a very poor design, especially for extended high load.
 
  • Rep+
Reactions: Daylight_Invader

·
Senior OCN Member
Joined
·
41,221 Posts
Software readings are not arcuate you need a digtal multi meter to get arcuate readings so you can ignore those numbers
 

·
Registered
Joined
·
1,699 Posts
My advice is to start underclocking and undervolting and see if it gets stable. If that doesn't help, you can probably rule out power delivery as an issue.
 

·
Registered
Joined
·
89 Posts
Discussion Starter · #6 ·
Under load a 5% variance in the 12 volt rail is considered acceptable. A good power supply will yield half that or less while under load. Unfortunately you did not mention how much load was on the GPU/s when you made your above screen shots, so I cannot comment on your PSU being okay or bad. Please let us know what kind of load the PSU was under when you took the screen shots.

FE cards have a problem with build quality. The cooler is poor in my opinion, Nvidia often put pads in the wrong place and have been known not to put a pad on the bottom VRAM chip, among other things. Just a very poor design, especially for extended high load.
I know, according to the ATX spec, but who knows how the 3090 would react with that 0.1V drop being pushed to the edge as it is ;-) PSU reports between 620W-700W load in that scenario, stable 12V line load on the PSU output side. MB sensor reports stable 12V on the input side throughout. Temps are not an issue here - I'm maintaining <90C on both cards (pads changes, heatsinks etc).

Software readings are not arcuate you need a digtal multi meter to get arcuate readings so you can ignore those numbers
Sure, but the correlation is there regardless - there's a drop on the 12V PCI-E line with both cards working.

My advice is to start underclocking and undervolting and see if it gets stable. If that doesn't help, you can probably rule out power delivery as an issue.
Yeah, that's next.

Are you using individual PCIE leads to the GPU or daisy chaining ?
2x single 8pin cables. Also tried with my Corsair 12pin cable (single connector, goes into two ports on the PSU) same stuff...
 

·
Facepalm
Joined
·
10,066 Posts
It's possible, but it also may have little to do with the PCIE Slot 12v. ALL of your 12v rails are reading on the slightly low side. And the GPU's internal input rails (that it gets from the main 12v rails) are down to 11.6v. Might be responsible for stability issues. Hard to really tell.
 

·
Registered
Joined
·
89 Posts
Discussion Starter · #8 ·
It's possible, but it also may have little to do with the PCIE Slot 12v. ALL of your 12v rails are reading on the slightly low side. And the GPU's internal input rails (that it gets from the main 12v rails) are down to 11.6v. Might be responsible for stability issues. Hard to really tell.
Thanks. There's something definitely fishy going on. Now, I'm completely stable at the exact same settings, same load, and PCI-E voltage stays stable at about 11.79-11.8V! The only thing different - a reboot. Looks like some good old Gigabyte magic...
 
  • Rep+
Reactions: littledonny

·
Registered
Joined
·
1,699 Posts
Thanks. There's something definitely fishy going on. Now, I'm completely stable at the exact same settings, same load, and PCI-E voltage stays stable at about 11.79-11.8V! The only thing different - a reboot. Looks like some good old Gigabyte magic...
Weird. Keep us posted, it's an interesting setup.
 

·
Registered
Joined
·
720 Posts
DP
 

·
Registered
Joined
·
4 Posts
When current draw goes up the voltage will droop as long as you stay above 11.4V you're PSU is within spec with both a 3080 & 3090 pulling power from your PSU you have to expect some Vdroop you may be able to counter it slightly with undervolting but you'll still see Vdroop
 

·
Registered
Joined
·
89 Posts
Discussion Starter · #13 ·
OK, so at this point I am pretty sure that the instabilities I have only happen when the PC comes out from hybrid sleep. Twice now instead of a "reboot on load after cold start from hybrid sleep" my PSU's OC/OV protection actually triggered. So, not quite sure whether to blame PSU, MB or GPUs, but something gets seriously wrong with power delivery in that case, it would seem.

I've now disabled "fast start-up" - after a "normal" restart/shutdown it looks like I'm completely stable. Still testing though.
 
1 - 13 of 13 Posts
Top