Previous page:
https://bitcointalk.org/index.php?topic=638915.msg20897855#msg20897855

Ethermine states that they have found and fixed the issue with DAG #128.

Really? My miner shows all data that it gets from the pool and I see that pool sends seed for epoch #128 sometimes.
It's easy to confirm: use different pool or different miner as I described above.

Again: DAG 128 epoch issue is related to ethermine.org pool only, it is not miner issue. It seems some of their servers use non-synced wallet. Ask pool support if they can fix this problem.
If you think that the reason is in my miner, try some other miner on this pool (to confirm that the issue is still present you may need to connect to the pool several times to catch bad server) or try different pool (to confirm that the issue is gone).
I'm contacting them right now.

I cannot install these drivers on Windows7 at all, now I have to reinstall Windows after several attempts. Try any different OpenCL miner to confirm that the issue is related to my miner, because the reason can be drivers.

I've checked some logs, the problem on pool side, pool sends different epochs for some reason. You can ask their support why it happens.

Yes, I had to use Ubuntu 16.04 for compiling v12.6 because GCC from Ubuntu 12.04 has some bugs and cannot compile my code correctly, I spent half a day until I found that the problem is not in my code but in old GCC.

The problem must be on pool side, but you can PM me the complete log file and I will check it to confirm.

v9.8 works with new driver. Voltage/clocks management fails in some cases because of bugs in drivers.
Right now I'm working on new version with x2 speed for LBC (for AMD cards).

These drivers must work better for Vega, though I was not able to install them for Vega FE. For Polaris I don't see any changes in speed for ZEC.


Unfortunately, I have no idea what you mean, I don't have enough experience in these Linux problems. I test my miners on ubuntu 16.04 and gpu-pro drivers, for old cards I use 15.12 drivers and Ubuntu 12.04, for NVidia I use Ubuntu 14.04 with CUDA8-compatible drivers, it works fine. Since you did not specify any details, may be you use some different Linux or drivers. My miners are for Windows mainly, I compile them for Linux because of a lot of users requests, I don't test all Linux builds and drivers because it can take forever.

Have you tried latest AB with latest AMD drivers for blockchain? It does not work for me (I don't see any changes in power consumption when I change voltage in AB), at least in Win10+RX580. So the problem is not in my miner only, it looks like another ADL issue:
https://bitcointalk.org/index.php?topic=1433925.msg20538901#msg20538901
I've contacted AMD and we will try to fix it.

I've checked Win10 + RX480, it seems it works, I see voltage changes in WattMan if I restart it. ADL does not report any errors. Though it does not mean that it works. Probabaly ADL in this beta version of drivers has some problems because I see a lot of reports about it.


I'm not going to add new coins right now, may be later.

Yes I will update this miner as soon as I finish new version of dual miner.

Ok I'll check it and include to next update if possible.

Already implemented since v9.6, check "History.txt" file for details.

I did not check it, but probably that guy just published closed beta drivers, so probably it's ok for the security.

Quick checking ccminer sources shows that GPU does not return the lowest hash value for nonces that were checked, so miner cannot show "best share" until at least one share is found. For solo mining it means that you won't see "best share" until you find a share that can solve the block, which may take a very long time.
Anyway, I will not argue, may be ccminer is better for solo mining, or for any kind of mining. However, I'm not going to change GPU kernels to get additional data about hashes that were checked because I still think it is useless work.

Ok I understand your point of view. Instead of correct calculations based on target diff and miner speed, you take best share for 24 hours and think that you can reach this target for next 24 hours. It's wrong in general, but I won't argue, your approach can work with some probability. I can add the info you need, but I will have to change all GPU kernels because they don't return share diff, so I won't do it just for you, sorry. Also I think ccminer will not show the best share until you find at least one share, if so, in solo mining it will not show the info you want too. I can make GPU compare diffs of all hashes to find the best and return best value even if this value is less than share target, but I will not do it because it's additional and useless work for GPU.

I was told that new version of drivers is almost ready, but it seems AMD changed plans and decided not to release special version of drivers just for miners. So we have to wait until new version of main drivers is released, DAG fix will be included there. I hope it will be soon...

You mine solo, so miner gets "share target = network diff", it means that every share that miner finds will solve a block. Now how it works: miner passes share target to GPU, if share found, GPU will return that share(s). GPU does not return the best share that was found, it would take additional time from GPU. Moreover, it is not necessary as you can calculate how often you will find shares in average. Press "s" and you will see total mining rate(speed) per second and share target. Divide "mining speed" / "Share Target" = "how many shares you will get in one second". For example, share target is 8000MH and mining speed is 80MH/s, it means that you will find 0.01 share per second, i.e. one share per 100 seconds, in average.
Checking best share you found in 5-15 minutes can be funny, but it will not help you in calculations.
PS. Mining solo depends on luck a lot, until you have at least 0.1% of network speed.