Topic: SILENTARMY v5: Zcash miner, 115 sol/s on R9 Nano, 70 sol/s on GTX 1070 - page 54. (Read 209286 times)
November 12, 2016, 02:09:28 PM
finally it work, win 10, 6 x 1070, doing up to 500 sol
November 12, 2016, 02:03:46 PM
new to this. how do you upload an image saved on ubuntu?
You mean a picture? Save it in any online service, like imgur (http://imgur.com/)
Then, insert the link to it here.
Got it?
November 12, 2016, 01:59:08 PM
I didn't remove +2, added another +2
+2 not need anymore for all platforms.COLL_DATA_SIZE_PER_TH best *5 for AMD and *6 (or *7) for NV Pascal.
November 12, 2016, 01:58:29 PM
new to this. how do you upload an image saved on ubuntu?
November 12, 2016, 01:51:55 PM
Can you share your changes ?
All default, no changes 
Quote
It is shared memory, and NR_SLOTS=12, so NR_SLOTS+2=14 bytes, and it is not aligned by 4 byte memory bank size... I tryed to fix it(make +2 bytes offset, seems getting another 1-2%, but difficult to confirm it.
I see no difference.Edit: NVidia no difference, but AMD RX480 faster by 1-1,5% when remove "+2".
Quote
gives +5% on 1070(confirmed).
You are right for GTX1070, went from 70 to 73 sols/sI didn't remove +2, added another +2
November 12, 2016, 01:46:14 PM
can anyone help me plz? 
I use ubuntu 14.04, 5x R9 280x
thaipham@thaipham280x:~/silentarmy$ ./silentarmy
Traceback (most recent call last):
File "./silentarmy", line 600, in
main()
File "./silentarmy", line 597, in main
Silentarmy(opts).run()
File "./silentarmy", line 282, in run
my_ensure_future(self.reconnect())
File "./silentarmy", line 47, in my_ensure_future
task = loop.create_task(coro)
AttributeError: 'UnixSelectorEventLoop' object has no attribute 'createtask'
I use ubuntu 14.04, 5x R9 280x
thaipham@thaipham280x:~/silentarmy$ ./silentarmy
Traceback (most recent call last):
File "./silentarmy", line 600, in
main()
File "./silentarmy", line 597, in main
Silentarmy(opts).run()
File "./silentarmy", line 282, in run
my_ensure_future(self.reconnect())
File "./silentarmy", line 47, in my_ensure_future
task = loop.create_task(coro)
AttributeError: 'UnixSelectorEventLoop' object has no attribute 'createtask'
November 12, 2016, 01:41:20 PM
Can you share your changes ?
All default, no changes Quote
It is shared memory, and NR_SLOTS=12, so NR_SLOTS+2=14 bytes, and it is not aligned by 4 byte memory bank size... I tryed to fix it(make +2 bytes offset, seems getting another 1-2%, but difficult to confirm it.
I see no difference.Edit: NVidia no difference, but AMD RX480 faster by 1-1,5% when remove "+2".
Quote
gives +5% on 1070(confirmed).
You are right for GTX1070, went from 70 to 73 sols/s 
November 12, 2016, 01:38:15 PM
Total 418.3 sol/s [dev0 59.9, dev1 61.2, dev2 39.4, dev3 58.3, dev4 66.6, dev5 62.9] 66 shares
Total 418.2 sol/s [dev0 58.3, dev1 60.2, dev2 38.6, dev3 57.7, dev4 65.8, dev5 63.6] 66 shares
Total 414.7 sol/s [dev0 58.8, dev1 60.6, dev2 35.0, dev3 59.9, dev4 63.1, dev5 64.2] 66 shares
Total 412.6 sol/s [dev0 58.0, dev1 61.0, dev2 31.0, dev3 57.8, dev4 62.8, dev5 66.4] 66 shares
running latest version 5 on win
all gpu are 470 4gb sapphire nitros EXCEPT dev 2 with elpida memory.
was running on 1st version it's fine but after upgraded to latest only dev 2 with elpida memory acting up...
any help ?
TIA
Total 418.2 sol/s [dev0 58.3, dev1 60.2, dev2 38.6, dev3 57.7, dev4 65.8, dev5 63.6] 66 shares
Total 414.7 sol/s [dev0 58.8, dev1 60.6, dev2 35.0, dev3 59.9, dev4 63.1, dev5 64.2] 66 shares
Total 412.6 sol/s [dev0 58.0, dev1 61.0, dev2 31.0, dev3 57.8, dev4 62.8, dev5 66.4] 66 shares
running latest version 5 on win
all gpu are 470 4gb sapphire nitros EXCEPT dev 2 with elpida memory.
was running on 1st version it's fine but after upgraded to latest only dev 2 with elpida memory acting up...
any help ?
TIA
November 12, 2016, 01:35:52 PM
guys what about this https://github.com/Optiminer/OptiminerZcash
November 12, 2016, 01:14:00 PM
I get 70sol/s on the 1070 with the latest SA kernel windows port
But zcash is currently being harvested by kernels that produce 200% more
hAPPY bd
But zcash is currently being harvested by kernels that produce 200% more
hAPPY bd
November 12, 2016, 01:11:12 PM
Latest version gives me 92-93 sol/s from modded RX480:
Quote
Total 361.5 sol/s [dev0 92.4, dev1 89.5, dev2 81.7, dev3 87.1] 307 shares
Total 361.8 sol/s [dev0 91.7, dev1 89.2, dev2 81.0, dev3 92.5] 308 shares
Total 358.8 sol/s [dev0 92.2, dev1 85.9, dev2 79.4, dev3 91.2] 310 shares
Total 361.8 sol/s [dev0 91.7, dev1 89.2, dev2 81.0, dev3 92.5] 308 shares
Total 358.8 sol/s [dev0 92.2, dev1 85.9, dev2 79.4, dev3 91.2] 310 shares
Can you share your changes ?
eXtremal
one question about this code
Code:
__local uchar *first_words = &first_words_data[(NR_SLOTS+2)*tlid];
And changing
Code:
#define COLL_DATA_SIZE_PER_TH (NR_SLOTS * 5)
to
Code:
#define COLL_DATA_SIZE_PER_TH (NR_SLOTS * 7)
gives +5% on 1070(confirmed).
November 12, 2016, 01:02:19 PM
Is it better to do --instances=2 or --instances=1?
What are people running with for this switch that gives the best results?
What are people running with for this switch that gives the best results?
November 12, 2016, 12:59:35 PM
How can i get Silentarmy running on windows?
https://mega.nz/#F!iwYQhI6S!2vl0D3jz3fH3-bZxHUurBgDownload cigwin 64bit when asked for packets search python3 interpreter install and run from start.bat included in zip file(don't forget to edit...or not
November 12, 2016, 12:59:06 PM
How can i get Silentarmy running on windows?
Download r5 from here:
mega:///#F!iwYQhI6S!2vl0D3jz3fH3-bZxHUurBg
and read windows.txt
November 12, 2016, 12:56:01 PM
How can i get Silentarmy running on windows?
November 12, 2016, 12:47:57 PM
latest win version with cygwin working fine
just wondering, is there some api?
just wondering, is there some api?
November 12, 2016, 12:38:04 PM
claymore vs silentarmy
CLAYMORE v4.0
-i 2
280X 155W 68 h/s
270x 41 h/s
SILENTARMY-v5-Cygwin64-r5
instances=2
280X 130W 72 sols
270x 44 sols
GTX 1060 3G 67W 63 sols
CLAYMORE v4.0
-i 2
280X 155W 68 h/s
270x 41 h/s
SILENTARMY-v5-Cygwin64-r5
instances=2
280X 130W 72 sols
270x 44 sols
GTX 1060 3G 67W 63 sols
November 12, 2016, 12:36:31 PM
just tried all versions of the windows port.. weird problem with 2 rigs... 1 rig is 6 RX 470's and it will only do 20sol/s on each card but when i installed on another rig thats exactly the same(OS/drivers/hardware) they do 76sol each... same with another set of rigs that are 290/290x mixed ...1 works awsome,1 very slow 23ish sols.. cant figure out what the difference is that is causing it..
November 12, 2016, 12:34:00 PM
Quote
yup...faster than SAv5. But I'm sure SA will catch up soon
Edit: maybe not as stable, just had a crash.
Yes it's faster, with 3 Rx 480 I've got from ~265 sol/s with silentarmy to 310 - 365 sol/s with this. But it crashed on GPU #0 after ~10 minutes and only a reset helped with it, so my vote is on silentarmy! You are doing great guys!
I had a rig crash and now it crashes each time I start optiminer. The rig just restarts. As I said previously, not stable or it asks too much to my cards...don't now
November 12, 2016, 12:33:51 PM
Any hint why the solver ist 2 times SLOWER in V5 compared to V4 ?
Regards
Try running with ./silentarmy --instances=1 option, it is faster on my nvidia cards with one instance
And replace main.c with this, then rebuild
http://s000.tinyupload.com/index.php?file_id=59090490226848298558
Sorry no same with the changed main.c :
Code:
:~/silentarmy$ ./sa-solver --nonces=5
Solving default all-zero 140-byte header
Building program
Hash tables will use 805.3 MB
Running...
Nonce 0000000000000000000000000000000000000000000000000000000000000000: 2 sols
Nonce 0100000000000000000000000000000000000000000000000000000000000000: 0 sols
Nonce 0200000000000000000000000000000000000000000000000000000000000000: 2 sols
Nonce 0300000000000000000000000000000000000000000000000000000000000000: 3 sols
Nonce 0400000000000000000000000000000000000000000000000000000000000000: 2 sols
Total 9 solutions in 1505.8 ms (6.0 Sol/s)
~/silentarmy.v4$ ./sa-solver --nonces=5
Solving default all-zero 140-byte header
Building program
Hash tables will use 805.3 MB
Running...
Nonce 0000000000000000000000000000000000000000000000000000000000000000: 2 sols
Nonce 0100000000000000000000000000000000000000000000000000000000000000: 0 sols
Nonce 0200000000000000000000000000000000000000000000000000000000000000: 2 sols
Nonce 0300000000000000000000000000000000000000000000000000000000000000: 3 sols
Nonce 0400000000000000000000000000000000000000000000000000000000000000: 2 sols
Total 9 solutions in 654.1 ms (13.8 Sol/s)
The difference is in the solver ..
regards
Anyone got an idea what's the cause of the speed decrease with V5 ?
./silentarmy --list
Devices on platform "NVIDIA CUDA":
ID 0: GRID K520
regards
I know it's not much - but I got ~ 15 wit V4 and only 50% with V5 .
or can't the miner be optinized for this chip ?
regards
Jump to: