[ANN] cudaMiner & ccMiner CUDA based mining applications [Windows/Linux/MacOSX]

[relm9]

Don't use T12x32, just let autotune find the right param. T12x32 was giving me CPU validation errors so probably is for you too.

[bathrobehero]

That hashrate is definitely not correct, but I can't help you further because as I see it the T kernel only works with single memory allocation (-m 1) which I can't use so all I can think of is trying different kernels.

[bigjme]

ok 2.69khash/s with auto. that seems more like it. but again, is there any way for me to get feedback so i know if its doing anything?

Ok so gpu-z is reporting 99% gpu load. 2gb of memory usage out of 3gb. Is that about right?

Its using roughly 55% tdp. So dropped my power by 125w. Just be nice to see some sort of feedback to see if it was doing anything. I guess I will see in the morning anyway if I have anything in my wallet

[bathrobehero]

What did autotune ended up with?

[Treggar]

Here's my experience using a Fermi (GTX570) for QQcoins

I just compiled for x64.  3 sessions going and barely making 30% gpu usage.

I let it run for 5 minutes to get the average hashrate as detected by the pool and it showed 0.2MH/s

[bigjme]

Autotune gets between 2.69 and 2.72khash/s

I will check the setting it uses tomorrow. Its almost 5am here so yeh :-p

Treggar your settings are wrong. Tell it to do auto and see what that does with 1 running. Its macing my gtx 780 out so yours shouldnt be doing that much. Unless that coin really is allowing much higher hashrates to yacoin. Im not sure

[Treggar]

It crashes with autotune.

[bigjme]

Im guessing because of it being an older gpu and maybe autotune isnt set up properly.

[Treggar]

Im guessing because of it being an older gpu and maybe autotune isnt set up properly.

Possibly, but it's still 100x what you're getting with a GTX780?  Maybe you should skip the autotune too

[relm9]

Yacoin has a much higher N factor than QQcoin. Higher the N factor, less kH, basically.

Anyway, you may be right about the amount of threads. With T36x3 I was able to bring up my hashrate to 98kH/s on a 780. I'm only seeing 6% GPU usage as well. However, if I open multiple instances of cudaminer it just halves the hashrate, and GPU usage stays the same.

Maybe something to do with the way memory is allocated on the Titan kernel?

[bathrobehero]

Autotune gets between 2.69 and 2.72khash/s

I will check the setting it uses tomorrow. Its almost 5am here so yeh :-p

Treggar your settings are wrong. Tell it to do auto and see what that does with 1 running. Its macing my gtx 780 out so yours shouldnt be doing that much. Unless that coin really is allowing much higher hashrates to yacoin. Im not sure

Autotune gave me K13x1.
But that config didn't used up all my vram so I went on a hunt. 13 times 1 is obviously 13, so I tried combinations close to 13 and so I ended up with 7x2 (=14, duh) and the results are amazing (for a non-Ti GTX 660, that is):



I had to get back to a 2-day old build though, becasue the newest is keep giving me CPU validation errors somewhat infrequently, but still.

[69charger]

So Yacoin is worth mining?

[bathrobehero]

So Yacoin is worth mining?

I don't know, I'm still in the honeymoon phase of bitcoin mining.


cbuchner1, I disovered a phenomenon regarding scantime (-s):

-s 1:                                                   -s 5:                                                   -s 60:


Is this really affecting the hashrate?
How does higher upper scantime limits bode with pool/solo mining?

[relm9]

It's weird that my Yacoin results (1.9kH/s) seem to be a lot lower than everyone else's. Are you guys with the 780 overclocking your cards? I have mine at +200.

Checked MSI Afterburner to make sure and my card isn't throttling, full clocks 99% usage.

Hm, I am mining on a pool and not solo, maybe that's why.

Edit: Nah, setup solo and same hash rates.

bathrobehero: that's interesting, I'm not seeing the same here though. Changing the scan time doesn't affect hash rate at all for me.

[see360]

Thank you for your work cbuchner1. I just sent another .33 LTC your way (around .65 in total now--that's about 3 days mining for me at 1050 khash/s, but I would have a lot less without cudaminer). 

Transaction IDs
0aedfb5a406ef286d1370465c0511959f4880c6afeba0aec03906c113e26a6b7
09443a293c0e260a826e4784d1abdca3c0e9dd6742d5fcb357c944044f924c63
0fd17d63d1a76773bfca83d9edb8cab143db2ef2715f5ae5b57f45a34921df8c

[coercion]

T12x32 is way to big for scrypt-jane. That's what you should use on regular scrypt coins. For yacoin that kind of config would require something like 50GB of GPU ram.

Autotune is not perfect. Look at the chart (run with -D to show debug info) then try configs that are just outside the populated area of the chart, until you've explored all the edges and found your limits. When you go too far it will print something like this: "GPU #0:Launch config 'T11x2' requires too much memory!"

I'm using T20x1 on my 780 and getting 3.4 kHash. T10x2 works well for me as well.

[relm9]

T12x32 is way to big for scrypt-jane. That's what you should use on regular scrypt coins. For yacoin that kind of config would require something like 50GB of GPU ram.

Autotune is not perfect. Look at the chart (run with -D to show debug info) then try configs that are just outside the populated area of the chart, until you've explored all the edges and found your limits. When you go too far it will print something like this: "GPU #0:Launch config 'T11x2' requires too much memory!"

I'm using T20x1 on my 780 and getting 3.4 kHash. T10x2 works well for me as well.

Thanks for the tips. I've tried some but T10x1 seems to be the best still. T20x1 doesn't perform well at all on my 780 - only 1.2 kH/s :\

Driver issue maybe? Running 331.92 here.

[dogelimic]

Can someone please explain this to me? It looks contradictory.

-C, --texture-cache   comma separated list of flags (0/1) specifying which of the CUDA devices shall use the texture cache for mining. Kepler devices will profit.

This says kepler devices will profit.

GPU #0: GeForce GTX 770 with compute capability 3.0
GPU #0: the 'K' kernel ignores the texture cache argument

My 770 is a kepler device, but cudaminer says that my kernel ignores the texture cache...

So which is it and how will my device profit from a launch option which is ignored?

[bigjme]

I cant use all my vram either. cbuchner1 what setring are you using to get 3.2kh/s? And is it using all available memory?

I just checked my yacoin. What is immature? I have 99.88 immature yacoins over night. Well in 5 hours but I dont know what it means

[coercion]

T12x32 is way to big for scrypt-jane. That's what you should use on regular scrypt coins. For yacoin that kind of config would require something like 50GB of GPU ram.

Autotune is not perfect. Look at the chart (run with -D to show debug info) then try configs that are just outside the populated area of the chart, until you've explored all the edges and found your limits. When you go too far it will print something like this: "GPU #0:Launch config 'T11x2' requires too much memory!"

I'm using T20x1 on my 780 and getting 3.4 kHash. T10x2 works well for me as well.

Thanks for the tips. I've tried some but T10x1 seems to be the best still. T20x1 doesn't perform well at all on my 780 - only 1.2 kH/s :\

Driver issue maybe? Running 331.92 here.

I'm running OS 10.9 on a hackintosh, so I don't know that I can help you out there. Although I did try it out on two gtx 670s at work today and they got 2 kH/s each with autotune with 331.20 on Linux. Are you running CUDA 5.5?

For those of you attempting to hash QQC or YBC or any other scrypt-N coin... you're probably going to need to modify some code and recompile. It looks to me like the time value used to calculate the N is hardcoded to Yacoin's start time. See GetNfactor() in scrypt-jane.cpp. edit: just reread the last two pages and I see that some of you have already done that. doh

Can someone please explain this to me? It looks contradictory.

-C, --texture-cache   comma separated list of flags (0/1) specifying which of the CUDA devices shall use the texture cache for mining. Kepler devices will profit.

This says kepler devices will profit.

GPU #0: GeForce GTX 770 with compute capability 3.0
GPU #0: the 'K' kernel ignores the texture cache argument

My 770 is a kepler device, but cudaminer says that my kernel ignores the texture cache...

So which is it and how will my device profit from a launch option which is ignored?

Are you using the latest version from git or a binary release? I see a commit message on Dec 28th that says "add back support for chunked memory allocation and texture cache to Kepler kernel. Slight speed-ups with -C 1 are seen." which implies he removed it at one point, possibly when upgrading to CUDA 5.5.

I cant use all my vram either. cbuchner1 what setring are you using to get 3.2kh/s? And is it using all available memory?

I just checked my yacoin. What is immature? I have 99.88 immature yacoins over night. Well in 5 hours but I dont know what it means

My yacoin config.
./cudaminer -u coercion.1 -p x -a scrypt-jane -o 127.0.0.1:9942 -l T21x1 -H 1

I just realized I can't run 21x1 since I pulled today. I ran that config all last night and now I can only run T20x1.

Immature means you can't spend them yet because they need to be included in more blocks so a fork doesn't screw you or anyone you transact with.