[ANN] cudaMiner & ccMiner CUDA based mining applications [Windows/Linux/MacOSX] - page 1128.

cbuchner1

hero member

Activity: 756

Merit: 502

Quote from: peacefulmind on April 11, 2013, 01:24:33 PM

Have a couple Titans clocked at 1GHZ and would be curious what they would do. People seem to have great result so far!

I have received reports that Titan support is presumably broken in the April 10th release. I would be interested to get a confirmation of that, or a report that it actually works. I am definitely willing to fix it, if it is really broken.

UPDATE: I put the April 9th release up for download again, for those who want to use it on their Titan systems.

Christian

Lacan82

sr. member

Activity: 247

Merit: 250

Quote from: peacefulmind on April 11, 2013, 01:24:33 PM

I don't want to re-read the whole thread - does this support stratum?

Have a couple Titans clocked at 1GHZ and would be curious what they would do. People seem to have great result so far!

have to use a proxy, but it runs great even with a proxy

peacefulmind

full member

Activity: 196

Merit: 100

I don't want to re-read the whole thread - does this support stratum?

Have a couple Titans clocked at 1GHZ and would be curious what they would do. People seem to have great result so far!

nst6563

sr. member

Activity: 252

Merit: 254

Quote from: cbuchner1 on April 11, 2013, 07:17:08 AM

I have been playing with options to improve system responsiveness. It involves choosing smaller launch configs and adding a few CPU sleep cycles before launching new CUDA work

Currently doing stratum mining for LTC on coinotron now, as wemineltc seems to have issues computing the correct payouts.

Would that slow down the hashrate? If so, could you make that an option that could be disabled?

Right now I've got 3 of these nvidia cards in my system and it's completely responsive. I'm using the onboard intel gpu for the primary display though so that may be why. Once I switched to that, my gpu0 shot up another 10kh/s and was even with gpu2.

Testarossa

newbie

Activity: 11

Merit: 0

CudaMiner 1st version
GTX 295
Windows 7
313.96 Driver
Stock settings
autotune
75 khash/sec

CudaMiner 10/04
GTX 295
Windows 7
313.96 Driver
Stock settings
autotune
56 khash/sec

Sad

cbuchner1

hero member

Activity: 756

Merit: 502

I have been playing with options to improve system responsiveness. It involves choosing smaller launch configs and adding a few CPU sleep cycles before launching new CUDA work

Currently doing stratum mining for LTC on coinotron now, as wemineltc seems to have issues computing the correct payouts.

Lacan82

sr. member

Activity: 247

Merit: 250

Quote from: xrobau on April 11, 2013, 04:38:34 AM

So with the GTX570, I'm getting 179-180KHash, while using the desktop. It's JUST on the border of being annoyingly laggy. But is working quite nicely.

--Rob

I'm assuming that is at the standard clockrate? I have pre-overclocked version. It is getting about 203

xrobau

member

Activity: 99

Merit: 10

Open Source Developer, Hardware Supplier

So with the GTX570, I'm getting 179-180KHash, while using the desktop. It's JUST on the border of being annoyingly laggy. But is working quite nicely.

--Rob

gchil0

newbie

Activity: 59

Merit: 0

BTW, thanks for the continued improvements to cudaminer! I tossed a litecoin your way.

navigator

sr. member

Activity: 362

Merit: 250

GTX 460 768mb SC getting 104kh/s with 4-10 version and 56x2 config.

gchil0

newbie

Activity: 59

Merit: 0

I've confirmed on another machine containing a K20 that the Titan kernel is broken. The regular kernel, though, shows significant improvement on my ancient 9600 GSO, the Tesla C1060's and the GTX 480's. Thanks!

wndrbr3d

hero member

Activity: 914

Merit: 500

Quote from: nst6563 on April 10, 2013, 11:15:00 PM

Anyone else run into problems when sli is enabled in the nvidia control panel?
On my setup when I launch cudaminer with sli enabled, it just sits there after it says starting 'x threads' and the whole system seems very sluggish (4.6Ghz i7-3770k w/24GB ram and ssd's....definately not used to sluggish)

If I disable SLI and then run cudaminer, it seems to work fine.

Is there any advantage to running in sli mode for mining?

There's an issue with the chunked memory allocation routine when running in SLI mode. Leave SLI disabled and start an instance for each GPU and you should be good to go.

nst6563

sr. member

Activity: 252

Merit: 254

Anyone else run into problems when sli is enabled in the nvidia control panel?
On my setup when I launch cudaminer with sli enabled, it just sits there after it says starting 'x threads' and the whole system seems very sluggish (4.6Ghz i7-3770k w/24GB ram and ssd's....definately not used to sluggish)

If I disable SLI and then run cudaminer, it seems to work fine.

Is there any advantage to running in sli mode for mining?

wndrbr3d

hero member

Activity: 914

Merit: 500

690GTX on 4-10:

autotune: 128x3 @ 161k/hash per GPU
no-autotune: 169x2 @ 158k/hash per GPU

chunked memory allocation still borked when left to run in SLI with both GPU's being fed by one instance, but runs fine if i fire up one instance per GPU.

670 with 4GB on 4-10:

autotune: 360x2 @ 169k/hash
no-autotune: 290x2 @ 165k/hash

it would seem that the trade off of more memory for fewer SP's in the 670 vs. one GPU in the 690 shows something interesting

grosminer

hero member

Activity: 718

Merit: 500

CudaMiner 2013-04-09
GTX 660Ti 2GB
Windows 7
314.14 WHQL Driver
GPU Core OC to 1320 MHz
autotune launch config settings 148x2
130-140 khash/sec

CudaMiner 2013-04-10
GTX 660Ti 2GB
Windows 7
314.14 WHQL Driver
GPU Core OC to 1320 MHz
autotune launch config settings 96x3
140-150 khash/sec

Great update!

Wole

newbie

Activity: 22

Merit: 0

560Ti 900 core clock:
2013-04-09: [2013-04-11 03:21:01] GPU #0: 107.78 khash/s with configuration 32x4
2013-04-10: [2013-04-11 03:32:00] GPU #0: 138.84 khash/s with configuration 64x2

Sweeet!

Lacan82

sr. member

Activity: 247

Merit: 250

Christian. is the -algo command suppose to work to switch? because it is skipping autotune and showing my gpu as (null)

chanzilla

newbie

Activity: 20

Merit: 0

Quote from: Number6 on April 10, 2013, 07:55:12 PM

Workaround for constant crash on exit.

Every time I went to exit the program (Ctrl-C) it would crash. Sometimes the driver would auto-recover, the screen would go black for a few seconds, then return, so I could answer the prompt and exit the program. Other times my driver would not auto-recover and I would get a black screen that lasted until I eventually hard rebooted my PC.

Anyway, I discovered that if you unplug or disable you network connection and wait for the timeout messages to appear, usually within 15 seconds, you can cleanly exit. While not an ideal solution, it sure beats taking a chance on getting the black screen and having to hard-reset your box, risking data loss, etc.

I agree with others that further CUDA optimizations would be the best use of the developer's time, so thought I would share this work-around for others who may also experience the "crash on-exits" problem.

Good tip. I just had to hard restart because this same issue. Most of the time the driver auto recovers.

cbuchner1

hero member

Activity: 756

Merit: 502

Seems my latest update gives the largest boost to Fermi architecture cards, and still some boost to Kepler devices.

nst6563

sr. member

Activity: 252

Merit: 254

This update is awesome!
I'm using 2 gtx560se's (fermi) and a gt430 (fermi)

gtx560se are each clocked at 962core/1924shader/2121mem

with cgminer I was topping out at 54kh/s
4/9/2013 build of cudaminer I was topping out at 76kh/s
4/10/2013 build of cuaminer I'm now topping out at 107kh/s

So all cards combined I'm cranking around 225kh/s with 2 x gtx560se and the gt430. Never thought I'd see them become useful.

Topic: [ANN] cudaMiner & ccMiner CUDA based mining applications [Windows/Linux/MacOSX] - page 1128. (Read 3426918 times)