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

[cbuchner1]

Could it be that the malloc, done once, was placed in a segment of GPU memory that didn't land in a high-performance configuration - and since it was only done once, it varies on each launch?

Single memory allocation vs. the default allocation scheme makes a difference on compute 1.x beause of an effect called "partition camping". You can try both variations - the -C 1 or 2 feature enforces single memory allocation mode.

[1713580033]

1713580033

[1713580033]

1713580033

[nakedman]

I tried building the building the cudaminer 417 from scratch and while doing ./configure I noticed this:

Code:

checking for curl-config... no
checking whether libcurl is usable... no
configure: error: Missing required libcurl >= 7.10.1
mads@Adria:~/cudaminer-src/cudaminer-src-2013.04.17$ curl --version
curl 7.26.0 (i486-pc-linux-gnu) libcurl/7.26.0 OpenSSL/1.0.1e zlib/1.2.7 libidn/1.25 libssh2/1.4.2 librtmp/2.3

Is that expected behavior? Because it looks a little wrong.

[cbuchner1]

Is that expected behavior? Because it looks a little wrong.

how about installing a libcurl developer package? libcurl4-dev, libcurl3-dev or libcurl-develop depending on your Linux distribution.

The thing about error messages is: they tell you what's wrong.

[lpshowboat0099]

Just want to say thanks for this program. Currently running at ~225 kH/s on my GTX 570 with -i 1 -l 30x7 and ~245 kH/s with interactive disabled. Card is OC'd to 900 on core clock and 1.05 V.

[termhn]

GTX 670 @1201/7008:

-i 0 -l 210x2
183 kH/s, 1880 MB

-i 0 -m 1
174 kH/s, 1264 MB, Autotuned 70x4

-i 0 -C 1 and -i 0 -C 2
184 kH/s, 1264 MB, Autotuned 56x5

-i 0 -m 1 -C 2
184 kH/s, 1264 MB, Autotuned 70x4

What values to use for 2 cards? Tried those with the lowest VRAM usage:
-d 0,1 -m 1,1 -C 2,2 -i 1,1 -l 56x5,56x5
but computer becomes very slow and eventually freezes or videodriver crashes.
It seems cudaMiner starts to use shared system memory, because VRAM is not enough.

Did you actually overclock your memory that much?! My stock mem clock is 3004..

[SubNoize]

580gtx getting between 235 and 265 with new build.
570gtx OC'd sitting on about 250 flat.

[FalconFour]

Well, I did a little A-B testing with pretty much every known option.

Everything described here is using the same version (2013-04-17 of cudaMiner), same driver (314.22), same card (9600 GT), same clock options (stock - verified to be running the same clocks all the time), same launch options (16x4). Just different OS and the options described:
- XP 32-bit, cache (-C) set to 0, 1, 2 = between 20-24khps - cache off seems to have best performance
- XP 32-bit, interactive turned on (1) = 16-18khps
- XP 32-bit, single-allocation turned on (1) = no difference, 20-22khps but only seems to impair performance
- XP 32-bit, autotune = 20-24khps depending on its mood (sometimes it picks 8x options, sometimes 16, but usually 16x4)
- XP 32-bit, dual-thread and half-workload (-t 2; -l 16x2,16x2; -i 0,0) = totals about 22khps (yep!), but crashes on Ctrl+C or if autotune is accidentally invoked by forgetting to add multiple options.
- Win7 x64, autotune = 10.8khps
- Win7 x64, interactive = 8-10khps
- Win7 x64, single-allocation = 10.8khps (initially hangs the screen but snaps back just before crashing - starts at "0.8khps")
- Win7 x64, cache 1 or 2 = no effect, still ~10khps

I really think you should dig out an XP CD and give it a try yourself - just have to install the nVidia drivers and the Visual C++ 2010 runtime, then launch it and see for yourself. Might just find a huge speed boost

Also, this testing was really tiresome since the display only updates once a minute (or if a result is found). Since the initial display is usually bogus (claiming very low figures at 4096 hashes right after launching), I have to wait a whole minute each time I play with background options (clocks, background apps, etc). Isn't there a way to make the display output the status more often? Couldn't find any related options in the "--help" notes...

[TiuraZ]

My GTX 460 does slightly more than 100kHash when overclocked - I think I've reached 110 kH.

I switched the interactive of and now I get over 100 khash. Didn't notice any effect on normal use.

[Mikanoshi]

Did you actually overclock your memory that much?! My stock mem clock is 3004..

3004 is 6008 effective frequency. I overclocked +500 = 3504(7008) for 24/7, it can actually go higher) I've seen 3700(7400) on these cards.

[nakedman]

I swapped to Windows in order to see which speeds I could get.

Using the -i flag apparently the program doesn't run for me. But just running it without any flags but o and O I get 144 k/hash with my 560 TI. I am not sure I quite understand the -l flag. What would a good setting for that be to a 560 TI?

[tallbikeguy]

I'd like to run this on Windows - I'm wondering, is everyone compiling this from source?  If not, where are you getting the executable?  The mega.co.nz address doesn't appear to be working.  Thanks!

[nakedman]

I'd like to run this on Windows - I'm wondering, is everyone compiling this from source?  If not, where are you getting the executable?  The mega.co.nz address doesn't appear to be working.  Thanks!

There is a link to a linux executable somewhere recent, page 17 or forward. I can upload the cudaminer to somewhere else than mega.co.nz for you if you like. Would you like the 32 or 64 bit? Just keep in mind that I'm not the creator or in any way affiliated with this project, so getting a MD5 sum of the package from someone else would be a good idea (Every random person could potentially be a bad guy).

[termhn]

Did you actually overclock your memory that much?! My stock mem clock is 3004..

3004 is 6008 effective frequency. I overclocked +500 = 3504(7008) for 24/7, it can actually go higher) I've seen 3700(7400) on these cards.

Ahh, thanks for the info... I hear mem clock doesn't matter as much as core clock... My max core is 1137, would it be worth it for me to OC mem?

[Schleicher]

The mega.co.nz address doesn't appear to be working.

It's working here, but only with Javascript enabled.
I uploaded it here:
http://www.sonstiges.org/xcoin/cudaminer-2013-04-17.zip

[cbuchner1]

meh. New GTX 560 Ti 448 core edition does 210 kHash with no overclocking, but as soon as I try to mine with just two cards (e.g. in conjunction with another 560 Ti) my power supply cuts out.

I just ordered a 1050W unit to supply enough juice to all 4 GPUs in my desktop.

Christian

[cbuchner1]

I swapped to Windows in order to see which speeds I could get.

Using the -i flag apparently the program doesn't run for me. But just running it without any flags but o and O I get 144 k/hash with my 560 TI. I am not sure I quite understand the -l flag. What would a good setting for that be to a 560 TI?

it's either -i 0 or -i 1, not just -i

144 kHash is OK for a 560Ti.  try passing -l 32x4 as launch argument, this might get you a few kHash more. If you feel like experimenting, also enable -C 1 or -C 2. Some people get 5-10 kHash more.

Christian

[nakedman]

meh. New GTX 560 Ti 448 core edition does 210 kHash with no overclocking

With what settings if I might ask?

[cbuchner1]

meh. New GTX 560 Ti 448 core edition does 210 kHash with no overclocking

With what settings if I might ask?

cudaminer.exe -d 0,3 -i 1,1 -l 28x8,32x4 -C 2,2

device 0 is the 560Ti 448 core
device 3 is the 560Ti regular

without the -i 1,1 my power supply cuts out immediately, you should use -i 0 for best performance

[Mikanoshi]

Ahh, thanks for the info... I hear mem clock doesn't matter as much as core clock... My max core is 1137, would it be worth it for me to OC mem?

Overclocking from 6GHz to 7GHz gives only 2-3kH/s
It's better for games, and even better for benchmarks.

[cbuchner1]

Well, I did a little A-B testing with pretty much every known option.

I really think you should dig out an XP CD and give it a try yourself - just have to install the nVidia drivers and the Visual C++ 2010 runtime, then launch it and see for yourself. Might just find a huge speed boost

Hmm, looks like old nVidia cards have serious issues with the WDDM driver model.