Thanks for the extensive explanation!
I mistyped, I'm using raw intensity and I'm on 13.12.
The same values, if set on the command line, work fine.
It hangs only when set live.
Currently I'm getting 3.350 Kh from a 7950 and 3.730 from an r9 290.
I mistyped, I'm using raw intensity and I'm on 13.12.
The same values, if set on the command line, work fine.
It hangs only when set live.
Currently I'm getting 3.350 Kh from a 7950 and 3.730 from an r9 290.
Can you post your exact command line, and then examples of what fails? also, how much memory your cards and system have as well as the number of shaders being reported by HWInfo or GPU-z (alternatively, you could launch yacminer with -D and -T and comb through the output, but that's messy - it will eventually get added to pretty output somewhere, sometime.)
rI has a similar limitation to xI - multiples of worksize are known to work, in-between values are questionable. I haven't been able to nail down what is affected and what is not - there are just too many variables. I thought it was just R7/R9 that was exempt, then thought it was driver - I know my R7 240, and R7 250 aren't limited, both on 13.12. My 7850M (unknown driver version) is not, but my 7850 desktop on 12.10 (I think) is limited. Also, no clue about what values can be valid under what driver version of linux...
However, in none of the cases has the computer or software hung (I compile and run for Windows only though). For me, if it finds a value it doesn't like, and can't enqueue it onto the openCL kernel, it disables the GPU. Setting rI back to a valid value and then re-enabling the GPU always gets it back to work for me.
here is the command line:
-R 6144,1856 --lookup-gap 3 --buffer-size 3264,2496 --gpu-engine 1100,1100 --gpu-memclock 1250,1250
what fails is changing -R "on the fly".
the first card is an r9 290 with 4GB and the second is a 7950 with 3GB (both failing), system has 8GB ram.
how can I see the number of shaders on linux?
when a card fails that way, if I try restarting, it goes "sick" and I need to reboot.
at this point, I believe it might be a driver issue (I'm using xedgers repository, which usually has beta drivers, but with ubuntu default ones it was much worse).
So, if you aren't specifying --worksize, it defaults to 256, and I wouldn't use -R 1856 (try -R 6144,1792 or throw a --worksize 256,64).
I really can't answer questions about linux - you'll have to google it, or, as I said, run it with -T -D and look through the output, it does get output somewhere in there.
So, the AMD supplied linux drivers don't work for mining? That really surprises me. I know Ubuntu is Debian based, and a colleague is running numerous machines on debian using YACMiner, and has been quite successful, even while still learning what he was doing.
)
