Mining configurations

RDashRDash Member Posts: 7
edited December 2015 in Mining
I wanted to share my mining config were I get the most out of my card it has been tested with various of cards. This mining config gives about 1.4 to 2.6 more MHash a second to my experience and it is tested on: R9 280X Toxic, Sapphire R9 280X Vapor-X and similar cards.

This config is tested on the following pools:
- Coinotron
- DwarfPool (My favorite)
- Nanopool (2nd favorite)

Personally i am using a Sapphire R9 280X Toxic screenshots are attached.



Be sure to update it with your own pool info!
setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_USE_SYNC_OBJECTS 1
setx GPU_MAX_HEAP_SIZE 100
ethminer.exe -F http://eth-eu.dwarfpool.com:80/0x7a8745f0cc9a4773e542782d15adc3a86ec5a237/Worker1 -G --cl-local-work 128 --cl-global-work 8192
Hopefully you guys can can help improving it.


Post edited by RDash on

Comments

  • ethfanethfan Member Posts: 458 ✭✭✭
    How much power is it using? Best measure is the Watts drawn at the wall by the card alone.
  • RDashRDash Member Posts: 7
    Personally I have not measured it, from my notice the card gets less heat. Next that I have undervolted it which barely affects the Hashrate.
  • oneshotkoneshotk Member Posts: 1
    I have the same Graphics card as RDash does, 2 of them in total. The haspower has no performance impact when doing several other things during mining. It's rock stable. As of the measurement, I use an heavily oc'ed i7 3960X CPU clocked at 4.8 GHZ. When my CPU (exceeds 200+ watt easily) is idle, I saw a measurement around 620 - 750 watt while both the R9 280X's are mining 50 Mhash/s total. I hope this is helpful.
  • estevaozabotestevaozabot Member Posts: 29
  • PansyfaustPansyfaust Member Posts: 29
    Nice, got a few extra hashes out of it.
  • zonozono Member Posts: 158 ✭✭
    what this actualy does "--cl-local-work 128 --cl-global-work 8192"?
    does it afect on hash speed improvement?

    thanks
  • EastwindEastwind Member Posts: 107
    The setting increases the speed about 5%.
  • RDashRDash Member Posts: 7
    @zono it increases it a bit, at some specific cards you can tweak the '--cl-local-work 128' to '256' but its not an amazing boost but it does help a bit ;)
  • chevrochevro Member Posts: 16
    What miner is that?
  • EastwindEastwind Member Posts: 107
    7970, on 0.94.1
  • anonymous95anonymous95 Member Posts: 248
    What will happend If I up to 20000 the --cl-global-work?
  • dlehenkydlehenky Member Posts: 2,249 ✭✭✭✭
    @anonymous95 Your OpenCL kernel will run longer, which diminishes the CPU overhead involved in running the kernels, to a point. With the default workgroup size of 64, your 20000 value for global work (a multiplier) will result in a OpenCL kernel that does 1,280,000 hashes before it completes. If you are getting 25 Mhashes/second (40 nanoseconds/hash), the kernel will run for ~52 ms. The default values give you a kernel that runs for ~10.8 ms. However, the kernel does not report a result until is has completed its entire run. So, if you were to find a result that would mine a block in the first, say, 5 ms of the kernel run, with a 52 ms kernel, you would be delayed by 47 ms before reporting that result. With the default settings, you would only be delayed 5.8 ms. A longer running kernel will also delay your starting to work on the next block, which may have come in during the current kernel run interval (someone else found a result for the block you're working on). There's a trade-off. Anyway, you can benchmark 'ethminer' with the '-M' option, using various values of '--cl-global-work' to see what values give you an improvement. Once you get to a certain value, depending on your OS and CPU performance, increasing '--cl-global-work' will have no discernible effect. You should run the benchmark for each value several times to get an accurate result to compare. Once you determine that value, there's no other benefit to increasing it. HTH.

    -Best Care
    David
  • dlehenkydlehenky Member Posts: 2,249 ✭✭✭✭
    @anonymous95 I mentioned this before on this forum, but '--cl-local-work' actually doesn't do what you would expect (if you knew what local work size really means to the OpenCL kernel that runs on the GPU). Suffice it to say that there's no difference between using '--cl-global-work 8192 --cl-local-work 128' and simply '--cl-global-work 16384'. The default value for '--cl-local-work' is 64, and that value is what's recommended by AMD. In addition, the actual OpenCL kernel code explicitly sets the local work size to 64, regardless of what you set it to on the command line. However, '--cl-local-work' does influence the value for global work size, which is what gets passed down to the GPU (it is the number of complete hashes the kernel will do in one run). Global work size is '--cl-local-work' (default 64) times '--cl-global-work' (default 4096). So, global work size will change if you change the default '--cl-local-work', but it is exactly the same as simply increasing the value of '--cl-global-work' by the same proportion.

    As far as your hash rate goes, pools are only giving you a very rough number. If you want to know what your GPU is actually capable of doing, to a close approximation of solo mining (no pool), run 'ethminer' in benchmark mode using '-M'. Run 'ethminer --help' to see all the 'ethminer' options.

    -Best Care
    David
  • ImAMiner?ImAMiner? Member Posts: 208 ✭✭
    edited February 2016
    Thanks @dlehenky that's interesting stuff. I actually use an older version of ethminer where, you can specify --gpu-workgroup-size xx --gpu-batch-size xx Any insight into how those work?

    I just want to point out that running ethminer in benchmark won't give you a real world approximation of how much ether you can mine because the DAG size is so much larger than the default size the benchmark runs at. @Genoil 's newest versions allow you to specify the DAG size in benchmark mode however.
  • dlehenkydlehenky Member Posts: 2,249 ✭✭✭✭
    @ImAMiner? First, the source code for the 'ethminer' version I have been running does not have those command line options, so I can't say what they do, but they look suspiciously like the newer '--cl-local-work' and '--cl-global-work', respectively. The OpenCL nomenclature is actually "local workgroup" and "global work group", the later being the sum total of the "work items" (in our case a work item is a complete hash), which could be termed the "batch size", I suppose. The other possibility for '--gpu-batch-size' is a "chunking size" for the DAG. There was an early effort in the original 'ethminer' to load the DAG down to the GPU in "chunks" and run the hash using the DAG in separate "chunks", but they never got it working and abandoned the code. Second, the benchmark gives very good results, unless you have a GPU that's affected by DAG size. Many GPUs are not affected, including mine, but I understand what you are saying. The benchmark uses the genesis DAG, which is 1 GB. Nvidia GPUs seem to be having more problems with the increased DAG size than AMD cards (although a few old ones do), which is why Genoil added that feature to the CUDA miner.

    -Best Care
    David
  • anonymous95anonymous95 Member Posts: 248
    I am getting 29mh/s last 6h

    With this settings = 29mhs/s
    setx GPU_MAX_ALLOC_PERCENT 100
    setx GPU_USE_SYNC_OBJECTS 1
    setx GPU_MAX_HEAP_SIZE 100
    --cl-local-work 256 --cl-global-work 15000

    But I dont understand 1 think.
    With default settings I get 26mh/s last 6h
    Only with ethminer.exe -F http://eth1.nanopool.org:8888/my poolxxx/miner1 -G I am getting 26mh/s

    If i Up the --cl-global-work 20000 I am getting more MH/s but I have more MS? So This mean with 29mh/s I am earning more slowly than 26mh/s?

    I hare Sapphire r9 390 8gb ddr5
  • suryawonosuryawono Member Posts: 63
    edited February 2016
    nice, my hashrate increase 10%

    @RDash anw, does --farm-recheck have big impact?
  • dlehenkydlehenky Member Posts: 2,249 ✭✭✭✭
    @suryawono '--farm-recheck' is the interval, in milliseconds, that the miner checks for new work. 500 is the default, but most pools recommend 200.

    -Best Care
    David
  • echo55echo55 Member Posts: 3
    i have no idea how u mining that high. i have same settings on 2x 280x gigabyte windforce and onlu 42 mh
  • chila09chila09 Member Posts: 1
    Hi, I don't find the version 15.8 of amd, someone has a link to download?
Sign In or Register to comment.