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

[zawawa]

you miner has good interface but its not fast enough

with 16 / 8 i get 820 on cryptonight (sgminer 920h/s)

if i go higher with intensity or worksize its not stable

Well, it is only optimized for a few cards so far...
Which card do you have?

i have rx 580 8gb cards...i use rawintensity in sgminer which is better for cryptonight from my experience

Does rawintensity make that much difference? Let me see...

[1715653440]

1715653440

[1715653440]

1715653440

[1715653440]

1715653440

[treanski]

you miner has good interface but its not fast enough

with 16 / 8 i get 820 on cryptonight (sgminer 920h/s)

if i go higher with intensity or worksize its not stable

Well, it is only optimized for a few cards so far...
Which card do you have?

i have rx 580 8gb cards...i use rawintensity in sgminer which is better for cryptonight from my experience

Does rawintensity make that much difference? Let me see...

rawintensity is the most important part for cryptonight...there are some magic numbers that increase hashrate very well

[preda]

rawintensity is the most important part for cryptonight...there are some magic numbers that increase hashrate very well



sharing is caring!!!!!!

[UnclWish]

rawintensity is the most important part for cryptonight...there are some magic numbers that increase hashrate very well



sharing is caring!!!!!!

I tried in GG many variants of rawintencity, but I can't achieve speed that GGS gives with globalwork size 32 on my RX 580 8Gb. Maybe on other cards rawintensity is needed, but not on RX 580.

zawawa, it's a bug in 1.1.0 alpha. GGS didn't save option "Disable promt before auto-start". Untill program opened option is enabled. Restarting GGS and it asking on start and option is disabled. I deleted GatelessGateSharp.sqlite and tune program again, but it didn't helps.

[Raimy]

I don't know if it has been reported already, but I get a lot of the following errors:

2017-11-18 16:27:25.9921 [1] Exception: The remote server returned an error: (429) Too Many Requests.   at System.Net.WebClient.DownloadDataInternal(Uri address, WebRequest& request)
   at System.Net.WebClient.DownloadString(Uri address)
   at GatelessGateSharp.MainForm.UpdateStatsWithLongPolling()

[nerdralph]

don't worry, you will earn only peanuts cause you chose the open source path 
i put a laughing smiley, but inside me i cry 

You know, that might be true, but right now - regardless of the income or lack thereof - I'm glad I open-sourced these, and I'd do it again.

Simply because of the fact that while the mining "community" may be full of over-entitled leeches that, if allowed, would take everything you could give, and disappear when you couldn't....

Well, shit - there seems to be ONE person here who understands the work that went into this shit, and actually values it. Even enough to give something back. Kinda motivating.

Seeing your keccak GCN asm code last year was a big motivation for me to learn GCN architecture and asm.  Although other priorities in life have kept me from releasing any asm kernels, some day I will, and they will be open source.

[zawawa]

I don't know if it has been reported already, but I get a lot of the following errors:

2017-11-18 16:27:25.9921 [1] Exception: The remote server returned an error: (429) Too Many Requests.   at System.Net.WebClient.DownloadDataInternal(Uri address, WebRequest& request)
   at System.Net.WebClient.DownloadString(Uri address)
   at GatelessGateSharp.MainForm.UpdateStatsWithLongPolling()

Ah, some pools don't like their API's to be accessed so often. Which pool is it?

[zawawa]

I tried raw intensities for CryptoNight, but I didn't get better hashrates.
I guess it wouldn't hurt to leave it like this, so I will let you guys figure out.

[zawawa]

Gateless Gate Sharp 1.1.1 alpha
https://github.com/zawawawa/GatelessGateSharp/releases/tag/v1.1.1-alpha

* Change "Intensity" to "Raw Intensity" for CryptoNight.
* Various bug fixes and improvements.

Lbry is next in line, I guess...

[zawawa]

This doesn't look all that complicated...

https://github.com/lbryio/lbrycrd-gpu/blob/master/algorithm/lbry.c

[zawawa]

Simple, huh.

Code:

static cl_int queue_lbry_kernel(struct __clState *clState, struct _dev_blk_ctx *blk, __maybe_unused cl_uint threads)
{
  cl_kernel *kernel = &clState->kernel;
  unsigned int num = 0;
  cl_ulong le_target;
  cl_int status = 0;

  le_target = *(cl_ulong *)(blk->work->target + 24);
  flip112(clState->cldata, blk->work->data);
  status = clEnqueueWriteBuffer(clState->commandQueue, clState->CLbuffer0, true, 0, 112, clState->cldata, 0, NULL, NULL);

  CL_SET_ARG(clState->CLbuffer0);
  CL_SET_ARG(clState->padbuffer8);
  num = 0;
  kernel = clState->extra_kernels;
  CL_SET_ARG_0(clState->padbuffer8);
  num = 0;
  
  CL_NEXTKERNEL_SET_ARG(clState->padbuffer8);
  CL_SET_ARG(clState->outputBuffer);
  CL_SET_ARG(le_target);

  return status;
}

Code:

{ "lbry", ALGO_LBRY, "", 1, 256, 256, 0, 0, 0xFF, 0xFFFFULL, 0x0000ffffUL, 2, 4 * 8 * 4194304, 0, lbry_regenhash, NULL, NULL, queue_lbry_kernel, gen_hash, NULL },

Code:

  else if(algorithm->type == ALGO_LBRY) readbufsize = 112;

Code:

  else if(work->pool->algorithm.type == ALGO_LBRY) nonce_pos = 108;

Code:

else if (pool->algorithm.type == ALGO_LBRY) {
		nonce = *((uint32_t *)(work->data + 108));
	}

[UnclWish]

zawawa, thanks for update! All works!

Interesting, that with the rawintensity I can receive the same speed as with the globalwork size in pre 1.1.1 versions, but not only with rawintensity equal to globalworksize*CU.
The same result as with RI 1152 (32*36) can be reached with rawintensity 1024, wich is can't be set via globalwork size on RX 580... 1024 can't be devided be 36...

It would be nice to set up intensity for each thread of GPUs, like it's made in XMR-Stak-AMD and, if I'm not mistaken, in Excavator too. It's possible to do? With this, it would be possible to tune more effective configs.

[Airen]

http://prntscr.com/hcfe03

In attempt of start I receive it...

[mfsmarques]

What is the best value for rawintensity? I tested 1152 and got the following:

RX 580 4GB - 770 H/s
RX 580 8GB - 860 H/s

[UnclWish]

What is the best value for rawintensity? I tested 1152 and got the following:

RX 580 4GB - 770 H/s
RX 580 8GB - 860 H/s

For 4Gb 1152 is too much. For 1152 it's need more than 4Gb videomemory. Try for 4Gb 1024.
For both 4Gb and 8Gb speed on 1024 and 1152 is equal. With 2 threads and 8 localwork size.

[zawawa]

http://prntscr.com/hcfe03

In attempt of start I receive it...

Which card are you using?
Can I see a screenshot of the dashboard?
You can just lower the intensity.

[Airen]

Which card are you using? - 580 8Gb
Can I see a screenshot of the dashboard? - Program can't start.
You can just lower the intensity. - Noup, because program can't start

[mfsmarques]

What is the best value for rawintensity? I tested 1152 and got the following:

RX 580 4GB - 770 H/s
RX 580 8GB - 860 H/s

For 4Gb 1152 is too much. For 1152 it's need more than 4Gb videomemory. Try for 4Gb 1024.
For both 4Gb and 8Gb speed on 1024 and 1152 is equal. With 2 threads and 8 localwork size.

I was using 1 thread, now 4GB - 850 H/s.
Thanks.

[Raimy]

I don't know if it has been reported already, but I get a lot of the following errors:

2017-11-18 16:27:25.9921 [1] Exception: The remote server returned an error: (429) Too Many Requests.   at System.Net.WebClient.DownloadDataInternal(Uri address, WebRequest& request)
   at System.Net.WebClient.DownloadString(Uri address)
   at GatelessGateSharp.MainForm.UpdateStatsWithLongPolling()

Ah, some pools don't like their API's to be accessed so often. Which pool is it?

It mainly looks to happen while mining on ethermine.org, but I have seen it on other pools as well.

[zawawa]

I don't know if it has been reported already, but I get a lot of the following errors:

2017-11-18 16:27:25.9921 [1] Exception: The remote server returned an error: (429) Too Many Requests.   at System.Net.WebClient.DownloadDataInternal(Uri address, WebRequest& request)
   at System.Net.WebClient.DownloadString(Uri address)
   at GatelessGateSharp.MainForm.UpdateStatsWithLongPolling()

Ah, some pools don't like their API's to be accessed so often. Which pool is it?

It mainly looks to happen while mining on ethermine.org, but I have seen it on other pools as well.

Did you try Version 1.1.1 alpha? Any differences?