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

[djm34]

hello everyone
currently i am running a gtx 690 and i cant get autotune to work for gpucoin setup
i am testing on pandacoin right now and autotune just crashes my comp everytime



i am running this one http://www.newegg.com/Product/Product.aspx?Item=N82E16814121636
with power target at 135%
gpu clock offset at +185 MHz
mem clock offset at +465 MHz

any help/working configs would be greatly appreciated

Edit: my command line,
cudaminer.exe --algo=scrypt:2048 -o http://127.0.0.1:6857 -u user -p pass

It's out of stock, that's why it isn't running...
Seriously, try to use -D with the autotune and do a control-C after 60 lines (or before), you should get your best config before.
Also it might be better to do the autotune only on one core

[worstplayer59]

hello everyone
currently i am running a gtx 690 and i cant get autotune to work for gpucoin setup
i am testing on pandacoin right now and autotune just crashes my comp everytime



i am running this one http://www.newegg.com/Product/Product.aspx?Item=N82E16814121636
with power target at 135%
gpu clock offset at +185 MHz
mem clock offset at +465 MHz

any help/working configs would be greatly appreciated

Edit: my command line,
cudaminer.exe --algo=scrypt:2048 -o http://127.0.0.1:6857 -u user -p pass

It's out of stock, that's why it isn't running...
Seriously, try to use -D with the autotune and do a control-C after 60 lines (or before), you should get your best config before.
Also it might be better to do the autotune only on one core

thanks!
got it to work using one core
choose config K8x24

edit cant get both cores though but 1s better then none

[jawtyng]

anyone tried a 750 non ti ? looking for the hashrates

With the scrypt algorithm, I am getting 255 KH/s max. That's with the MSI GAMING N750 TF 1GD5/OC card. My overclock settings are +135 MHz on the core clock and +125 MHz on the memory clock. My cudaminer settings are "-i 0 -C 1 -H 1 -l T4x24 -m 1". I am using the 02/18/2014 release of cudaminer and I haven't tried any other algorithms because I have been mining only since Thursday.

[Severaz]

anyone tried a 750 non ti ? looking for the hashrates

With the scrypt algorithm, I am getting 255 KH/s max. That's with the MSI GAMING N750 TF 1GD5/OC card. My overclock settings are +135 MHz on the core clock and +125 MHz on the memory clock. My cudaminer settings are "-i 0 -C 1 -H 1 -l T4x24 -m 1". I am using the 02/18/2014 release of cudaminer and I haven't tried any other algorithms because I have been mining only since Thursday.

ok that seems about what i thought it would be.. its a good option if the prices of the ti's keep going up looks like there all sold out on newegg now.. thanks for the info

[bacon_wrangler]

anyone tried a 750 non ti ? looking for the hashrates

With the scrypt algorithm, I am getting 255 KH/s max. That's with the MSI GAMING N750 TF 1GD5/OC card. My overclock settings are +135 MHz on the core clock and +125 MHz on the memory clock. My cudaminer settings are "-i 0 -C 1 -H 1 -l T4x24 -m 1". I am using the 02/18/2014 release of cudaminer and I haven't tried any other algorithms because I have been mining only since Thursday.

You should try the latest version, which has been tuned for Maxwell.

[jawtyng]

Thanks for the tip. I just tried the newest release of cudaminer. My max hash rate increased from 255 KH/s to 262 KH/s. Thanks also to Christian for all the work he has done on cudaminer.

[gfgrimm]

Hi people !

Something very intriguing happened to me a few hours ago. I'm mining Dopecoin an alt that uses scrypt.
I'm using the 02-04-14 version with auto-tune. The most I got in all my tries was around 200kh/s with my gtx 670.

I restarted cudaminer without looking at the output and was surprised to hear the fan of my GPU working a lot faster.
I look at the cudaminer window and was shocked to see it was mining at 340mh/s !
I scrolled back to see what config auto-tune used and it was k14x28. I checked on the pool stats to see if the pool was getting the same hash/sec and it was.

Usually my gpu is at 63°C and this time it was at 85°C. I thought I was a jerk not to have noticed I could mine a lot faster earlier and decided to solo mine instead of pool mining. Setup a new .bat with -l k14x28 and got the usual 200kh/s again.

I could not get the 340mh/s again (solo or pool mining)... Tried a lot of times to have auto-tune figure it out again but it did not :/

The gpu usage was the same when mining at 340 or 200mh/s only temp changed.

Did that already happen to anyone else ? How can I get this mining speed again ??

You using lowercase or uppercase k? scrypt needs uppercase K.

[ivanlabrie]

Stanr010, did you enable frame rate targets in precision x?

If so, disable that... 

[usualsuspect]

Hmm 173khash/so is all I can get out of my 780

are those the right numbers for a 780 cause my asus 680 gives me 356 kh and those are with stock settings

[cbuchner1]

Hmm 173khash/so is all I can get out of my 780

are those the right numbers for a 780 cause my asus 680 gives me 356 kh and those are with stock settings

bigjme, a crash of a CUDA kernel, the GPU clocks are sometimes throttled to half of the peak speeds. A reboot may fix it. You can check your clocks with GPU-z...

[jstabb]

Just an FYI as it seems getting more than 6 cards running in one system has been problematic for folks.

I'm now running (7) EVGA GTX 750Ti FTWs on an ASRock Fatil1ty Z87 Killer motherboard (Win7 x64). I had some problems getting normal risers to work (required jumping the vga presence pin), but USB risers have worked flawlessly without needing a jumper. Just make sure to disable the onboard video. If anyone wants more details, I'm happy to share.

[ManIkWeet]

Just an FYI as it seems getting more than 6 cards running in one system has been problematic for folks.

I'm now running (7) EVGA GTX 750Ti FTWs on an ASRock Fatil1ty Z87 Killer motherboard (Win7 x64). I had some problems getting normal risers to work (required jumping the vga presence pin), but USB risers have worked flawlessly without needing a jumper. Just make sure to disable the onboard video. If anyone wants more details, I'm happy to share.

Images, images is all we desire :3

[cbuchner1]

Just an FYI as it seems getting more than 6 cards running in one system has been problematic for folks.

I'm now running (7) EVGA GTX 750Ti FTWs on an ASRock Fatil1ty Z87 Killer motherboard (Win7 x64). I had some problems getting normal risers to work (required jumping the vga presence pin), but USB risers have worked flawlessly without needing a jumper. Just make sure to disable the onboard video. If anyone wants more details, I'm happy to share.

Nice. I've got the same board. And I've been thinking to sell of my 3 780 Ti cards to build a more power efficient rig. 15 kHash/s @ Yacoin isn't cutting it. And 10 powered USB cable risers are already ordered.

Christian

[proxlol]

Sorry for asking a potentially beaten-to-death question, the search function isn't great on this forum :/

Looking for kernel advice for a 780 Ti mining Vertcoin. Best I've come up with so far is:

cudaminer.exe -o __________  -u ____ -p ____ --algo=scrypt:2048 -i 0 -l T15x16 -H 2 -m 1
pause

pulling 370khash @ 1200/1914

---

can I be doing better with something else?

[djm34]

Sorry for asking a potentially beaten-to-death question, the search function isn't great on this forum :/

Looking for kernel advice for a 780 Ti mining Vertcoin. Best I've come up with so far is:

cudaminer.exe -o __________  -u ____ -p ____ --algo=scrypt:2048 -i 0 -l T15x16 -H 2 -m 1
pause

pulling 370khash @ 1200/1914

---

can I be doing better with something else?

Actually I can't... sound pretty good to me...

[cbuchner1]

can I be doing better with something else?

Yeah, for Vertcoin there is a bit of a performance cliff, which you can see firsthand when running
autotune with the -D flag. At some warp number (dependent on the block count) the performance
drops drastically. So the ideal configurations for normal scrypt with warp numbers at the kernel's
limit (x24) don't work.

I think this must be due to saturating/overloading the memory controller.

Have you ever tried autotuning the lower case "t" kernel for Vertcoin? This used to be the fastest
kernel before nVidia submitted something better. This "t" kernel implements two different memory
access schemes. One is called SIMPLE (used for Yacoin), the other is called ANDERSEN (used for scrypt).

We should experiment using SIMPLE also for scrypt with N>=2048. Those with access to the source code
and a working comipler environment can already play with swapping ANDERSEN for SIMPLE in the kernel launches (the places with three brackets like <<< >>>) to check if there are any benefits to be had.

Christian

[proxlol]

Actually I can't... sound pretty good to me...

Alrighty then

Yeah, for Vertcoin there is a bit of a performance cliff, which you can see firsthand when running
autotune with the -D flag. At some warp number (dependent on the block count) the performance
drops drastically. So the ideal configurations for normal scrypt with warp numbers at the kernel's
limit (x24) don't work.

I think this must be due to saturating/overloading the memory controller.

Have you ever tried autotuning the lower case "t" kernel for Vertcoin? This used to be the fastest
kernel before nVidia submitted something better.

Christian

Ahhhh alright then. I'm not going to pretend I know a great deal about these matters, but I did find it interesting that my config above was pulling a higher hash than all the usual x24 variations.

Nope I haven't tried that yet. I'll go lookup the autotune argument in your readme and see what I come up with.

One thing is did notice with my above config in particular is that there is no performance gain with -H 1 compared to -H 2. This is pretty unusual in my experience with your miner; usually you can expect a decent gain using -H 1 at the cost of some system responsiveness. Not complaining about anything haha, just something I noticed

[Dotcommie]

For some reason my graphics driver crashes when autotuning with higher lookup-gaps. Not immediately, but when it´s almost done.

the crashing is due to a time-out... haven't found a way to solve it yet. Use the -D option to at least get some readings before the crash.

The timeout settings can be changed in the registry:
http://www.microsoft.com/whdc/device/display/wddm_timeout.mspx

I've been having timeout errors and other errors like Cuda Error 30(unknown error) after anywhere from 30min to 9hrs with the geforce 750tis, I'm trying it out after changing the registry settings to disable it.  Hopefully that's the fix because it's super annoying and makes all these cards worthless at the moment.  I've just been testing them on my main pc which runs Windows 8.1, so maybe win7 that i'm putting on my other machines with play nicer.

Thanks for all your time and effort on this Christian!  (I'd donate but the small pool I've been on is 400% over estimates, but still has yet to get a block)

PS- Does it normallly list your folder structure on errors? D:\Christian\Documents\Visual Studio....etc?

[Finnminer]

Just an FYI as it seems getting more than 6 cards running in one system has been problematic for folks.

I'm now running (7) EVGA GTX 750Ti FTWs on an ASRock Fatil1ty Z87 Killer motherboard (Win7 x64). I had some problems getting normal risers to work (required jumping the vga presence pin), but USB risers have worked flawlessly without needing a jumper. Just make sure to disable the onboard video. If anyone wants more details, I'm happy to share.

Thanks, just the info I was looking for! Just ordered one.

[djm34]

can I be doing better with something else?

Yeah, for Vertcoin there is a bit of a performance cliff, which you can see firsthand when running
autotune with the -D flag. At some warp number (dependent on the block count) the performance
drops drastically. So the ideal configurations for normal scrypt with warp numbers at the kernel's
limit (x24) don't work.

I think this must be due to saturating/overloading the memory controller.

Have you ever tried autotuning the lower case "t" kernel for Vertcoin? This used to be the fastest
kernel before nVidia submitted something better. This "t" kernel implements two different memory
access schemes. One is called SIMPLE (used for Yacoin), the other is called ANDERSEN (used for scrypt).

We should experiment using SIMPLE also for scrypt with N>=2048. Those with access to the source code
and a working comipler environment can already play with swapping ANDERSEN for SIMPLE in the kernel launches (the places with three brackets like <<< >>>) to check if there are any benefits to be had.

Christian

In which file is this ?