Topic: [ANN] TeamRedMiner - Optimized Win/Linux AMD GPU Miner for lyra2z and phi2 - page 40. (Read 85791 times)
September 23, 2018, 11:31:49 AM
Is there anyway you can make this miner work on the AMDGPU Pro PAL driver? Specifically version 18.30? The issue is Cryptonight full rate hashing is not really supported until that version. This miner is the only one that must use ROCm. You have support for AMDGPU Pro but it is only legacy OpenCL support on the driver and not PAL support. Can you please support PAL or I guess Windows but I'd rather just have PAL support so my builds just work... at this point I need to have an image for all other miners then a seperate ROCm image for this miner.
September 23, 2018, 08:00:05 AM
API support please...
September 23, 2018, 06:14:53 AM
You can do solo extraction support on these algorithms. I think you will have so many more fans.
I have been searching for a long time than you can get a solo for Lyra2z. but only cpuminer-opt-3.8.8.1 for today could connect directly to the wallet!
September 21, 2018, 05:08:48 PM
todxx - good to see you back on the forum.
Looking forward to seeing what magic you can do to improve performance on my Vega56 cards. I've been using tdxminer 0.2.2.2 since release I think!
Thanks.
Looking forward to seeing what magic you can do to improve performance on my Vega56 cards. I've been using tdxminer 0.2.2.2 since release I think!
Thanks.
September 21, 2018, 12:36:25 PM
Oh well, always nice to bump the thread I guess - some little perk of replying...
September 21, 2018, 10:46:06 AM
any plan for windows?
I'm shocked sometimes by the inability of people who can't be bothered to read even two posts above their own, where their question would have been answered. Why even bother to respond? They probably can't even read the reply.
September 21, 2018, 09:37:03 AM
any plan for windows?
September 21, 2018, 09:22:51 AM
@dallase
Hmm, that's surprising. Will look into it as soon as I can find some time.
Hmm, that's surprising. Will look into it as soon as I can find some time.
More often than not after a "connection was closed by the pool" event it picks right back up.
Code:
[2018-09-21 09:08:05] Pool lux.pickaxe.pro share accepted.
[2018-09-21 09:08:07] Pool lux.pickaxe.pro received new job. (job_id: 530)
[2018-09-21 09:08:09] Stats GPU 0 - phi2: 5.558Mh/s (5.550Mh/s)
[2018-09-21 09:08:15] Pool lux.pickaxe.pro connection was closed by the pool.
[2018-09-21 09:08:18] Pool lux.pickaxe.pro successfully subscribed.
[2018-09-21 09:08:20] Pool lux.pickaxe.pro authorization succeeded.
[2018-09-21 09:08:20] Pool lux.pickaxe.pro set difficulty to 1.000000
[2018-09-21 09:08:20] Pool lux.pickaxe.pro extranonce subscribe succeeded.
[2018-09-21 09:08:23] Pool lux.pickaxe.pro received new job. (job_id: 52f)
[2018-09-21 09:08:27] Pool lux.pickaxe.pro share accepted.
[2018-09-21 09:08:30] Pool lux.pickaxe.pro received new job. (job_id: 530)
[2018-09-21 09:08:30] Pool lux.pickaxe.pro share accepted.
[2018-09-21 09:08:30] Pool lux.pickaxe.pro share accepted.
So I'm not sure what happens when it does not like the log I showed earlier. I've seen it happen 3 times now.
September 20, 2018, 11:38:05 AM
If I were to guess, Windows release will be end of next week (7-10 days from now). There are a few things we need to take care of before a win release. We also knew that this week would see very little progress as we've both been very busy.
September 20, 2018, 10:33:54 AM
Ahem...
...Windows?
...Windows?
September 20, 2018, 09:34:39 AM
@heavyarms1912 So this miner isn't using GCN Assembly?
Raw assembly is an integral part of our toolchain, yes. Don't know why it matters to the end user though, hash is hash?
Raw assembly is an integral part of our toolchain, yes. Don't know why it matters to the end user though, hash is hash?
Because I recall you mentioned Vega 64 do ~10 MHs and RX580 ~6Mhs on lyra2z ?
September 20, 2018, 09:26:00 AM
@heavyarms1912 So this miner isn't using GCN Assembly?
Raw assembly is an integral part of our toolchain, yes. Don't know why it matters to the end user though, hash is hash?
Raw assembly is an integral part of our toolchain, yes. Don't know why it matters to the end user though, hash is hash?
September 20, 2018, 08:34:41 AM
@keksik
Can't say I have anything good to say given the limited info, it sounds like you have one rig running fine on ethos 1.3.2, the other one not so much, with identical installs and similar hardware setups? And other miners work fine on the problematic rig?
Can't say I have anything good to say given the limited info, it sounds like you have one rig running fine on ethos 1.3.2, the other one not so much, with identical installs and similar hardware setups? And other miners work fine on the problematic rig?
Hi Thank you for your time looking at my comment, i was able to fix my problem as i forgot to set pcie gen to auto
rx 570 are hashing 3,1 mhs on 1250mhz core thanks again and looking forward for more improvements (read more hash for our cards
) September 20, 2018, 03:32:44 AM
@MinedTangerine
As I read the code on their github, the new LBK3 hash the VTL devs have created is a simple chain of Bmw, Blake, Keccak. My assumption was that it would contain some Lyra2 aspect as well, this from the leading "L" in the algo name. I can't find any indications of that being the case though. In this form, it should be trivial for anyone who knows the sgminer codebase to create a forked version supporting LBK3. The hashes involved are also difficult to optimize above the best publicly available OpenCL versions.
I could be missing something though, just did a very quick code browse of their github? Otherwise, the bottom line is that there's currently no point spending time on this algo for us as part of this miner, we can't add enough value to justify a dev fee, and it will be completely slaughtered by fpgas if/when it becomes interesting enough. The upcoming Acorns from GPUHoarder will most probably be capable of pipelining the full algo on its Artix-7 chip.
Out of curiousity, I've reached out to neithender on the VTL discord for final clarification though.
As I read the code on their github, the new LBK3 hash the VTL devs have created is a simple chain of Bmw, Blake, Keccak. My assumption was that it would contain some Lyra2 aspect as well, this from the leading "L" in the algo name. I can't find any indications of that being the case though. In this form, it should be trivial for anyone who knows the sgminer codebase to create a forked version supporting LBK3. The hashes involved are also difficult to optimize above the best publicly available OpenCL versions.
I could be missing something though, just did a very quick code browse of their github? Otherwise, the bottom line is that there's currently no point spending time on this algo for us as part of this miner, we can't add enough value to justify a dev fee, and it will be completely slaughtered by fpgas if/when it becomes interesting enough. The upcoming Acorns from GPUHoarder will most probably be capable of pipelining the full algo on its Artix-7 chip.
Out of curiousity, I've reached out to neithender on the VTL discord for final clarification though.
Thank you a lot for the input, I really appreciate it. Good findings and interesting. I did not check the code I admit. Hm its not the best choice of algos that they have decided to use but ok. I don't expect that algo to pick any major interest soon. But if they deliver what they promised it could get attention later.
Otherwise I agree it does not seem worth of your time especially if it cannot be optimized. Your work on the lyra2z kernel is amazing btw. You showed the green camp what reds are capable of.
September 20, 2018, 03:29:35 AM
The real winner in the Lyra2 race will be whoever releases an optimized miner for the AMD R9 series. There is still a LOT of that gen cards out there and I for one would tip/fee-mine with my old R9 cards to support that effort. Any plans for this?
R9 cards are not profitable anymore unless you have really cheap electricity. I have a few R9 380 and a R9 290x. And while 290x is a beast it eats power like no tommorow.
September 19, 2018, 10:06:03 PM
So this miner isn't using GCN Assembly?
September 19, 2018, 05:48:14 PM
The real winner in the Lyra2 race will be whoever releases an optimized miner for the AMD R9 series. There is still a LOT of that gen cards out there and I for one would tip/fee-mine with my old R9 cards to support that effort. Any plans for this?
September 19, 2018, 03:26:04 PM
@keksik
Can't say I have anything good to say given the limited info, it sounds like you have one rig running fine on ethos 1.3.2, the other one not so much, with identical installs and similar hardware setups? And other miners work fine on the problematic rig?
Can't say I have anything good to say given the limited info, it sounds like you have one rig running fine on ethos 1.3.2, the other one not so much, with identical installs and similar hardware setups? And other miners work fine on the problematic rig?
September 19, 2018, 02:22:52 PM
i tried ethos 1.3.2 on another rig, with asus mining expert and 9 gpu (mix 570/580) and it works a bit creepy.. 1-2.4mhs per card and i am using 1200 (950) / 900, so it should work without problem...any advice?
September 19, 2018, 02:21:17 PM
@MinedTangerine
As I read the code on their github, the new LBK3 hash the VTL devs have created is a simple chain of Bmw, Blake, Keccak. My assumption was that it would contain some Lyra2 aspect as well, this from the leading "L" in the algo name. I can't find any indications of that being the case though. In this form, it should be trivial for anyone who knows the sgminer codebase to create a forked version supporting LBK3. The hashes involved are also difficult to optimize above the best publicly available OpenCL versions.
I could be missing something though, just did a very quick code browse of their github? Otherwise, the bottom line is that there's currently no point spending time on this algo for us as part of this miner, we can't add enough value to justify a dev fee, and it will be completely slaughtered by fpgas if/when it becomes interesting enough. The upcoming Acorns from GPUHoarder will most probably be capable of pipelining the full algo on its Artix-7 chip.
Out of curiousity, I've reached out to neithender on the VTL discord for final clarification though.
As I read the code on their github, the new LBK3 hash the VTL devs have created is a simple chain of Bmw, Blake, Keccak. My assumption was that it would contain some Lyra2 aspect as well, this from the leading "L" in the algo name. I can't find any indications of that being the case though. In this form, it should be trivial for anyone who knows the sgminer codebase to create a forked version supporting LBK3. The hashes involved are also difficult to optimize above the best publicly available OpenCL versions.
I could be missing something though, just did a very quick code browse of their github? Otherwise, the bottom line is that there's currently no point spending time on this algo for us as part of this miner, we can't add enough value to justify a dev fee, and it will be completely slaughtered by fpgas if/when it becomes interesting enough. The upcoming Acorns from GPUHoarder will most probably be capable of pipelining the full algo on its Artix-7 chip.
Out of curiousity, I've reached out to neithender on the VTL discord for final clarification though.
Jump to: