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

[djm34]

is it a new coin the RAD... I like the name...
Just solo mined my first MRC block while doing some test. 

I was wondering if the warp calculation with the Z kernel in windows could be a bit too conservative.
Autotune gives me with my 780ti (no monitor attached) 635 warp, however I checked different config and I got something 40-50khash/s more with 43x16 which corresponds to 688warp.
I was using before that 42x15 and was doing around 610khash/s, now I get 665khash/s
And it seems to work ok since it found a block...

[1713983340]

1713983340

[1713983340]

1713983340

[1713983340]

1713983340

[1713983340]

1713983340

[bathrobehero]

In windows there's a 12% memory backoff value in place in cudaMiner to avoid out of memory crashes. If your card has no monitor attached you can get away with changing that value to 1% or 2% which increases performance as cudaMiner uses more VRAM - if autotune doesn't crash, that is.

[13G]

UTC mining on GTX TITAN @ 930Mhz

550khash/s with parameters -d 0 -i 0 -H 2 -C 2 -m 1 -l Z14x24

[sambiohazard]

Hi guys I have got a new Asus GTX 660Ti. I am using 12-18 version with follwing batch file

cudaminer.exe -H 2 -m 1 -i 1 -l K7x32 -o stratum+tcp://asia.middlecoin.com:3333 -O 1AqywTTPox5W8Vk15JbPMfkXvXXCAztxxQ:123

For this I am getting 255 khash/s. Is this good for no overclock and 12-18 version.

Also what launch config I can use with X,Y,Z kernels with new release expected this week? Thanks for the answers.

Anyone please

[fiorex]

Hi guys, do you know the ideal configuration for a NVidia Gtx765m for cudaminer?
I see 220khash/s during autotune but the real power of miner is 110 khash/s
Why?
my batch file:

cudaminer.exe -t 1 -o stratum+tcp://eu.multipool.us:3352 -u fiorex.1 -p x -i 0

Thanks!

[ManIkWeet]

I compiled the latest GitHub version and must say that -for me- the x64 executable is -still- a little bit slower than the old one.
~4.26khash/s on the old one.
~4.15khash/s on the newest one.
Now that might seem like a small difference, but every little bit counts
That is with the following parameters: -m 1 -l T128x2 -b 1024 -L 4 -i 1 --algo=scrypt-jane -o http://127.0.0.1:8108 -u user -p password
I also noticed that pooled mining prints a higher hashrate...

[djm34]

In windows there's a 12% memory backoff value in place in cudaMiner to avoid out of memory crashes. If your card has no monitor attached you can get away with changing that value to 1% or 2% which increases performance as cudaMiner uses more VRAM - if autotune doesn't crash, that is.

Looking through the code, no backoff at all would give only 661 warps, while apparently the card is running well at 688.
No monitor attached the card use only 108mb over 3072mb of vram, idle
While running at 688warps it uses 2946mb. (688x4+108 = 2860mb<2946, what do I forget ?)

edit: wrong calculation 635/0.88 is not 661 rather 681

[cbuchner1]

In windows there's a 12% memory backoff value in place in cudaMiner to avoid out of memory crashes. If your card has no monitor attached you can get away with changing that value to 1% or 2% which increases performance as cudaMiner uses more VRAM - if autotune doesn't crash, that is.

to be more specific, the 12% backoff is a "better safe than sorry" feature to avoid evil crashes during autotune with the nVidia WDDM drivers. I've had weird stuff happen, like single digit hash rates and total system lockups when taking too much memory. So I chose a pretty high default backoff percentage on Windows to avoid any kind of trouble.

Christian

[cbuchner1]

I have typed up the README file for today's release, but there are a couple of
TODO's remaining in the code

-allow the miner to work through N-factor changes by adjusting the lookup-gap
-support lookup-gap in the F (Fermi) kernel
-remove spinlock_kernel.cu and legacy_kernel.cu as they are no longer of interest
-rename the kernels according to the description in the README (no worries most
 of the old names like L,X,Y,Z will keep working because I allow for alias names)

We are set for a release very late in the day (Berlin time)

[Thirtybird]

In windows there's a 12% memory backoff value in place in cudaMiner to avoid out of memory crashes. If your card has no monitor attached you can get away with changing that value to 1% or 2% which increases performance as cudaMiner uses more VRAM - if autotune doesn't crash, that is.

to be more specific, the 12% backoff is a "better safe than sorry" feature to avoid evil crashes during autotune with the nVidia WDDM drivers. I've had weird stuff happen, like single digit hash rates and total system lockups when taking too much memory. So I chose a pretty high default backoff percentage on Windows to avoid any kind of trouble.

Christian

I've had this happen in YACMiner as well - and what I've found is happening is the amount of memory being allocated overflows the dedicated memory on the card and rolls over to dynamic memory (think of swapping memory to disk), AKA system memory.  On one of my systems, when this happens, it rolls over to system ram, which doesn't have enough available, so Windows swaps it out to disk.  Think about how slow hashing YAC on your hard drive would be!  It's easier to pull the power cord than recover from that. 

The only way I've been able to see this is by monitoring the memory using GPUz while launching - if you have any thoughts, would love to try to detect when that happens in code and throttle the settings back.

[Prima Primat]

Hi guys, do you know the ideal configuration for a NVidia Gtx765m for cudaminer?
I see 220khash/s during autotune but the real power of miner is 110 khash/s
Why?
my batch file:

cudaminer.exe -t 1 -o stratum+tcp://eu.multipool.us:3352 -u fiorex.1 -p x -i 0

Thanks!

Try -l K4x32
Then tweak with -H 2, -H 1 or -H 0

[djm34]

In windows there's a 12% memory backoff value in place in cudaMiner to avoid out of memory crashes. If your card has no monitor attached you can get away with changing that value to 1% or 2% which increases performance as cudaMiner uses more VRAM - if autotune doesn't crash, that is.

to be more specific, the 12% backoff is a "better safe than sorry" feature to avoid evil crashes during autotune with the nVidia WDDM drivers. I've had weird stuff happen, like single digit hash rates and total system lockups when taking too much memory. So I chose a pretty high default backoff percentage on Windows to avoid any kind of trouble.

Christian

I've had this happen in YACMiner as well - and what I've found is happening is the amount of memory being allocated overflows the dedicated memory on the card and rolls over to dynamic memory (think of swapping memory to disk), AKA system memory.  On one of my systems, when this happens, it rolls over to system ram, which doesn't have enough available, so Windows swaps it out to disk.  Think about how slow hashing YAC on your hard drive would be!  It's easier to pull the power cord than recover from that.  

The only way I've been able to see this is by monitoring the memory using GPUz while launching - if you have any thoughts, would love to try to detect when that happens in code and throttle the settings back.

by the way, is there a way to reset the cards (or the driver) without rebooting ?
From time to time, something "bad" happen, the driver doesn't crash but the card stay in some bad state running at very low clock speed. So far the only way I found has been to reboot, but is there some better way of doing this ?

[sambiohazard]

Hi guys I have got a new Asus GTX 660Ti. I am using 12-18 version with follwing batch file

cudaminer.exe -H 2 -m 1 -i 1 -l K7x32 -o stratum+tcp://asia.middlecoin.com:3333 -O 1AqywTTPox5W8Vk15JbPMfkXvXXCAztxxQ:123

For this I am getting 255 khash/s. Is this good for no overclock and 12-18 version.

Also what launch config I can use with X,Y,Z kernels with new release expected this week? Thanks for the answers.

Anyone please

Bump

[fiorex]

Hi guys, do you know the ideal configuration for a NVidia Gtx765m for cudaminer?
I see 220khash/s during autotune but the real power of miner is 110 khash/s
Why?
my batch file:

cudaminer.exe -t 1 -o stratum+tcp://eu.multipool.us:3352 -u fiorex.1 -p x -i 0

Thanks!

Try -l K4x32
Then tweak with -H 2, -H 1 or -H 0

i try -l K4x32 but i'm at 130khash/s... what can i do also?
Thanks!!

[djm34]

Hi guys I have got a new Asus GTX 660Ti. I am using 12-18 version with follwing batch file

cudaminer.exe -H 2 -m 1 -i 1 -l K7x32 -o stratum+tcp://asia.middlecoin.com:3333 -O 1AqywTTPox5W8Vk15JbPMfkXvXXCAztxxQ:123

For this I am getting 255 khash/s. Is this good for no overclock and 12-18 version.

Also what launch config I can use with X,Y,Z kernels with new release expected this week? Thanks for the answers.

Anyone please

Bump

Look at the spreadsheet on the first page

[sambiohazard]

Spreadsheet says nothing about X,Y,Z kernels.

[bathrobehero]

by the way, is there a way to reset the cards (or the driver) without rebooting ?
From time to time, something "bad" happen, the driver doesn't crash but the card stay in some bad state running at very low clock speed. So far the only way I found has been to reboot, but is there some better way of doing this ?

I found a tool though, called GFX clock fixer (I only found the download link and not the site I found it on) It said something about using a neat trick to reset the driver.

Use at your own risk, obviously. Oh, and if it's not doing it's magic and the window just stays there, fire up another instance of it.

[sambiohazard]

Hi bathrobehero, I got a 660Ti on your recommendation. Thanks. Can you tell me if 255 Khash/s is good enough for stock clocks?

[bathrobehero]

Hi bathrobehero, I got a 660Ti on your recommendation. Thanks. Can you tell me if 255 Khash/s is good enough for stock clocks?

Hi, depends what you're mining   I sent a PM.

[tacojohn]

I have 4 GPUs plugged into one system.  When I run one process per GPU, I get very stable results.  When I let cudaminer spawn threads to simultaneously drive 2 or more GPUs, I get sporadic cpu validation errors and about a 50% rejection rate from the pool.

Anybody else seeing this generally?  

I'm running the head revision, 64-bit linux.