Only Using 3 of 4 GPU works

[thorwaldhoffmann]

Hello together.

We got a new workstation in the office for rendering stuff (what else )

We tried to render different scenes and got some problems, crashes, freezes, etc. using it as a slave didn't show any problems or I just didn't recognize them.

For animations we got problems rendering over night, or after like 3-5 hours.

before I do a big documentation of what we tried and what problems we got, there are maybe some of you have a simular setup without problems.

Threadripper 2950x
Gigabyte X399 Designare
128 GB RAM
4x 2080 ti
Win 10
C4D R20
Octane V4 (tried 4.02.1-R3, 4.03, 4.04)
Driver (tried 417.22, 417.71, 419.17)

Anyone who have simular or maybe same CPU+Mainboard+GPU setup?

Using only 3 cards works fine. Deactivating one of them (not one specific card).

And what about single Octane Bench results: Is it normal to have one card that has lower OB? I got 307, 309, 311 and 244. Testing all separate (OB 4.00c).

I'm at the end of my little knowledge.

[firstperson]

even if all the cards are the same make and model they will have variations in temp at full load, and clock speed thus resulting in slightly different OB scores.
sometimes i get a card that is 20+ lower score and i can see its running hotter at full load (maybe due to the cooler not being as efficient or just the GPU itself idk) and i will exchange a card like that for new one.

if you have trouble getting all four cards to render without crashing could be a power issue, whats your PSU?

[thorwaldhoffmann]

even if all the cards are the same make and model they will have variations in temp at full load, and clock speed thus resulting in slightly different OB scores.
sometimes i get a card that is 20+ lower score and i can see its running hotter at full load (maybe due to the cooler not being as efficient or just the GPU itself idk) and i will exchange a card like that for new one.

if you have trouble getting all four cards to render without crashing could be a power issue, whats your PSU?

CORSAIR 1600W AX1600i Modular is the PSU.
It should have enough power but do you think there could be a problem?

This night I tried to render a 4K interior image.
for testing purpose I really cranked up the samples (to 36000) to have 10 hours of rendertime. after some hours I got cuda error 702 and cinema crashed (see screenshot)
the 4th card totally "died" and came back up after restart (had this several times now).
The 4th card is not showing anything in MSI Afterburner anymore and cant be regonized by "Corsair Link 4" Tool. Restart helped and the card is back.

See several screenshots of crashes.
When I had driver 417.22 I got total system crashes and no chance for screenshots.

Would be great if anyone got ideas for solving.

[akihikotv]

i used to have a problem like this.... did you check your temps??? maybe they dont have enough room to take air...

No Air space

No Air Flow

-*----

Solution

My Solution

Now I dont have any problems with crashes rendering in Octane

--------

Akizilla BattleStation

i7 7820x
64GB ram
Msi Plus x299
3 1080Ti (2x Asus Strix OC + 1 Zotac Amp Edition)
1000w EVGA Gold

[thorwaldhoffmann]

hi akihikotv,

thanks for your input.
temps are no problem i think. its cooled by eiswolf cooling system.

and the problem is just there with Threadripper + 4x 2080 ti setup.

any other ideas? didn't figure out to solve the problem yet.

thanks!

[glimpse]

Have You tried to look through those posts on gigabyte forums I've pointed You on facebook? =)

[thorwaldhoffmann]

hi tom,

thanks for the thread (just to collect it here and maybe it helps others: http://forum.gigabyte.us/thread/1864/gi ... sXmi8Ghhi4)
I will show this to my IT-guy as well. We deleted the driver via DDU and installed 417.22 now to test it further.
I also need to test it when I use the machine with 4 cards as a slave and see if something happens. The problem is, that I planned to use the machine as my main workstation and add more of threadripper+quad gpu systems as slaves but I stopped my plans obviously.
Lets wait until I did a 5-6 hour animation render-test now. I'll keep you updated.

[zxctnt]

Hi friend,i have seen your post on Facebook group.And i have the similar rig system with you. Currently I don‘t know it is the problem of the AMD platform+20 series card hardware rig. Or it is the driver and software issue.

Having seen many posts have issue with x399 motherboard with 128GB RAM and 4x20 series cards. So i think this is not an accident. Hope some friends to find out the original factor of these issues.

This is the post of my issue.
viewtopic.php?f=86&t=71534