Topic: CCminer M7 (XCN) by djm34, fixed + optimized for cuda 8 and new cards by PALLAS - page 19. (Read 52620 times)
August 31, 2016, 07:13:25 AM
So any progress to get a windows build pallas? It would be good to have one with the new improvements you have made.
August 31, 2016, 07:03:18 AM
New commit: "1070 hashrate improvements"
Still not on par with 980Ti, but it's an improvement ;-)
Should affect 1060 and 1080 as well.
Still not on par with 980Ti, but it's an improvement ;-)
Should affect 1060 and 1080 as well.
i have one mystic moment with a tsiv fork:
i have 3 PC, all are Win7x64
960stock+970OC1506/7000 = works 24/7 with no problem 362.0 driver
4x970 (same OC, same bios) = works 3-4 hrs, than stops to send shares (see no visible mistakes)
3pscAMD (at Claymore) + 970 (same OC, same bios) = works 8-12 hrs may be more, than same stop (Claymore still work normally)
i'm set drivers from 362.0 till the lates - same bug
may be you have an advice.
Thanks
August 31, 2016, 05:46:36 AM
New commit: "1070 hashrate improvements"
Still not on par with 980Ti, but it's an improvement ;-)
Should affect 1060 and 1080 as well.
Still not on par with 980Ti, but it's an improvement ;-)
Should affect 1060 and 1080 as well.
August 31, 2016, 04:24:19 AM
Just wanted to let you guys know that I've been working on an important part of the algo (the multiplication routines) and used many different approaches to try and improve them. Unfortunately I couldn't came with any measurable hashrate increase: the code was already pretty good. I have some ideas left to try and will let you all know.
but is certain that the 1070 need improvement, it is even slower than a 980ti, maybe it need some "code trick" like the famous killer groestl from christian
yes that's still a mistery to solve, I think I'll focus on that next.
August 31, 2016, 04:15:00 AM
Just wanted to let you guys know that I've been working on an important part of the algo (the multiplication routines) and used many different approaches to try and improve them. Unfortunately I couldn't came with any measurable hashrate increase: the code was already pretty good. I have some ideas left to try and will let you all know.
but is certain that the 1070 need improvement, it is even slower than a 980ti, maybe it need some "code trick" like the famous killer groestl from christian
August 31, 2016, 03:54:33 AM
Just wanted to let you guys know that I've been working on an important part of the algo (the multiplication routines) and used many different approaches to try and improve them. Unfortunately I couldn't came with any measurable hashrate increase: the code was already pretty good. I have some ideas left to try and will let you all know.
August 30, 2016, 02:07:11 PM
So where to get a windows build?
only old one avaliable (can be found in my prev posts)
August 30, 2016, 09:57:46 AM
So where to get a windows build?
August 27, 2016, 12:20:42 PM
Nethash was too high even before this second pump.
That is normal @bittrex@randomcoin, xcn@polo doing this hmm...
Everyone sees it is manipulated but why?
How much does 970 on xcn to be on topic?
That is normal @bittrex@randomcoin, xcn@polo doing this hmm...
Everyone sees it is manipulated but why?
How much does 970 on xcn to be on topic?
with a tsiv 8900-9300-9800 Kh at my OC 1506 one
August 27, 2016, 10:47:11 AM
Nethash was too high even before this second pump.
That is normal @bittrex@randomcoin, xcn@polo doing this hmm...
Everyone sees it is manipulated but why?
How much does 970 on xcn to be on topic?
That is normal @bittrex@randomcoin, xcn@polo doing this hmm...
Everyone sees it is manipulated but why?
How much does 970 on xcn to be on topic?
August 27, 2016, 10:32:31 AM
peaked at 600. Now back to 450.
The mining hashrate has doubled as well.
Total network hashrate 5505.05 GHASH/s
The mining hashrate has doubled as well.
Total network hashrate 5505.05 GHASH/s
Looks as large farm owners migrates from LBRY, case i see LBRY network hashrate decreasing
August 27, 2016, 03:24:51 AM
peaked at 600. Now back to 450.
The mining hashrate has doubled as well.
Total network hashrate 5505.05 GHASH/s
The mining hashrate has doubled as well.
Total network hashrate 5505.05 GHASH/s
August 27, 2016, 12:51:17 AM
XCN again, wtf is going on?
something must always be pumped it's how the things work here
Haha, that's how it goes. But I just can't understand this one.
it's like the feathercoin pump, it will return where it was soon ennough, just a p&d
August 26, 2016, 02:57:20 PM
Yeah price is up again, well over 350 sat.
I'm hope, but not sure that it will be for a long time
may be we will see impruved ccminer in a bin? ;-)
of at least more stable than old tsiv fork...
August 26, 2016, 12:11:41 PM
Yeah price is up again, well over 350 sat.
August 26, 2016, 11:48:39 AM
XCN again, wtf is going on?
something must always be pumped it's how the things work here
Haha, that's how it goes. But I just can't understand this one.
August 26, 2016, 11:20:26 AM
XCN again, wtf is going on?
something must always be pumped it's how the things work here
August 26, 2016, 10:45:20 AM
XCN again, wtf is going on?
August 26, 2016, 08:08:05 AM
SP - your M7 miner not working correctly - when you planned fix it ?
I tried to build Pallas github for windows with cuda 6.5, and it doesn't seem to work. (90% rejects on the pool). Cuda 8.0 was crashing when compiling.I don't have time to look at it now.
I have investigated some more.
The cuda 6.5 32bit build hash is failing when startnonce is bigger than 0x00ffffff. So if the highbyte is filled with something else than 0. It produce the wrong hash.
I enabled cpu mining verification.
reproducable in benchmark mode by adding this line on the top:
pdata[29] = 0x023445566;
( const uint32_t FirstNonce = pdata[29] )
If it work , also it works on windows 10 x64 ? .
August 26, 2016, 12:01:52 AM
SP - your M7 miner not working correctly - when you planned fix it ?
I tried to build Pallas github for windows with cuda 6.5, and it doesn't seem to work. (90% rejects on the pool). Cuda 8.0 was crashing when compiling.I don't have time to look at it now.
I have investigated some more.
The cuda 6.5 32bit build hash is failing when startnonce is bigger than 0x00ffffff. So if the highbyte is filled with something else than 0. It produce the wrong hash.
I enabled cpu mining verification.
reproducable in benchmark mode by adding this line on the top:
pdata[29] = 0x023445566;
( const uint32_t FirstNonce = pdata[29] )
Jump to: