Topic: [ANN] cudaMiner & ccMiner CUDA based mining applications [Windows/Linux/MacOSX] - page 999. (Read 3426921 times)
January 20, 2014, 06:07:45 PM
Is vertcoin any profitable? I mean I'm not sure how much it says, but the average hashrate per miners of the biggest pool is over 700 kH/s.
January 20, 2014, 05:53:19 PM
Tried Vertcoin just now, but I am getting CPU validation errors with the latest git build.
2048 is the current N value right? This is what I'm using for settings:
--algo=scrypt:2048 -i 0 -d 1 -H 0 -l T12x32
2048 is the current N value right? This is what I'm using for settings:
--algo=scrypt:2048 -i 0 -d 1 -H 0 -l T12x32
thank you for noticing
in the CPU validation code I just changed
j = 32 * (X[16] & 1023);
into
j = 32 * (X[16] % N);
so it should work correctly for ANY N, even non powers of 2.
Christian
Awesome, it's working now
January 20, 2014, 05:52:06 PM
Tried Vertcoin just now, but I am getting CPU validation errors with the latest git build.
2048 is the current N value right? This is what I'm using for settings:
--algo=scrypt:2048 -i 0 -d 1 -H 0 -l T12x32
2048 is the current N value right? This is what I'm using for settings:
--algo=scrypt:2048 -i 0 -d 1 -H 0 -l T12x32
thank you for noticing
in the CPU validation code I just changed
j = 32 * (X[16] & 1023);
into
j = 32 * (X[16] % N);
so it should work correctly for ANY N, even non powers of 2.
Christian
January 20, 2014, 05:46:42 PM
Tried Vertcoin just now, but I am getting CPU validation errors with the latest git build.
Debug output:
2048 is the current N value right? This is what I'm using for settings:
--algo=scrypt:2048 -i 0 -d 1 -H 0 -l T12x32
Debug output:
Quote
Hash: 000005656bc0b01bf4aac839405df06c36fae7cdcae8c66b96118058bd5c46a6
Target: 00007fff80000000000000000000000000000000000000000000000000000000
[2014-01-20 14:43:25] GPU #1: GeForce GTX 780 result does not validate on CPU (i=1534, s=0)!
[2014-01-20 14:43:27] DEBUG: hash <= target
Hash: 00007baba609dfb7bc1c16fa272a513215145695ab99a6ce49b61f4918d88d73
Target: 00007fff80000000000000000000000000000000000000000000000000000000
[2014-01-20 14:43:27] GPU #1: GeForce GTX 780 result does not validate on CPU (i=1354, s=1)!
[2014-01-20 14:43:29] DEBUG: hash <= target
Hash: 00003473ba7d4b4082634e13221c5895c6e8fdf2206a22004d6c97b1d7942181
Target: 00007fff80000000000000000000000000000000000000000000000000000000
[2014-01-20 14:43:29] GPU #1: GeForce GTX 780 result does not validate on CPU (i=1581, s=1)!
[2014-01-20 14:43:29] DEBUG: hash <= target
Hash: 000024ba094bf02988538219f98d45fa2ba131b1107b783fcb13e6c4a3efe342
Target: 00007fff80000000000000000000000000000000000000000000000000000000
[2014-01-20 14:43:29] GPU #1: GeForce GTX 780 result does not validate on CPU (i=595, s=1)!
Target: 00007fff80000000000000000000000000000000000000000000000000000000
[2014-01-20 14:43:25] GPU #1: GeForce GTX 780 result does not validate on CPU (i=1534, s=0)!
[2014-01-20 14:43:27] DEBUG: hash <= target
Hash: 00007baba609dfb7bc1c16fa272a513215145695ab99a6ce49b61f4918d88d73
Target: 00007fff80000000000000000000000000000000000000000000000000000000
[2014-01-20 14:43:27] GPU #1: GeForce GTX 780 result does not validate on CPU (i=1354, s=1)!
[2014-01-20 14:43:29] DEBUG: hash <= target
Hash: 00003473ba7d4b4082634e13221c5895c6e8fdf2206a22004d6c97b1d7942181
Target: 00007fff80000000000000000000000000000000000000000000000000000000
[2014-01-20 14:43:29] GPU #1: GeForce GTX 780 result does not validate on CPU (i=1581, s=1)!
[2014-01-20 14:43:29] DEBUG: hash <= target
Hash: 000024ba094bf02988538219f98d45fa2ba131b1107b783fcb13e6c4a3efe342
Target: 00007fff80000000000000000000000000000000000000000000000000000000
[2014-01-20 14:43:29] GPU #1: GeForce GTX 780 result does not validate on CPU (i=595, s=1)!
2048 is the current N value right? This is what I'm using for settings:
--algo=scrypt:2048 -i 0 -d 1 -H 0 -l T12x32
January 20, 2014, 05:44:56 PM
@cbuchner1 I have compiled a few versions of CudaMiner (along with instructions) for Mac OS X users: http://www.johnchapman.net/cudaminer/
I plan to compile an OS X version for each future release (as long as they compile ;-)). Thanks for the great work!
I plan to compile an OS X version for each future release (as long as they compile ;-)). Thanks for the great work!
Wow! This is great. I put a link into the first posting.
Awesome. :-)
January 20, 2014, 05:37:03 PM
"Seems Kepler Kernal has better memory allocation then the Titan for Scrypt-Jane.
don't think so. Features for the Titan kernel have been brought to the same level as for Kepler. Just -C 1 and 2 aren't needed for Titan, as caching is automatic and always implied.
Christian
January 20, 2014, 05:17:33 PM
I think Titans have a problem allocating all that memory.
Quoted from the Scrypt-jane spreadsheet:
"GPU Memory usage: 2883 MB"
"Seems Kepler Kernal has better memory allocation then the Titan for Scrypt-Jane. Texture Cache set to -C 1 throws an error indicating it fails over to -C 2 but if you launch a -C2 the hashrate is nearly halved."
Edit: Also "Current Titan kernal appears not to allow much more then this before driver soft crashes currently."
Quoted from the Scrypt-jane spreadsheet:
"GPU Memory usage: 2883 MB"
"Seems Kepler Kernal has better memory allocation then the Titan for Scrypt-Jane. Texture Cache set to -C 1 throws an error indicating it fails over to -C 2 but if you launch a -C2 the hashrate is nearly halved."
Edit: Also "Current Titan kernal appears not to allow much more then this before driver soft crashes currently."
Could this be a 32-bit limitation? Maybe we need a working 64-bit version?
January 20, 2014, 05:03:46 PM
You are the best!
I will report how vertcoin performs.
I will report how vertcoin performs.
Getting about 81 kh/s on a 770M (910MHZ) vs 214 kh/s on a R9 270X (which was getting 450 kh/s on "normal" scrypt and 2.9 kh/s on YAC) - Impressive
January 20, 2014, 04:56:14 PM
You are the best!
I will report how vertcoin performs.
I will report how vertcoin performs.
the current github is not reaching the speeds of the 2013-12-18 release, which is kind of a mystery to me.
So your vertcoin mining might be a bit slower than expected.
Fixing this will be another important milestone before the next binary release.
Christian
January 20, 2014, 04:52:24 PM
You are the best!
I will report how vertcoin performs.
I will report how vertcoin performs.
January 20, 2014, 04:37:57 PM
I see that someone has "released cudaminer" for vertcoin
https://bitcointalksearch.org/topic/m.4620453
at least with a better ratio than the buggy AMD implementation vs scrypt - would have been gentlemanly if they had tipped their hat to Christian for all his effort.
https://bitcointalksearch.org/topic/m.4620453
at least with a better ratio than the buggy AMD implementation vs scrypt - would have been gentlemanly if they had tipped their hat to Christian for all his effort.
I believe the CPU based validation code has some kind of hardcoded scratchpad buffer size for N=1024. This needs to be fixed.
EDIT:
oh look what's there !
Code:
scratch[32768]
EDIT2: okay, a likely fix is now on github. mine away!
January 20, 2014, 04:28:43 PM
Any news on the vertcoin issue?
latest github version still gives me the following error:
Unhandled exception at 0x000007F679AD8A64 in cudaminer.exe: Stack cookie instrumentation code detected a stack-based buffer overrun.
Seems something related to cpu memory management.
Crashes with Win and Linux.
I know there are maybe more important things to do but vertcoin is already looking very promising.
latest github version still gives me the following error:
Unhandled exception at 0x000007F679AD8A64 in cudaminer.exe: Stack cookie instrumentation code detected a stack-based buffer overrun.
Seems something related to cpu memory management.
Crashes with Win and Linux.
I know there are maybe more important things to do but vertcoin is already looking very promising.
Not sure which issue you are referring to ? The post linked to in the post above (yours) has a working version for windows - there was a slight change for vertcoin apparently.
January 20, 2014, 04:19:11 PM
Any news on the vertcoin issue?
latest github version still gives me the following error:
Unhandled exception at 0x000007F679AD8A64 in cudaminer.exe: Stack cookie instrumentation code detected a stack-based buffer overrun.
Seems something related to cpu memory management.
Crashes with Win and Linux.
I know there are maybe more important things to do but vertcoin is already looking very promising.
latest github version still gives me the following error:
Unhandled exception at 0x000007F679AD8A64 in cudaminer.exe: Stack cookie instrumentation code detected a stack-based buffer overrun.
Seems something related to cpu memory management.
Crashes with Win and Linux.
I know there are maybe more important things to do but vertcoin is already looking very promising.
January 20, 2014, 04:16:59 PM
I see that someone has "released cudaminer" for vertcoin
https://bitcointalksearch.org/topic/m.4620453
at least with a better ratio than the buggy AMD implementation vs scrypt - would have been gentlemanly if they had tipped their hat to Christian for all his effort.
https://bitcointalksearch.org/topic/m.4620453
at least with a better ratio than the buggy AMD implementation vs scrypt - would have been gentlemanly if they had tipped their hat to Christian for all his effort.
January 20, 2014, 03:29:24 PM
That one was crashing my nvidia driver (gtx 560 ti), im guessing it crashes more people... :/
January 20, 2014, 03:23:58 PM
Also, does memory amount on gpu matter?
For scrypt? Not so much, but for scrypt-jane it does matter.
Is Scrypt jane heavily dependent on memory available? If so how are 6GB Titans performing?
I think Titans have a problem allocating all that memory.
Quoted from the Scrypt-jane spreadsheet:
"GPU Memory usage: 2883 MB"
"Seems Kepler Kernal has better memory allocation then the Titan for Scrypt-Jane. Texture Cache set to -C 1 throws an error indicating it fails over to -C 2 but if you launch a -C2 the hashrate is nearly halved."
Edit: Also "Current Titan kernal appears not to allow much more then this before driver soft crashes currently."
Sorry im very noob into this, would be nice if anyone would post a working version for Gpu/Windows 64 
I believe one was shared a couple of pages back.
I now have 30 kHash/s working for me. I think I am pulling more than a kilowatt of power now. Eeek. I am global warming.
More like global warning..... to ATI/AMD!
........ok, I stop.
January 20, 2014, 03:12:33 PM
Sorry im very noob into this, would be nice if anyone would post a working version for Gpu/Windows 64
January 20, 2014, 03:08:15 PM
I now have 30 kHash/s working for me. I think I am pulling more than a kilowatt of power now. Eeek. I am global warming.
January 20, 2014, 03:05:06 PM
@cbuchner1 I have compiled a few versions of CudaMiner (along with instructions) for Mac OS X users: http://www.johnchapman.net/cudaminer/
I plan to compile an OS X version for each future release (as long as they compile ;-)). Thanks for the great work!
I plan to compile an OS X version for each future release (as long as they compile ;-)). Thanks for the great work!
Wow! This is great. I put a link into the first posting.
January 20, 2014, 03:04:39 PM
Is Scrypt jane heavily dependent on memory available? If so how are 6GB Titans performing?
I think no one has tested this yet.
But the lookup gap makes things much less critical now on cards with lesser memory.
Christian
Jump to: