Topic: [XMR] JCE Miner Cryptonight/forks, now with GPU! - page 28. (Read 90841 times)

From b5 version I noticed that miner needs much more time than 5 minutes to reach max stable speed on heavy. It's about 40-50 minutes. It's make difficult to test...

EDIT: With b-x strange situation with speed - it's became max, than works with max speed some time, and than it start to fluctuate again as on warm-up and didn't stabilize again...

I see no difference with the hybrid version over standard version, neither better or worse
CPU usage is virtually nil, I think I dont understand configuration for the hybrid version.
Experimenting with one Vega 64 on 8 core xeon, no 1x risers.

I repeat again. Variation was set correctly. In miner window all shares was marked as stale and all was rejected by pool. It was ~1-2 hours. Pool statistics also shows performance drop (on pool website). After some time all come back to normal. Other rigs with srbminer were working without a problem. Nvidia equihash miner that was working at the same time on this mixed vega-nvidia rig also had to problem.

I think it was some devfee switching problem with your miner. My bad, I have no logs (( .

no, my miner doesn't produce only stale shares. and how do you know they are stale? if they're all rejected so you chose the wrong --variation
no pool would let you mine for 2h with only bad shares. if that's just about the message in jce "stale, may be rejected" and then it gets accepted, so it's a good share.

bx that's the standalone .exe in Github (gpu project)

speed and bad shares were expected, not a big deal, but i really expected a more stable hashrare... at least it was more stable on my own rigs.

Sorry, where is the B-X version?

--variation property was set correctly.

EDIT2: 2nd - Bad shares appears on Pitcairn chip too. Speed on RX 580 8Gb is so unstable as on b5 and b6.

Miner has wrong determined your coin. Set coin variation manully. For heavy algo add --variation 5 to command line.

lost almost 2 hours mining of my vega rig using b5 and my favorite heavy coin and pool. Only stale shares (((. other rigs were using srbminer 1.7.0 with the same connectivity settings and had no problems.

Ok, thanks! Let's try it now...

EDIT:
1st impressions - As you said - bad shares on RX appears, but not so much... Speed on Pitcairn is a bit faster. Speed on RX 580 8Gb looks like on b5... Speed stability is still watching...

@migo77 : here you gave the linux cpuid, what was detected by JCE linux?
In all case you have great control on CPU:

Simple: --t N to force N threads, and --archi to set architecture (e.g. ryzen, athlon64, core2 ...)
Advanced: use manual config (examples are provided in the .zip)

So far most of my users are Windows CPU miners, GPU is far less (i've more competitors there) and Linux almost negligible

For GPU mining, minimum is Windows 8.1, Win7 is ok for CPU.

Online is the B-X version (not packaged, just the .exe)

JCE GPU miner designed to work on Windows 8+. Windows 7 not supported.

not working on windows 7 x64


no problem with other miners with win 7, and no problem with JCE with win10

Hi JCE, nice work! I read here you are missing linux cpu miners. I'm one of them and have some questions to ask. Is this the right place? I'm mining mostly on older xeon DP servers and some exotic setups which are wrongly detected by your miner like this:

processor   : 11
vendor_id   : AuthenticAMD
cpu family   : 16
model      : 8
model name   : Six-Core AMD Opteron(tm) Processor 2435
stepping   : 0
cpu MHz      : 2600.000
cache size   : 512 KB
physical id   : 1
siblings   : 6
core id      : 5
cpu cores   : 6
apicid      : 21
initial apicid   : 13
fpu      : yes
fpu_exception   : yes
cpuid level   : 5
wp      : yes
flags      : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm 3dnowext 3dnow constant_tsc rep_good nopl nonstop_tsc extd_apicid eagerfpu pni monitor cx16 popcnt lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt hw_pstate vmmcall npt lbrv svm_lock nrip_save pausefilter
bugs      : tlb_mmatch fxsave_leak sysret_ss_attrs null_seg amd_e400
bogomips   : 5200.28
TLB size   : 1024 4K pages
clflush size   : 64
cache_alignment   : 64
address sizes   : 48 bits physical, 48 bits virtual
power management: ts ttp tm stc 100mhzsteps hwpstate


Thank you, kind regards Milan

Man, you are frying your cards - 1+ volts, fans 3000+...and all that for some more 20-30h !

After other some days of test for me JCE is faster than any other miner on old cards like 370 with 2 gb.
Can get stable on v8 515 hs at 1150 gpu 1500 core. Also on lite, fast, v7 algo is ok.

But surprising for me is on my rx 470 gpu, first step sapphire reference can get with 0.35b stable 840 hs on heavy with 1250 gpu / 1950 mem.
Some time is stable at more than 900 hs but after drop at ~840.

If jce solved the problem with hashrate stability was great.

470 RIG



370 RIG

One theory is higher worksize.  Perform compute operations on a larger set of data.
You can achieve lower power consumption on xmr-stak or jce/srb too with higher worksize but your hashrate gets affected.  Possibly, it's combination of GCN assembly optimizations + larger worksize.

competition is good

Had the stable heavy 1920h/s on V64 with the latest Vega driver. What is interesting that stable 1920 hashrate I could catch with several miner restarts when some first starts were unstable and lower. But when caught it was working stable till I manually stop it. 12H nonstop tested with 2 last gpu releases. But then I decided to test different drivers and now I cannot reproduce previous 1920. Only 1820 available. Could miner reserve some extra memory? Or maybe there's a way to forbid win10 using video memory for its needs.
I've noticed that teamred uses less mem with higher V8 hashes than SRB and JCE. Thats strange imo.

It's great news! I'm waiting b-x version to test. And it's great that you have thoughts how to decrease power consumption on v8. I believe in you! I like your miner and use it.