Quote from: vmozara on July 07, 2018, 02:03:27 PM

I am trying to set the JCE miner on my home PC, that is Vega 64 and Ryzen 1700

When I do the "auto" my CPU is hashing 600+ H/s, but when i remove auto and put -c config.txt to use the GPU as well, no matter what config i try, my CPU always ends up hashing 300-400 or even less. I am trying to mine CN7. Anybody have some good configuration for 8-core ryzen to copy paste into the config for CN7 mining?

Question 2. How to see why is the difference between blue hashrate and effective hashrate? Only few times so far I have seen stale rejected share or duplicate, but sometimes there is quite the difference between blue and effective hashrate and I would like to see if I have some hardware errors or something

Question 3: Why is it not recommended to run the log file?

Question 4: I am mining on moneroocean pool, how to know if the miner is mining properly if the coin is switched by the pool?

I started the rigs more than 40 hours ago and so far there was 0 issues with stability. Whenever I had srbminer uptime for more than few days, some cards would lose few H/s, up to hundred or two. Here, so far, perfect. I didn't even do any tuning, justa copy paste of config with 1936 intensity that fellow forumer put here, and 0 issues with best hash rate I have ever seen from a miner.

Is there any chance you get a Vega card for better development?

Wow, you're such a Vega-holic

1. You probably left the default example of CPU in the config.txt that mine little, as a fail safe example. Lermite here kindly gave the best known config for Ryzen 1700. Take a look at the Github doc, there are more example. The CPU part of JCE GPU is 100% the same than in JCE CPU. Hence why the binary is as big.

2. Mathematically, Effective should converge to average blue hashrate minus 2% (-0.9% fee, ~ -1% stale shares, the Stale loss is higher on Nicehash). But since the blue is an instant hashrate and Effective is the sum of found share divided by the uptime, at one precise point of time, they may be very different. They tend on the long term. To check for errors, enable the log and search for "Rejected".

3. No hard reason, just that is prevents your hard drive to go to Sleep mode. That's why i disable log by default, otherwise there's no impact at all on performances.

4. If you get the shares pool-side, so all is good. Otherwise when you select the wrong fork, there's a long message from JCE to give config advices. If the pool switch coin in the middle of a session, which may happen once per year, so JCE will be kicked. Because it generates, compiles, optimizes and obfuscates its OpenCL for a very precise mining session, hotswitch is absolutly impossible.
I don't plan to add this feature that would be useful once or twice per year. And even auto-switch coins like ETN may need manual reconfigure on automatic switcher like XMRig. No motivation to implement this.

5. My approach to make rigs is:
* Use one single 150A tweaked industrial 12V power block instead of normal PSUs. Pros: cheap, fanless, has control over voltage. Cons: no software power-on/off, either all or none rigs can be up at a time, needs some soldering.
* Use decade-old bases, I love low-voltage core2-quad + DDR2, Bulldozer +DDR3 is good too. DDR4 is so expensive Sad

* Use quad- or penta- RX550 rather than a vega for the same hashrate. Pro: cheaper, consumes less power. Cons: one full rig is equal to just one vega.

So at this stage i've no Vega to planned for buy, but if i get good feedbacks from my improved Heavy code, I may try to get one. Even a used loud blower one would do the job.

Quote from: samvicky on July 07, 2018, 05:24:33 PM

Sincerely your miner is best as it's the one gives me highest hashrate.pls add gpu temp , backup pools (just in case) and the ability to automatically reboot the rig in case of gpu failure or miner crash. Wink

Thanks bro.
All this is somehow planned but i've still a lot of other features to add.

version 0.31b GPU done, now testing. I rewrote my Heavy algo (and no, not with that imperdin-thing, again, my code is completely different) and got something like 0.4% on my HD7950... I lack the most popular cards (RX570, 580 and Vega) to optimize on, i'm relying on bling optims for now Sad

edit: i wanted to add support for ITA fork, a fork of Haven, and noticed they changed

Code:

idx0 = (~d) ^ q

into

Code:

idx0 = ~(d ^ q)

which... is exactly the same*. That ITA coin is in the hand of real experts, no reason to panic. Roll Eyes

* binary not and xor are both commutative and reversible. Like -a*b is -(a*b) or a*(-b), binary ~d^q is ~(d^q)