-
Notifications
You must be signed in to change notification settings - Fork 27
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
so slow hashrate at AMD RX580 4gb Ubuntu #20
Comments
@khunpoum Sorry, that speed is the best it can do. There is a better one made by @zjazz https://github.com/zjazz/zjazz_amd_miner |
@fancyIX, there is one more pattern for lyra2 algo: https://github.com/KL0nLutiy/sgminer-kl/blob/master/kernel/lyra2REv2.mdz.cl, maybe it will be faster? |
@KL0nLutiy, nice, will try. Is "lyra2REv2.mdz.cl" faster than "lyra2rev2.cl" in that folder? If so, it will be great |
@fancyIX, yes it is faster, also you could use wolf-echo and wolf-skein from https://github.com/KL0nLutiy/sgminer-kl/blob/master/kernel/x16.cl, it's much faster. |
|
@KL0nLutiy now it's much faster, but still not good enough. Any idea how to push it further? |
What about the lycl pattern? Is it slower than mdz? |
@KL0nLutiy It's a little bit slower. |
@fancyIX, compiled that branch, but could not start miner failed during building binary with no errors. setx GPU_FORCE_64BIT_PTR=0 sgminer.exe -k phi2 -o stratum+tcp://yiimp.eu:8332 -u LfV448zDbNezd8EEVGgVcju2qJNtyfC1n8.worker2 -p donate -I 18 -w 64 -g 1 -d 1 --log-file log.txt |
@KL0nLutiy what does log.txt say? |
@KL0nLutiy "Failed to ADL_Overdrive5_ODPerformanceLevels_Get". Maybe it's irrelevant but you can try disable ADL |
@fancyIX, no it is always like this |
@KL0nLutiy I tried it and it can run on my vega 56 on ubuntu 18.04. Maybe you forgot to make clean or remove .bin file? |
@KL0nLutiy Or, maybe my sgminer cannot run with your env at all, not specific to that branch. Did you try master branch? |
Master branch is working I test it with rx 580 8gb on windows, no .bin files in the folder |
@fancyIX, could you please attech compiled miner with lycl pattern? |
Windows or Linux? I guess windows version, right?
Regards!
Changdao from cell phone
… On Aug 17, 2018, at 4:07 AM, KL0nLutiy ***@***.***> wrote:
@fancyIX, could you please attech compiled miner with lycl pattern?
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub, or mute the thread.
|
@KL0nLutiy I tried windows build but it cannot run, exits during compiling. Linux build works for me. Strange. |
@fancyIX, same behavior |
May I know why you need compiled #12 branch? I will try to figure out why
it cannot run on windows.
…On Fri, Aug 17, 2018 at 11:03 AM KL0nLutiy ***@***.***> wrote:
@fancyIX <https://github.com/fancyIX>, same behavior
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#20 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AeZiJ3s2J6XD4QBoqBZ9gEa7SRwHBLK8ks5uRwWFgaJpZM4V84P4>
.
|
@KL0nLutiy I just realized that branch is behind and cannot work with newest luxcore. Fixed. You can try newest code now. |
Just an idea for speed improvment, maybe split mdz kernel into 3 part like lycl pattern |
Also midstate for cubehash can be added |
@KL0nLutiy Thanks for the suggestion. That code is good for lyra2rev2, but seems not good for lyra2, since lyra2 requires 4x more memory of the matrix |
@KL0nLutiy about that new pattern, did you get chance to try it? It should be better than the original mdz pattern since more local memory is used to store matrix instead of global mem. But I don't know if it's better in lyra2 case. I will try it anyway. BTW, where did you find it? |
@fancyIX, no I did not try it, find compiled version on chinese forum. |
Now hashrate of RX 580 is about 1.3Mhs and Vega 64 is about 2.6Mhs |
No long "so slow" |
~800 khs at RX570/580 4gb
so slow hashrate, what is wrong?
amdgpu-pro 17.40
sgminer compiled from commit 844f29b (current master branch)
The text was updated successfully, but these errors were encountered: