Bitcoin Forum
February 22, 2017, 01:27:40 PM *
News: Latest stable version of Bitcoin Core: 0.13.2  [Torrent]. (New!)
 
   Home   Help Search Donate Login Register  
Pages: « 1 2 3 4 5 6 7 8 9 10 11 [12] 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 ... 1142 »
  Print  
Author Topic: [ANN] cudaMiner & ccMiner CUDA based mining applications [Windows/Linux/MacOSX]  (Read 2994133 times)
SubNoize
Jr. Member
*
Offline Offline

Activity: 47


View Profile
April 14, 2013, 07:03:16 AM
 #221

Can someone explain this "64x2" "S27x3,28x4" thing to me or point me in the right direction on reading up on it??

I have a 580gtx and i'm trying to figure out the best set up
1487770060
Hero Member
*
Offline Offline

Posts: 1487770060

View Profile Personal Message (Offline)

Ignore
1487770060
Reply with quote  #2

1487770060
Report to moderator
1487770060
Hero Member
*
Offline Offline

Posts: 1487770060

View Profile Personal Message (Offline)

Ignore
1487770060
Reply with quote  #2

1487770060
Report to moderator
1487770060
Hero Member
*
Offline Offline

Posts: 1487770060

View Profile Personal Message (Offline)

Ignore
1487770060
Reply with quote  #2

1487770060
Report to moderator
Advertised sites are not endorsed by the Bitcoin Forum. They may be unsafe, untrustworthy, or illegal in your jurisdiction. Advertise here.
1487770060
Hero Member
*
Offline Offline

Posts: 1487770060

View Profile Personal Message (Offline)

Ignore
1487770060
Reply with quote  #2

1487770060
Report to moderator
SubNoize
Jr. Member
*
Offline Offline

Activity: 47


View Profile
April 14, 2013, 07:16:52 AM
 #222

also gettings lots of crashes after like 20 minutes of use.
hammz
Member
**
Offline Offline

Activity: 80



View Profile
April 14, 2013, 07:53:49 AM
 #223

I stopped using it.  Can't connect to the pools I want to, and kattare connects but keeps giving massive stales.

back to CGminer for me.
cancrena
Newbie
*
Offline Offline

Activity: 9


View Profile
April 14, 2013, 08:02:52 AM
 #224

Can someone explain this "64x2" "S27x3,28x4" thing to me or point me in the right direction on reading up on it??

I have a 580gtx and i'm trying to figure out the best set up

well i did not figure out the meaning, but if you run it and let it autotune, it will choose automatically what's the best one (then you can add the flag in the batch file, like -l 112x2 for me)
dabster
Newbie
*
Offline Offline

Activity: 16


View Profile
April 14, 2013, 08:06:32 AM
 #225

This is what I am getting with my gtx 680

[2013-04-14 03:56:23] GPU #0:  186.17 khash/s with configuration  80x2
[2013-04-14 03:56:23] GPU #0: using launch configuration  80x2

[2013-04-14 04:06:23] GPU #0: GeForce GTX 680, 1448960 hashes, 181.06 khash/s
[2013-04-14 04:06:23] accepted: 8/11 (72.73%), 181.06 khash/s (yay!!!)
[2013-04-14 04:07:03] GPU #0: GeForce GTX 680, 7424000 hashes, 181.87 khash/s
[2013-04-14 04:07:32] DEBUG: got new work in 131 ms
[2013-04-14 04:07:35] DEBUG: hash <= target


Not bad.
Notanon
Sr. Member
****
Offline Offline

Activity: 378


Pastor of Muppets


View Profile
April 14, 2013, 08:22:57 AM
 #226

Trying out the 10/4 release, it's working better than the initial release and am getting about 7.9kh/s on my GT 220M, which is 33% better than using Scryptminer. Wasn't expecting miracles given how old the card in my laptop is, but it is a nice boost all the same. Grin

Edit: and the 14/4 release is giving me 8.54kh/s.

AMHash
ASICMINERROCKMINER ● Purchase from: AMHash (20Th/s min) ● Havelock (1Gh/s min)
Cloud-mining contracts: 0.0012 BTC per Gh ● Maintenance fee: $0.001551 per Gh per day ● Upto 6% Christmas Bonus

cbuchner1
Hero Member
*****
Offline Offline

Activity: 742


View Profile
April 14, 2013, 10:25:39 AM
 #227

also gettings lots of crashes after like 20 minutes of use.

I've seen one crash after letting it run through the night. This is actually the first time I've had cudaminer seen crashing during regular operation.
Weirds me out.

cbuchner1
Hero Member
*****
Offline Offline

Activity: 742


View Profile
April 14, 2013, 10:30:15 AM
 #228

Can someone explain this "64x2" "S27x3,28x4" thing to me or point me in the right direction on reading up on it??

I have a 580gtx and i'm trying to figure out the best set up

well i did not figure out the meaning, but if you run it and let it autotune, it will choose automatically what's the best one (then you can add the flag in the batch file, like -l 112x2 for me)

112x2 means it throws 112 blocks at CUDA, and each consists of 2 warps. A warp is a group of 32 threads.

So in total it computes 112*2*32 = 7168 hashes in parallel in a single CUDA kernel launch.

And because the scrypt scratchpad is 131072 bytes long, this would consume 7168*131072 bytes of memory
on the card. That's about 917 MB.
cbuchner1
Hero Member
*****
Offline Offline

Activity: 742


View Profile
April 14, 2013, 10:49:16 AM
 #229

Has anyone successfully to connected to the weminteltc.com pool?

I also tried just http://eu.wemineltc.com:3333

I downloaded their stratum proxy executable, it connects, but only solo mines packets.

Not sure what I need to populate for the connection address at this point.


They've just updated their windows stratum binary to default to scrypt. Previously you had to explicitly instruct it to use scrypt for proof of work.
Also note that it opens local port 8332

It's kind of cool to have a stratum proxy that requires zero setup. Just click the EXE and you're good to point cudaminer at it.

So you need to direct cudaminer to -o http://localhost:8332 -O yourworkername.1:yourpassword

If you used port 9332 and your litecoin app ran in -server mode, then yes, you might have been solo mining. Wink
cbuchner1
Hero Member
*****
Offline Offline

Activity: 742


View Profile
April 14, 2013, 11:04:24 AM
 #230

Trust me, you want to compile this for 32 bit.

Compilation instruction for Fedora:
http://www.patrickmin.com/linux/tip.php?name=bitcoin_litecoin

On ubuntu and debian, you need the
g++-multilib and ia32-lib packages installed and possibly the libcurl4-dev:i386 package as well

Christian
SubNoize
Jr. Member
*
Offline Offline

Activity: 47


View Profile
April 14, 2013, 11:17:58 AM
 #231

Can someone explain this "64x2" "S27x3,28x4" thing to me or point me in the right direction on reading up on it??

I have a 580gtx and i'm trying to figure out the best set up

well i did not figure out the meaning, but if you run it and let it autotune, it will choose automatically what's the best one (then you can add the flag in the batch file, like -l 112x2 for me)

112x2 means it throws 112 blocks at CUDA, and each consists of 2 warps. A warp is a group of 32 threads.

So in total it computes 112*2*32 = 7168 hashes in parallel in a single CUDA kernel launch.

And because the scrypt scratchpad is 131072 bytes long, this would consume 7168*131072 bytes of memory
on the card. That's about 917 MB.


cheers! are there guides for these things? or is it just trial and error? Sorry i don't want to annoy you too much but do you know what would be some settings to throw at a 580gtx to see any improvement or should i just let the auto tune take care of it and not worry?

Also is there any information i can give you when it crashes?
to3m
Newbie
*
Offline Offline

Activity: 26


View Profile
April 14, 2013, 11:43:52 AM
 #232

also gettings lots of crashes after like 20 minutes of use.

I've seen one crash after letting it run through the night. This is actually the first time I've had cudaminer seen crashing during regular operation.
Weirds me out.



I got a std::bad_alloc from uint32x4_t::operator new after leaving it for about 8-9 hours. The process had used 1.7GB so I assume there's a memory leak somewhere... I think one problem is that in scanhash_scrypt, the 2 Xx4 buffers are allowed to leak.

--Tom
cbuchner1
Hero Member
*****
Offline Offline

Activity: 742


View Profile
April 14, 2013, 12:12:45 PM
 #233

Why can't it be 64-bit? I can only chroot to gentoo-prefix (which does not support multilib) to compile cudaminer.

Mostly due to lack of testing opportunity (at the moment),  and different priorities of the developer.

You could try disabling the "tex" kernels (comment them out, don't call them) to see if it builds in 64bits.
The texture caching feature was a failure, anyway.

But there's no guarantee that it will compute correctly in 64 bits. At least the CPU will validate this now.

cbuchner1
Hero Member
*****
Offline Offline

Activity: 742


View Profile
April 14, 2013, 12:13:45 PM
 #234

I got a std::bad_alloc from uint32x4_t::operator new after leaving it for about 8-9 hours. The process had used 1.7GB so I assume there's a memory leak somewhere... I think one problem is that in scanhash_scrypt, the 2 Xx4 buffers are allowed to leak.

That explains my crash that I saw in the morning. I will be fixing that pronto.

UPDATE!  the leak fix version is posted

Some latest performance stats:
GTX 460  -l28x4   100kHash/s
GTX 560Ti -l32x4  144kHash/s

Now if I tear out the power hungry GTX 260 and put in a GT 640, I should be able to hit close to 300kHash and NOT blow out my PSU.
Notanon
Sr. Member
****
Offline Offline

Activity: 378


Pastor of Muppets


View Profile
April 14, 2013, 12:18:15 PM
 #235

Just switched to mining Mincoins on the vircurpool pool and the hash rate on the GT 220M is now averaging around 9.8kh/s. interesting that it's a bit higher compared to the Ozcoin Litecoin pool.

AMHash
ASICMINERROCKMINER ● Purchase from: AMHash (20Th/s min) ● Havelock (1Gh/s min)
Cloud-mining contracts: 0.0012 BTC per Gh ● Maintenance fee: $0.001551 per Gh per day ● Upto 6% Christmas Bonus

cbuchner1
Hero Member
*****
Offline Offline

Activity: 742


View Profile
April 14, 2013, 12:25:31 PM
 #236

cheers! are there guides for these things? or is it just trial and error? Sorry i don't want to annoy you too much but do you know what would be some settings to throw at a 580gtx to see any improvement or should i just let the auto tune take care of it and not worry?

Trust the autotune. Try tuning a couple of times, note down the configs and speeds that it found - and then pick the one that consistently puts out the highest kHash.

This is how I just found my 32x4 and 28x4 configurations for my GTX 560Ti and GTX 460 cards.

The thing is, with every cudaminer update, these ideal configurations can change. Even overclocking your card can move the ideal configuration to some other "sweet spot".

Christian
nst6563
Sr. Member
****
Offline Offline

Activity: 252


View Profile
April 14, 2013, 01:29:48 PM
 #237

I definately like the speed improvement on the last update.  I went from a total hashrate of 230kh/s (2 gtx560se's and a gt430) to 264kh/s.

Unfortunately the hashrate of my gtx260 went down by 10kh/s.  I plan on replacing it soon anyway.

Was I helpful? BTC: 1B5SzPsBDuGLvGYuWw5b4Po8MaFSGck1Dt    LTC: LSQEM9re967ds5ZTHxs6SUvmKVpB5KdMwZ    DOGE:  DS3HN4s4YainUsiwrFcAoXa4e3WiSaM74P
cbuchner1
Hero Member
*****
Offline Offline

Activity: 742


View Profile
April 14, 2013, 01:39:06 PM
 #238

Unfortunately the hashrate of my gtx260 went down by 10kh/s.  I plan on replacing it soon anyway.

I will see what I can do about restoring the performance of the "S" kernels... Also I am trying to fix the driver crash on CTRL-C

Old vs. New.

GT 640 4GB 128bit GDDR3 memory
vs.
GTX 260 896MB 448-bit memory

Both do about same kHash/s, but difference in power draw is factor 3 or 4.
My PC is now doing 280 kHash/sec on 3 GPUs. Wink


cbuchner1
Hero Member
*****
Offline Offline

Activity: 742


View Profile
April 14, 2013, 04:50:13 PM
 #239

Hey, the CTRL-C issue has been fixed. Grab the update from the first posting.

Don't directly close the DOS box with the X button though, that doesn't leave my program enough time to do cleanup.

The CPU utilization is not really predictable. Sometimes it's just 10%, other times  it goes up to 40%. Weird.


TiWu
Newbie
*
Offline Offline

Activity: 25


View Profile
April 14, 2013, 05:06:55 PM
 #240

GTX 570 Club3D
Win 8 x64 - 314.22 WHQL driver

  • 2013-04-09: 120x2 150 kh/s
  • 2013-04-10: 30x7  175 kh/s
  • 2013-04-14: 30x7  210 kh/s (update 2, crashes on CTRL-C)
  • 2013-04-14: 30x8  220 kh/s (update 3)

Autotune for latest (update 3) version, no debug:
Code:
[2013-04-14 18:57:01] GPU #0: GeForce GTX 570 with compute capability 2.0
[2013-04-14 18:57:01] GPU #0: interactive: 0, tex-cache: 0, single-alloc: 0
[2013-04-14 18:57:01] GPU #0: Performing auto-tuning (Patience...)
[2013-04-14 18:58:26] GPU #0:  222.38 khash/s with configuration  30x8
[2013-04-14 18:58:26] GPU #0: using launch configuration  30x8
[2013-04-14 18:58:26] GPU #0: GeForce GTX 570, 7680 hashes, 0.09 khash/s
[2013-04-14 18:58:27] GPU #0: GeForce GTX 570, 7680 hashes, 119.87 khash/s
[2013-04-14 18:58:37] GPU #0: GeForce GTX 570, 2242560 hashes, 221.51 khash/s
[2013-04-14 18:58:37] accepted: 1/1 (100.00%), 221.51 khash/s (yay!!!)

Shutdown issue is handled nicely too.
Code:
[2013-04-14 19:00:37] accepted: 13/14 (92.86%), 217.86 khash/s (yay!!!)
Ctrl-C
[2013-04-14 19:00:38] workio thread dead, waiting for workers...
[2013-04-14 19:00:38] worker threads all shut down
Terminate batch job (Y/N)?
Pages: « 1 2 3 4 5 6 7 8 9 10 11 [12] 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 ... 1142 »
  Print  
 
Jump to:  

Sponsored by , a Bitcoin-accepting VPN.
Powered by MySQL Powered by PHP Powered by SMF 1.1.19 | SMF © 2006-2009, Simple Machines Valid XHTML 1.0! Valid CSS!