Topic: CCminer(SP-MOD) Modded NVIDIA Maxwell / Pascal kernels. - page 1219. (Read 2347588 times)
yeah, looks like something has broken in my latest @github. I get some does not validate on the cpu. They are rare @x11. But they are present on standard clocks.. Need to investigate..
ERROR RATE:
Since adopting the latest, improved, ccminer, the error rate has gone from 15-20% rejects to less than 1% rejects. Currently, my rigs run the latest git for 1.5.0, tpruvot's version. Hash rates have improved significantly, also. -scryptr
Since adopting the latest, improved, ccminer, the error rate has gone from 15-20% rejects to less than 1% rejects. Currently, my rigs run the latest git for 1.5.0, tpruvot's version. Hash rates have improved significantly, also. -scryptr
Release 10 works perfectly. No problems whatsoever.
Hashrates (750Ti O/C)
X11: 3000 kH/s
X13: 2320 kH/s
X15: 2000 kH/s
Hashrates (750Ti O/C)
X11: 3000 kH/s
X13: 2320 kH/s
X15: 2000 kH/s
What are your clocks with the 750Ti?
I can get only 2950-2970KH/s with 1246MHz Core / 1580MHz Memory
1210MHz Core / 1600MHz Memory
dont press ctrl c 
else its this kind of small problems i try to fix for my final v1.5.0, i reinstalled ODBG to be able to trace that ... I stop merging SP and Klaus improvements for the moment (except bugs)
I want to make a "final" api 1.2 and a stable base for others, and to be able to correctly compare our changes.. djm34 seems to be back on the scene
else its this kind of small problems i try to fix for my final v1.5.0, i reinstalled ODBG to be able to trace that ... I stop merging SP and Klaus improvements for the moment (except bugs)I want to make a "final" api 1.2 and a stable base for others, and to be able to correctly compare our changes.. djm34 seems to be back on the scene
It looks like the ostrich putting its head in the ground technic to avoid looking at the problem...
dont press ctrl c else its this kind of small problems i try to fix for my final v1.5.0, i reinstalled ODBG to be able to trace that ... I stop merging SP and Klaus improvements for the moment (except bugs)
I want to make a "final" api 1.2 and a stable base for others, and to be able to correctly compare our changes.. djm34 seems to be back on the scene
else its this kind of small problems i try to fix for my final v1.5.0, i reinstalled ODBG to be able to trace that ... I stop merging SP and Klaus improvements for the moment (except bugs)I want to make a "final" api 1.2 and a stable base for others, and to be able to correctly compare our changes.. djm34 seems to be back on the scene
ccminer SP_ release 11 with 750ti average (out of 6 cards):
X11: 2870 Khash
X13: 2280 Khash
The Ctrl+C bug is still present.
I also noticed that while mining at NiceHash I now see a green tick below the extranonce subscription column. Previously it was a red cross. Is this related to the optimisation of ccminer?
X11: 2870 Khash
X13: 2280 Khash
The Ctrl+C bug is still present.
I also noticed that while mining at NiceHash I now see a green tick below the extranonce subscription column. Previously it was a red cross. Is this related to the optimisation of ccminer?
polanskiman--
The hash improvements on X11, X13, Xetc, are due to DJM34, sp_, Epsylon3, and tsiv. The extranonce compliance and several non-kernel functional improvements are due to Epsylon3; his focus appears to be more on improving control functions.
What do you mean by the "control-c bug"? -scryptr
DJM34 was the latest dev to have worked and produced a stable release with significant improvements. FYI the very latest improvement of ccminer was done by SP_ (~ +200 Khash) - Still beta as I understand it though.
In order to quit ccminer you can proceed by pressing Ctrl+c. In SP_'s release of ccminer there seems to be a bug where Ctrl+c will lead to the crashing of ccminer.
This happens:
Then pressing Y or N produces the same result: ccminer closes
I never EVER had this problem with DJM34's release.
ccminer SP_ release 11 with 750ti average (out of 6 cards):
X11: 2870 Khash
X13: 2280 Khash
The Ctrl+C bug is still present.
I also noticed that while mining at NiceHash I now see a green tick below the extranonce subscription column. Previously it was a red cross. Is this related to the optimisation of ccminer?
X11: 2870 Khash
X13: 2280 Khash
The Ctrl+C bug is still present.
I also noticed that while mining at NiceHash I now see a green tick below the extranonce subscription column. Previously it was a red cross. Is this related to the optimisation of ccminer?
polanskiman--
The hash improvements on X11, X13, Xetc, are due to DJM34, sp_, Epsylon3, and tsiv. The extranonce compliance and several non-kernel functional improvements are due to Epsylon3; his focus appears to be more on improving control functions.
What do you mean by the "control-c bug"? -scryptr
DJM34 was the latest dev to have worked and produced a stable release with significant improvements. FYI the very latest improvement of ccminer was done by SP_ (~ +200 Khash) - Still beta as I understand it though.
In order to quit ccminer you can proceed by pressing Ctrl+c. In SP_'s release of ccminer there seems to be a bug where Ctrl+c will lead to the crashing of ccminer.
This happens:
ccminer SP_ release 11 with 750ti average (out of 6 cards):
X11: 2870 Khash
X13: 2280 Khash
The Ctrl+C bug is still present.
I also noticed that while mining at NiceHash I now see a green tick below the extranonce subscription column. Previously it was a red cross. Is this related to the optimisation of ccminer?
X11: 2870 Khash
X13: 2280 Khash
The Ctrl+C bug is still present.
I also noticed that while mining at NiceHash I now see a green tick below the extranonce subscription column. Previously it was a red cross. Is this related to the optimisation of ccminer?
polanskiman--
The hash improvements on X11, X13, Xetc, are due to DJM34, sp_, Epsylon3, and tsiv. The extranonce compliance and several non-kernel functional improvements are due to Epsylon3; his focus appears to be more on improving control functions.
What do you mean by the "control-c bug"? -scryptr
I am referring to SP_'s release 11 of ccminer. I know the devs who contributed to ccminer... Thanks for the lesson though
DJM34 was the latest dev to have worked and produced a stable release with significant improvements. FYI the very latest improvement of ccminer was done by SP_ (~ +200 Khash) - Still beta as I understand it though.
In order to quit ccminer you can proceed by pressing Ctrl+c. In SP_'s release of ccminer there seems to be a bug where Ctrl+c will lead to the crashing of ccminer.
ccminer SP_ release 11 with 750ti average (out of 6 cards):
X11: 2870 Khash
X13: 2280 Khash
The Ctrl+C bug is still present.
I also noticed that while mining at NiceHash I now see a green tick below the extranonce subscription column. Previously it was a red cross. Is this related to the optimisation of ccminer?
X11: 2870 Khash
X13: 2280 Khash
The Ctrl+C bug is still present.
I also noticed that while mining at NiceHash I now see a green tick below the extranonce subscription column. Previously it was a red cross. Is this related to the optimisation of ccminer?
polanskiman--
The hash improvements on X11, X13, Xetc, are due to DJM34, sp_, Epsylon3, and tsiv. The extranonce compliance and several non-kernel functional improvements are due to Epsylon3; his focus appears to be more on improving control functions.
What do you mean by the "control-c bug"? -scryptr
ccminer SP_ release 11 with 750ti average (out of 6 cards):
X11: 2870 Khash
X13: 2280 Khash
The Ctrl+C bug is still present.
I also noticed that while mining at NiceHash I now see a green tick below the extranonce subscription column. Previously it was a red cross. Is this related to the optimisation of ccminer?
X11: 2870 Khash
X13: 2280 Khash
The Ctrl+C bug is still present.
I also noticed that while mining at NiceHash I now see a green tick below the extranonce subscription column. Previously it was a red cross. Is this related to the optimisation of ccminer?
Moved the nist5 final test into the skein, removed a kernalcall. Small nist5 boost. 100-200KHASH. (750ti)
Prices are at all time low, but to increase profit you need to mine with a algoswitcher.
And if you live in a cold country like me, you can use the heat to heat your house for free.
And if you live in a cold country like me, you can use the heat to heat your house for free.
cpu(system): 30 watt
6x750ti: 40-50 watt each
x11: 3000 x 6 = 18000 khash
you lose about 5% to 10% khash due to stale, rejected, new work ...
(don't forget all the fees until you have fiat in your hands)
so in reality its something like:
16500 khash for 300 watt (0.15$)
put these values into what to mine:
best x11: was MINUS 0.25 cent each day
darkcoin: MINUS 0.46 cent each day
6x750ti: 40-50 watt each
x11: 3000 x 6 = 18000 khash
you lose about 5% to 10% khash due to stale, rejected, new work ...
(don't forget all the fees until you have fiat in your hands)
so in reality its something like:
16500 khash for 300 watt (0.15$)
put these values into what to mine:
best x11: was MINUS 0.25 cent each day
darkcoin: MINUS 0.46 cent each day
On github klaus_T has commited some improvements in BMW and shavite, need to check them out and merge.
Oops. Looks like I have added a bug in shavite.I see that you are working. Nice.
Please add a pull request when you are done so I can add all the improvements into mine.
Decimal intensity is nice. I believe we can tweak all cards with this setting.
My gainward(750ti) was doing +30KHASH after the setting the intensity between 19 and 20. I think around 19.525 is the optimal setting for this card on x11.
My gainward(750ti) was doing +30KHASH after the setting the intensity between 19 and 20. I think around 19.525 is the optimal setting for this card on x11.
i added support for decimal intensities :p
On github klaus_T has commited some improvements in BMW and shavite, need to check them out and merge.
Oops. Looks like I have added a bug in shavite.
The maxwell cards use very little power, so it is still possible to make a few bucks with the latest optimalizations. With overclocking the release 11 will do 3MHASH@40-50 watt. a 280x will make 4MHASH @ 150 watt with the badman mod(stolen kernal opensource). I think I can push the 750ti up to 3.5 or perhaps 4. Only half of the kernals are improved, and there are still some easy pickings. On github klaus_T has commited some improvements in BMW and shavite, need to check them out and merge. I also want to work on Quark. Seems one of the most profitable to mine at the moment.
You did say you would stop publishing when you got to 10MH/s on 980, remember?
After I went opensource, I have got 0.01 BTC in donation from one miner. I have released a miner that is 30% faster on the 980, and get nothing in return. Come on guys, Donate some Beers at least ?
I picked up a couple of 280x now from a tired miner. Got them for $150 a piece. Time to move on to OpenCL.
The sad reality is that miners have to ROI as well and even with your miner the profitability is just not there - and I hate to say it, but - open source miners spread like wildfire so nobody is really ahead of anyone. I only tested your miner in case I'll need them but I try to avoid the X-series. I don't like the weird massive hashrate spikes, the 3-3.5GH/s rentable "rigs" and the incoming ASICs. And when you have farms that are being retired every day, people won't really feel like donating.
I got 2 DRK's Yesterday from someone. Thanks
in 5 years they might be worth $1000 each.
in 5 years they might be worth $1000 each.
...so I sent you 4000$ in drk before you gone open source... yeah, generous me
joke aside. I did not forget your project and will send out some drk again.
I'm not the x11 mining guy but appreciate all the development that keeps our hobby alive.
And it is really great to see that you and epsylon complement each other.
True, its not easy to donate much in times of thin profit (mostly none in my case of fucking high energy costs)... but if everybody would donate a tiny little bit, everything would be fine for both sides.
It should be clear for everybody that no donation means no more development on the long run.
Its an easy equation: Why should somebody investing hours of work part his development for nothing and lose his own advantage?
Only for 5 guys saying some warm thank you and only two of them donating while 500 greedy others just using the source?
I'm not sure if I would be that generous.
Jump to: