Topic: CCminer(SP-MOD) Modded NVIDIA Maxwell / Pascal kernels. - page 1098. (Read 2347601 times)

Hmm I think its not at all , compared to AMD - Nvidia rules the hash...

nist5 is kinda slow (((

Just tested nist5 on my mixed rig ( 970 +750ti) with release50.

Without using i or g parameters reaches 27.8MH/s for 970 and 9.78MH/s for 750ti.
I tried these kind of combinations then:
-g 2 -i 19 --cpu-priority 3
-g 2 -i 19.5 --cpu-priority 3
-g 2 -i 20 --cpu-priority 3
-g 4 -i 19.5 --cpu-priority 3
-g 5 -i 19.5 --cpu-priority 3

The result is almost same: first 400-500 shares the speed is much more fast , but then slowly falls below the quoted speeds...

Hi Folks,

I'm seeing this:
 
"cuda error in func 'x11_sind512_cpu_init' at line 641 : invalid texture reference."


Can anybody help me fix this?

I believe that would be a YES. Different setting for different cards. Could be wrong, but it's what I would do. Oh wait, sorry, I know that you can run one instance of your miner, by using the command switches (-d). Look at your readme file. It's in there. Personally, I would run different instances of the miner tho, so if one card craps out, you can close the miner to fix that card while the rest keep going. Otherwise you shut them all down to work on just one. It's all a preference thing. Hope that helps. Done.

Did I miss the answer to that question or did it go unnoticed?
Do you need to run two ccminers when you have a mix of GPUs (750Ti & 970 for example)?

Yes,,,yes,,,yes! THIS is what I've been trying to get an answer to for these last several days. How do I get a workaround for this? More to the point...I need an example of the .bat file to set it up.

Thank you

edit; if ya go back about 9 post pages, I asked about this. I guess because I haven't gotten very involved, until lately, nobody thought to look at my posting much. I can't write code but I do know when something ain't right. Question...just what does the workio thread do? Is it software (code) related, hardware (vid card), or internet speed (not sure of tech term) related? If I knew what that did, I could experiment. Don't remember who, but somebody posted a example of the .bat loop file but I could not get it to work. Just running a simple one now but playing with the switches. My current file (I change and fiddle with it tho);

ccminer -a quark -o stratum+tcp://yaamp.com:4033 -u LV4oV0xyLWu6pntuq97L0PheeavJsALJYo -p mHung   -i 20.1 -r -R=3 --cpu-priority 3  

It's actually not running too bad. Still get the random workio thread disconnect tho. Sooo...how do I force a restart of ccminer. The vid card & driver doesn't seem to crash but things slow down, then...crap, nothing, workio disconnect. If I'm sitting here (which is a lot of the time) I just close and restart ccminer. I'd really like a set it, forget it, have a nice day kinda thing tho. I could make a beer run and it would still be mining or go to the beach (which is where I'm headed in a bit). Ya know. OK...done editing!  Well almost...I really would like a solution. I've forgotten more than I remember, so I leave things up to the ppl in the know...you guys. You're a great group, mostly, and I know eventually the info comes out.  Thank You very much for any replies. Bye, for a few hours...heading out for some fun in the sun.

true - hence the reason for the question ...

we will have two spare machines at the end of the week when the farm rebuild is finished ( existing rebuild - not the new build ) ...

ill setup one centos and one sl machine to test with cuda 7 ...

if all works and is stable - the ccminer-spmod v50 compile will take place after that ...

if that builds ok - testing phase will go for a few days after that ...

i may pull one of the workers down ( 6 x 750ti oc lp ) and plugin a new hdd to it and test that way first ...

there are only 4 more machines to be rebuilt to f20x64 currently to get back to mining again - until the full farm rebuild to custom frames ... that should finish tomorrow - so the rest of the week to get all the other jobs sorted ( including the setup and testing of the multi-algo donation links ) ...

then the centos / scientific linux build will take place ... and the compile and testing of ccminer-spmod v50 ...

i can let you know of the progress ( or regress ) of the builds if you like ...

#crysx

I just tried one test compile of ccminer-1.5.48 and there were no issues. I don't have either centos7 or cuda7
running on bare metal yet so I couldn't test the compiled executable.

With the delays in releasing cuda on fedora 21 and the short fedora life cycle it only leaves a small usable
window to build a supported fedora based miner.

This is a disconnect from the pool. Happens when there is a lot of traffic. You need to use a bat file  wait and loop.

or use a minercontrol software like "miner control"

@sp

Reject rate on lyra2 is still high on release 50. I am only about 81% accepted. I have no special settings in the bat and using Windows 7 64bit.

I have  improved groestl-coin.

New hashrates

970: 20.5-22.5 MHASH
960: 13.5-15 MHASH
750ti: 7,4-8.5 MHASH

Every time Ctrl-C submitted results in ccminer crash:


Notebook Geforce 840M with standard start parameters only (-o -u -p) on Nicehash algo quark.

EDIT: edited algorithm

what are your command parameters? ...

ccminer.exe xxx xxx xxx ? ...

#crysx

I am getting this error and the non-responding program window then ccminer stops.
this is strange because i havnt had any issues with release 48
changing i or g did not help
Free Image Hosting

is it in a case? ... or an open rig? ...

#crysx

thats actually strange - as yaamp seems to be more of an issue than westhash with regards to rejects / drops in connections ...

only an hour ago - we couldnt access the site at all ...

could it be a firewall issue somewhere? ... just a stab in the dark there ...

#crysx

not sure about the 970 - but the 750ti oc lp cards have no issue ...

#crysx

yes - i do understand that ...

all of the servers we have are RHEL based - scientific linux - that have the same support lifespan ...

unfortunately NONE of them have the bleeding edge tech that fedora has ...

then again - there really isnt any need to have that for miners when the new ones are built ( custom framed ones with 12 cards per rig ) ...

i used to use centos AGES ago until the kernel itself refused to run with more than 1 280x in it ...

thats all fixed now - and we mostly have nvidia cards in the farm - so may try it out ... both scientific linux AND centos ...

for now though - the rebuild of the current farm needs to finish as we are missing almost 1/3rd the hashrate because they are down ...

does cuda 7 work well - compiling and stability and all? ...

#crysx

Hey all,

Thanks for the input for my .bat. I had already done everything mentioned so I have no flippin idea. Something is kicking me in nice/west. Works fine in yaamp. I'll work on it.

Thank you