Author

Topic: 4th GPU dying - Sapphire R9 280X Dual-x (Read 1479 times)

newbie
Activity: 24
Merit: 0
February 11, 2014, 10:04:09 AM
#10
I acctually found out the undervolting was the problem... or Sapphire Trixx rather. reseting clocks etc made everything stable so it's trixx thats crashing my card

You were having the problem after under volting the cards?

Yea. I reseted them with TRIXX again and rebooted computer and everything has been running fine

Be careful whenever you undervolt.

Careful about what? Never underclocked before only overclocked so new to this business ^^
hero member
Activity: 630
Merit: 502
February 11, 2014, 02:00:38 AM
#9
Flashing a custom BIOS with modified clock speeds, voltage and fan profile is the most reliable method in my experience.
full member
Activity: 238
Merit: 100
February 11, 2014, 01:51:50 AM
#8
I acctually found out the undervolting was the problem... or Sapphire Trixx rather. reseting clocks etc made everything stable so it's trixx thats crashing my card

You were having the problem after under volting the cards?

Yea. I reseted them with TRIXX again and rebooted computer and everything has been running fine

Be careful whenever you undervolt.
newbie
Activity: 24
Merit: 0
February 10, 2014, 05:26:59 PM
#7
I acctually found out the undervolting was the problem... or Sapphire Trixx rather. reseting clocks etc made everything stable so it's trixx thats crashing my card

You were having the problem after under volting the cards?

Yea. I reseted them with TRIXX again and rebooted computer and everything has been running fine
full member
Activity: 238
Merit: 100
February 10, 2014, 02:03:02 PM
#6
I acctually found out the undervolting was the problem... or Sapphire Trixx rather. reseting clocks etc made everything stable so it's trixx thats crashing my card

You were having the problem after under volting the cards?
newbie
Activity: 24
Merit: 0
February 10, 2014, 05:01:14 AM
#5
I acctually found out the undervolting was the problem... or Sapphire Trixx rather. reseting clocks etc made everything stable so it's trixx thats crashing my card
legendary
Activity: 952
Merit: 1000
February 10, 2014, 01:13:07 AM
#4
You haven't undervolted?

900-100W from wall seems impossible for 4x 280x. I run 1100 from wall with undervolts from 1.200v to 1.143v on each card.

Depends on how far you undervolt. Some of our Gigabyte 280x Rev2 could undervolt as low as 1.01V @ 1040/1500, and our 5 GPU rig would pull 1250W at the wall. If you were to take out one of the GPUs, I bet you it would pull just a hair over 1000W, maybe 1020W?
sr. member
Activity: 400
Merit: 250
February 09, 2014, 07:24:18 PM
#3
You haven't undervolted?

900-100W from wall seems impossible for 4x 280x. I run 1100 from wall with undervolts from 1.200v to 1.143v on each card.

newbie
Activity: 24
Merit: 0
February 09, 2014, 04:57:24 AM
#2
Buuump
newbie
Activity: 24
Merit: 0
February 08, 2014, 09:27:47 PM
#1
Heya!

Just expanded and added an exact same setup. Everytime I start cgminer after 2-3mins the 4th GPU dies - I get the message "driver has stopped responding and has recovered" and the 4th GPU becomes a CPU miner. If I disconnect 4th GPU the same thing happens to the 3rd one - which is fine when 4th is connected....

Windows 8.1
Corsair AX 1200i Digital 1200W 80+
Sapphire Radeon R9 280X 3GB DUAL-X OC x 4
Corsair 4GB (1x4GB) CL9 1600Mhz VENGEANCE
Intel Pentium G3220 3.0 GHz (Haswell)
ASUS Z87-PRO ATX
120GB Sandisk SSD

14.1 Driver - Tested with 13.25 aswell no improvements.
CCC disabled
Reformated comp. no fix
Pulling 900-1000W from wall

setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_USE_SYNC_OBJECTS 1
cgminer --scrypt -o stratum+tcp://amsterdam.middlecoin.com:3333 -u 1Ff5EdwbfSon8GZkVUyKLkte5mMuixi2tx -p x --failover-only -o stratum+tcp://middlecoin.com:3333 -u 1Ff5EdwbfSon8GZkVUyKLkte5mMuixi2tx -p x  --lookup-gap 2 --thread-concurrency 8193 -g 2 -I 13 -w 256 --auto-fan --gpu-fan 30-75 --temp-cutoff 90 --temp-overheat 85 --temp-target 72 --gpu-memclock 1500 --gpu-engine 1020 --gpu-powertune 15 --expiry 1 --scan-time 1 --queue 0

Help me out please!
Jump to: