CCminer(SP-MOD) Modded GPU kernels.

[theotherme]

considering, I never worked on quark, your comparison is irrelevant as usual

Your Lyra2v2 kernal is slow on the gtx 950 and the gtx 960. I have improved it. It is faster on all the maxwell models. My modded kernals are 100-200KHASH bether on the 750tis'
Around 5%.
It's enough to prove that my work is faster.

you just tweaked kernel parameters (which were adjustable by the user in my version btw), so no there is no real difference...
(except the huge overclock on your cards)

No. I have done more. My kernel compiles down to 110 registers and no spillbytes. Yours is 213 regs. Do a filecompare and check for yourself.

doesn't make any difference, you should know it... as long as there is enough registers (and it is clearly not 213 reg either... but hey I am happy to keep your thread on top of the mining forum... so you can get donation... for a last time)

Typical troll. Disprove a argument > switch the topic. Do it over and over again till the person you're arguing with gives up.

I like how you talk about broken software and yours is, then you say 'it doesn't matter cause a earlier release wasn't broken'. Same thing could be said about SPs software when things break. But oh it's different because it's a different version of cuda? That's like having software that only works on XP when everyone is using W7 or W10, then stating it works on XP so everyone should stay on it. GG hypocrite

"Typical troll. Disprove a argument > switch the topic. Do it over and over again till the person you're arguing with gives up."
yep and you are an expert on that matter, as your little text prove it... 

[theotherme]

I just think there should be no such discussion.

People wonna donate to sp_? Do it.
People wonna pledge djm34? Do it.
People wonna buy private AMD kernels from me? Do it! :-D

I've done so much opensource that I'll never feel guilty for selling some software ;-)

well sure, you can always make a donation to me  (djm34... since I use my other name)
however don't do it as pledging for something, as I have no real plan to release anything
for the moment (will let you know if I decide to release something don't worry  )

djm34.

[becker]

So sad it appears the fellowship of rigs has broken, turning on each other to prove who has the ultimate kernel of power.

[sp_]

SP_
thank you for all your work and continued effort ... it is truly rare to find in this day and age!
donation sent

Thanks alot. Everytime I publish a opensource speedup, the private kernal developers are complaining.
I don't care about it. Let's leave it.

I just gave you 5% in the quark algo for free on compute 5.2 cards. And I've got some beers in donations.. Give me some more and I will optimize more. simple.

[theotherme]

So sad it appears the fellowship of rigs has broken, turning on each other to prove who has the ultimate kernel of power.

the fellowship in on #ccminer channel and is doing rather well (even though there isn't much activities except from epsylon3 and his cpuminer) and for info we asked several time sp to join in (and he just ignored us...   so...) but it is clear that he would like to rule us all (well... good luck with that  )

[bathrobehero]

Offtopic: I picked up few mini 970s which only have one 8 pin connector and got surprised by how loud these are.
These are about 10% slower and about 10-15 °C hotter (at 90% fan speed mind you) than the 3 fan Windforce edition cards (at 70% fan speed). And collectively, they sound like a blowdryer.
I knew these would be louder and slower but not this much. Thought it could be useful info for others.

[joblo]

Offtopic: I picked up few mini 970s which only have one 8 pin connector and got surprised by how loud these are.
These are about 10% slower and about 10-15 °C hotter (at 90% fan speed mind you) than the 3 fan Windforce edition cards (at 70% fan speed). And collectively, they sound like a blowdryer.
I knew these would be louder and slower but not this much. Thought it could be useful info for others.

Thanks for that. I had been eying this card as well but was concerned about heat given the compact size.
The EVGA 960 shorty looks interesting though for smaller enclosures.

[pokeytex]

Contemplating selling my (9) 750ti's, (3) 970's, and (1) 960 because I can't find anything worth mining anymore.  Is there some hidden gem I don't know about?

[antonio8]

Offtopic: I picked up few mini 970s which only have one 8 pin connector and got surprised by how loud these are.
These are about 10% slower and about 10-15 °C hotter (at 90% fan speed mind you) than the 3 fan Windforce edition cards (at 70% fan speed). And collectively, they sound like a blowdryer.
I knew these would be louder and slower but not this much. Thought it could be useful info for others.

Thanks for that. I had been eying this card as well but was concerned about heat given the compact size.
The EVGA 960 shorty looks interesting though for smaller enclosures.

Those are pretty decent and quite. they even make a 4GB version. Although I don't have them in an enclosure.

[bensam1231]

Offtopic: I picked up few mini 970s which only have one 8 pin connector and got surprised by how loud these are.
These are about 10% slower and about 10-15 °C hotter (at 90% fan speed mind you) than the 3 fan Windforce edition cards (at 70% fan speed). And collectively, they sound like a blowdryer.
I knew these would be louder and slower but not this much. Thought it could be useful info for others.

Yup, HTPC card... I would never buy one of those. I contemplated buying blowers because they move heat away from components better, but opted not to as it doesn't seem to matter much. Maybe if I had all my cards in a case that'd be different.

[chrysophylax]

Contemplating selling my (9) 750ti's, (3) 970's, and (1) 960 because I can't find anything worth mining anymore.  Is there some hidden gem I don't know about?

where are you located? ...

#crysx

[dominuspro]

Contemplating selling my (9) 750ti's, (3) 970's, and (1) 960 because I can't find anything worth mining anymore.  Is there some hidden gem I don't know about?

Only ETH in linux.

Just tested my rig with my old usb linux installation and it seems OK. Also the 750ti should be somewhere where it was. Getting 110MH with 980 +3x970 +960 +750ti. Power at the wall shows exactly 900W.

In windows I got 96MH without 750ti. But there is also a memory OC of all the cards 750ti=6300Mhz, 960=7500Mhz, others to 7700Mhz. I somehow succedeed to put the cards into the higher power state and OC also the memory. I couldn't find a way in windows for that.
 

[serg_25]

Contemplating selling my (9) 750ti's, (3) 970's, and (1) 960 because I can't find anything worth mining anymore.  Is there some hidden gem I don't know about?

Only ETH in linux.

Just tested my rig with my old usb linux installation and it seems OK. Also the 750ti should be somewhere where it was. Getting 110MH with 980 +3x970 +960 +750ti. Power at the wall shows exactly 900W.

In windows I got 96MH without 750ti. But there is also a memory OC of all the cards 750ti=6300Mhz, 960=7500Mhz, others to 7700Mhz. I somehow succedeed to put the cards into the higher power state and OC also the memory. I couldn't find a way in windows for that.
 

Hi dominuspro
I think you are wrong about 750Ti. I also tested this card a week ago, and can prove that TLB is definitely an issue. Yep, in benchmark mode i got 18 mh\s from 2 750Ti cards. but during benchmark DAG file is aroung 1G only, but if you will try real mining it will be more that 1G, as a result i got only 5mh\s from 2 cards on real pool.
But 980/970/960 still can mine well.
If you want to recheck my test - just remove all your hi-end cards and leave 750 Ti only in your rig for test.

P.S. let me know if you will have different result.
P.S.S that's on linux of course.

[pokeytex]

Contemplating selling my (9) 750ti's, (3) 970's, and (1) 960 because I can't find anything worth mining anymore.  Is there some hidden gem I don't know about?

where are you located? ...

#crysx

New Jersey, USA

[chrysophylax]

Contemplating selling my (9) 750ti's, (3) 970's, and (1) 960 because I can't find anything worth mining anymore.  Is there some hidden gem I don't know about?

where are you located? ...

#crysx

New Jersey, USA

kool ...

bit far for me to come and get ...

#crysx

[dominuspro]

Contemplating selling my (9) 750ti's, (3) 970's, and (1) 960 because I can't find anything worth mining anymore.  Is there some hidden gem I don't know about?

Only ETH in linux.

Just tested my rig with my old usb linux installation and it seems OK. Also the 750ti should be somewhere where it was. Getting 110MH with 980 +3x970 +960 +750ti. Power at the wall shows exactly 900W.

In windows I got 96MH without 750ti. But there is also a memory OC of all the cards 750ti=6300Mhz, 960=7500Mhz, others to 7700Mhz. I somehow succedeed to put the cards into the higher power state and OC also the memory. I couldn't find a way in windows for that.
 

Hi dominuspro
I think you are wrong about 750Ti. I also tested this card a week ago, and can prove that TLB is definitely an issue. Yep, in benchmark mode i got 18 mh\s from 2 750Ti cards. but during benchmark DAG file is aroung 1G only, but if you will try real mining it will be more that 1G, as a result i got only 5mh\s from 2 cards on real pool.
But 980/970/960 still can mine well.
If you want to recheck my test - just remove all your hi-end cards and leave 750 Ti only in your rig for test.


P.S. let me know if you will have different result.
P.S.S that's on linux of course.

You are right. I cannot enable only 750ti with cuda-devices param, but I can disable the 750ti and I loose 7MH. So 750ti is hashing around 7MH. On the same clocks(1426/6300) it used to be over 11MH some months ago. 960 with 2GB is not loosing that much. It is still around 11MH.

The 980+3x970 combo increased hashrate with mem oc from 86MH(Windows) to 92MH(Linux).

The power limit is still on the factory settings on all gpus.

A screenshot if someone finds it somehow usefull:

[sambiohazard]

A little something for the devs, they dont have nvidia miners for launch

https://bitcointalk.org/index.php?topic=1290358.0

Do read this too

https://bitcointalk.org/index.php?topic=1290358.msg13258018#msg13258018

or may be they already contacted you.

[ZeroFossilFuel]

Starting a couple of weeks ago one of my computers that I use for mining while it's otherwise idle suddenly started giving me an "out of memory" error. I presume it came on the heels of a CUDA update for the Xubuntu 15.04 that I run on that machine. The only way I can get it to start mining is from a cold boot with sudo. Any subsequent restarts of the miner without a cold boot generate the out of memory error. The other machine I use is fat, dumb and happy. Both have EVGA GTX 750's, non-super clocked.

The machine that works is running Xubuntu 14.04, CUDA 7.5-18 and ccminer 1.5.74
The machine giving me grief runs Xubuntu 15.04, CUDA 6.5.14-1 and ccminer 1.5.74

I may try pushing the 2nd machine to CUDA 7.5 like the first but it seems to me, based on the releases I see and recommended pairings, that it should be the other way around!

Pokeytex, I'm accepting donations! 

[joblo]

Starting a couple of weeks ago one of my computers that I use for mining while it's otherwise idle suddenly started giving me an "out of memory" error. I presume it came on the heels of a CUDA update for the Xubuntu 15.04 that I run on that machine. The only way I can get it to start mining is from a cold boot with sudo. Any subsequent restarts of the miner without a cold boot generate the out of memory error. The other machine I use is fat, dumb and happy. Both have EVGA GTX 750's, non-super clocked.

The machine that works is running Xubuntu 14.04, CUDA 7.5-18 and ccminer 1.5.74
The machine giving me grief runs Xubuntu 15.04, CUDA 6.5.14-1 and ccminer 1.5.74

I may try pushing the 2nd machine to CUDA 7.5 like the first but it seems to me, based on the releases I see and recommended pairings, that it should be the other way around!

Pokeytex, I'm accepting donations! 

I get that on a couple of my cards (970, 750ti) but not on others of the same type. One is in a Windows system the other Linux.
Both systems running ccminer 1.5.74 compiled with cuda 6.5 but has been happening with previous releases. I have not tested
with cuda 7.5. The problem only manifests itself on startup, the more I switch algos, the more likely to occur up to several times
per day. Another symptom is low hashrate due to degraded GPU state. Out of memory error occurs mostly wth neoscrypt. Some
algos seem less affected, ie neoscrypt may crash but quark might run. Everything back to normal after a reboot. Updated drivers
many times with no effect.

I concluded it's a card problem.

In your case try swapping the 750s to confirm.

[sp_]

A bug in cudamalloc. On windows you need to add 16gig of virtuaL memory.