[ANN] Opensource primecoin GPU miner beta-testing started! (Crossplatform)

Marketplace Primecoin
41

[quote=“primer10, post:37, topic:2223”][code]
[1] 2.032G, average: 2.035G
[2] 1.994G, average: 2.016G

  • speed: 4.026G, average: 4.050G

    • chains/1: 18242910 38004.886/sec
    • chains/2: 1390497 2896.779/sec
    • chains/3: 162057 337.608/sec
    • chains/4: 12720 26.499/sec
    • chains/5: 1071 2.231/sec
    • chains/6: 70 0.146/sec
    • chains/7: 3 0.006/sec 22.499/hour

[/code][/quote]
Now it maximum for Radeon R9 280X, wait speedups… coming soon =)

18-04-2014 00:29:00: Error JSON result is error
May be invalid authorization parameters... show your primecoind (or primecoin-qt) and xpmclminer command line arguments.
42

[quote=“eXtremal.ik7, post:41, topic:2223”][quote=“primer10, post:37, topic:2223”][code]
[1] 2.032G, average: 2.035G
[2] 1.994G, average: 2.016G

  • speed: 4.026G, average: 4.050G

    • chains/1: 18242910 38004.886/sec
    • chains/2: 1390497 2896.779/sec
    • chains/3: 162057 337.608/sec
    • chains/4: 12720 26.499/sec
    • chains/5: 1071 2.231/sec
    • chains/6: 70 0.146/sec
    • chains/7: 3 0.006/sec 22.499/hour

[/code][/quote]
Now it maximum for Radeon R9 280X, wait speedups… coming soon =)[/quote]

+1 !

can’t wait!! btw, mine’s 7950 not 280x

43

Hi eXtremal.ik7,

First of all thanks for your work and for releasing it as open-source.

I suggest you to add the following to your README.md:

[ul][li]A XPM address for donations so we may chip something to you when we start getting some blocks.[/li]
[li]libncurses as dependency. I was unable to compile without it.[/li]
[li]The -DOPENCL_INCLUDE_DIRS=/opt/AMDAPP/include flag for cmake. It seems most of us need it (myself included).[/li]
[li]The instruction to extract libamdocl64.so form Catalyst 14.03 and replace it in /opt/AMDAPP/lib/x86_64/libamdocl64.so or /usr/lib/fglrx/libamdocl64.so or whatever other paths distros may use.[/li]
[li]License (MIT? GPL?).[/li][/ul]

I am using your miner nicely in my rig with 3x R9 270X running Lubuntu 13.10. Processor is an Intel Core i7 930 @ 2.80 GHz.
In case it is of any use to you here goes the result of the benchmark: http://pastebin.com/55bBthyh I just copied the results for GPU 1, all the others were similar.
Cards are with core at 1100 MHz and memory at 1480 MHz.

Running for real it looks like this:

[code][1] 1.371G, average: 1.382G
[2] 1.359G, average: 1.348G
[3] 1.368G, average: 1.376G

  • speed: 4.099G, average: 4.107G

    • chains/1: 44334921 38718.338/sec
    • chains/2: 3392948 2963.111/sec
    • chains/3: 396160 345.972/sec
    • chains/4: 31064 27.129/sec
    • chains/5: 2544 2.222/sec
    • chains/6: 182 0.159/sec
    • chains/7: 13 0.011/sec 40.871/hour[/code]

Thanks again for all the work!

44

yes, primecoind already listen:

ubuntu@ubuntu:~/xpm/xpmminer/xpmminer/build$ netstat -nap | grep prime (Not all processes could be identified, non-owned process info will not be shown, you would have to be root to see it all.) tcp 0 0 0.0.0.0:9911 0.0.0.0:* LISTEN 15025/primecoind tcp 0 0 127.0.0.1:9912 0.0.0.0:* LISTEN 15025/primecoind

45

Just in case it is useful for anyone, here goes my start-up script for R9 270X:

#!/bin/sh export DISPLAY=:0 export GPU_MAX_ALLOC_PERCENT=100 export GPU_USE_SYNC_OBJECTS=1 sudo aticonfig --adapter=all --od-enable sudo aticonfig --adapter=all --odsc=1120,1500 sudo amdconfig --adapter=all --odcc xpmclminer -u user -p pass -w <XPM addr>

Now waiting for built-in overclock controls and pooled mining ::slight_smile:

46

Ок, I will try to build miner with NVidia SDK, but I have no GeForce GPUs to full test it.
Try replace line 264:
-uint32_t inverseModulosCurrent[primesPerThread];
+uint32_t inverseModulosCurrent[WeaveDepth / GroupSize];[/quote]
Anybody had any luck getting this to run with an NVIDIA card? Tried the above, which gives me the following error:

[code]OpenCL platform available: NVIDIA Corporation
:2:26: warning: unknown ‘#pragma OPENCL EXTENSION’ - ignored
#pragma OPENCL EXTENSION cl_amd_printf : enable
^
:50:3: warning: implicit declaration of function ‘printf’ is invalid in C99
printf("%08X %08X %08X %08X ", Reg.x, Reg.y, Reg.z, Reg.w);
^
:265:26: error: openCL does not support variable length arrays
uint32_t inverseModulos[primesPerThread]; // <— large buffer for global memory placing
^

Error: can’t compile OpenCL kernel /usr/local/share/xpmminer/kernel.cl[/code]

47

[quote=“eXtremal.ik7, post:41, topic:2223”]

18-04-2014 00:29:00: Error JSON result is error

May be invalid authorization parameters… show your primecoind (or primecoin-qt) and xpmclminer command line arguments.[/quote]

First, After all, it works! so nice~, thank you! (following is detail )

3*amd r290: default clock
[1] 3.170G, average: 3.137G
[2] 2.488G, average: 2.473G
[3] 2.399G, average: 2.374G
 * speed: 8.057G, average: 7.984G

   * chains/1: 78889089 75126.111/sec
   * chains/2: 6022605 5735.329/sec
   * chains/3: 702747 669.226/sec
   * chains/4: 55511 52.863/sec
   * chains/5: 4376 4.167/sec
   * chains/6: 316 0.301/sec
   * chains/7: 20 0.019/sec 68.566/hour
   * chains/8: 3 0.003/sec 10.285/hour

Things goes so strange… After checked all parameters, recompile again, it still print “json error”, then i git pull newest code and compiled, other error occurred, speed always keep 0.000G, so i git reset 530462c606d52a60e822c8cae33f94243d1f36bf, rm all build stuff, recompile it, it works! i have no idea what happens between it at all!

BTW: another issue: the GPU temperature is so high, because the fan speed is not adjust automatically,

[code]aticonfig --odgt --odgc --adapter=all

Adapter 0 - Supported device 67B1
Sensor 0: Temperature - 79.00 C

Adapter 1 - Supported device 67B1
Sensor 0: Temperature - 90.00 C

Adapter 2 - Supported device 67B1
Sensor 0: Temperature - 85.00 C

Adapter 0 - Supported device 67B1
Core (MHz) Memory (MHz)
Current Clocks : 975 1250
Current Peak : 975 1250
Configurable Peak Range : [300-1500] [150-2000]
GPU load : 97%

Adapter 1 - Supported device 67B1
Core (MHz) Memory (MHz)
Current Clocks : 946 1250
Current Peak : 947 1250
Configurable Peak Range : [300-1500] [150-2000]
GPU load : 100%

Adapter 2 - Supported device 67B1
Core (MHz) Memory (MHz)
Current Clocks : 947 1250
Current Peak : 947 1250
Configurable Peak Range : [300-1500] [150-2000]
GPU load : 100%
[/code]
@this temperature, FAN speed only reach 33 percentage:

[code]ubuntu@ubuntu:~/xpm/xpmminer/xpmminer/build$ env DISPLAY=:0.0 aticonfig --pplib-cmd “get fanspeed 0”
Fan speed query:
Query Index: 0, Speed in percent
Result: Fan Speed: 33%

ubuntu@ubuntu:~/xpm/xpmminer/xpmminer/build$ env DISPLAY=:0.1 aticonfig --pplib-cmd “get fanspeed 0”
Fan speed query:
Query Index: 0, Speed in percent
Result: Fan Speed: 31%

ubuntu@ubuntu:~/xpm/xpmminer/xpmminer/build$ env DISPLAY=:0.2 aticonfig --pplib-cmd “get fanspeed 0”
Fan speed query:
Query Index: 0, Speed in percent
Result: Fan Speed: 33%[/code]

by now, i forced set fanspeed to 70%, the GPU temp keep ~75celsuis.

We know the temp control function will added soon, eXtremal.ik7 thank you for your work!

48

Latest version builds on NVidia and can run some benchmarks (including mining benchmark), mining not working.

3*amd r290: default clock [1] 3.170G, average: 3.137G [2] 2.488G, average: 2.473G [3] 2.399G, average: 2.374G
Use latest version, performance drop on multi-gpu configurations fixed (all your 3 GPU will be show ~3100G). If you can't start miner (all counters show 0), try restart primecoind.
Now waiting for built-in overclock controls and pooled mining
For overclocking and temperature & fan speed control - start cgminer in parallel with wrong pool address.. it will be not mining, but GPU control functions still works...
49

Anyone found a block yet with this ?

I think i found one, saw the wallet notification but disappeared in the transactions list … orphaned?

50

[quote=“eXtremal.ik7, post:48, topic:2223”]Use latest version, performance drop on multi-gpu configurations fixed (all your 3 GPU will be show ~3100G). If you can’t start miner (all counters show 0), try restart primecoind.

For overclocking and temperature & fan speed control - start cgminer in parallel with wrong pool address… it will be not mining, but GPU control functions still works…[/quote]

will try the cgminer method! btw, do we need to reach the cgminer main display where we can set the gpu control?

51
btw, do we need to reach the cgminer main display where we can set the gpu control?
For initial setup GPU & memory clocks, fan speed, powertune - no, but I not test control features, initial setup enough for me.
52

[quote=“eXtremal.ik7, post:48, topic:2223”]

3*amd r290: default clock
[1] 3.170G, average: 3.137G
[2] 2.488G, average: 2.473G
[3] 2.399G, average: 2.374G

Use latest version, performance drop on multi-gpu configurations fixed (all your 3 GPU will be show ~3100G). If you can’t start miner (all counters show 0), try restart primecoind.[/quote]

yeah, just as you said, average speed per card is about 3.1G :slight_smile:

[code][1] 2.620G, average: 3.194G
[2] 3.239G, average: 3.177G
[3] 3.067G, average: 3.020G

  • speed: 8.927G, average: 9.391G[/code]
53

Some statistic: primorials in last 25000 blocks:

primorials[0] = 0 primorials[1] = 0 primorials[2] = 1 primorials[3] = 3 primorials[4] = 17 primorials[5] = 70 primorials[6] = 227 primorials[7] = 408 primorials[8] = 728 primorials[9] = 1112 primorials[10] = 1765 primorials[11] = 5430 primorials[12] = 6656 primorials[13] = 4520 primorials[14] = 2314 primorials[15] = 1144 primorials[16] = 391 primorials[17] = 136 primorials[18] = 42 primorials[19] = 16 primorials[20] = 7 primorials[21] = 6 primorials[22] = 2 primorials[23] = 2 primorials[24] = 0
It seems, Claymore and primegpu uses primorials 11-13.. 448-bit arithmetic implementation is wrong way. I think they uses 356 arithmetic (or 320-bit on first steps) and more fast sieve implementation. With primorial 12 xpmclminer shows 4.700G speed on Radeon R9 290X but finds less sub-chains (1-9)... May be reason is a different sieve configuration.
54

[quote=“eXtremal.ik7, post:53, topic:2223”]Some statistic: primorials in last 25000 blocks:

primorials[0] = 0 primorials[1] = 0 primorials[2] = 1 primorials[3] = 3 primorials[4] = 17 primorials[5] = 70 primorials[6] = 227 primorials[7] = 408 primorials[8] = 728 primorials[9] = 1112 primorials[10] = 1765 primorials[11] = 5430 primorials[12] = 6656 primorials[13] = 4520 primorials[14] = 2314 primorials[15] = 1144 primorials[16] = 391 primorials[17] = 136 primorials[18] = 42 primorials[19] = 16 primorials[20] = 7 primorials[21] = 6 primorials[22] = 2 primorials[23] = 2 primorials[24] = 0
It seems, Claymore and primegpu uses primorials 11-13.. 448-bit arithmetic implementation is wrong way. I think they uses 356 arithmetic (or 320-bit on first steps) and more fast sieve implementation. With primorial 12 xpmclminer shows 4.700G speed on Radeon R9 290X but finds less sub-chains (1-9)... May be reason is a different sieve configuration.[/quote]

should we be using 12 as well? what is the impact of less sub chains?

55
should we be using 12 as well? what is the impact of less sub chains?
Now - not, my sieve implementation too slow now for use primorial 12. If you want increase performance, add this arguments:
--extensions-num 11 --weave-depth 8960
56

[quote=“eXtremal.ik7, post:55, topic:2223”]If you want increase performance, add this arguments:

--extensions-num 11 --weave-depth 8960
[/quote]

Wow indeed! It is like a cheat code! increased from 4G to 4.7G

Thanks!!

57

so no support for 32 bit OS at all ? :frowning:

58

??? I think we can compile 32bits, no?

59

Btw what do the 2 params mean?

60

switching rigs from BAMT to PIMP, tested xpmminer with PIMP, running fine :slight_smile: