CCminer(SP-MOD) Modded GPU kernels.

[5w00p]

Valid suggestions.

Tried the lowered clocks. No difference.

Release 44 works perfectly. But 45 and 46 just plain suck on my machine.

I should tell you that the PSU is fine. I have even given the GPU it's own 550W PSU.
And the GPU is watercooled. Doesn't go over ~42C under load.

[1714939108]

1714939108

[ryen123]

Release 46 doesn't perform well on my machine, (GTX 970 / Win 8.1 x64) mining Quark algo, no settings.

The power consumption is a sawtooth waveform. It loads the card to ~120% (overclocked) and then down to basically 0%. Over and over. Up down, up down, up down.

Mining XMR the power consumption might fluctuate a little, and might not load all the way near 100%, but it is at least never below 60%.

I don't know why this happens and I know sp_ works hard on it and most you guys are on linux, but it sucks for me.

You lowered the power limit back to 100%?

[5w00p]

(~42C was mining XMR, quark algo pushing the temp to ~50C)


I am getting ~15Mh/sec reported mining quark algo with release 44 on GTX 970 @ 1496MHz/3505Mhz gpu/mem.

No parameters set.

[sp_]

(~42C was mining XMR, quark algo pushing the temp to ~50C)
I am getting ~15Mh/sec reported mining quark algo with release 44 on GTX 970 @ 1496MHz/3505Mhz gpu/mem.
No parameters set.

Did you try the 3 exe files in the archieve(release 46)?

ccminer50.exe
ccminer52.exe
ccminer.exe

ccminer52.exe should be best for your card.

[chrysophylax]

This is the best release so far. Single thread on 750Ti is rock solid at 6766 kh/s [quark], showing the same and higher accepted hashrate stats on westhash. With -g 4 -i 22 comes close to single thread performance but not consistent.

wow - you are getting almost 1MH per card more than we are ... on stock clocks? ... how? ...

#crysx

[Star65]

I have the same issue as 5w00p on both 750ti's (Asus and Palit). PSU 1000W. The same on Quark and x11, on ccminer-1.6.2-git-tpruvot too,
but on dmd-gr algo all ok!

[tbearhere]

Yeah, when I start quark mining with -g my speed is less (28M for 2x970), compared to 31M when starting without any settings (only -r and -R). release 46

Thats why I never ues those settings ...I just overclock.

[fenomenhaa]

This is the best release so far. Single thread on 750Ti is rock solid at 6766 kh/s [quark], showing the same and higher accepted hashrate stats on westhash. With -g 4 -i 22 comes close to single thread performance but not consistent.

wow - you are getting almost 1MH per card more than we are ... on stock clocks? ... how? ...

#crysx

I think with moded bios, i moded my asus's bios too, but power comsp. increase %10-15,about hashrate just increase the  %3-4 .I decided to return stock bios. 1400 core clock gives 6600 on quark

[ryen123]

This is the best release so far. Single thread on 750Ti is rock solid at 6766 kh/s [quark], showing the same and higher accepted hashrate stats on westhash. With -g 4 -i 22 comes close to single thread performance but not consistent.

wow - you are getting almost 1MH per card more than we are ... on stock clocks? ... how? ...

#crysx

Mine is EVGA 750TI FTW ACX @ 1404 (Boost) + 3198 Mem

[tbearhere]

This is the best release so far. Single thread on 750Ti is rock solid at 6766 kh/s [quark], showing the same and higher accepted hashrate stats on westhash. With -g 4 -i 22 comes close to single thread performance but not consistent.

wow - you are getting almost 1MH per card more than we are ... on stock clocks? ... how? ...

#crysx

Mine is EVGA 750TI FTW ACX @ 1404 (Boost) + 3198 Mem

ryen123
Are your 750ti's stable at 1404 boost ..I mean do you have any crashes at all? Mine love 1350 boost  that's their sweet spot.

[ltc_bilic]

I have the same issue as 5w00p on both 750ti's (Asus and Palit). PSU 1000W. The same on Quark and x11, on ccminer-1.6.2-git-tpruvot too,
but on dmd-gr algo all ok!

I think the code has drastically changed regarding the load/power consumption factors....Here are my observations about it:
- I always run binary releases from sp directly,
- on quark my miner of choice was old sp mod 3.1 release,
- have 1 comp for testing and two rigs 6x750 ti gigabytes all OC,
- 3.1 release gave me best hashrate until latest release 4.6 which gave significant boost,
- look at graph recorded with nvidia inspector, running different miners on same machine,
- each miner was running quark for about one minute,...didn't count exactly
- but pay attention to gpu load/power consumption:


gifs upload

- see how the graph on 3.1 miner is almost flat!...while newest release (4.6) is quite an improvement regarding the older (4.4)

[ryen123]

ryen123
Are your 750ti's stable at 1404 boost ..I mean do you have any crashes at all? Mine love 1350 boost  that's their sweet spot.

Yeah it's stable. Will crash if +10 more on the core. So its the limit for mine.

[chrysophylax]

This is the best release so far. Single thread on 750Ti is rock solid at 6766 kh/s [quark], showing the same and higher accepted hashrate stats on westhash. With -g 4 -i 22 comes close to single thread performance but not consistent.

wow - you are getting almost 1MH per card more than we are ... on stock clocks? ... how? ...

#crysx

I think with moded bios, i moded my asus's bios too, but power comsp. increase %10-15,about hashrate just increase the  %3-4 .I decided to return stock bios. 1400 core clock gives 6600 on quark

well that explains it then ...

...

#crysx

[chrysophylax]

This is the best release so far. Single thread on 750Ti is rock solid at 6766 kh/s [quark], showing the same and higher accepted hashrate stats on westhash. With -g 4 -i 22 comes close to single thread performance but not consistent.

wow - you are getting almost 1MH per card more than we are ... on stock clocks? ... how? ...

#crysx

Mine is EVGA 750TI FTW ACX @ 1404 (Boost) + 3198 Mem

nice boost - and stable? ...

edit - never mind ... read further and you already answered that ...

#crysx

[Slava_K]

On last commit i see speed down on quark 5%...

[go6ooo1212]

Personally I can't gain improvement on qubit with r.46 instead of 45. I also can't use specified compute 50 or 52 , because I have both 750ti and 970 on my rig...

[Grim]

You guys all use the latest driver?

[5w00p]

You guys all use the latest driver?

I do: 350.12 WHQL Win 8.1 x64

[chrysophylax]

You guys all use the latest driver?

the latest in the rpmfusion repo ...

#crysx

[sp_]

I've managed to optimize skein 10% faster.

edit:

submitted to github.

Skein is now the 2nd most profitable algo to mine on maxwell

edit:

10% on the 750ti
22% on the gtx 970 (from 192MHASH to 235MHASH standard clocks)