I have now fixed the groestl speedup 2 on github. Klaus_t has been optimizing quark, so I am waiting for a pull request, and then build release 31.
Topic: CCminer(SP-MOD) Modded NVIDIA Maxwell / Pascal kernels. - page 1185. (Read 2347597 times)
Quark:
v28= ~11,160 kh/s
v30= ~11,441 kh/s
x11:
v28= ~5,730 kh/s
v30= ~5,850 kh/s
lyra2:
v28= ~1,350 kh/s
v30= ~1,350 kh/s
(2 750Ti's, no overclock)
v28= ~11,160 kh/s
v30= ~11,441 kh/s
x11:
v28= ~5,730 kh/s
v30= ~5,850 kh/s
lyra2:
v28= ~1,350 kh/s
v30= ~1,350 kh/s
(2 750Ti's, no overclock)
Try groestl or diamondgroestl.
You can see on the commit on github that this is groestl speedup part 1. I have part 2 soon ready for checkin, but it is currently mixing the bits, and produce wrong results. Lyra is not using the bitslice groestl (killer groestl)
so my improvements will not have an effect. But I guess if I swap the implementation, Lyra2 will get a boost as well. Did you try that DJM34?
About the spreadcoinminer:
I will integrate and optimize TSIV's spreadcoin implementation into the latest fork of ccminer. I will send out beta versions for a fee of 0.1 BTC. The beta will be a windows executable.
I might send out more than one exe in the testing phase, if I manage to optimize more.
I will publish the sourcecode after one month of betatesting. (When I publish the sourcecode, the spreadcoin will spread bether, and secure the coin)
The current speed will be announced when the exefile is released. Hopefully next weekend 25-26 january.
The estimated speed increase is 30-40% on the 980 cards
If you want a early seat, you can start donating to my BTC or DRK adress in my signature.The current speed will be announced when the exefile is released. Hopefully next weekend 25-26 january.
For the SpreadX11 opt. miner beta:
d1bf1798d350a590b056758975cae0a9cc3a724a083c48cc3d26f05e03758c39
Just waiting to get it and compensate my 2x750Ti lost.
The latest checkin breaks the hash. So please wait. Tried to solve it, but went to the pub instead cheers
cheershehehe - all good - still mining for a few more beers for you mate ...
have you checked how its going with yaamp / quark? ... happy with what you see with the latest build and hashrates? ...
njoi ...
#crysx
The latest checkin breaks the hash. So please wait. Tried to solve it, but went to the pub instead cheers
cheers I just submitted groestl speedup part 2. quark is up another 30KHASH. x11 around 15-20KHASH. x13 -10KHASH
Also submitted 7KHASH more in x14 making it a total for x14 around 16KHASH boost.
All on the 750ti.
edit: perhaps a bit early checkin groestl coin is broken :/
Also submitted 7KHASH more in x14 making it a total for x14 around 16KHASH boost.
All on the 750ti.
edit: perhaps a bit early checkin groestl coin is broken :/
sp - the miner seems to running stable since i set them mining quark last night ...
is it wise to stay mining with the previous compile - or compile a new one with these increases and restart the quark mining donation?
im just worried about stability at the moment - thats all ...
what do you think?
#crysx
I just submitted groestl speedup part 2. quark is up another 30KHASH. x11 around 15-20KHASH. x13 -10KHASH
Also submitted 7KHASH more in x14 making it a total for x14 around 16KHASH boost.
All on the 750ti.
edit: perhaps a bit early checkin groestl coin is broken :/
Also submitted 7KHASH more in x14 making it a total for x14 around 16KHASH boost.
All on the 750ti.
edit: perhaps a bit early checkin groestl coin is broken :/
SP_ v30-
Both rigs increased their hash rates by about 900kh/s each mining quark. --scryptr
EDIT: Both rigs still running steady, nearing 1000 accepts each at about 1% error rate. --scryptr
Both rigs increased their hash rates by about 900kh/s each mining quark. --scryptr
EDIT: Both rigs still running steady, nearing 1000 accepts each at about 1% error rate. --scryptr
anyone knows how, with visual studio, to get the "release" directory with less crap in it (I still need the ptx though...) ?
Shift-delete the crap
Euh.. Nuke it?
anyone knows how, with visual studio, to get the "release" directory with less crap in it (I still need the ptx though...) ?
Shift-delete the crap
anyone knows how, with visual studio, to get the "release" directory with less crap in it (I still need the ptx though...) ?
Shift-delete the crap
anyone knows how, with visual studio, to get the "release" directory with less crap in it (I still need the ptx though...) ?
Quark:
v28= ~11,160 kH/s
v30= ~11,441 kH/s
x11:
v28= ~5,730 kH/s
v30= ~5,850 kH/s
lyra2:
v28= ~1,350 kH/s
v30= ~1,350 kH/s
Mjollnir:
v30=~21,000 MH/s
(2 750Ti's, no overclock)
v28= ~11,160 kH/s
v30= ~11,441 kH/s
x11:
v28= ~5,730 kH/s
v30= ~5,850 kH/s
lyra2:
v28= ~1,350 kH/s
v30= ~1,350 kH/s
Mjollnir:
v30=~21,000 MH/s
(2 750Ti's, no overclock)
The new gtx 960 might fail on the default settings because the intensity is set to high for compute 5.2 devices. How much memory are they planning to ship with the new cards?
The fix will be to set the intensity manually with the -i parameter. f.eks -i 19
The fix will be to set the intensity manually with the -i parameter. f.eks -i 19
For people with mixed cards (or thinking about mixing in a 960) in their rigs, the -i parameter is not a per card setting, so they'll either need to run seperate ccminer instances or hope the same -i parameter works across cards
. Just something for people to think about. Quark:
v28= ~11,160 kh/s
v30= ~11,441 kh/s
x11:
v28= ~5,730 kh/s
v30= ~5,850 kh/s
lyra2:
v28= ~1,350 kh/s
v30= ~1,350 kh/s
(2 750Ti's, no overclock)
v28= ~11,160 kh/s
v30= ~11,441 kh/s
x11:
v28= ~5,730 kh/s
v30= ~5,850 kh/s
lyra2:
v28= ~1,350 kh/s
v30= ~1,350 kh/s
(2 750Ti's, no overclock)
Try groestl or diamondgroestl.
You can see on the commit on github that this is groestl speedup part 1. I have part 2 soon ready for checkin, but it is currently mixing the bits, and produce wrong results. Lyra is not using the bitslice groestl (killer groestl)
so my improvements will not have an effect. But I guess if I swap the implementation, Lyra2 will get a boost as well. Did you try that DJM34?
I realized Lyra2 wasn't part of the speed up, it just happened that the algo was profitable for like five minutes on yaamp, and I happen to be in front of my computer while it happened. Since I hadn't posted any Lyra2 rates in a while (ever?) thought I would do it, just in case someone was interested.
Quark:
v28= ~11,160 kh/s
v30= ~11,441 kh/s
x11:
v28= ~5,730 kh/s
v30= ~5,850 kh/s
lyra2:
v28= ~1,350 kh/s
v30= ~1,350 kh/s
(2 750Ti's, no overclock)
v28= ~11,160 kh/s
v30= ~11,441 kh/s
x11:
v28= ~5,730 kh/s
v30= ~5,850 kh/s
lyra2:
v28= ~1,350 kh/s
v30= ~1,350 kh/s
(2 750Ti's, no overclock)
Try groestl or diamondgroestl.
You can see on the commit on github that this is groestl speedup part 1. I have part 2 soon ready for checkin, but it is currently mixing the bits, and produce wrong results. Lyra is not using the bitslice groestl (killer groestl)
so my improvements will not have an effect. But I guess if I swap the implementation, Lyra2 will get a boost as well. Did you try that DJM34?
The new gtx 960 might fail on the default settings because the intensity is set to high for compute 5.2 devices. How much memory are they planning to ship with the new cards?
The fix will be to set the intensity manually with the -i parameter. f.eks -i 19
The fix will be to set the intensity manually with the -i parameter. f.eks -i 19
Quark:
v28= ~11,160 kh/s
v30= ~11,441 kh/s
x11:
v28= ~5,730 kh/s
v30= ~5,850 kh/s
lyra2:
v28= ~1,350 kh/s
v30= ~1,350 kh/s
(2 750Ti's, no overclock)
v28= ~11,160 kh/s
v30= ~11,441 kh/s
x11:
v28= ~5,730 kh/s
v30= ~5,850 kh/s
lyra2:
v28= ~1,350 kh/s
v30= ~1,350 kh/s
(2 750Ti's, no overclock)
Try groestl or diamondgroestl.
You can see on the commit on github that this is groestl speedup part 1. I have part 2 soon ready for checkin, but it is currently mixing the bits, and produce wrong results. Lyra is not using the bitslice groestl (killer groestl)
so my improvements will not have an effect. But I guess if I swap the implementation, Lyra2 will get a boost as well. Did you try that DJM34?
About the spreadcoinminer:
I will integrate and optimize TSIV's spreadcoin implementation into the latest fork of ccminer. I will send out beta versions for a fee of 0.1 BTC. The beta will be a windows executable.
I might send out more than one exe in the testing phase, if I manage to optimize more.
I will publish the sourcecode after one month of betatesting. (When I publish the sourcecode, the spreadcoin will spread bether, and secure the coin)
The current speed will be announced when the exefile is released. Hopefully next weekend 25-26 january.
The estimated speed increase is 30-40% on the 980 cards
If you want a early seat, you can start donating to my BTC or DRK adress in my signature.The current speed will be announced when the exefile is released. Hopefully next weekend 25-26 january.
any possibility of a linux x64 version for beta as well sp? ...
ill donate ( and i assume most will ) for the cause of improving the spreadcoin miner ... but a windows version is useless to me ...
anyway around that - or are you just keeping for the windows based systems? ...
it just simply means that 30 days of no testing on my end ...
#crysx
Looks like pool rejects are higher making #30 less efficient then #29. quark
how many rejects and which pool? I was on yaamp for 1.5 hours, with one reject ("reject reason: Job not found" right after a block change), which seemed fine to me.
wrong its good ....great
sorry Quark:
v28= ~11,160 kh/s
v30= ~11,441 kh/s
x11:
v28= ~5,730 kh/s
v30= ~5,850 kh/s
lyra2:
v28= ~1,350 kh/s
v30= ~1,350 kh/s
(2 750Ti's, no overclock)
v28= ~11,160 kh/s
v30= ~11,441 kh/s
x11:
v28= ~5,730 kh/s
v30= ~5,850 kh/s
lyra2:
v28= ~1,350 kh/s
v30= ~1,350 kh/s
(2 750Ti's, no overclock)
Try groestl or diamondgroestl.
You can see on the commit on github that this is groestl speedup part 1. I have part 2 soon ready for checkin, but it is currently mixing the bits, and produce wrong results. Lyra is not using the bitslice groestl (killer groestl)
so my improvements will not have an effect. But I guess if I swap the implementation, Lyra2 will get a boost as well. Did you try that DJM34?
About the spreadcoinminer:
I will integrate and optimize TSIV's spreadcoin implementation into the latest fork of ccminer. I will send out beta versions for a fee of 0.1 BTC. The beta will be a windows executable.
I might send out more than one exe in the testing phase, if I manage to optimize more.
I will publish the sourcecode after one month of betatesting. (When I publish the sourcecode, the spreadcoin will spread bether, and secure the coin)
The current speed will be announced when the exefile is released. Hopefully next weekend 25-26 january.
The estimated speed increase is 30-40% on the 980 cards
If you want a early seat, you can start donating to my BTC or DRK adress in my signature.The current speed will be announced when the exefile is released. Hopefully next weekend 25-26 january.
Looks like pool rejects are higher making #30 less efficient then #29. quark
how many rejects and which pool? I was on yaamp for 1.5 hours, with one reject ("reject reason: Job not found" right after a block change), which seemed fine to me.
Jump to: