Author

Topic: 5xxx series cards with a 7xxx series card (Read 742 times)

full member
Activity: 206
Merit: 100
http://canadianbit.com/ - Let's Mine Canadian!
November 13, 2013, 07:56:42 PM
#13
I have tried to get a 5770 and 7950 to work in same instance of cgminer but it usually freezes or crashes after random periods of time. It seems different cards running in same mining software instance can cause problems in cgminer etc...  I have had better luck, like you, running each card in a separate instance (separate cgminers for each card for example).  Just thought I would throw out some experience I have had with running 2 diff cards in a system.  From what I have read (and experienced) it is better to have two of the same model cards to do that.  Good luck!  I have also used MultiMiner software but it also freezes if two different cards using both bfgminer and/or cgminer at same time.
hero member
Activity: 585
Merit: 500
November 13, 2013, 06:21:15 PM
#12
thanks - i've now configured 3 seperate instances of bfgminer for each of my gpu's... (as my 7950 is my primary card i've set the intensity to 15)
sr. member
Activity: 333
Merit: 250
"Raven's Cry"
November 13, 2013, 07:35:06 AM
#11
add this [ --device 1 --remove-disabled ]

--device (the card you want to use in each instance, e.g 0/1/2)
hero member
Activity: 585
Merit: 500
November 13, 2013, 05:37:40 AM
#10
i have 4GB in the system and 3 GPUs.

If i run the miner at 21712 my AMD driver crashes for the 58xx series cards

The 58xx cards don't have enough ram to use 21712 (and I never said to use 21712).  I said to use that on the 7950.  I presume you already have working settings for the existing cards.

Yeah the cards were working fine... how do i run separate instances of cgminer/bfgminer? whenever i run the miner it automatically pulls in all the GPUs.
hero member
Activity: 642
Merit: 500
November 12, 2013, 07:45:50 PM
#9
i have 4GB in the system and 3 GPUs.

If i run the miner at 21712 my AMD driver crashes for the 58xx series cards

The 58xx cards don't have enough ram to use 21712 (and I never said to use 21712).  I said to use that on the 7950.  I presume you already have working settings for the existing cards.
hero member
Activity: 585
Merit: 500
November 12, 2013, 07:00:41 PM
#8
i have 4GB in the system and 3 GPUs.

If i run the miner at 21712 my AMD driver crashes for the 58xx series cards
hero member
Activity: 642
Merit: 500
November 12, 2013, 05:25:28 PM
#7
It potentially will fix it, from my rough understanding that error comes from not having enough memory. Running that environment variable lets your miner use all of the ram available, so higher TCs are possible.



Correct.  The OpenCL driver will only claim a certain amount of memory by default.  It has to be overridden.  Run this before cgminer:

Code:
setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_USE_SYNC_OBJECTS 1

If it still fails, you don't have enough RAM in that machine (not likely though if there's only a few cards in it).
sr. member
Activity: 364
Merit: 250
November 12, 2013, 05:19:36 PM
#6
It potentially will fix it, from my rough understanding that error comes from not having enough memory. Running that environment variable lets your miner use all of the ram available, so higher TCs are possible.

hero member
Activity: 585
Merit: 500
November 12, 2013, 05:16:25 PM
#5
for the 7950, use this:

--thread-concurrency 21712 --worksize 256 -g 1 -I 20

Also, you'll need to set the GPU_MAX_ALLOC_PERCENT environment variable to 100 (instructions vary per OS).  Use a 9:8 ratio for engine:memclock.

i get this error when i set it to 21712

[2013-11-12 21:15:22] Maximum buffer memory device 0 supports says 536870912
[2013-11-12 21:15:22] Your scrypt settings come to 1422917632

is that what the GPU_MAX_ALLOC_PERCENT fixes?

I'm running win 7
hero member
Activity: 642
Merit: 500
November 12, 2013, 04:45:16 PM
#4
for the 7950, use this:

--thread-concurrency 21712 --worksize 256 -g 1 -I 20

Also, you'll need to set the GPU_MAX_ALLOC_PERCENT environment variable to 100 (instructions vary per OS).  Use a 9:8 ratio for engine:memclock.
hero member
Activity: 585
Merit: 500
November 12, 2013, 04:32:06 PM
#3
I have a gigabyte 7950 3GB version - hash rate is around 300 - accepted is 74 with 180 HW errors.
sr. member
Activity: 420
Merit: 250
November 12, 2013, 01:18:50 PM
#2
I have bought myself a 7950 for mining to add to my 5xxx series mining rig (replaced two 5770s with a 7950)

I've got the card working fine BUT i'm getting a lot of HW errors with the -thread-concurrency setting... i was expecting quite a performance increase from my 7950 but only seem to be getting about the same as what one of my 5770s was getting.

Current settings: -g 1 -w 256 -I 15,18,18 --auto-fan --auto-gpu --temp-target 70

I have tried a --thread-concurrency 8000 and 6142 and I get a lot of HW errors and the hash rate seems much lower than it should be...

Please post the stats you are getting as well as what version card it is.
hero member
Activity: 585
Merit: 500
November 12, 2013, 12:43:29 PM
#1
I have bought myself a 7950 for mining to add to my 5xxx series mining rig (replaced two 5770s with a 7950)

I've got the card working fine BUT i'm getting a lot of HW errors with the -thread-concurrency setting... i was expecting quite a performance increase from my 7950 but only seem to be getting about the same as what one of my 5770s was getting.

Current settings: -g 1 -w 256 -I 15,18,18 --auto-fan --auto-gpu --temp-target 70

I have tried a --thread-concurrency 8000 and 6142 and I get a lot of HW errors and the hash rate seems much lower than it should be...
Jump to: