nice work. it hasnt took you long to catch up with Claymore. (need a thumbs up emoji)
How far do you think you can think you can push the RX cards? Do you have a dev fee included in your miner or are you just going on donations?
Topic: Gateless Gate Sharp 1.3.8: 30Mh/s (Ethash) on RX 480! - page 188. (Read 214431 times)
I just figured out a way to give GG the final boost it needs to catch up with Claymore's etc.
All I need to do is to divide each round of Wagner's into subrounds according to the number of slots in each row.
The basic idea is to put rows with fewer slots together to reduce the size of the read cache for slots and to increase occupancy.
This is almost guaranteed to work as there shouldn't be too much over head.
All I need to do is to divide each round of Wagner's into subrounds according to the number of slots in each row.
The basic idea is to put rows with fewer slots together to reduce the size of the read cache for slots and to increase occupancy.
This is almost guaranteed to work as there shouldn't be too much over head.
And - 0.3 miner not working with 280x: crash right after start with error "sa-solver.exe has stopped working" for each gpu.
That's good to know. I suspect the culprit is the "volatile" keyword.
Let me fix that issue as well.
Tested it with OC-ed 1070
2 instances 300sol/s
nicehash 410sol/s
If I up the threads it give only 0sol/s
looks like windows build use some CPU 20% for instance on i5 CPU
2 instances 300sol/s
nicehash 410sol/s
If I up the threads it give only 0sol/s
looks like windows build use some CPU 20% for instance on i5 CPU
Not bad, not bad. You just need to increase the number of instances for high-end NVIDIA cards, then.
For some unknown reasons, WIN32 threading does not work with NVIDIA drivers. Strange, strange...
Try with instances instead of threads.
Yeah, "--instances 2" did work with GTX 1060 and is now the default setting for the NVIDIA version.
I should mention this somewhere in the documentation...
And - 0.3 miner not working with 280x: crash right after start with error "sa-solver.exe has stopped working" for each gpu.
1070 micron 2050/4400/75%tdp, cpu g1840, Win8.1
With single card - it is possible get up to 290 s/s, but...it has high cpu usage - load at ~90%. With 3 cards I can get ~275 s/s for one, and with 6 cards - only ~220.
Tried with different threads/instances parameters and all has high cpu usage.
With single card - it is possible get up to 290 s/s, but...it has high cpu usage - load at ~90%. With 3 cards I can get ~275 s/s for one, and with 6 cards - only ~220.
Tried with different threads/instances parameters and all has high cpu usage.
Thanks for the report. I guess the current CPU usage fix does not work too well.
I will look into this problem.
1070 micron 2050/4400/75%tdp, cpu g1840, Win8.1
With single card - it is possible get up to 290 s/s, but...it has high cpu usage - load at ~90%. With 3 cards I can get ~275 s/s for one, and with 6 cards - only ~220.
Tried with different threads/instances parameters and all has high cpu usage.
With single card - it is possible get up to 290 s/s, but...it has high cpu usage - load at ~90%. With 3 cards I can get ~275 s/s for one, and with 6 cards - only ~220.
Tried with different threads/instances parameters and all has high cpu usage.
Tested it with OC-ed 1070
2 instances 300sol/s
nicehash 410sol/s
If I up the threads it give only 0sol/s
looks like windows build use some CPU 20% for instance on i5 CPU
2 instances 300sol/s
nicehash 410sol/s
If I up the threads it give only 0sol/s
looks like windows build use some CPU 20% for instance on i5 CPU
Not bad, not bad. You just need to increase the number of instances for high-end NVIDIA cards, then.
For some unknown reasons, WIN32 threading does not work with NVIDIA drivers. Strange, strange...
Try with instances instead of threads.
Yeah, "--instances 2" did work with GTX 1060 and is now the default setting for the NVIDIA version.
I should mention this somewhere in the documentation...
Tested it with OC-ed 1070
2 instances 300sol/s
nicehash 410sol/s
If I up the threads it give only 0sol/s
looks like windows build use some CPU 20% for instance on i5 CPU
2 instances 300sol/s
nicehash 410sol/s
If I up the threads it give only 0sol/s
looks like windows build use some CPU 20% for instance on i5 CPU
Not bad, not bad. You just need to increase the number of instances for high-end NVIDIA cards, then.
For some unknown reasons, WIN32 threading does not work with NVIDIA drivers. Strange, strange...
Try with instances instead of threads.
Tested it with OC-ed 1070
2 instances 300sol/s
nicehash 410sol/s
If I up the threads it give only 0sol/s
looks like windows build use some CPU 20% for instance on i5 CPU
2 instances 300sol/s
nicehash 410sol/s
If I up the threads it give only 0sol/s
looks like windows build use some CPU 20% for instance on i5 CPU
Not bad, not bad. You just need to increase the number of instances for high-end NVIDIA cards, then.
For some unknown reasons, WIN32 threading does not work with NVIDIA drivers. Strange, strange...
Tested it with OC-ed 1070
2 instances 300sol/s
nicehash 410sol/s
If I up the threads it give only 0sol/s
looks like windows build use some CPU 20% for instance on i5 CPU
2 instances 300sol/s
nicehash 410sol/s
If I up the threads it give only 0sol/s
looks like windows build use some CPU 20% for instance on i5 CPU
My wife is now constantly complaining about my forgetfulness and absent-mindedness, but that's OK...
my wife does the same, and i dont code, i just mine.......
btw those long walks with wife do wonders
Yeah, I need to slow down a little bit. Not too much, though.
After Equihash, I would really like to look into Ethash and CryptoNight.
cryptonight for nvidia would be awesome, it is totaly untouched for the moment , latest miners comes from 2014 (tsiv/klaus) and underperforming alot comparing to amd.
Awesome job so far...your miner is now just as fast as optiminers (if you factor in his dev fee speed he is still slightly faster, but regardless this is still nice work). 470s are now doing 170 with this miner (and your only 40S/s away from claymour). Will def send some btc your way.
Thanks! That would help me to buy more equipment for testing.
I really want to test this miner with high-end cards.
My wife is now constantly complaining about my forgetfulness and absent-mindedness, but that's OK...
my wife does the same, and i dont code, i just mine.......
btw those long walks with wife do wonders
Yeah, I need to slow down a little bit. Not too much, though.
After Equihash, I would really like to look into Ethash and CryptoNight.
I just uploaded a new release with precompiled binaries for Windows:
https://github.com/zawawawa/gatelessgate
188 sol/s on RX 480 and 165 sol/s on GTX 1060 3GB.
Pretty good given the fact that I haven't touched a GCN assembler or PTX.
I am also thinking about swapping gatelessgate.py with sgminer-gm.
We will see how that goes.
https://github.com/zawawawa/gatelessgate
188 sol/s on RX 480 and 165 sol/s on GTX 1060 3GB.
Pretty good given the fact that I haven't touched a GCN assembler or PTX.
I am also thinking about swapping gatelessgate.py with sgminer-gm.
We will see how that goes.
Awesome job so far...your miner is now just as fast as optiminers (if you factor in his dev fee speed he is still slightly faster, but regardless this is still nice work). 470s are now doing 170 with this miner (and your only 40S/s away from claymour). Will def send some btc your way.
My wife is now constantly complaining about my forgetfulness and absent-mindedness, but that's OK...
my wife does the same, and i dont code, i just mine.......
btw those long walks with wife do wonders
My wife is now constantly complaining about my forgetfulness and absent-mindedness, but that's OK...
my wife does the same, and i dont code, i just mine.......
btw those long walks with wife do wonders
Code:
In file included from main.c:33:0:
main.c: In function ‘tests’:
main.c:1858:9: error: ‘NR_ROWS_LOG’ undeclared (first use in this function)
assert(NR_ROWS_LOG >= 12);
I simply removed the function "tests" and the call to it and it compiles and runs...
I just fixed it after a long morning walk with my wife. Thanks for letting me know!
Code:
In file included from main.c:33:0:
main.c: In function ‘tests’:
main.c:1858:9: error: ‘NR_ROWS_LOG’ undeclared (first use in this function)
assert(NR_ROWS_LOG >= 12);
I simply removed the function "tests" and the call to it and it compiles and runs...
Jump to: