Running into trouble w/ SRBminer version 1.6.8 on my mining rig with 12 rx580 gpus (driver 8.15.1 or 8.16.1). Miner is now failing to start 1-3 gpus (gpus showing 0 hashes) after receiving jobs from pool. This was not a problem with SRB 1.6.7 on driver 18.3.4.
Miner works fine if I split into two jobs of 6 gpus, each. However, this messes with the algo switching. Anyone seeing anything similar or have a fix? Reducing intensity or reinstalling drivers doesn't work. Perhaps there's a timeout issue when attaching a larger number of GPUs under the latest drivers?
Log trace
[2018-10-20 15:00:54] json_receive: {"id":1,"jsonrpc":"2.0","error":null,"result":{"status":"OK"}}
[2018-10-20 15:00:54] miner_result: Pool accepted result 0x00000905
[2018-10-20 15:01:56] hashrate: GPU0: 0 H/s [BUS:1]
[2018-10-20 15:01:56] hashrate: GPU1: 828 H/s [BUS:2]
[2018-10-20 15:01:56] hashrate: GPU2: 834 H/s [BUS:3]
[2018-10-20 15:01:56] hashrate: GPU3: 830 H/s [BUS:4]
[2018-10-20 15:01:56] hashrate: GPU4: 830 H/s [BUS:5]
[2018-10-20 15:01:56] hashrate: GPU5: 829 H/s [BUS:6]
[2018-10-20 15:01:56] hashrate: GPU6: 828 H/s [BUS:9]
[2018-10-20 15:01:56] hashrate: GPU7: 1022 H/s [BUS:10]
[2018-10-20 15:01:56] hashrate: GPU8: 907 H/s [BUS:12]
[2018-10-20 15:01:56] hashrate: GPU9: 826 H/s [BUS:13]
[2018-10-20 15:01:56] hashrate: GPU10: 890 H/s [BUS:14]
[2018-10-20 15:01:56] hashrate: GPU11: 907 H/s [BUS:15]
[2018-10-20 15:01:56] hashrate: Total: 9531 H/s
[2018-10-20 15:02:11] watchdog: GPU0 [BUS: 1] hashing speed is 0 H/S
What do you mean by "if I split into two jobs of 6 gpus" ? Does it mean that it works when you split gpu's into two groups and run two instances of SRBMiner?
If so, we've even got a mixed rig of Vegas and RX580's to work with algo switching by having a different directory of configs for each group. It works with my rx580 rig too (which I created the test case). In our case I used my own proxy to handle the algo switching. Technique should work without the proxy but here's the link:
https://github.com/hesido/cryptonote-switching-relay-proxyWhat you do is, for each group of cards, have a separate config directory, algos text pointing to those different config directories, and a separate pool.txt. You can use a notepad++ to do file search and replace as necessary in the config directory for easier manipulation of the separate config texts.
This way, the rx580's hash at their max and vega's hash at their max, by hashing at the best algo which can be different.