Miner- Optiminer/Zcash GPU miner v1.7!

Update:
Just tested Claymore V9.3 on Linux with 7 RX480’s and it behaves exactly the same as optiminer on both rigs. GPU2 reduces Sol/s over time and hangs, this is also the 7th PCIe slot on both MB. This points to OS, AMD Driver, or BIOS. I have tried Legacy and UEFI installs and fiddled with the BIOS settings for the PCI bus.

I am leaning toward Linux/BIOS as the issue since its so difficult to install the OS with all 7 GPU’s recognized.

I’m running with both GPU2’s disabled, since removing them to install in another rig would require a complete reinstall of the OS and reconfiguration on these two rigs.

I have not tried full Ubuntu desktop and have been working with the 16.04 Server version. Should not make a difference but I will try and post the results.

I am finalizing a new release with much improved hash rates! Stay tuned!

Version 1.2.0 released!

Change log:

  • Fix memory leek and potential race condition.
  • Improved speed, 5-15% higher hash rate, depending on the device.

This version brings kernels customized for specific devices. Currently, GCN1.1 and GCN1.2 devices are supported that use the fglrx driver. For some of those devices the new kernels are not used automatically. If you see the message “Use ‘experimental-kernel’ flag to unlock a more optimized version that is still experimental for this device.” it means the device should be supported but it has not been tested yet. To try the new kernel just add --experimental-kernel flag to the command line.

Thanks… with 4x 290x and 2x 290 on claymore 9.3 getting 2000s/s and 1730w
with optiminer 1680w and 1900s/s

What bios modifications do you use?

1120 1125
1150 1125

@Optiminer

I don’t get it. You mean that GCN 4th gen devices like RX 470 are not “natively” supported yet; maybe just your wording?
Just want to know if I should update RX 470 rigs or not.
GCN1.1 and 1.2 are R9 390 and R9 380 cards, as far as I remember.

Yes, RX470/480 devices do not get the benefits of device-specific optimizations yet. It’s still a bit faster than v1.1.0 on those as well though.

At last, faster than Claymore’s!

5% gain (2854 Sols, Llinux - 10X R9-290X unmoded, instead of 2725 -2.5% fees) = 7.5% faster total. The big gain is at cards with Elpida memory, which now are very near Hynix speed.

Healthy, Happy, and Prosperous New Year!

Happy New Year! My Nanos went from 275 s/s on v0.6 to 333 s/s on v1.2 (both net after devfee), with a 3% power drop as the cherry on top. Great work! v0.6 was rock solid for me. I was away for 11 days, and all 20 6-GPU rigs ran non-stop - not a single hiccup. I hope v1.2 matches that stability! I can’t get Claymore v9.3 for linux to run stable on my rigs, and the net hash rate was -20 s/s, compared to opti v1.2, but both devs are producing quality miners, for sure.

@Optiminer Having a problem with v1.2, fglrx 15.12 driver, 6x Nano rigs with optimized kernel: a GPU will go to 0% utilization/idle clock for no apparent reason. It doesn’t hang the rig, though, as I run one optiminer per card, and I can simply restart the miners, rather than reboot, to recover, but I’m losing hash rate until I detect it (manually, at the moment). I don’t know why the watchdog isn’t triggering.

Edit/Update: The miner process for one or more GPUs is simply quitting, without the watchdog triggering. Ubuntu 14.04.3, btw.

Do you already use the provided mining script? I recommend to run the miner in a loop to catch such cases where it crashes. Should be rare though.

Not too rare: I’ve had 5 GPU processes quit this morning (again, I run one optiminer process per GPU, which has proven time and again to be more stable). I’m already running with “-i 3” for stability reasons, even though I do see about a 2% better hash rate with the miner running with auto-detect intensity (5 in my case). I have a cron job now checking for dropped miner processes and restarting the miners (not the rig). If that proves to be a workable solution, I’m fine with it.

@dlehenky What is the CPU load on the machine? Is your network stable? Do all miners quit at the same time?

CPU load isn’t a problem. My 6-GPU rigs use about 50% of 1 thread (2core x 2thread i3 @3.6GHz). The network is very stable. I’m using SSL connections to flypool. No, all miners do not quit at the same time, just one here, one there. Keep in mind that I ran v0.6 for days on end with no issues whatsoever with 20 rigs. The problems I’m having now are specific to v1.2.

Thanks for the additional info. I do not have much idea what it could be right now but will keep it in my head.

Is there a reason why starting the miner in a loop from a shell script does not work for you that you use a cron job?

The cron job is cleaner, to me, than 123 (# of GPUs I’m running) shell script loops. Obviously, the real solution is for the miner not to quit, especially without some information about the reason, but I realize you have to deal with many configuration variables, both HW and SW, which makes it particularly difficult to test all the possibilities.

I second a potential problem on GCN 1.2 card, in my case ASUS Strix R9 380. It simply quits.
It crashes so frequently on Optiminer 1.1 and 1.2, fglrx-update, Ubuntu 15.10 that I had to switch it to eth.
It was working stably on 0.6.0, if it is of any help.
I initially thought that it was due to me running a non-canonical version (not 14.04 or 16.04), but reading @dlehenky made me think otherwise.
BTW, no problems with either R7 370 or RX 470 on Opti 1.2.

For those that get crashes under Windows, try starting it with --force-generic-kernel flag.