|
andrucrypt (OP)
Member
Offline
Activity: 719
Merit: 49
|
|
October 30, 2018, 11:38:48 PM |
|
HIVEOS I am getting same error both 10 x rx 580 and 12 x rx 560 rigs. Whats missing? I can't open the screen, but I guess you are missing dependencies, just run this: apt install libuv1 libmicrohttpd10
|
|
|
|
andrucrypt (OP)
Member
Offline
Activity: 719
Merit: 49
|
|
October 30, 2018, 11:44:21 PM |
|
Is it possible for you to optimize the hawaii kernels any further? The best I could get for a non-reference 290x was 2.9 mh/s for hmq1725 for example, somewhere between a 460 and 470. And 2.14 mh/s on x22i though I haven't tried 12.6 yet. My reference 290x will probably fair better but not by much. Thanks.
Old hardware like Hawaii needs special techniques, and without having those cards I can't do a lot here... Anyway, did you try different opencl-launch parameters? Old cards like higher worksize(e.g. try 20x256 instead of 20x128 and so on) and maybe better to use lower intensity or higher(that number 20)
|
|
|
|
ETminer
Newbie
Offline
Activity: 3
Merit: 0
|
|
October 31, 2018, 12:43:32 AM |
|
For x22i, I'm seeing mid 3000 kH/s for both rx570s and rx Vegas across a variety of intensity/worksize settings. I would have expected to get more hash out of a Vega. Is anyone seeing anything similar or doing better with this algo?
card cclck/v mclck/v int/work hash/s ------------------------------------------------------------- rx570 8GB 1250/900mv 2100/860mV 19x0 3400 Vega 56 1527/930mv 910/910mv 21x256 3300 Vega 64 1632/950mv 1025/930mv 21x256 3800
Power consumption is pretty good, though. Getting 800W at the wall with 3 Vegas and 2 rx570s.
--ET
|
|
|
|
hizzy2023
Newbie
Offline
Activity: 30
Merit: 0
|
|
October 31, 2018, 12:46:57 AM |
|
Why is Linux version in .7z format? How to unpack that?
|
|
|
|
alucard20724
|
|
October 31, 2018, 01:12:37 AM Last edit: October 31, 2018, 02:06:25 AM by alucard20724 |
|
For x22i, I'm seeing mid 3000 kH/s for both rx570s and rx Vegas across a variety of intensity/worksize settings. I would have expected to get more hash out of a Vega. Is anyone seeing anything similar or doing better with this algo?
card cclck/v mclck/v int/work hash/s ------------------------------------------------------------- rx570 8GB 1250/900mv 2100/860mV 19x0 3400 Vega 56 1527/930mv 910/910mv 21x256 3300 Vega 64 1632/950mv 1025/930mv 21x256 3800
Power consumption is pretty good, though. Getting 800W at the wall with 3 Vegas and 2 rx570s.
--ET
I get 5500 on my vega 64 with stock voltages and mem clock set to 1000 with 18x0 two threads with version 0.12.5.1.. haven't tried with 12.6 yet with those settings update: i just rebenched my vega64 with stock voltages, mem clock set to 1000 with 18x0 two threads with version 12.5.1 i get 5500 hash. with 12.6 with same settings i get 3850 hash... both done with win10 pro x64, and 18.10.2 drivers with hbcc set to off. i ran a test earlier with hbcc on and was slightly less. less than 1%. I keep hbcc off now because of cvn8
|
|
|
|
jtmillican
Newbie
Offline
Activity: 31
Merit: 0
|
|
October 31, 2018, 01:34:22 AM |
|
Not sure how to set this up on HiveOS. Asking me for a URL to miner download. Is one available? Any advice appreciated.
|
|
|
|
JK2206
Newbie
Offline
Activity: 2
Merit: 0
|
|
October 31, 2018, 04:31:51 AM |
|
I've got a rig of mainly rx580s but also rx460 when I try recommended --opencl-threads 3 --opencl-launch 18x128 or 20x128 the rx580s run great but my rx460 dies stops hashing shows n/a and have to reboot the rig to bring it back up if I don't use any extra parameters the rx460 hashes but the rest don't seem to be optimized
|
|
|
|
justoiurii
Newbie
Offline
Activity: 103
Merit: 0
|
|
October 31, 2018, 06:27:34 AM |
|
WOW! Well done, Master! Was 18,8 Mh, now - 21,3. Locking the same. RX480, bc driver
|
|
|
|
andrucrypt (OP)
Member
Offline
Activity: 719
Merit: 49
|
|
October 31, 2018, 06:32:28 AM |
|
For x22i, I'm seeing mid 3000 kH/s for both rx570s and rx Vegas across a variety of intensity/worksize settings. I would have expected to get more hash out of a Vega. Is anyone seeing anything similar or doing better with this algo?
card cclck/v mclck/v int/work hash/s ------------------------------------------------------------- rx570 8GB 1250/900mv 2100/860mV 19x0 3400 Vega 56 1527/930mv 910/910mv 21x256 3300 Vega 64 1632/950mv 1025/930mv 21x256 3800
Power consumption is pretty good, though. Getting 800W at the wall with 3 Vegas and 2 rx570s.
--ET
Looks like there some specific thing with Vega, working to fix that. Thanks for tests!
|
|
|
|
andrucrypt (OP)
Member
Offline
Activity: 719
Merit: 49
|
|
October 31, 2018, 06:34:14 AM |
|
I've got a rig of mainly rx580s but also rx460 when I try recommended --opencl-threads 3 --opencl-launch 18x128 or 20x128 the rx580s run great but my rx460 dies stops hashing shows n/a and have to reboot the rig to bring it back up if I don't use any extra parameters the rx460 hashes but the rest don't seem to be optimized
What algo? Not all can run 3 threads. Also, most of algo's need higher voltaqe for gpu's, so give that info too please.
|
|
|
|
andrucrypt (OP)
Member
Offline
Activity: 719
Merit: 49
|
|
October 31, 2018, 06:35:42 AM |
|
Not sure how to set this up on HiveOS. Asking me for a URL to miner download. Is one available? Any advice appreciated.
Check OP post, there is a link HiveOS asks you
|
|
|
|
hizzy2023
Newbie
Offline
Activity: 30
Merit: 0
|
|
October 31, 2018, 02:51:17 PM |
|
i get Error CL_OUT_OF_HOST_MEMORY when calling clCreateCommandQueueWithProperties in red when i try custom hiveos installer and try to run it. Any ideas?
|
|
|
|
dragonmike
|
|
October 31, 2018, 03:00:26 PM Last edit: October 31, 2018, 03:39:39 PM by dragonmike |
|
I've got 7 GPUs on some rigs. If I don't specify --opencl-devices=7, the first GPU will hash but without the thread, intensity and worksize parameters I assign. It'll just use defaults. If I do specify --opencl-devices=7, miner will not initialise and keep looping the start procedure.Bleh, ignore me, I typed the opencl-config parameters twice in the same batch file
|
|
|
|
melloyellow
|
|
October 31, 2018, 03:50:47 PM |
|
Is it possible for you to optimize the hawaii kernels any further? The best I could get for a non-reference 290x was 2.9 mh/s for hmq1725 for example, somewhere between a 460 and 470. And 2.14 mh/s on x22i though I haven't tried 12.6 yet. My reference 290x will probably fair better but not by much. Thanks.
Old hardware like Hawaii needs special techniques, and without having those cards I can't do a lot here... Anyway, did you try different opencl-launch parameters? Old cards like higher worksize(e.g. try 20x256 instead of 20x128 and so on) and maybe better to use lower intensity or higher(that number 20) I have played around with all the settings. In general they do seem to like 256 but with 2 threads and yes intensity around 20 is best. Sucks, I know these cards are much more capable than the speeds I can do now but I can't write code. Thanks for providing the kernels and thinking of us older guys at least. I'll keep posting my hashrates as I go along, maybe someone will come up with better settings... hmq1725 : 2.9 mh/s x22i : 2.3 mh/s x16r : 5 mh/s estimated, due to the random order
|
|
|
|
andrucrypt (OP)
Member
Offline
Activity: 719
Merit: 49
|
|
October 31, 2018, 05:48:51 PM |
|
Is it possible for you to optimize the hawaii kernels any further? The best I could get for a non-reference 290x was 2.9 mh/s for hmq1725 for example, somewhere between a 460 and 470. And 2.14 mh/s on x22i though I haven't tried 12.6 yet. My reference 290x will probably fair better but not by much. Thanks.
Old hardware like Hawaii needs special techniques, and without having those cards I can't do a lot here... Anyway, did you try different opencl-launch parameters? Old cards like higher worksize(e.g. try 20x256 instead of 20x128 and so on) and maybe better to use lower intensity or higher(that number 20) I have played around with all the settings. In general they do seem to like 256 but with 2 threads and yes intensity around 20 is best. Sucks, I know these cards are much more capable than the speeds I can do now but I can't write code. Thanks for providing the kernels and thinking of us older guys at least. I'll keep posting my hashrates as I go along, maybe someone will come up with better settings... hmq1725 : 2.9 mh/s x22i : 2.3 mh/s x16r : 5 mh/s estimated, due to the random order Today I will release new build with x22i improvements, and I hope it will increase hashrate a lot for Hawaii, same as for Vega cards
|
|
|
|
anilkilic
Newbie
Offline
Activity: 8
Merit: 0
|
|
October 31, 2018, 06:51:44 PM |
|
i get Error CL_OUT_OF_HOST_MEMORY when calling clCreateCommandQueueWithProperties in red when i try custom hiveos installer and try to run it. Any ideas?
I am getting this error too. I can't open the screen, but I guess you are missing dependencies, just run this: apt install libuv1 libmicrohttpd10 no its CL_OUT_OF_HOST_MEMORY when calling clCreateCommandQueueWithProperties and i already have this dependecies
|
|
|
|
andrucrypt (OP)
Member
Offline
Activity: 719
Merit: 49
|
|
October 31, 2018, 07:34:27 PM |
|
0.12.7 beta - fixed low hashrate on Vega cards - more improvements to x22i - now higher intensity can be set
NOTE: don't forget to check updated OPTIMAL SETTINGS section with new optimal parameters for this version anilkilic, hizzy2023 try this new build, it should consume less memory.
|
|
|
|
cantr
Member
Offline
Activity: 124
Merit: 10
|
|
October 31, 2018, 09:25:24 PM |
|
Firstly; thank you for new version 0.12.7 beta.
tried and worked well! Regards.
But my hashrate for Sapphire Nitro RX570 4 Gb Hynix - x22i - 3,1 Mhs
I am waiting information abouth best oc & other settings.
|
|
|
|
andrucrypt (OP)
Member
Offline
Activity: 719
Merit: 49
|
|
October 31, 2018, 09:28:27 PM |
|
Firstly; thank you for new version 0.12.7 beta.
tried and worked well! Regards.
But my hashrate for Sapphire Nitro RX570 4 Gb Hynix - x22i - 3,1 Mhs
I am waiting information abouth best oc & other settings.
Did you try optimal parameters I have shared in OP? Also this is not bad result, if you are running your gpu with some undervolt/underclock. My tests are done for stock clocks and modified timings for memory.
|
|
|
|
|