Hi JCE-Miner!
I'm following your project from very begging because of your excellent CPU implementation and i can say only great work!
I also have few mining rigs but only one is using windows to power my 4 Vega 56 and it will be just great to see Linux version of miner implemented to SimpleOS.
For the fees don't fell bad abut them I`m more than happy to pay a % for software that i actually use I even have proposal to change fee behavior when using GPU miner in way that all CPU work will go as dev fee... it will not be much probably as my CPU on Vega rig is doing like ~40Hs on new XMR algo and fee % from one GPU is like 10 times more but still i want simple config for my GPU miner that will allocate my CPU to your account so give me configuration with proper wallet and i will switch this today when i go to update miner. Same idea for LINUX version of GPU for default all CPU hash power should be by default a donation it is not much for us miners but it can be a motivation for you for further development.
Best Regards
Topic: [XMR] JCE Miner Cryptonight/forks, now with GPU! - page 39. (Read 90814 times)
To stabilize speed on 580 8G try to set beta to 16.
Testing JCE 33b5 on RX580 8GB cards (x8) with heavy using win 10 x64 with Adrenalin 18.6.1.
Similar findings as above, big issue with warm-up, only a few cards will hash at full speed, others just don't want to wake up
The speed seems better by about 5-10%, although the miner behaves the same as if you would set too much intensity, when you only get a few cards hashing at full speed.
Similar findings as above, big issue with warm-up, only a few cards will hash at full speed, others just don't want to wake up
The speed seems better by about 5-10%, although the miner behaves the same as if you would set too much intensity, when you only get a few cards hashing at full speed.
oh i tried without --archi but it´s the same error, i don´t know what´s happening the miner is using multi hash but only x2 and even with x2 there´s 16mb l3 cache left for the 20 threads and 8mb for the 24 threads i want to try x3 but the -c parameter always gives this error.
Please edit the start.sh and check there's no more --auto Inside. If yes, remove it.
JCE, you should edit the topic name to contain miner version, and also big news like this, to attract more people
Good idea, i'll put some hashrate numbers, like Cast and TeamRed didJCE, great work! see, you can do anything, just need some motivation! 
Thanks, you treator who switched to TR day one! 
Just a joke, and i lacked conviction rather than motivation. Conviction i wasn't already at hardware max, and yes, that's what TR provided. Again, the simultaneous releases are not coincidential, and this time i release first!
I need to update some doc, and notably to state for the new CN-Heavy speed and fees. The old doc still says that's 0.9%
I just want to give some feedback with new version 0.33b5 :
1. Performance definitely awesome.
2. Fee 2.1% I don't really mind about that.
3. RX 580 8 GB Most of card didn't reach peak hash rate and very difficult to reach peak hash rate.
4. RX Vega instantly reach peak hash rate without warm up.
So over all RX Vega will use 0.33b5 and RX 580 8 GB still use 0.33b4 because in 0.33b4 most of card reach peak hash rate easily.
hope you can improve warm up on 0.33b5 like in 0.33b4.
1. Performance definitely awesome.
2. Fee 2.1% I don't really mind about that.
3. RX 580 8 GB Most of card didn't reach peak hash rate and very difficult to reach peak hash rate.
4. RX Vega instantly reach peak hash rate without warm up.
So over all RX Vega will use 0.33b5 and RX 580 8 GB still use 0.33b4 because in 0.33b4 most of card reach peak hash rate easily.
hope you can improve warm up on 0.33b5 like in 0.33b4.
Similar behavior.
Not all cards reach max hashrate.
Tests on polaris at 1250 core, 2000-2000 memory
Tests on Vegas 1410 core, 950-1100 memory
I did get a boost from
8gb polaris cards 1150-1180 to 1240-1260 H/s (~7%)
4gb polaris cards got a boost from 800 to 840 H/s (~5-6%)
Reference Vega 56 modded got 1750 up from 1500 H/s (~17%)
Vega 56 Pulse (Hynix) 1600 up from ~1400 before H/s (~18.5%)
@JCE-Miner great work on cn-heavy.
Anyone?
How much you get with 280x?
How much you get with 280x?
I have 3 R9 280x running 1180/1500 +20 powertune and get 390h/s on two and 440h/s on the other on CN Heavy. JCE 33b5 didn't increase by much - maybe 5%. But wow my RX cards are nearly 30% faster than 33b. And JCE was already the fasted CN miner by far. Some are pretty close in miner-reported hashrate, but at the pool JCE is 10-20% better than all others. I'm using BETA blockchain drivers. Still the fasted for me.
I just want to give some feedback with new version 0.33b5 :
1. Performance definitely awesome.
2. Fee 2.1% I don't really mind about that.
3. RX 580 8 GB Most of card didn't reach peak hash rate and very difficult to reach peak hash rate.
4. RX Vega instantly reach peak hash rate without warm up.
So over all RX Vega will use 0.33b5 and RX 580 8 GB still use 0.33b4 because in 0.33b4 most of card reach peak hash rate easily.
hope you can improve warm up on 0.33b5 like in 0.33b4.
1. Performance definitely awesome.
2. Fee 2.1% I don't really mind about that.
3. RX 580 8 GB Most of card didn't reach peak hash rate and very difficult to reach peak hash rate.
4. RX Vega instantly reach peak hash rate without warm up.
So over all RX Vega will use 0.33b5 and RX 580 8 GB still use 0.33b4 because in 0.33b4 most of card reach peak hash rate easily.
hope you can improve warm up on 0.33b5 like in 0.33b4.
Similar behavior.
Not all cards reach max hashrate.
Tests on polaris at 1250 core, 2000-2000 memory
Tests on Vegas 1410 core, 950-1100 memory
I did get a boost from
8gb polaris cards 1150-1180 to 1240-1260 H/s (~7%)
4gb polaris cards got a boost from 800 to 840 H/s (~5-6%)
Reference Vega 56 modded got 1750 up from 1500 H/s (~17%)
Vega 56 Pulse (Hynix) 1600 up from ~1400 before H/s (~18.5%)
@JCE-Miner great work on cn-heavy.
Breaking change
When I started JCE-GPU, my competitors where SRB and Claymore. The JCE fees were aligned with SRB (0.9%) on purpose.
Now the game has changed, Claymore is obsolete and the lead in performances is TeamRed. I'm getting on-par with him on v8, but i still lacked a decent implementation for Heavy-class to completely outperform SRB, leave it behind and have only one real competitor. I'm like Stalin, i avoid having two opponents at the same time.
Dedicated to UnclWish who asked for it for a long time, here comes...
JCE GPU 0.33b5 with +20% performance on Heavy/Haven/Tube
Tested on both HD7000 and RX cards, with perfs in the range +15% to +25%
My HD7950 jumped from 335 to 415, my RX560 from 340 to 400
The bad news is that the fees are now aligned with TeamRed, at 2.1% (still a bit below their 2.5%) but only on Heavy-class algos, where the extra perf deserve them. On CN and CN-light, that's still 0.9%. If mixed with CPU, it's a weighted average, as before.
0.33b5 GPU
When I started JCE-GPU, my competitors where SRB and Claymore. The JCE fees were aligned with SRB (0.9%) on purpose.
Now the game has changed, Claymore is obsolete and the lead in performances is TeamRed. I'm getting on-par with him on v8, but i still lacked a decent implementation for Heavy-class to completely outperform SRB, leave it behind and have only one real competitor. I'm like Stalin, i avoid having two opponents at the same time.
Dedicated to UnclWish who asked for it for a long time, here comes...
JCE GPU 0.33b5 with +20% performance on Heavy/Haven/Tube
Tested on both HD7000 and RX cards, with perfs in the range +15% to +25%
My HD7950 jumped from 335 to 415, my RX560 from 340 to 400
The bad news is that the fees are now aligned with TeamRed, at 2.1% (still a bit below their 2.5%) but only on Heavy-class algos, where the extra perf deserve them. On CN and CN-light, that's still 0.9%. If mixed with CPU, it's a weighted average, as before.
0.33b5 GPU
Quote
Huge performance boost for Heavy/HVX/Tube
Respective fees increased to 2.1%
Very light perf boost for CN-v8
Respective fees increased to 2.1%
Very light perf boost for CN-v8
I just want to give some feedback with new version 0.33b5 :
1. Performance definitely awesome.
2. Fee 2.1% I don't really mind about that.
3. RX 580 8 GB Most of card didn't reach peak hash rate and very difficult to reach peak hash rate.
4. RX Vega instantly reach peak hash rate without warm up.
So over all RX Vega will use 0.33b5 and RX 580 8 GB still use 0.33b4 because in 0.33b4 most of card reach peak hash rate easily.
hope you can improve warm up on 0.33b5 like in 0.33b4.
Please give the full command line you used so I can understand. Just replace your wallet (parameter -u) by XXX to stay anonymous
./start.sh --archi skylake -o XXXXXXX -u XXXXXXXXXXX -p X -c config.txt
Code:
"cpu_threads_conf" :
[
{ "cpu_architecture" : "auto", "affine_to_cpu" : 0, "use_cache" : true, "multi_hash":2 },
{ "cpu_architecture" : "auto", "affine_to_cpu" : 1, "use_cache" : true, "multi_hash":2 },
....
{ "cpu_architecture" : "auto", "affine_to_cpu" :23, "use_cache" : true, "multi_hash":2 },
]
See the "auto" tokens in my example? That's where you can put "skylake" instead.[
{ "cpu_architecture" : "auto", "affine_to_cpu" : 0, "use_cache" : true, "multi_hash":2 },
{ "cpu_architecture" : "auto", "affine_to_cpu" : 1, "use_cache" : true, "multi_hash":2 },
....
{ "cpu_architecture" : "auto", "affine_to_cpu" :23, "use_cache" : true, "multi_hash":2 },
]
But the performance change will be null or negligible on such a fast CPU, the manual architecture has an impact on slow CPUs like Core2 or Athlon. Better keep "auto".
Lermite kindly provided optimal config for RX570, it may be good on RX580 too.
Using 100% memory is good for CN-Heavy but not always for CN-v8
RX570 4G
Code:
"gpu_threads_conf":
[
{ "mode": "GPU", "worksize": 8, "alpha": 64, "beta": 8, "index": 0, "multi_hash": 976 },
{ "mode": "GPU", "worksize": 8, "alpha": 64, "beta": 8, "index": 0, "multi_hash": 976 },
]
[
{ "mode": "GPU", "worksize": 8, "alpha": 64, "beta": 8, "index": 0, "multi_hash": 976 },
{ "mode": "GPU", "worksize": 8, "alpha": 64, "beta": 8, "index": 0, "multi_hash": 976 },
]
RX570 8G
Code:
"gpu_threads_conf":
[
{ "mode": "GPU", "worksize": 8, "alpha": 64, "beta": 8, "index": 1, "multi_hash": 1008 },
{ "mode": "GPU", "worksize": 8, "alpha": 64, "beta": 8, "index": 1, "multi_hash": 1008 },
]
[
{ "mode": "GPU", "worksize": 8, "alpha": 64, "beta": 8, "index": 1, "multi_hash": 1008 },
{ "mode": "GPU", "worksize": 8, "alpha": 64, "beta": 8, "index": 1, "multi_hash": 1008 },
]
JCE, you should edit the topic name to contain miner version, and also big news like this, to attract more people
I read the last few pages and couldn't find any info but anyone here manage to run the Tahiti or the Hawaii GPUs with this miner?
The Tahiti being the 7950 7970 280X and Hawaii being 290 390 295
So far the Pitcairn speeds are impressive basically what I get with my Tahiti with XMR Stak.
The Tahiti being the 7950 7970 280X and Hawaii being 290 390 295
So far the Pitcairn speeds are impressive basically what I get with my Tahiti with XMR Stak.
Anyone?
How much you get with 280x?
Tested with my home pc, easily pulled 1800h/s with 1400 core /1100 mem vega 64 with following:
{ "mode" : "GPU", "worksize" : 8, "alpha" : 128, "beta" : 8, "gamma" : 8, "delta" : 4, "epsilon" : 4, "zeta" : 4, "index" : 0, "multi_hash":960 },
{ "mode" : "GPU", "worksize" : 8, "alpha" : 128, "beta" : 8, "gamma" : 8, "delta" : 4, "epsilon" : 4, "zeta" : 4, "index" : 0, "multi_hash":960 },
The consumption of the card seemed far less than cn v8. Probably 2000 can be reached with proper tuning of the card.
however i can not test any further, as i just managed to have my rigs running smoothly and tomorrow I am embarking the ship for next 30 days. This ground breaking update comes 1 or 2 days too late for me, I could have switched the rigs to heavy algo and save some power. Now it is too risky for me for testing - any trouble and i can not repair my rigs for 30 days. Well, next time!
JCE, great work! see, you can do anything, just need some motivation!
{ "mode" : "GPU", "worksize" : 8, "alpha" : 128, "beta" : 8, "gamma" : 8, "delta" : 4, "epsilon" : 4, "zeta" : 4, "index" : 0, "multi_hash":960 },
{ "mode" : "GPU", "worksize" : 8, "alpha" : 128, "beta" : 8, "gamma" : 8, "delta" : 4, "epsilon" : 4, "zeta" : 4, "index" : 0, "multi_hash":960 },
The consumption of the card seemed far less than cn v8. Probably 2000 can be reached with proper tuning of the card.
however i can not test any further, as i just managed to have my rigs running smoothly and tomorrow I am embarking the ship for next 30 days. This ground breaking update comes 1 or 2 days too late for me, I could have switched the rigs to heavy algo and save some power. Now it is too risky for me for testing - any trouble and i can not repair my rigs for 30 days. Well, next time!
JCE, great work! see, you can do anything, just need some motivation!
Breaking change
When I started JCE-GPU, my competitors where SRB and Claymore. The JCE fees were aligned with SRB (0.9%) on purpose.
Now the game has changed, Claymore is obsolete and the lead in performances is TeamRed. I'm getting on-par with him on v8, but i still lacked a decent implementation for Heavy-class to completely outperform SRB, leave it behind and have only one real competitor. I'm like Stalin, i avoid having two opponents at the same time.
Dedicated to UnclWish who asked for it for a long time, here comes...
JCE GPU 0.33b5 with +20% performance on Heavy/Haven/Tube
Tested on both HD7000 and RX cards, with perfs in the range +15% to +25%
My HD7950 jumped from 335 to 415, my RX560 from 340 to 400
The bad news is that the fees are now aligned with TeamRed, at 2.1% (still a bit below their 2.5%) but only on Heavy-class algos, where the extra perf deserve them. On CN and CN-light, that's still 0.9%. If mixed with CPU, it's a weighted average, as before.
0.33b5 GPU
Ok, thanx! Let's try new version... Settings is the same as in prev versions, I hope... When I started JCE-GPU, my competitors where SRB and Claymore. The JCE fees were aligned with SRB (0.9%) on purpose.
Now the game has changed, Claymore is obsolete and the lead in performances is TeamRed. I'm getting on-par with him on v8, but i still lacked a decent implementation for Heavy-class to completely outperform SRB, leave it behind and have only one real competitor. I'm like Stalin, i avoid having two opponents at the same time.
Dedicated to UnclWish who asked for it for a long time, here comes...
JCE GPU 0.33b5 with +20% performance on Heavy/Haven/Tube
Tested on both HD7000 and RX cards, with perfs in the range +15% to +25%
My HD7950 jumped from 335 to 415, my RX560 from 340 to 400
The bad news is that the fees are now aligned with TeamRed, at 2.1% (still a bit below their 2.5%) but only on Heavy-class algos, where the extra perf deserve them. On CN and CN-light, that's still 0.9%. If mixed with CPU, it's a weighted average, as before.
0.33b5 GPU
Quote
Huge performance boost for Heavy/HVX/Tube
Respective fees increased to 2.1%
Very light perf boost for CN-v8
Respective fees increased to 2.1%
Very light perf boost for CN-v8
https://github.com/jceminer/cn_gpu_miner/blob/master/README.md#troubleshooting
last question
You have an Intel IPG/GPU in your computer, don't you?
last question
You have an Intel IPG/GPU in your computer, don't you?
yea i was reading notes. Figured that out. was editing my own post
.
https://github.com/jceminer/cn_gpu_miner/blob/master/README.md#troubleshooting
last question
You have an Intel IPG/GPU in your computer, don't you?
last question
You have an Intel IPG/GPU in your computer, don't you?
Quote
Do you have any miner for Linux at the moment ?? I have a few 7970 cards but only have ubuntu 14-16 as OS.
For CPU yes, but not GPU, and this is very unlikely regarding how little time i can allow to Linux versions. Even the CPU versions are less often updated than the Windows ones. 0.33b5 GPU
Huge performance boost for Heavy/HVX/Tube
Respective fees increased to 2.1%
Very light perf boost for CN-v8
Huge performance boost for Heavy/HVX/Tube
Respective fees increased to 2.1%
Very light perf boost for CN-v8
Still get the same abort
Abort was called at 348 line in file:
d:\qb\workspace\19992\src\vpg-compute-neo\runtime\os_interface\windows\wddm\wddm.cpp
EDIT: Nvm just read that iGPU are detected too now. Had to change gpu indexes in config file.
Breaking change
When I started JCE-GPU, my competitors where SRB and Claymore. The JCE fees were aligned with SRB (0.9%) on purpose.
Now the game has changed, Claymore is obsolete and the lead in performances is TeamRed. I'm getting on-par with him on v8, but i still lacked a decent implementation for Heavy-class to completely outperform SRB, leave it behind and have only one real competitor. I'm like Stalin, i avoid having two opponents at the same time.
Dedicated to UnclWish who asked for it for a long time, here comes...
JCE GPU 0.33b5 with +20% performance on Heavy/Haven/Tube
Tested on both HD7000 and RX cards, with perfs in the range +15% to +25%
My HD7950 jumped from 335 to 415, my RX560 from 340 to 400
The bad news is that the fees are now aligned with TeamRed, at 2.1% (still a bit below their 2.5%) but only on Heavy-class algos, where the extra perf deserve them. On CN and CN-light, that's still 0.9%. If mixed with CPU, it's a weighted average, as before.
0.33b5 GPU
When I started JCE-GPU, my competitors where SRB and Claymore. The JCE fees were aligned with SRB (0.9%) on purpose.
Now the game has changed, Claymore is obsolete and the lead in performances is TeamRed. I'm getting on-par with him on v8, but i still lacked a decent implementation for Heavy-class to completely outperform SRB, leave it behind and have only one real competitor. I'm like Stalin, i avoid having two opponents at the same time.
Dedicated to UnclWish who asked for it for a long time, here comes...
JCE GPU 0.33b5 with +20% performance on Heavy/Haven/Tube
Tested on both HD7000 and RX cards, with perfs in the range +15% to +25%
My HD7950 jumped from 335 to 415, my RX560 from 340 to 400
The bad news is that the fees are now aligned with TeamRed, at 2.1% (still a bit below their 2.5%) but only on Heavy-class algos, where the extra perf deserve them. On CN and CN-light, that's still 0.9%. If mixed with CPU, it's a weighted average, as before.
0.33b5 GPU
Quote
Huge performance boost for Heavy/HVX/Tube
Respective fees increased to 2.1%
Very light perf boost for CN-v8
Respective fees increased to 2.1%
Very light perf boost for CN-v8
Lermite kindly provided optimal config for RX570, it may be good on RX580 too.
Using 100% memory is good for CN-Heavy but not always for CN-v8
RX570 4G
RX570 8G
For my 2 RX 580 8G cards best multi_hash parameters for v8 are multiples of 96. Such as 1104, 1200 etc. Higher - speeder. Best is 1680, 1776 and 1872.Using 100% memory is good for CN-Heavy but not always for CN-v8
RX570 4G
Code:
"gpu_threads_conf":
[
{ "mode": "GPU", "worksize": 8, "alpha": 64, "beta": 8, "index": 0, "multi_hash": 976 },
{ "mode": "GPU", "worksize": 8, "alpha": 64, "beta": 8, "index": 0, "multi_hash": 976 },
]
[
{ "mode": "GPU", "worksize": 8, "alpha": 64, "beta": 8, "index": 0, "multi_hash": 976 },
{ "mode": "GPU", "worksize": 8, "alpha": 64, "beta": 8, "index": 0, "multi_hash": 976 },
]
RX570 8G
Code:
"gpu_threads_conf":
[
{ "mode": "GPU", "worksize": 8, "alpha": 64, "beta": 8, "index": 1, "multi_hash": 1008 },
{ "mode": "GPU", "worksize": 8, "alpha": 64, "beta": 8, "index": 1, "multi_hash": 1008 },
]
[
{ "mode": "GPU", "worksize": 8, "alpha": 64, "beta": 8, "index": 1, "multi_hash": 1008 },
{ "mode": "GPU", "worksize": 8, "alpha": 64, "beta": 8, "index": 1, "multi_hash": 1008 },
]
If I can ask, which value do you have set for gpu MHz, mem MHz and voltage?
Now I'm on 1680 for multihash (above this value after some time it crash...) and 1150MHz/2000MHz/850mv for GPU setting... since a couple of hours it works without crashing
I'm getting in the same card from 910 to 960MH/s... I do not understand, however, why the value move up and down so much
I read the last few pages and couldn't find any info but anyone here manage to run the Tahiti or the Hawaii GPUs with this miner?
The Tahiti being the 7950 7970 280X and Hawaii being 290 390 295
So far the Pitcairn speeds are impressive basically what I get with my Tahiti with XMR Stak.
The Tahiti being the 7950 7970 280X and Hawaii being 290 390 295
So far the Pitcairn speeds are impressive basically what I get with my Tahiti with XMR Stak.
Lermite kindly provided optimal config for RX570, it may be good on RX580 too.
Using 100% memory is good for CN-Heavy but not always for CN-v8
RX570 4G
RX570 8G
For my 2 RX 580 8G cards best multi_hash parameters for v8 are multiples of 96. Such as 1104, 1200 etc. Higher - speeder. Best is 1680, 1776 and 1872. Using 100% memory is good for CN-Heavy but not always for CN-v8
RX570 4G
Code:
"gpu_threads_conf":
[
{ "mode": "GPU", "worksize": 8, "alpha": 64, "beta": 8, "index": 0, "multi_hash": 976 },
{ "mode": "GPU", "worksize": 8, "alpha": 64, "beta": 8, "index": 0, "multi_hash": 976 },
]
[
{ "mode": "GPU", "worksize": 8, "alpha": 64, "beta": 8, "index": 0, "multi_hash": 976 },
{ "mode": "GPU", "worksize": 8, "alpha": 64, "beta": 8, "index": 0, "multi_hash": 976 },
]
RX570 8G
Code:
"gpu_threads_conf":
[
{ "mode": "GPU", "worksize": 8, "alpha": 64, "beta": 8, "index": 1, "multi_hash": 1008 },
{ "mode": "GPU", "worksize": 8, "alpha": 64, "beta": 8, "index": 1, "multi_hash": 1008 },
]
[
{ "mode": "GPU", "worksize": 8, "alpha": 64, "beta": 8, "index": 1, "multi_hash": 1008 },
{ "mode": "GPU", "worksize": 8, "alpha": 64, "beta": 8, "index": 1, "multi_hash": 1008 },
]
Jump to: