Gateless Gate Sharp 1.3.8: 30Mh/s (Ethash) on RX 480!

[Pete_X]

I know, I know... CryptoNight-Heavy is working now.
So much work for so little money, though. *sigh*

Congrats ! Speedy and splendid work!

Did some fast test runs ... after playing a bit with the intensity it looks just great ! ->

1114h/s Sapp 480 8GB Samsung
1007h/s Sapp 580 8GB Hynix

Probably there is more - just needs some time to tweak 

[1714934549]

1714934549

[1714934549]

1714934549

[1714934549]

1714934549

[1714934549]

1714934549

[zawawa]

Zawawa, you should add a field on memory timings page, where you put the entire memory timing (since i got some good timings) and the miner put every value on it respective box ^^ (kinda of a importing timings features)

on my saphire pulse 580 8gb, i got black screen on every algo with your timings

The default timings are for Samsung GDDR5. I will add more configurations.

[zawawa]

I know, I know... CryptoNight-Heavy is working now.
So much work for so little money, though. *sigh*

Congrats ! Speedy and splendid work!

Did some fast test runs ... after playing a bit with the intensity it looks just great ! ->

1114h/s Sapp 480 8GB Samsung
1007h/s Sapp 580 8GB Hynix

Probably there is more - just needs some time to tweak 

Looks like you figured out how to use the new GGS.

[zawawa]

Zawawa, need money ?
CN beside the fork isn't profitable, ethereum the same.
Please make a X16r miner. All the algos are under sgminer. its needed a good kernel, and you have the best one, and good setup  may be even memory timing on the flybetween the algos..
All my Nvidia are on x16r or x16s.
Most of the AMd miners are looking for alternatives but an gtx 1060 easily beat an RX 580 .


I forgot: please fix: SQL ( we don't have I7 on the rigs and  its too slow, at least give the option of NOT using it I prefer to rewrite everything if I have to, instead its always restarting  even if I selected not to do so) and memory paging some people doesn't use english and the paging isn't on C:

I will get to these issues. X16R sound like a good idea. RX 580 should outperform GTX 1060 by a large margin, at least. Let me see...

[zawawa]

So I just need to port this as a starter, then:

https://github.com/aceneun/sgminer-gm-x16r/blob/x16r/kernel/x16r.cl

[zawawa]

Hmm...

Code:

  { "x16r", ALGO_X16R, "", 1, 256, 256, 0, 0, 0xFF, 0xFFFFULL, 0x0000ffffUL, 32, 8 * 16 * 4194304, 0, x16r_regenhash, NULL, queue_x16r_kernel, gen_hash, append_x13_compiler_options, enqueue_x16r_kernels },

Code:

static cl_int queue_x16r_kernel(struct __clState *clState, struct _dev_blk_ctx *blk, __maybe_unused cl_uint threads)
{
  cl_kernel *kernel;
  unsigned int num;
  cl_ulong le_target;
  cl_int status = 0;
  uint8_t hashOrder[X16R_HASH_FUNC_COUNT];

  le_target = *(cl_ulong *)(blk->work->device_target + 24);
  flip80(clState->cldata, blk->work->data);
  x16r_getalgolist(&clState->cldata[4], hashOrder);

  status = clEnqueueWriteBuffer(clState->commandQueue, clState->CLbuffer0, true, 0, 80, clState->cldata, 0, NULL, NULL);
  if (status != CL_SUCCESS)
    return -1;

  for (int i = 0; i < X16R_HASH_FUNC_COUNT; i++) {
    kernel = &clState->extra_kernels[i];
    CL_SET_ARG_0(clState->padbuffer8);
  }

  kernel = &clState->extra_kernels[hashOrder[0] + X16R_HASH_FUNC_COUNT];
  num = 0;
  CL_SET_ARG(clState->CLbuffer0);
  CL_SET_ARG(clState->padbuffer8);

  kernel = &clState->kernel;
  num = 0;
  CL_SET_ARG(clState->padbuffer8);
  CL_SET_ARG(clState->outputBuffer);
  CL_SET_ARG(le_target);

  return status;
}

static cl_int enqueue_x16r_kernels(struct __clState *clState,
                                   size_t *p_global_work_offset, size_t *globalThreads, size_t *localThreads)
{
  cl_int status;
  uint8_t hashOrder[X16R_HASH_FUNC_COUNT];

  x16r_getalgolist(&clState->cldata[4], hashOrder);

  status = clEnqueueNDRangeKernel(clState->commandQueue,
      clState->extra_kernels[hashOrder[0] + X16R_HASH_FUNC_COUNT],
      1, p_global_work_offset,
      globalThreads, localThreads, 0, NULL, NULL);
  if (unlikely(status != CL_SUCCESS)) {
    applog(LOG_ERR, "Error %d: Enqueueing kernel onto command queue. (clEnqueueNDRangeKernel)", status);
    return status;
  }

  for (int i = 1; i < X16R_HASH_FUNC_COUNT; i++) {
    status = clEnqueueNDRangeKernel(clState->commandQueue,
        clState->extra_kernels[hashOrder[i]],
        1, p_global_work_offset,
        globalThreads, localThreads, 0, NULL, NULL);
    if (unlikely(status != CL_SUCCESS)) {
      applog(LOG_ERR, "Error %d: Enqueueing kernel onto command queue. (clEnqueueNDRangeKernel)", status);
      return status;
    }
  }

  status = clEnqueueNDRangeKernel(clState->commandQueue,
      clState->kernel,
      1, p_global_work_offset,
      globalThreads, localThreads, 0, NULL, NULL);
  if (unlikely(status != CL_SUCCESS)) {
    applog(LOG_ERR, "Error %d: Enqueueing kernel onto command queue. (clEnqueueNDRangeKernel)", status);
    return status;
  }

  return 0;
}

[murgorx]

I am getting a "Please enter at least one valid wallet address" when I try to start mining with 1.3.2 prealpha! Claymore's etn miner is working without a problem... what is causing this issue?

EDIT: This is what's in the Log section.

2018-04-07 20:52:26.0240 [1] Gateless Gate Sharp 1.3.2 prealpha started.
2018-04-07 20:52:27.0397 [1] Number of Devices: 5
2018-04-07 20:52:27.1491 [1] Successfully initialized AMD Display Library.
2018-04-07 20:52:27.1491 [1] Number of ADL Adapters: 80
2018-04-07 20:52:27.7742 [1] ADL_Adapter_AdapterInfo_Get() returned error code 0
2018-04-07 20:52:27.7742 [1] Failed to initialize NVIDIA Management Library.
2018-04-07 20:52:27.7898 [1] Exception in UpdateControls(): The given key was not present in the dictionary.   at System.ThrowHelper.ThrowKeyNotFoundException()
   at System.Collections.Generic.Dictionary`2.get_Item(TKey key)
   at GatelessGateSharp.MainForm.UpdateControls()
2018-04-07 20:52:37.3846 [1] Loading settings from C:\Users\Murgor Alexander\AppData\Roaming\GatelessGateSharp\GatelessGateSharp.sqlite...
2018-04-07 20:52:41.3382 [1] Loaded settings.
2018-04-07 20:52:41.3538 [1] Invalid argument: C:\Program Files\Zawawa Software LLC\Gateless Gate Sharp\GatelessGateSharp.exe
2018-04-07 20:52:47.3858 [13] Running garbage collection...
2018-04-07 20:52:47.3858 [13] Memory used before collection: 19MB
2018-04-07 20:52:47.4169 [13] Memory used before collection: 14MB
2018-04-07 20:52:49.3860 [15] API Listener started.

[zawawa]

I am getting a "Please enter at least one valid wallet address" when I try to start mining with 1.3.2 prealpha! Claymore's etn miner is working without a problem... what is causing this issue?

EDIT: This is what's in the Log section.

2018-04-07 20:52:26.0240 [1] Gateless Gate Sharp 1.3.2 prealpha started.
2018-04-07 20:52:27.0397 [1] Number of Devices: 5
2018-04-07 20:52:27.1491 [1] Successfully initialized AMD Display Library.
2018-04-07 20:52:27.1491 [1] Number of ADL Adapters: 80
2018-04-07 20:52:27.7742 [1] ADL_Adapter_AdapterInfo_Get() returned error code 0
2018-04-07 20:52:27.7742 [1] Failed to initialize NVIDIA Management Library.
2018-04-07 20:52:27.7898 [1] Exception in UpdateControls(): The given key was not present in the dictionary.   at System.ThrowHelper.ThrowKeyNotFoundException()
   at System.Collections.Generic.Dictionary`2.get_Item(TKey key)
   at GatelessGateSharp.MainForm.UpdateControls()
2018-04-07 20:52:37.3846 [1] Loading settings from C:\Users\Murgor Alexander\AppData\Roaming\GatelessGateSharp\GatelessGateSharp.sqlite...
2018-04-07 20:52:41.3382 [1] Loaded settings.
2018-04-07 20:52:41.3538 [1] Invalid argument: C:\Program Files\Zawawa Software LLC\Gateless Gate Sharp\GatelessGateSharp.exe
2018-04-07 20:52:47.3858 [13] Running garbage collection...
2018-04-07 20:52:47.3858 [13] Memory used before collection: 19MB
2018-04-07 20:52:47.4169 [13] Memory used before collection: 14MB
2018-04-07 20:52:49.3860 [15] API Listener started.

If you want to use custom pools, you need to choose "Custom Pools" on the dashboard.

[OhMyHolyToken]

I choose Crypto-heavy, sumo.Fairpool, and it doesn't works "Custom pool settings are invalide"...

I used those settings before without any problem.

[skablast]

That can be a start but..

In reality that  is the best available in term of building  but NOT in term of algos.
The kernel accept  --thread-concurrency 1792, -g2 that is the best setup for RX580 8gb samsung but the Algos compiled aren't the fastest available on github.
The one from Brianmct have faster algos but doesn't accept all the sgminer command


What the "dev" have done at least for cc miner/Cuda, is just to look around and finding faster Algos version on github  thus increasing the overall speed, for the moment.

Nice guys, expecially OCMINER but they don't call themselves DEV . give a look at his repository that I think have the fastest versions of algos also for sgminer. I don't know for sure because I haven't compiled his version yet and probably I will never do.

If you need testing, I am here. You will be in pole position if your average for rx580 is > 7.437 that is the best I achieved with an rx580

But just go to : Ravencalc.xyz and see what the average people hashrate per card is.

there are 7 coin at the moment using x16r or x16s, but increasing

[murgorx]

If you want to use custom pools, you need to choose "Custom Pools" on the dashboard.

OH! I did not see that there's an option "Custom Pools" in there, lol. Okay, since I changed that now the miner is running at about the same speed as Claymore's miner with the only difference, that the power withdrawal with your miner is 520-525 STABLE! when with Claymore's miner it jumps from 670-720 constantly
As I am looking at the miner, for the last 5 minutes the speed is 885-7 on all cards and it does not move with more than .2-3. Very, very stable, thank you!

The only problem with the miner, as with all other versions is the following: When I get from the dashboard to the other sections - custom pools/devices/log, the two graphs are always bugging and they just freeze They do not show any activity. With the latest version, I don't even get to see them! They are all white.

[theodrim]

1007h/s Sapp 580 8GB Hynix

Are your hynix is H5GQ8H24MJ ? If yes, what timings are you using for it (and intensity settings for that matter)? I've only managed to push ~840 h/s at 1250 core 2250 mem clocks.

[zawawa]

I choose Crypto-heavy, sumo.Fairpool, and it doesn't works "Custom pool settings are invalide"...

I used those settings before without any problem.

Could I take a look at the "Custom Pools" tab page? Did you enter a password?

[zawawa]

That can be a start but..

In reality that  is the best available in term of building  but NOT in term of algos.
The kernel accept  --thread-concurrency 1792, -g2 that is the best setup for RX580 8gb samsung but the Algos compiled aren't the fastest available on github.
The one from Brianmct have faster algos but doesn't accept all the sgminer command


What the "dev" have done at least for cc miner/Cuda, is just to look around and finding faster Algos version on github  thus increasing the overall speed, for the moment.

Nice guys, expecially OCMINER but they don't call themselves DEV . give a look at his repository that I think have the fastest versions of algos also for sgminer. I don't know for sure because I haven't compiled his version yet and probably I will never do.

If you need testing, I am here. You will be in pole position if your average for rx580 is > 7.437 that is the best I achieved with an rx580

But just go to : Ravencalc.xyz and see what the average people hashrate per card is.

there are 7 coin at the moment using x16r or x16s, but increasing

As long as I have a working version, further optimizations should be a smooth sail. We will see...

[zawawa]

If you want to use custom pools, you need to choose "Custom Pools" on the dashboard.

OH! I did not see that there's an option "Custom Pools" in there, lol. Okay, since I changed that now the miner is running at about the same speed as Claymore's miner with the only difference, that the power withdrawal with your miner is 520-525 STABLE! when with Claymore's miner it jumps from 670-720 constantly
As I am looking at the miner, for the last 5 minutes the speed is 885-7 on all cards and it does not move with more than .2-3. Very, very stable, thank you!

The only problem with the miner, as with all other versions is the following: When I get from the dashboard to the other sections - custom pools/devices/log, the two graphs are always bugging and they just freeze They do not show any activity. With the latest version, I don't even get to see them! They are all white.

Glad to hear the miner is stable. As for the graphs, are there any exceptions in the log at the startup?

[OhMyHolyToken]

I choose Crypto-heavy, sumo.Fairpool, and it doesn't works "Custom pool settings are invalide"...

I used those settings before without any problem.

Could I take a look at the "Custom Pools" tab page? Did you enter a password?

X in password resolve the problem.

But the hashrate of Crypto-heavy is twice lower, is that normal?

[murgorx]

Glad to hear the miner is stable. As for the graphs, are there any exceptions in the log at the startup?

This is what I got from the start up:

2018-04-07 21:38:51.0016 [1] Gateless Gate Sharp 1.3.2 prealpha started.
2018-04-07 21:38:51.8767 [1] Number of Devices: 5
2018-04-07 21:38:51.9704 [1] Successfully initialized AMD Display Library.
2018-04-07 21:38:51.9704 [1] Number of ADL Adapters: 80
2018-04-07 21:38:52.5486 [1] ADL_Adapter_AdapterInfo_Get() returned error code 0
2018-04-07 21:38:52.5486 [1] Failed to initialize NVIDIA Management Library.
2018-04-07 21:38:52.5642 [1] Exception in UpdateControls(): The given key was not present in the dictionary.   at System.ThrowHelper.ThrowKeyNotFoundException()
   at System.Collections.Generic.Dictionary`2.get_Item(TKey key)
   at GatelessGateSharp.MainForm.UpdateControls()
2018-04-07 21:39:02.5653 [1] Loading settings from C:\Users\Murgor Alexander\AppData\Roaming\GatelessGateSharp\GatelessGateSharp.sqlite...
2018-04-07 21:39:06.9095 [1] Loaded settings.
2018-04-07 21:39:06.9095 [1] Invalid argument: C:\Program Files\Zawawa Software LLC\Gateless Gate Sharp\GatelessGateSharp.exe
2018-04-07 21:39:10.7693 [1] Launching miner(s) for Custom Pool 0...
2018-04-07 21:39:11.0818 [14] Miner thread for Device #0 started.
2018-04-07 21:39:11.0818 [14] NiceHash mode is off.
2018-04-07 21:39:11.0975 [14] Loaded BinaryKernels\Ellesmere_cryptonight_8.bin for Device #0.
2018-04-07 21:39:11.1131 [14] Built cryptonight program for Device #0.
2018-04-07 21:39:11.1131 [14] Build options: -O5 -IKernels -DWORKSIZE=8
2018-04-07 21:39:11.1131 [14] Switching to Monero Version 1
2018-04-07 21:39:11.2537 [15] Miner thread for Device #0 started.
2018-04-07 21:39:11.2537 [15] NiceHash mode is off.
2018-04-07 21:39:11.2537 [15] Loaded BinaryKernels\Ellesmere_cryptonight_8.bin for Device #0.

Afterwards it just repeats the last 7-8 rows for all gpu's and the hash reports kicks in.

[cryptoinvestor_x]

If the miner crashes during optimization is there anyway to stop it on the next launch? After rebooting and re-opening moves down the list and crashes once again.

[zawawa]

If the miner crashes during optimization is there anyway to stop it on the next launch? After rebooting and re-opening moves down the list and crashes once again.

This is by design as the optimizer pushes the hardware to the limits.
I will add a dialog box so that you can abort the optimizer.

[zawawa]

Glad to hear the miner is stable. As for the graphs, are there any exceptions in the log at the startup?

This is what I got from the start up:

2018-04-07 21:38:51.0016 [1] Gateless Gate Sharp 1.3.2 prealpha started.
2018-04-07 21:38:51.8767 [1] Number of Devices: 5
2018-04-07 21:38:51.9704 [1] Successfully initialized AMD Display Library.
2018-04-07 21:38:51.9704 [1] Number of ADL Adapters: 80
2018-04-07 21:38:52.5486 [1] ADL_Adapter_AdapterInfo_Get() returned error code 0
2018-04-07 21:38:52.5486 [1] Failed to initialize NVIDIA Management Library.
2018-04-07 21:38:52.5642 [1] Exception in UpdateControls(): The given key was not present in the dictionary.   at System.ThrowHelper.ThrowKeyNotFoundException()
   at System.Collections.Generic.Dictionary`2.get_Item(TKey key)
   at GatelessGateSharp.MainForm.UpdateControls()
2018-04-07 21:39:02.5653 [1] Loading settings from C:\Users\Murgor Alexander\AppData\Roaming\GatelessGateSharp\GatelessGateSharp.sqlite...
2018-04-07 21:39:06.9095 [1] Loaded settings.
2018-04-07 21:39:06.9095 [1] Invalid argument: C:\Program Files\Zawawa Software LLC\Gateless Gate Sharp\GatelessGateSharp.exe
2018-04-07 21:39:10.7693 [1] Launching miner(s) for Custom Pool 0...
2018-04-07 21:39:11.0818 [14] Miner thread for Device #0 started.
2018-04-07 21:39:11.0818 [14] NiceHash mode is off.
2018-04-07 21:39:11.0975 [14] Loaded BinaryKernels\Ellesmere_cryptonight_8.bin for Device #0.
2018-04-07 21:39:11.1131 [14] Built cryptonight program for Device #0.
2018-04-07 21:39:11.1131 [14] Build options: -O5 -IKernels -DWORKSIZE=8
2018-04-07 21:39:11.1131 [14] Switching to Monero Version 1
2018-04-07 21:39:11.2537 [15] Miner thread for Device #0 started.
2018-04-07 21:39:11.2537 [15] NiceHash mode is off.
2018-04-07 21:39:11.2537 [15] Loaded BinaryKernels\Ellesmere_cryptonight_8.bin for Device #0.

Afterwards it just repeats the last 7-8 rows for all gpu's and the hash reports kicks in.

I guess I need to find out what the heck is causing this exception... Thanks!