Topic: Claymore's Dual Ethereum AMD+NVIDIA GPU Miner v15.0 (Windows/Linux) - page 1194. (Read 6590565 times)

Greetings,

I am trying to resolve a system-hanging-requiring-reboot issue, the ETH/SIA mining runs fine for a minute or so, then does the GPU hangs in OpenCL call thing.  These are stock video cards with no overclocking, Claymore 5.1, Beta 2. Ubuntu 15.10.  It runs fine with ethminer, and has 8Gb.  I saw a previous answer to this type of problem to be add memory, but that doesn't seem to me to be the problem.  The shown hash rate isn't the problem, I have adjusted the SIA intensity up and down.  Thanks much in advance for any insight. 

GPU #0 recognized as Radeon 280X/380X
GPU #1 recognized as Radeon 280X/380X
GPU #2 recognized as Radeon 390
GPU #3 recognized as Radeon 290X
GPU #4 recognized as Radeon 290X



ETH - Total Speed: 41.449 Mh/s, Total Shares: 1, Rejected: 0, Time: 00:01
ETH: GPU0 5.401 Mh/s, GPU1 0.000 Mh/s, GPU2 13.438 Mh/s, GPU3 11.316 Mh/s, GPU4 11.294 Mh/s
SC - Total Speed: 2486.920 Mh/s, Total Shares: 4, Rejected: 0
SC: GPU0 324.062 Mh/s, GPU1 0.000 Mh/s, GPU2 806.296 Mh/s, GPU3 678.943 Mh/s, GPU4 677.619 Mh/s
WATCHDOG: GPU 1 hangs in OpenCL call, exit
WATCHDOG: GPU 1 hangs in OpenCL call, exit


����������������������������������������������������������������ͻ
� Claymore's Dual ETH + DCR/SC AMD GPU Miner v5.1 Beta �
����������������������������������������������������������������ͼ

ETH: 1 pool is specified
Main Ethereum pool is us-east1.nanopool.org:9999
SC: 1 pool is specified
Main Siacoin pool is sia-us-east1.nanopool.org:9980

Well, while writing this I've found the way. It's sad, but it works.

Assume you have 4 GPUs onboard.

1. Run CDM with -mode 1 option. Let it mine for at least 5 min to reach stable temperatures
2. Write down current temperature values with -mode 1 option. Let's say 87 88 92 90
3. Stop CDM and run it with -mode 1-123 option. As you can see, you run only one GPU in the dual mode. Wait 1-2 min
4. Compare current temperatures with ones, written on the step 2. One of them should be higher since the dual mode consumes more power and produces more heat. Let's say, you see 87 93 92 90. It means that GPU0 should be GPU1. Write down this match: 1XXX
5. Stop CDM and run it with -mode 1-023 option. Here you find GPU1. Imagine, that you see something like 87 88 92 95. Well, GPU1 should be GPU3. Write down this match: 13XX
6. Repeat steps 3-4 until you detect all GPUs. Let's say, the match string is 1302.
7. At the end just put the match string to -di option. In our case it should be -di 1302.
8. Enjoy

I'm adding "-di detect" option right now to avoid such weird actions to detect correct order.

I'm sorry, if it looks like this. I just want to use your fantastic miner as much as possible, paying you 35Mh of my hash power. To do it effectively, I have to understand these temperature questions. I need a working method how to obtain SPEEDS/TEMPS list matching.

Well, while writing this I've found the way. It's sad, but it works.

Assume you have 4 GPUs onboard.

1. Run CDM with -mode 1 option. Let it mine for at least 5 min to reach stable temperatures
2. Write down current temperature values with -mode 1 option. Let's say 87 88 92 90
3. Stop CDM and run it with -mode 1-123 option. As you can see, you run only one GPU in the dual mode. Wait 1-2 min
4. Compare current temperatures with ones, written on the step 2. One of them should be higher since the dual mode consumes more power and produces more heat. Let's say, you see 87 93 92 90. It means that GPU0 should be GPU1. Write down this match: 1XXX
5. Stop CDM and run it with -mode 1-023 option. Here you find GPU1. Imagine, that you see something like 87 88 92 95. Well, GPU1 should be GPU3. Write down this match: 13XX
6. Repeat steps 3-4 until you detect all GPUs. Let's say, the match string is 1302.
7. At the end just put the match string to -di option. In our case it should be -di 1302.
8. Enjoy

Now you ask me why AB does not match the miner. I'm not developer of AB. I have no idea how it enumerates cards.

Claymore 5.2 crashes for me anytime I send epool.txt or config.txt from ethman v1.5 Windows 10

With debug enabled last line is:


File gets downloaded just fine

So MSI AB uses the list of GPU different from the miner's temperature list! ((( How could it be? And... What to do then?

All other options work with miner's GPU list, i.e. after applying "-di" option.

Well, It's mostly clear now. The last (I hope) question. If I've changed the order of GPU with -di option, how other options  work? If I set "-mode 1-01", then these "01" are numbers before applying -di option or after that?

Even if you set "-di 1" miner will show "GPU0" anyway. But this GPU0 will be GPU1 (i.e. second GPU) in OpenCL list. With "-di" you don't change order of temps list, you change order in miner GPUs list that is taken from OpenCL list of GPUs. The list of temperatures is always the same.
Afterburner does not have to use OpenCL to mine, so it can match temps and GPUs. Miner has to use two lists: for temperatures control and for OpenCL work.
Again: "-di" change the list of GPUs in miner. For example:
1. You don't specify "-di". Miner directly takes entire list of GPUs from OpenCL list:
OpenCL          Miner
GPU0         GPU0
GPU1         GPU1
GPU2         GPU2
GPU3         GPU3

2. You specified "-di 1":
OpenCL          Miner
GPU1         GPU0

So miner displays "GPU0" because it is the first GPU in its list. But this is GPU1 in OpenCL list.

3. You specified "-di 3210":
OpenCL          Miner
GPU3         GPU0
GPU2         GPU1
GPU1         GPU2
GPU0         GPU3

So miner takes GPU3 from OpenCL list and adds it to its list of GPUs, so it will be GPU0 for miner. Then it takes GPU2 from OpenCL and saves it as "GPU1", and so on.
"-di ABCD" (A, B, C, D are some numbers) will cause GPU_A from OpenCL list becomes GPU0 in miner, GPU_B from OpenCL list becomes GPU1 in the miner, etc.
So you can renumber cards as you want to match temps list. The only possible problem if you don't want to assign all cards to miner, in this case sometimes you won't be able to renumber GPUs in a proper way. I think about adding another option "-map" for that, but it seems it will make things even more complex to understand.

Does it work with the RX 480?

what is the benefit if you use SIA Stratum protocol?

I'm looking at it, unfortunately there is no source, no documentation so I have to do some rev engineering again...

A pity that there is no consensus with those protocols.. maybe we should start a "pool consortium" and find a common way of writing/integrating stratums...

I would love to use this miner @ suprnova's pool.
It would be nice if you could make that happen ocminer, your pool would benefit highly with all the extra hashpower from the eth(etc)/sia miners using the perfect Claymore's miner, no doubt!
For now I have to use another pool, but I would like to go back to yours!