CCminer(SP-MOD) Modded GPU kernels.

[sp_]

Seems to be a bug in the extranounce code for multiple threads. The hashrate is dropping after the pool is changing the extranonce

[CapnBDL]

Seems to be a bug in the extranounce code for multiple threads. The hashrate is dropping after the pool is changing the extranonce

I'm seeing the same. I've got a couple of hours to play, so I'll try some different switches and see if it gets better.

edit; did you see my post below? Kidda gets me when it runs for so long, /w no determined timing, and pulls the fault. Never had that happen before. I did add some things to the network, but it should handle it. Just strange!!

[sp_]

I have fixed it now. I have made a new switch. '-e' that will switch of the extranonce subscription.
If the 'g' switch is on '-e' will automaticly be activated...

This is not a bugfix, but just a workaround.

[sp_]

New world record for quark on the 750ti

yaamp.com:4033 quark block 1621936
GPU #1: GeForce GTX 750 Ti, 6992
GPU #0: GeForce GTX 750 Ti, 9895
GPU #0: GeForce GTX 750 Ti, 8825
GPU #1: GeForce GTX 750 Ti, 6487


-g 5 -i 22 standard clocks.

[rednoW]

New world record for quark on the 750ti

yaamp.com:4033 quark block 1621936
GPU #1: GeForce GTX 750 Ti, 6992
GPU #0: GeForce GTX 750 Ti, 9895
GPU #0: GeForce GTX 750 Ti, 8825
GPU #1: GeForce GTX 750 Ti, 6487


-g 5 -i 22 standard clocks.

too good to be true

[sp_]

New world record for quark on the 750ti
yaamp.com:4033 quark block 1621936
GPU #1: GeForce GTX 750 Ti, 6992
GPU #0: GeForce GTX 750 Ti, 9895
GPU #0: GeForce GTX 750 Ti, 8825
GPU #1: GeForce GTX 750 Ti, 6487
-g 5 -i 22 standard clocks.

too good to be true

Build and try. Seems to be around +0.5MHASH pr 750ti average on the pool.

Nicehash is currently hashing 16GHASH of quark. At one point the quark hashrate was higher than x11.

[CapnBDL]

I have fixed it now. I have made a new switch. '-e' that will switch of the extranonce subscription.
If the 'g' switch is on '-e' will automaticly be activated...

This is not a bugfix, but just a workaround.

What does -e do and if I'm using -g I don't need it or what?

[rednoW]

Build and try. Seems to be around +0.5MHASH pr 750ti average on the pool.

Nicehash is currently hashing 16GHASH of quark. At one point the quark hashrate was higher than x11.

I do ... but build process takes some time
Will see ... new charts on nicehash will be helpful

[CapnBDL]

I have fixed it now. I have made a new switch. '-e' that will switch of the extranonce subscription.
If the 'g' switch is on '-e' will automaticly be activated...

This is not a bugfix, but just a workaround.

What does -e do and if I'm using -g I don't need it or what?

Also ...is it uploaded to git?

[sp_]

Submitted some more fixes. Less rejects on the pool.

I will not build release 50 yet. I need to fix the bugs first.

[sp_]

You can see the list of commits here:

https://github.com/sp-hash/ccminer/commits/windows

If you want to run the latest beta you need to build it yourself.

[rednoW]

Also ...is it uploaded to git?

yes, latest sources from git show no extranonce subscribe support on the pool when using -g parameter

[CapnBDL]

You can see the list of commits here:

https://github.com/sp-hash/ccminer/commits/windows

If you want to run the latest beta you need to build it yourself.

@sp-

Aww Man....I don't do code...compile...etc. But I can run the dog-snot outa a Win .exe.

edit; Ohh..... I've got some x11 shares comin your way. That is your .bat isn't it?
                  (hmmm...I'm impressed. Consistent 33s kh)
edit2; Aww...hell...sorry sp, meant to send those shares to you. Thought it was your x11.bat. Somebody got some shares!!!

[scryptr]

New world record for quark on the 750ti
yaamp.com:4033 quark block 1621936
GPU #1: GeForce GTX 750 Ti, 6992
GPU #0: GeForce GTX 750 Ti, 9895
GPU #0: GeForce GTX 750 Ti, 8825
GPU #1: GeForce GTX 750 Ti, 6487
-g 5 -i 22 standard clocks.

too good to be true

Build and try. Seems to be around +0.5MHASH pr 750ti average on the pool.

Nicehash is currently hashing 16GHASH of quark. At one point the quark hashrate was higher than x11.

COMMIT 788--

I cloned git and built 788 10 minutes ago.  Mining Quark with "-g 5 -i 22", the hash rates appear realistic, ranging from 5.7Mh/s to 6.6Mh/s per card, with variation.  30 mining threads are initiated at launch!  Build 782 produced a steady 6.3Mh/s to 6.4Mh/s per card, single threaded.  However my rig is getting about 37.3Mh/s with this configuration, instead of 38.4Mh/s with build 782, single-threaded.  Acceptance rate is about 90% for the 6x750ti FTW rig running build 788.

Mining single-threaded with build 788 brought the rig back to about 38.2Mh/s, but reaching full speed seems to take more time, as observed with build 782.  Acceptance at the pool is about 95%.

If the miner is started with the "-g" switch, extranonce subscription is turned off.  This is verified at the pool.       --scryptr

[rednoW]

For me no luck with -g again. Running without parameters gives significantly better results. But miner shows realistic rate with -g now, it's good.

[sp_]

Turning off extranonce is not a good fix to the problem. Connection is reset all the time and we get a higher reject rate. I need to insert code that reset's the gpu's when a extranode request is changed. The current gpu code needs all the threads to use the same extranodesize.

For the 750 1mb -g 5 -22 will not work as it needs 1,8 GIG of memory.

[antonio8]

@sp_

Just finally ran the re-compiled version for lyra2 and the rejected rate looks to be a lot better.

[rednoW]

It looks like last commit has 10% rejects "job not found" even running without -g with extranonce subscribe enabled. nicehash ...

[Grout]

It looks like last commit has 10% rejects "job not found" even running without -g with extranonce subscribe enabled. nicehash ...

I compiled commit d064c7 (on linux) and I see no performance difference on Quark compared to yesterday's version (didn't write down the commit, sorry), without specifying -g 5 -i 22. 11380kH/s average for my 2 750Ti.
When I specify -g 5 -22, I see no performance improvement, but I get about 10-15% reject: Job not found errors on Nicehash.

[chrysophylax]

will get testing done later this afternoon and see how it goes with quark ...

testing will be done on westhash - stock clocks - single miner - one on donation address driectly - not through link ( as the link forces extranonce subscription which is what it should be - otherwise we get a higher rejection share rate ) and one on the donation link stratum+tcp://donate.granitecoinx.com:7919/ ...

can setup up two machines that are the same build - and run one with -g -i parameters while the other runs just the -i parameter ...

what are the best parameters for either? ...

i need half a day before we can start compile / testing ...

can we clone git and build it from git sp? ...

edit - pushed one worker through the donation link with standard -r -R settings ( no -i ) because it was just easy and took 30 seconds to do ...

ill have the other worker setup later today with the -g settings and see how they compare ...

let me know if there is any other parameters that need to be set on the -g worker ...

btw - this is on yaamp ...

tanx ...

#crysx