Topic: [XMR] JCE Miner Cryptonight/forks, now with GPU! - page 47. (Read 90841 times)
It's strange... But, why if I manual set cpu_architecture to "vishera" (I have FX8320) then miner use assambly "generic_aes", but if I set cpu_architecture to "auto", then miner use assambly "generic_aes_avx"? I think with avx it's better?
Hi,
I'm using JCE especially on older systems without AES-NI, because this miner is without competition here. Perhaps some may want to know, what's up with cn/2 alias Cryptonight v8 here. The extra instructions seem to be hard work for the oldies. An AMD Opteron 1381 dropped from 70 H/s to 50 H/s, two Opteron 2374 HE (working with 6 Threads) dropped from 120 H/s to 70 H/s
Is there a chance for further optimizations or is all done what can be done with non-AES cpus?
I'm using JCE especially on older systems without AES-NI, because this miner is without competition here. Perhaps some may want to know, what's up with cn/2 alias Cryptonight v8 here. The extra instructions seem to be hard work for the oldies. An AMD Opteron 1381 dropped from 70 H/s to 50 H/s, two Opteron 2374 HE (working with 6 Threads) dropped from 120 H/s to 70 H/s
Is there a chance for further optimizations or is all done what can be done with non-AES cpus?
Normal, i missed the deadline, so long. I still think I did the right thing by making the CPU version, JCE was already a CPU-only miner when v7 came out, turning into GPU-only would have been strange.
wait your GPU version
your CPU version is good
Normal, i missed the deadline, so long. I still think I did the right thing by making the CPU version, JCE was already a CPU-only miner when v7 came out, turning into GPU-only would have been strange.
JCE,
I switched to srbminer until the gpu version is ready. Sorry man!
I switched to srbminer until the gpu version is ready. Sorry man!
Monero has just switched to v8
My miner works fine on v7, and on v8 CPU (but v8 GPU is broken, I know, i couldn't have finished it in time).
I'll redo some CPU tests on real Monero pools now it has switched.
Tested right now by double-clicking the start.bat in the CPU package (which mine XMR v8 on the donation wallet)
So far, it just works.
note: since i cannot dev anymore on algos, being away of my rigs or any decent AMD gpu, i'm working on the algo autoswitch, Monero-Ocean style. It will be available on CPU version first.
My miner works fine on v7, and on v8 CPU (but v8 GPU is broken, I know, i couldn't have finished it in time).
I'll redo some CPU tests on real Monero pools now it has switched.
Quote
regardless on pool and algo (v7/v8)
can you give an example of pool and algo you experienced a fail on? so I can test my side, at least on CPUQuote
20:31:54 | CPU Thread 0 finds a Share, value 5000
20:31:54 | Accepted by the pool in 108 ms.
20:31:54 | Accepted by the pool in 108 ms.
Tested right now by double-clicking the start.bat in the CPU package (which mine XMR v8 on the donation wallet)
So far, it just works.
note: since i cannot dev anymore on algos, being away of my rigs or any decent AMD gpu, i'm working on the algo autoswitch, Monero-Ocean style. It will be available on CPU version first.
Only "rejects" and "Invalid job id" regardless on pool and algo (v7/v8)
Hoped it works...
(vishera+5*baffin)
updated: works on cpu! had to install 18.6.1
Hoped it works...
(vishera+5*baffin)
updated: works on cpu! had to install 18.6.1
hi jes'
i had faith when i see 33b
i had faith when i see 33b
Quote
hum version 0.33B en v8 ,on rig 5x570 2800h/s?
in v7 :4500h/s in v8 with xmrig 4100h/s
in v7 :4500h/s in v8 with xmrig 4100h/s
Same answer as before: v8 GPU in JCE is very unfinished, and either fails to start or mines very slow. I just ran out of time and now i've no access to any decent dev env to finish it. As crazy as it sounds, I cannot even run my own GPU miner, i'm writting right now from a decade old laptop, in the middle of nowhere. I put as many warnings as I could on the Github page, and cannot do anything more for now. I may provide non-openCL related fixes meanwhile, but the v8 will remain bad until november. I won the race against time for CPU, but lost for GPU
The only good thing is that the perf problem is consistent for all who could have make it run, and same for the Pitcairn hang at startup, so once i'm back to my rigs, fixing should be easy.
Quote
and my rig freezes.
If you're trying v8 GPU, just give up until I release a decent version in ~2 weeks. The current code is just not usable as-is. Sorry for such a long delay. You can still use the CPU part that runs fine.If that's on V7 or other, use a manual configuration and lower the multi_hash parameter. Lowering parameter worksize from 8 to 4 sometimes helps too.
Whats the recommended driver? im using 18.9.3 and my rig freezes.
i have mix gpu in my system, 3 x 580, 2 x 570, 1 x 470, 1 x r9 Fury and 2 x Nvidia 1070.
Using win 1709, any one can help me with freeze problem.
i have mix gpu in my system, 3 x 580, 2 x 570, 1 x 470, 1 x r9 Fury and 2 x Nvidia 1070.
Using win 1709, any one can help me with freeze problem.
You should try 18.6.1
Whats the recommended driver? im using 18.9.3 and my rig freezes.
i have mix gpu in my system, 3 x 580, 2 x 570, 1 x 470, 1 x r9 Fury and 2 x Nvidia 1070.
Using win 1709, any one can help me with freeze problem.
i have mix gpu in my system, 3 x 580, 2 x 570, 1 x 470, 1 x r9 Fury and 2 x Nvidia 1070.
Using win 1709, any one can help me with freeze problem.
hi jes'
hum version 0.33B en v8 ,on rig 5x570 2800h/s?
in v7 :4500h/s in v8 with xmrig 4100h/s
i use the same config.txt:
{ "mode" : "GPU", "worksize" : 4, "alpha" : 64, "beta" : 8, "gamma" : 4, "delta" : 4, "epsilon" : 4, "zeta":4, "index" : 0, "multi_hash":880 },
{ "mode" : "GPU", "worksize" : 4, "alpha" : 64, "beta" : 8, "gamma" : 4, "delta" : 4, "epsilon" : 4, "zeta":4, "index" : 0, "multi_hash":880 },
?
hum version 0.33B en v8 ,on rig 5x570 2800h/s?
in v7 :4500h/s in v8 with xmrig 4100h/s
i use the same config.txt:
{ "mode" : "GPU", "worksize" : 4, "alpha" : 64, "beta" : 8, "gamma" : 4, "delta" : 4, "epsilon" : 4, "zeta":4, "index" : 0, "multi_hash":880 },
{ "mode" : "GPU", "worksize" : 4, "alpha" : 64, "beta" : 8, "gamma" : 4, "delta" : 4, "epsilon" : 4, "zeta":4, "index" : 0, "multi_hash":880 },
?
@Uaciuganadu:
What GPU do you use?
Don't forget to add parameter --variation 15 to mine v8, in case you hadn't did it.
Pitcairn: other already reported the Pitcairn compile fails, I documented it on the Github page, but cannot fix yet, as I said, i'm away from my rigs until november. That fork is a very bad timing for me. I barery had time to test the OpenCL, and had just none to optimize.
Even when it works, the performance of the code, which is not mine but the naive reference one, is very very bad, as expected.
What GPU do you use?
Don't forget to add parameter --variation 15 to mine v8, in case you hadn't did it.
Pitcairn: other already reported the Pitcairn compile fails, I documented it on the Github page, but cannot fix yet, as I said, i'm away from my rigs until november. That fork is a very bad timing for me. I barery had time to test the OpenCL, and had just none to optimize.
Even when it works, the performance of the code, which is not mine but the naive reference one, is very very bad, as expected.
Same GPU`s as before 2Vega and one rx580, but the 33 version is broken for me on both v7 and v8.
Code:
13:49:44 | Pool changes Difficulty to 64000.
Abort was called at 318 line in file:
D:\qb\workspace\19992\src\vpg-compute-neo\runtime\os_interface\windows\wddm.cpp
Press any key to continue . . .
I will have to switch to something else till November or when you get the time to debug the code again.
@Uaciuganadu:
What GPU do you use?
Don't forget to add parameter --variation 15 to mine v8, in case you hadn't did it.
Pitcairn: other already reported the Pitcairn compile fails, I documented it on the Github page, but cannot fix yet, as I said, i'm away from my rigs until november. That fork is a very bad timing for me. I barery had time to test the OpenCL, and had just none to optimize.
Even when it works, the performance of the code, which is not mine but the naive reference one, is very very bad, as expected.
What GPU do you use?
Don't forget to add parameter --variation 15 to mine v8, in case you hadn't did it.
Pitcairn: other already reported the Pitcairn compile fails, I documented it on the Github page, but cannot fix yet, as I said, i'm away from my rigs until november. That fork is a very bad timing for me. I barery had time to test the OpenCL, and had just none to optimize.
Even when it works, the performance of the code, which is not mine but the naive reference one, is very very bad, as expected.
Same GPU`s as before 2Vega and one rx580, but the 33 version is broken for me on both v7 and v8.
Code:
13:49:44 | Pool changes Difficulty to 64000.
Abort was called at 318 line in file:
D:\qb\workspace\19992\src\vpg-compute-neo\runtime\os_interface\windows\wddm.cpp
Press any key to continue . . .
I will have to switch to something else till November or when you get the time to debug the code again.
Tested v8 on 64-bit CPU's. On AMD FX8320 and Core i3 6100... Noticed significant drop in speed compared with v7 with the same config... About 25-30% less...
obvious new monero algo uses more compute. rx 560 2gb will shine here, where 550 will struggle. optimum miners making use of compute cores utilization efficiently will result in more hash for less electricity.
Anyway, sounds like a big thing for FPGAs but also acorn (afaik acorn accelerators does computational work way faster? as in more power savings for v8)
Anyway, sounds like a big thing for FPGAs but also acorn (afaik acorn accelerators does computational work way faster? as in more power savings for v8)
I know it's not because of the pool 
I added a big warning in the Github doc to state the GPU v8 is just not usable as-is, it rarely works, and when it does, it's very very slow.
I'm wondering if it was a good idea to provide such an unfinished version. I had choice between being 2 weeks late or provide a broken code. Two bad choices.
I could have reversed the status by making the GPU version first, and disable the double-check, but the CPU version is what defined JCE as the fastest miner so I gave it the priority.
online is the
GPU 0.33b
that's a bunch of little fixes i planned before, and it fixes nothing about the bad v8
I added a big warning in the Github doc to state the GPU v8 is just not usable as-is, it rarely works, and when it does, it's very very slow.
I'm wondering if it was a good idea to provide such an unfinished version. I had choice between being 2 weeks late or provide a broken code. Two bad choices.
I could have reversed the status by making the GPU version first, and disable the double-check, but the CPU version is what defined JCE as the fastest miner so I gave it the priority.
online is the
GPU 0.33b
Quote
No more fees on paused CPUs
The GPU init on Windows 7 now displays a clear message instead of garbage symbols
Warmup restarts after a long pause on GPU (because of pause or disconnect)
The GPU init on Windows 7 now displays a clear message instead of garbage symbols
Warmup restarts after a long pause on GPU (because of pause or disconnect)
that's a bunch of little fixes i planned before, and it fixes nothing about the bad v8
@Uaciuganadu:
What GPU do you use?
Don't forget to add parameter --variation 15 to mine v8, in case you hadn't did it.
Pitcairn: other already reported the Pitcairn compile fails, I documented it on the Github page, but cannot fix yet, as I said, i'm away from my rigs until november. That fork is a very bad timing for me. I barery had time to test the OpenCL, and had just none to optimize.
Even when it works, the performance of the code, which is not mine but the naive reference one, is very very bad, as expected.
What GPU do you use?
Don't forget to add parameter --variation 15 to mine v8, in case you hadn't did it.
Pitcairn: other already reported the Pitcairn compile fails, I documented it on the Github page, but cannot fix yet, as I said, i'm away from my rigs until november. That fork is a very bad timing for me. I barery had time to test the OpenCL, and had just none to optimize.
Even when it works, the performance of the code, which is not mine but the naive reference one, is very very bad, as expected.
Get similar error. --variation 15 is set.
Tested on killallasics.moneroworld.com pool which is v8 ready.
Pool worked on xmr-stak.
@Uaciuganadu:
What GPU do you use?
Don't forget to add parameter --variation 15 to mine v8, in case you hadn't did it.
Pitcairn: other already reported the Pitcairn compile fails, I documented it on the Github page, but cannot fix yet, as I said, i'm away from my rigs until november. That fork is a very bad timing for me. I barery had time to test the OpenCL, and had just none to optimize.
Even when it works, the performance of the code, which is not mine but the naive reference one, is very very bad, as expected.
What GPU do you use?
Don't forget to add parameter --variation 15 to mine v8, in case you hadn't did it.
Pitcairn: other already reported the Pitcairn compile fails, I documented it on the Github page, but cannot fix yet, as I said, i'm away from my rigs until november. That fork is a very bad timing for me. I barery had time to test the OpenCL, and had just none to optimize.
Even when it works, the performance of the code, which is not mine but the naive reference one, is very very bad, as expected.
Version GPU 0.33a didn't compiles kernel for v8 for Pitcairn 270X 4Gb card. No errors, just hangs on building kernel for thread 1.
For other algos all wroks fine...
For other algos all wroks fine...
Jump to: