CCminer(SP-MOD) Modded GPU kernels.

[sp_]

I might add bitcredits later.
is this your code dga?
https://github.com/dave-andersen/cudapts

Yup.  I think I committed an optimization or two to the cudaptsx version (x == xpassthrough, for ypool):
https://github.com/dave-andersen/cudaptsx/blob/master/xptMiner/gpuhash.cu
It's not a fast algorithm for doing the collision finding.  I wrote it pre-Maxwell, so computing the SHA was the bottleneck and it wasn't worth optimizing to do something faster.  With Maxwell, the sha is fast enough that random global memory accesses will be very bad.

Looks like alot of work to integrate into ccminer, but it would be interesting to do a maxwell mod.

[joblo]

Interesting is that I reach more hash with 55r, instead of56 on 750TI OCed

My test was on stock clock and Linux. But on second look my reference was 1.5.54.
Could your card be throttling more? I have a Win box with a 750ti OCed +100 I'll be
testing later once I get v56 compiled for Windows. I'll let you know what I find.

[go6ooo1212]

Interesting is that I reach more hash with 55r, instead of56 on 750TI OCed

My test was on stock clock and Linux. But on second look my reference was 1.5.54.
Could your card be throttling more? I have a Win box with a 750ti OCed +100 I'll be
testing later once I get v56 compiled for Windows. I'll let you know what I find.

Idk how actually see if its throttling , but with previous couple of releases , has no prob

[bensam1231]

https://www.cryptopia.co.nz/Exchange?market=FLAX_BTC

New algo
fast profitable and free gpu miner

10MHASH on the gtx980

no amd miner

Yeah, just looked at that. Looks like there is only .5 BTC worth of buy orders on there. Which means the price will tank the moment someone starts mining it with any sort of oumph.

CHC has like 2BTC worth of buy orders up.

[Slava_K]

C11 only boooos on suprnova!

[joblo]

Interesting is that I reach more hash with 55r, instead of56 on 750TI OCed

My test was on stock clock and Linux. But on second look my reference was 1.5.54.
Could your card be throttling more? I have a Win box with a 750ti OCed +100 I'll be
testing later once I get v56 compiled for Windows. I'll let you know what I find.

Idk how actually see if its throttling , but with previous couple of releases , has no prob

I don't know of an actual thorttling indicator. Drops in the clock freq is how I uaually tell.

[myagui]

I don't know of an actual thorttling indicator. Drops in the clock freq is how I uaually tell.

Your keyboard is throttling badly! 

[sp_]

C11 only boooos on suprnova!

seems to fail. I will look into it.

[go6ooo1212]

Thats with the new release and +130 clock / +300 mem on AB:

[joblo]

I don't know of an actual thorttling indicator. Drops in the clock freq is how I uaually tell.

Your keyboard is throttling badly! 

LOL. That's a pretty high error rate, I should get that checked.

[sp_]

C11 only boooos on suprnova!

seems to fail. I will look into it.

I have fixed it now. The source and the .7z exe file is updated to the corrected version

[joblo]

Thats with the new release and +130 clock / +300 mem on AB:

Widely varying performance among the boards. 3 & 5 are the winners and 0 the loser.
Assuming they are identical boards you have a good starting point to troubleshoot
why GPU0 is so much slower. How are your temps? One annoyance I've found is the default
fan profile is too slow to react. The clock starts throttling back before the fan ramps up.

[go6ooo1212]

The cards are identical yes. The temp is between 60-65 degrees of Celsius. I think the hash drops on some of the cards when I connect through the Teamviewer...

[joblo]

I've noticed the default intensity is too high for some cards/algos. Ive also noticed some odd behaviour
with intensity. Quark would initially run with -i 23.5. The next time i started it it failed at -i 23.5 but
ran at 23. Then 23 failed and I lowered it to 22.5.  970 + 780ti.

The preceding was on a Windows machine while I was playing with the intensity parm. When it failed
the error was out of memory. I left it at that but days later after an unrelated reboot I gave it another
try and it worked. The default intensity issues seem to be caused by my degraded system.

I have now seen similar behaviour on Linux and with different algos. I had a profit switch script that had
beed running for a day or so then on one switch, out of memory. Any attempt to start ccminer after that
resulted in the out of memory error. A reboot fixed it.

In both cases it appears to be gradual degradation leading me to suspect a slow memory leak. I don't think
this is a new issue in 1.5.55. Instead it has likely been around for a while but has only become apparent
because of an increase in pool switches following the SAK price crash. I suspect every ccminer kill is
leaking a little mem.

If anyone sees out of memory errors try rebooting.

[joblo]

The cards are identical yes. The temp is between 60-65 degrees of Celsius. I think the hash drops on some of the cards when I connect through the Teamviewer...

Of course, GPU0 has a display attached, Doh!

I'm getting 6680 on 750ti, GPU OC +100, temp below 60C. Reported clock speed is a saw tooth,
slow rise to 1420 then sudden drop to 1375. I think that's throttling.

[hashbrown9000]

@sp_ , how does the c11 algo compare with quark algo power-wise?  safe to keep the 750ti overclocked ~ +100 MHz?

[hashbrown9000]

there is definitely some sort of memory leak going on with GPU 0 in a multi-rig system, at least after quitting mining, compiling a new miner, and restarting mining with the newest miner. 

on c11 algo, GPU 0 will only display 1/3 the hashrate of other good cards in the system, but it looks like the card is crashed b/c my temp monitor for card 0 shows a very low temp as it's not mining anything.  i've verified this on two different rigs now using the same procedure.  on the third rig, I rebooted after a new miner compilation, and so far the reboot seems to have fixed this.

[joblo]

there is definitely some sort of memory leak going on with GPU 0 in a multi-rig system, at least after quitting mining, compiling a new miner, and restarting mining with the newest miner. 

on c11 algo, GPU 0 will only display 1/3 the hashrate of other good cards in the system, but it looks like the card is crashed b/c my temp monitor for card 0 shows a very low temp as it's not mining anything.  i've verified this on two different rigs now using the same procedure.  on the third rig, I rebooted after a new miner compilation, and so far the reboot seems to have fixed this.

Funny you mention that. I had noticed one of my cards hashing quark at about 1/3 of the expected rate.
Restarting ccminer fixed it. It may have happened another time but not enough to show a pattern.

Breaking news: As I write this I just noticed one of my rigs hashing quark slowly. Both cards were hashing at
about 1/2 the usual. Killing ccminer did not work. It failed to connect to the pool with endless retry
messages but no error.

I rebooted and it's back to normal.

[hashbrown9000]

interesting.  i'm on the 865 build.  seems like its not only c11 algo.  my oc settings are about +100 MHz on the GPU and none on the MEM

[AliMan]

Release 56 has finally made be break 100 mh/s on quark. Virtual beers for everyone!