Screenshot from ethpool for a share from the xbox miner
Unfortunately, discovered that this miner cant work full capcatiy without Xbox ID release since UWP games only get 1/4 of GPU. Miner should work very well AS IS with Xbox ID or any "Full Release" that can fully use GPU
~3 MH/s on Xbox Series S. Feasible to get 5 MH/s with minor changes to current code. With major changes to HLSL code, 15-20MHS should be possible, but needs a Shader code wizard. Currently Unknown MH/s On Series X / Xbox One
The code is also pretty bad and has a lot of old code in it since I made this for research purposes. Maybe someone could clean it up. But at the end of the day all that matters is the HLSL shader code if you want to make this feasible (increase the MH/s)
Ethereum Wallet: 0x53C58a76a9E702efC8298E7F29f322Cd2e59847E \ \ \ \ \ \ \
The 3 MH/s is closer to 1.5 MH/s for most pools due to bug #1 (bondary condition) in the bugs section. If your pool uses a boundary stricter than 00000000XXXXXXXXXXXXX then its fine
Some epochs don't work, ~ 1/5. If you try running and find that it has ~ .32 MH/s Try running the test code (read the testing section) which should have the full ~ 3 MH/s on Series S or ~ 50% your max hash rate if you run on your local GPU ( My Nvidia 2070 gets 17 MH/s for example). Then wait till the next epoch (or fix the bug if your a real G!). Someone would need to find out why some epochs don't work to make a viable miner... :( . My theory is some seeds are too large (ie the seed for epoch 439 is too large) and causes an out of memory error in the HLSL shader code. The error is definitely in the HLSL code, but it's just a PITA to debug HLSL.
HLSL based upon code from: https://github.com/gyf304/ethash-hlsl with some bug fixes
Win 10
Visual Studio 2019
VS C++ and WinRT packages ( should ask if you want to install these after git cloning)
Xbox Series S, Series X or Xbox One. (Might work with others? :shrug: )
if you have any problems make an issue and I'll try to help
First you will need a Microsoft Developer account to use dev mode on your Xbox. Unfortunately... this used to cost $10 and now costs $19. Really dumb IMO. Then you will need to put your Xbox in developer mode. This article covers the above 2 steps well. https://www.howtogeek.com/703443/how-to-put-your-xbox-series-x-or-s-into-developer-mode/
Then you will need to connect via WIFI or build the appx package and put on xbox. This doc covers it well https://docs.microsoft.com/en-us/windows/uwp/xbox-apps/development-environment-setup
Lastly on your xbox, make sure to set the miner to "Game" mode. "App" Mode has limited GPU access
NOTE: Someone could also make the wallet editable and export the miner to appx, then people dont need to build the miner to run on Xbox, just download the app and go to dev mode. Its actually really easy, I just didn't do it because the miner is non-feasible due to Bugs #1 and #2. Its almost feasible though!
Edit std::string wallet = "0x53C58a76a9E702efC8298E7F29f322Cd2e59847E";
~ line 9 of MainPage.cpp to your wallet
Change pool by commenting out iResult = getaddrinfo("us1.ethpool.org", "3333", &hints, &result);
~ line 112 in start_sock()
in MainPage.cpp and uncomment the pool you want to try or find their URL/ Port etc. Most of the pools commented out should work when uncommented but I dont remember, because some don't ( and some like nanopool have a very high boundary which is almost unusable at 3 MH/s.
1.The Bound for the miner is a hash that begins with 00000000XXXXXXXXXXXXX (8 preceding 0s example: 00000000b9de3b6e15924387fde5fa0a13a2fb4725a30ffbb597e4721da28f36). However, most pools have a less strict hash than this so you will miss out on blocks significantly hurting your return from the pool. The fix is to add a boundary checker to ETHashMine.hlsl ~ line 469. Currently it is
if (concat[0].x == 0) {
where concat[0] stores the 8 first digits, so only if there all zero do we report a solution.
this should be
if (concat[0].x < boundary)
but since concat is in little endian this is actually super non-trivial (at least to me. Its probably trivial to some bit master).
Some epochs are 50x slower... NO idea why. Probably some error in the HLSL code, which on error GPU code takes the maximum path to return which could be 50x slower. An example error epoch is epoch 435 (seed = 37f0818a24a483c5bd9c28e7b455358ccfe14a11e3504f5290946f9e3582775c ). On my 2070 RTX , I get 17 MH/s on epoch 434 and 436 and .32 MH/s on epoch 435, around 50x slower...
in MainPage::MainPage()
change miner = DXMiner(0);
to miner = DXMiner(YOUR_GPU_OR_CPU_NUM)
where the number is the index number from the dropdown from the device tab from the GUI when you run it. Sorry I hardcoded this for SPEEEDDDDD and since GPU is always index 0 so it should only cause trouble if you have more than 1
MainPage.cpp creates a DXMiner object from DXMiner.cpp. Then MainPage.cpp connects via socket to a pool. With the current epoch etc. info from the pool, DXMiner then creates DAG in ETHASHGenerateDataset.hlsl. Then MainPage.cpp starts a new thread for the DXMiner to mine while connected to the pool. In the new thread, The DXMiner.mine() function is then called repeatedly and DXMiner.nonce is updated after each run of DXMiner.mine(). Each Call of DXMiner.mine calls ETHashMine.hlsl to see if there is solutions. If a solution is found, the MainPage.cpp sees the change in DXMiner.solutions and sends it to the pool.
Open Mainpage.cpp
and comment out or uncomment
miner.set_test_vars();
miner.mine_forever();
in MainPage::MainPage()
function to change the miner to run on test data so you can test speed, vs actually connect to a pool and mine.
These lines make the miner mine forever before even showing the GUI, so you cant test speed changes without even loading the GUI.
If you edit the HLSL shader code in ETHash.hlsi, ETHashGenerateDataset.hlsl and ETHashMine.hlsl you should be able to get speed improvements if your a better graphics programmer than I am. The real key is to debug what is going on in the xbox (GPU usage) which is locked to Xbox Creator program members. If you could see where the shader code is inefficient (ex. is it inefficient in UAV memory vs SRV memory vs thread size vs warp size) then update to meet max GPU performance