HaloGenius
Jr. Member
Offline
Activity: 251
Merit: 4
|
|
December 22, 2021, 07:41:30 AM |
|
1.35 CUDA 11.4 build hangs on start with segmentation fault. Also archive contains some footprints of upx
|
Hive OS - mining platform #1. Register with promo code MINING4ALL and get 10$ to your balance
|
|
|
sp_ (OP)
Legendary
Offline
Activity: 2954
Merit: 1087
Team Black developer
|
|
December 22, 2021, 09:04:58 AM |
|
All that hard work stealing open source code, making some improvements,
This miner is original work. The stratum code is written from scratch, and have low hits on the antivirus compared to the other miners out there. The latest gminer:
|
|
|
|
|
th3dark1nSid3me
Newbie
Offline
Activity: 15
Merit: 0
|
|
December 22, 2021, 11:09:32 AM |
|
Just saw that 1.35 is available is SMOS. I gave it a try in a 8 x 3060 Ti FHR RIG. Same config as 1.32. Reported hashrate in SMOS is lower. 437.xx MH/s in 1.32 went to 415.xx MH/s in 1.35. The thing is that poolside it now seems more correct since 1.32 reported 437.xx MH/s was never the same as current or average. Tried 1.33 and had the same results as 1.35.
Was the 1.32 and previous versions over reporting the hashrate or is it something else? My cards have gone sub 60 MH/s now and I don't like it.
|
|
|
|
sp_ (OP)
Legendary
Offline
Activity: 2954
Merit: 1087
Team Black developer
|
|
December 22, 2021, 11:12:44 AM |
|
Try --dagintensity 8 (or find the best setting 1 to 9) (short option -K 9)
We have done a change in the threads per block configuration when using dagintensity. This was done to be able to run on higher clocks.
|
|
|
|
HaloGenius
Jr. Member
Offline
Activity: 251
Merit: 4
|
|
December 22, 2021, 11:25:09 AM |
|
1.35 CUDA 11.4 build hangs on start with segmentation fault.
Uploaded new 11.4 build. Please try again. Now seems it works. At least mining started and shares accepted by pool )
|
Hive OS - mining platform #1. Register with promo code MINING4ALL and get 10$ to your balance
|
|
|
Dojo76
Newbie
Offline
Activity: 30
Merit: 0
|
|
December 22, 2021, 12:22:48 PM |
|
Thanks, will you release cuda 11.4 build soon? The 11.5 build is faster, and early x-mas holiday now. So might not be buildt in a few days. The team is Tired. 10 months of hard work completed.. All that hard work stealing open source code, making some improvements, and then profiting off of it without giving back. Must be tiring. Is TBM based on an open source project? It seems like there guys are doing their own coding, they are very promising and adding value some to the scene. As soon as the latest version is released in Hive OS I'll test it again
|
|
|
|
th3dark1nSid3me
Newbie
Offline
Activity: 15
Merit: 0
|
|
December 22, 2021, 02:00:34 PM |
|
Try --dagintensity 8 (or find the best setting 1 to 9) (short option -K 9)
We have done a change in the threads per block configuration when using dagintensity. This was done to be able to run on higher clocks.
This is my config -d -K 1 --dagintensity 1 --xintensity 284 -a ethash -H eu1.ethermine.org -p 4444 -r 40 -y 20 -t 120 -W ETH_WALLET -w $rigName -P x --maintenance-wallet MAINTENANCE_WALLET --maintenance-percent 5 As I understood dagintensity should be lower in higher clocks. Has this changed in new version?
|
|
|
|
sp_ (OP)
Legendary
Offline
Activity: 2954
Merit: 1087
Team Black developer
|
|
December 22, 2021, 02:27:03 PM Last edit: December 22, 2021, 08:51:28 PM by sp_ |
|
This is my config
-d -K 1 --dagintensity 1 --xintensity 284 -a ethash -H eu1.ethermine.org -p 4444 -r 40 -y 20 -t 120 -W ETH_WALLET -w $rigName -P x --maintenance-wallet MAINTENANCE_WALLET --maintenance-percent 5
As I understood dagintensity should be lower in higher clocks. Has this changed in new version?
yes. The dagintensy has changed. 1 is too low. try 2 or 4 or more. Previously the dagintensity only set the config for the dag creation, but now we use this setting for the solver as well. Default is 0 (9 on nvidia)
Is TBM based on an open source project? It seems like there guys are doing their own coding, they are very promising and adding value some to the scene.
TBM is a new miner and not based on any other project. I am an opensource dev and have contributed to various crypto projects since 2014.
|
|
|
|
Alexstabilini
Jr. Member
Offline
Activity: 42
Merit: 2
|
|
December 22, 2021, 03:28:28 PM |
|
This is my config
-d -K 1 --dagintensity 1 --xintensity 284 -a ethash -H eu1.ethermine.org -p 4444 -r 40 -y 20 -t 120 -W ETH_WALLET -w $rigName -P x --maintenance-wallet MAINTENANCE_WALLET --maintenance-percent 5
As I understood dagintensity should be lower in higher clocks. Has this changed in new version?
yes. The dagintensy has changed. 1 is too low. try 2 or 4 or more. Previously the dagintensity only set the config for the dag creation, but now we use this setting for the solver as well. Default is 0 (9 on nvidia) Hi! My rig ( 2x1060+9xp104+3060ti+4000) works only with dagintensity 1 in 1.35 any suggestions to solve the problem? alex
|
|
|
|
sp_ (OP)
Legendary
Offline
Activity: 2954
Merit: 1087
Team Black developer
|
|
December 22, 2021, 03:44:48 PM |
|
Hi! My rig ( 2x1060+9xp104+3060ti+4000) works only with dagintensity 1 in 1.35 any suggestions to solve the problem? alex
Dag verification failed? you can try to halve the xintensity and use --dagintensity 2 13 cards need alot of virtual memory in the rig
|
|
|
|
KillrBee
Jr. Member
Offline
Activity: 139
Merit: 3
|
|
December 22, 2021, 04:11:56 PM |
|
Thanks, will you release cuda 11.4 build soon? The 11.5 build is faster, and early x-mas holiday now. So might not be buildt in a few days. The team is Tired. 10 months of hard work completed.. All that hard work stealing open source code, making some improvements, and then profiting off of it without giving back. Must be tiring. I guess they're really playing a good long game, since they keep generating new features and slowly fixing bugs, which based on your comments, I assume they're generating on purpose to be sneaky and underhanded about some GNU violation? Even if they did take open source code and modify it for commercial purposes, the software is better than any others out there and there are a lot worse things in this world to get upset about. The worst is that they have incorporated an open source project and built out and around it, which is common practice. Even creating a closed-source project based on the modification of the core open source code is not exactly uncommon, and there are far more flagrant violators of the GNU agreement for you to gripe about; how about elastic search, they _IPO'd_ on modified open source software which wasn't shared with the community. I am a single point of reference for the validity of this software, it works very well for me. I'm not talking about the client generated hash numbers - I agree that there are a number of cases where the client generously over-reports the hash rate, but my actual at-pool share rate calculations still results in a hash rate of 132-140 for my 3090 and 65-67 for my 6#00 series AMD cards, and I am hoping for more improvement in the AMD cards through personal tuning and software improvements -- and this part of the software can't be open source stolen, because no other miner comes anywhere close in terms of performance, and no other miner comes even remotely close in terms of the methods required to tune the software to obtain such high at pool results. I feel sad about the many obviously desperately unhappy people in the mining software community, because there are so many other things to direct focus and energy on than spite and undirected anger; live your life the way you wish, but trust me, there are happier and more satisfying ways to enjoy our lives.
|
|
|
|
sp_ (OP)
Legendary
Offline
Activity: 2954
Merit: 1087
Team Black developer
|
|
December 22, 2021, 05:37:56 PM Last edit: December 22, 2021, 05:48:54 PM by sp_ |
|
my actual at-pool share rate calculations still results in a hash rate of 132-140 for my 3090 and 65-67 for my 6#00 series AMD cards, and I am hoping for more improvement in the AMD cards through personal tuning and software improvements -- and this part of the software can't be open source stolen, because no other miner comes anywhere close in terms of performance, and no other miner comes even remotely close in terms of the methods required to tune the software to obtain such high at pool results.
Thanks for your support. This miner might be harder to setup correctly, but if you do, you will be rewarded. I plan to create a new improved autotune in the next release which will check 72 kernels/launchconfigs instead of 8 today. Playing with the --dagintensity in v1.35 can actually gain a little bit on some cards. The default 9 is not always the fastest. The modification is not done on amd cards yet, but can be implemented. I might seperate the tweak into a new setting.
|
|
|
|
wasarianta
Newbie
Offline
Activity: 40
Merit: 0
|
|
December 22, 2021, 06:21:46 PM |
|
Thanks for your support. This miner might be harder to setup correctly, but if you do, you will be rewarded.
I plan to create a new improved autotune in the next release which will check 72 kernels/launchconfigs instead of 8 today. Playing with the --dagintensity in v1.35 can actually gain a little bit on some cards. The default 9 is not always the fastest. The modification is not done on amd cards yet, but can be implemented. I might seperate the tweak into a new setting.
72 kernels ...you kill me... sir ...u are already did with --xintensity xxxxx haha.. autotune will help so...... much and will you make autotune kernels be dynamic something like --xintensity [-1] or at least at periodic time or epoch change or whatever haha no idea.. what now --dagintensity in v1.35 are have effect .. well i will find out..
|
|
|
|
sp_ (OP)
Legendary
Offline
Activity: 2954
Merit: 1087
Team Black developer
|
|
December 22, 2021, 06:54:45 PM Last edit: December 22, 2021, 07:22:05 PM by sp_ |
|
autotune will help so...... much and will you make autotune kernels be dynamic something like --xintensity [-1] or at least at periodic time or epoch change or whatever haha no idea..
what now --dagintensity in v1.35 are have effect .. well i will find out..
The problem with the current autotune is that it runs for a too short time. Ideally it should run for minutes and not seconds. The problem is that the hashrate is reduced abit when trying out bad configurations. We don't have all the cards and models, so our solution is to create a program that will try to calcuate the best settings automaticly, or let you configure it yourself. Nothing beats Human testing. V1.35 have 4 tuning options. Different cards have different optimal settings. Different power settings have different optimal settings. Different pools have different optimal settings. --kernel (0-9) --xintensity (0-10000) / (0-2500 amd) --tweak (0-8) --dagintensity (0-9) / (0-8 amd) So tune and be rewarded
|
|
|
|
Wolf9466
Newbie
Offline
Activity: 22
Merit: 2
|
|
December 23, 2021, 11:45:05 PM |
|
Again, you're lying and then dodging the issue. I didn't say you copied gminer - what I said was, you copied the Ethash OpenCL kernel.
|
|
|
|
sp_ (OP)
Legendary
Offline
Activity: 2954
Merit: 1087
Team Black developer
|
|
December 24, 2021, 12:49:55 PM |
|
I feel sad about the many obviously desperately unhappy people in the mining software community, because there are so many other things to direct focus and energy on than spite and undirected anger; live your life the way you wish, but trust me, there are happier and more satisfying ways to enjoy our lives.
Well written. Merry x-mas.
|
|
|
|
wasarianta
Newbie
Offline
Activity: 40
Merit: 0
|
|
December 25, 2021, 08:07:52 AM |
|
sir... guide me for 3060ti lhr in hiveos for v1.35 extra parameters --lhr-unlock, --xintensity, --dagintensity?? CUDA 11.4 working?
thanks
|
|
|
|
sp_ (OP)
Legendary
Offline
Activity: 2954
Merit: 1087
Team Black developer
|
|
December 25, 2021, 09:08:24 AM Last edit: December 25, 2021, 09:21:19 AM by sp_ |
|
Only lhr-unlock and tweak works in the lhr mode. It is made for cuda 11.5 and the latest driver, but should work on cuda 11.4 as well. Expected speed on the 3060ti is 42mhash with 60% power and 7900 memory. This is with --lhr-unlock [1] Hiveos is using the short options, so you have to find the correct letters from the short options. more info here: https://github.com/sp-hash/TeamBlackMiner/blob/main/FAQ.md
|
|
|
|
th3dark1nSid3me
Newbie
Offline
Activity: 15
Merit: 0
|
|
December 25, 2021, 10:16:55 AM |
|
autotune will help so...... much and will you make autotune kernels be dynamic something like --xintensity [-1] or at least at periodic time or epoch change or whatever haha no idea..
what now --dagintensity in v1.35 are have effect .. well i will find out..
The problem with the current autotune is that it runs for a too short time. Ideally it should run for minutes and not seconds. The problem is that the hashrate is reduced abit when trying out bad configurations. We don't have all the cards and models, so our solution is to create a program that will try to calcuate the best settings automaticly, or let you configure it yourself. Nothing beats Human testing. V1.35 have 4 tuning options. Different cards have different optimal settings. Different power settings have different optimal settings. Different pools have different optimal settings. --kernel (0-9) --xintensity (0-10000) / (0-2500 amd) --tweak (0-8) --dagintensity (0-9) / (0-8 amd) So tune and be rewarded I really appraise TBM efforts and hope you'll become the next big thing. Since all the mentioned options vastly change the behavior of a RIG, could you mention recommended settings for all the mainstream cards or recommended low and high values for different cards along with how a setting changes the behavior of the card?
|
|
|
|
|