DiabloMiner GPU Miner

[kindle]

Thanks diablo I guess I will have to make do with the hash rate till amd comes out with the new sdk for 6990. Thanks for your great miner, I am using it for the dual 6990. It hovers around 1ghash/s and its at -f 7777777. Hahah I tested and the numbers seem to slow down the hash rate when its lesser than -f 17 but anything more it hits 1ghash/s constant. Which is nice =)
GPU Max at 93% with powerplay maxed out

It generally maxes out at your OS's kernel's scheduler's timeslice length. Generally, this is 100hz or 1000hz. Higher than -f 1000 makes no sense. Not only that, my miner has an automatic lower limit as to not send too little work to the GPU. The actual -f maximum may be much lower.

Wait I dont quite get the part OS kernel scheduler timeslice length. What I do face is a huge drop in hash/s when i go lower than -f 15, at -f 2 its almost 600mhash/s so thats why i up the -f number to go it to run at 1ghash

[1713507533]

1713507533

[1713507533]

1713507533

[1713507533]

1713507533

[1713507533]

1713507533

[DiabloD3]

Thanks diablo I guess I will have to make do with the hash rate till amd comes out with the new sdk for 6990. Thanks for your great miner, I am using it for the dual 6990. It hovers around 1ghash/s and its at -f 7777777. Hahah I tested and the numbers seem to slow down the hash rate when its lesser than -f 17 but anything more it hits 1ghash/s constant. Which is nice =)
GPU Max at 93% with powerplay maxed out

It generally maxes out at your OS's kernel's scheduler's timeslice length. Generally, this is 100hz or 1000hz. Higher than -f 1000 makes no sense. Not only that, my miner has an automatic lower limit as to not send too little work to the GPU. The actual -f maximum may be much lower.

Wait I dont quite get the part OS kernel scheduler timeslice length. What I do face is a huge drop in hash/s when i go lower than -f 15, at -f 2 its almost 600mhash/s so thats why i up the -f number to go it to run at 1ghash

Basically, the GPU babysitting threads go to sleep between OpenCL executions. The OS's kernel can only change threads so fast. Thus, having OpenCL executions that are shorter than the length of one timeslice means the GPU is waiting for the babysitting threads to catch up. However, as I said, the minimum kernel length is tuned to your hardware, it doesn't seem you can ever reach the timeslice length.

As for -f 1 looking slower, its because at extremely long execution lengths, it tends to screw over the OS's clock in as that it jumps large periods to catch up to the real world. This is why, when using -f 1, you use the second number my miner outputs and run it for 10 minutes before you trust the number.

Linux doesn't suffer this nearly as much as Windows does.

[kindle]

Hi diablo, thanks for the prompt reply. I have a plan now and that is to mix my dual 6990 with the 5850 + 5970. Hopefully windows will recognise both as opencl and no need to run xfire. =P

[Cerebrum]

Does DiabloMiner support the Long Polling protocol that DeepBit has introduced? If not, do you have any plans to implement it?

[DiabloD3]

Does DiabloMiner support the Long Polling protocol that DeepBit has introduced? If not, do you have any plans to implement it?

I don't see value in the long polling protocol. Its much easier to just use a dedicated Websockets protocol. I plan on introducing this in the future.

[DiabloD3]

Update: I've added mtrlt's bat script for Windows. No longer do you have to run the whole Java command by hand.

[DiabloD3]

Update: I've added -z, so you can now change the size of the in kernel loop. The number is the exponent of power of two, 0 turns it off. Default is -z 10 (1024) on Radeon 5xxx/6xxx, -z 0 on everything else.

[Randall Flagg]

Github repo, issue tracker, and wiki
Newest binary, always updated to match newest git revision

Works on AMD Radeon 5xxx and up (SDK 2.1 or 2.4 recommended, avoid 2.2 and 2.3), and Nvidia 8xxx and up (make sure you have their newest driver).
...

Note, in addition to the previous notes: OSX users, you need at least OSX 10.6 (Snow Leopard) to run OpenCL apps of any kind.

Hi, i run seach about my problem without luck...

i have 2 questions

1)
I have those error in my console :

Code:

[20/04/11 12:08:27] Block 290 found on ATI Radeon HD 5770 (#1)
[20/04/11 12:10:40] ERROR: Invalid block found on ATI Radeon HD 5770 (#1), possible driver or hardware issue
[20/04/11 12:10:47] Block 291 found on ATI Radeon HD 5770 (#1)

So i run the latest version of Diablo, with mac os 10.6.6, what can be the error ?

2)
Also, with the default setting (-w 64) i had 70 M/hash sec, i have tried with -w 128, now i have 105 Mhash, but i am supposed to be at over 150... i have tried -w 128, 192, 256 without any luck (the program don`t start)

thanks !

[DiabloD3]

Github repo, issue tracker, and wiki
Newest binary, always updated to match newest git revision

Works on AMD Radeon 5xxx and up (SDK 2.1 or 2.4 recommended, avoid 2.2 and 2.3), and Nvidia 8xxx and up (make sure you have their newest driver).
...

Note, in addition to the previous notes: OSX users, you need at least OSX 10.6 (Snow Leopard) to run OpenCL apps of any kind.

Hi, i run seach about my problem without luck...

i have 2 questions

1)
I have those error in my console :

Code:

[20/04/11 12:08:27] Block 290 found on ATI Radeon HD 5770 (#1)
[20/04/11 12:10:40] ERROR: Invalid block found on ATI Radeon HD 5770 (#1), possible driver or hardware issue
[20/04/11 12:10:47] Block 291 found on ATI Radeon HD 5770 (#1)

So i run the latest version of Diablo, with mac os 10.6.6, what can be the error ?

2)
Also, with the default setting (-w 64) i had 70 M/hash sec, i have tried with -w 128, now i have 105 Mhash, but i am supposed to be at over 150... i have tried -w 128, 192, 256 without any luck (the program don`t start)

thanks !

Apple wrote their own OpenCL implementation for OSX instead of using far more optimized ones supplied by AMD or Nvidia. Using OSX means you lose about 20% of your speed right off the bat. Try toying with -z 0 in combination with -w 128.

There is pretty much no way to fix this, Apple loves screwing customers.

[Randall Flagg]

Apple wrote their own OpenCL implementation for OSX instead of using far more optimized ones supplied by AMD or Nvidia. Using OSX means you lose about 20% of your speed right off the bat. Try toying with -z 0 in combination with -w 128.

Unrecognized option: -z

So i have the binary and the option isn't there.  I had to download the source and compiled it.   I have no speed increase at all...

btw Will be great to have more documentation about what the switch does exactly and how to compile it, i had to dig for that

[DiabloD3]

Apple wrote their own OpenCL implementation for OSX instead of using far more optimized ones supplied by AMD or Nvidia. Using OSX means you lose about 20% of your speed right off the bat. Try toying with -z 0 in combination with -w 128.

Unrecognized option: -z

So i have the binary and the option isn't there.  I had to download the source and compiled it.   I have no speed increase at all...

btw Will be great to have more documentation about what the switch does exactly and how to compile it, i had to dig for that

I only recently added it. Make sure you have the absolute newest binary. Try using somewhere between -z 0 and -z 20, you might get speed increases, you might not. I don't think OSX's impl has bitalign, so you're losing a lot of performance and there might be no way to fix it.

[Jim Hyslop]

Hi,

I'm trying to get Diablo Miner running on my Debian box. I'm using the latest DiabloMiner.zip, ATI drivers v11.3 and the 2.4 SDK. When I start DiabloMiner, it displays two lines of text, then crashes my logon session. I think the second line says "Connecting to (somethingorother)" but it disappears too quickly to tell.

I have two Radeon 6870 cards.

Any help would be appreciated!

[DiabloD3]

Hi,

I'm trying to get Diablo Miner running on my Debian box. I'm using the latest DiabloMiner.zip, ATI drivers v11.3 and the 2.4 SDK. When I start DiabloMiner, it displays two lines of text, then crashes my logon session. I think the second line says "Connecting to (somethingorother)" but it disappears too quickly to tell.

I have two Radeon 6870 cards.

Any help would be appreciated!

You're supposed to start DiabloMiner in a terminal. It is not a GUI app.

[Jim Hyslop]

You're supposed to start DiabloMiner in a terminal. It is not a GUI app.

I did start it from a terminal.

[DiabloD3]

You're supposed to start DiabloMiner in a terminal. It is not a GUI app.

I did start it from a terminal.

Please post the entire error, then.

[Jim Hyslop]

You're supposed to start DiabloMiner in a terminal. It is not a GUI app.

I did start it from a terminal.

Please post the entire error, then.

I can't - the terminal displays two messages, which look normal, then disappears along with my login session. :-(

I just checked the syslog, and there are no errors logged, but I did notice on login I get some warnings:

Code:

Apr 24 18:43:08 griphook AptDaemon: INFO: Initializing daemon
Apr 24 18:43:31 griphook acpid: client 9131[0:0] has disconnected
Apr 24 18:43:31 griphook acpid: client connected from 10969[0:0]
Apr 24 18:43:31 griphook acpid: 1 client rule loaded
Apr 24 18:43:32 griphook gnome-session[10994]: atk-bridge-WARNING: AT_SPI_REGISTRY was not started at session startup.
Apr 24 18:43:32 griphook gnome-session[10994]: atk-bridge-WARNING: IOR not set.
Apr 24 18:43:32 griphook gnome-session[10994]: atk-bridge-WARNING: Could not locate registry
Apr 24 18:43:32 griphook gdm-simple-greeter[11010]: Gtk-WARNING: /scratch/build-area/gtk+2.0-2.20.1/gtk/gtkwidget.c:5628: widget not within a GtkWindow

/var/log/Xorg.0.log has some interesting, and possibly useful info:

Code:

(EE) fglrx(1): Multiview is only supported with acceleration; this screen will now shutdown.
(EE) fglrx(1): PreInit failed
(II) fglrx(1): === [xdl_x750_atiddxPreInit] === end
(EE) fglrx(0): Failed shutdown interrupts. Error -9
(II) UnloadModule: "fglrx"
(II) UnloadModule: "fb"
(II) UnloadModule: "fglrxdrm"
(II) UnloadModule: "fglrxdrm"
(--) Depth 24 pixmap format is 32 bpp
(EE) fglrx(0): atiddxDriScreenInit failed, GPS not been initialized. 
(WW) fglrx(0): ***********************************************************
(WW) fglrx(0): * DRI initialization failed                               *
(WW) fglrx(0): * kernel module (fglrx.ko) may be missing or incompatible *
(WW) fglrx(0): * 2D and 3D acceleration disabled                         *
(WW) fglrx(0): ***********************************************************

Code:

(II) Module amdxmm: vendor="X.Org Foundation"
        compiled for 1.4.99.906, module version = 2.0.0
(EE) fglrx(0): XMM failed to open CMMQS connection.(EE) fglrx(0): 
(EE) fglrx(0): XMM failed to initialize

Code:

(WW) fglrx(0): board is an unknown third party board, chipset is supported
(EE) fglrx(0): ACPI: DRM connection failed
(WW) fglrx(0): Hasn't establisted DRM connection
(II) fglrx(0): [FB] MC range(MCFBBase = 0xf00000000, MCFBSize = 0x40000000)
(WW) fglrx(0): No DRM connection for driver fglrx.

and this one:

Code:

(II) fglrx(0): Modeline "1440x900"x0.0  136.75  1440 1536 1688 1936  900 903 909 942 -hsync +vsync (70.6 kHz)

Backtrace:
0: /usr/bin/Xorg (xorg_backtrace+0x28) [0x4acd58]
1: /usr/bin/Xorg (0x400000+0x61a09) [0x461a09]
2: /lib/libpthread.so.0 (0x7f7c7c125000+0xef60) [0x7f7c7c133f60]
3: /usr/lib/xorg/modules/drivers/fglrx_drv.so (xdl_x750_swlDriOpenConnection+0x25) [0x7f7c784a28d5]
4: /usr/lib/xorg/modules/drivers/fglrx_drv.so (0x7f7c780bb000+0x3d7150) [0x7f7c78492150]
5: /usr/lib/xorg/modules/drivers/fglrx_drv.so (0x7f7c780bb000+0x3d3d82) [0x7f7c7848ed82]
6: /usr/bin/Xorg (0x400000+0x48c84) [0x448c84]
7: /usr/bin/Xorg (0x400000+0x25c35) [0x425c35]
8: /lib/libc.so.6 (__libc_start_main+0xfd) [0x7f7c7ac29c4d]
9: /usr/bin/Xorg (0x400000+0x257e9) [0x4257e9]
Segmentation fault at address 0x8

Fatal server error:
Caught signal 11 (Segmentation fault). Server aborting


Please consult the The X.Org Foundation support 
         at http://wiki.x.org
 for help. 
Please also check the log file at "/var/log/Xorg.0.log" for additional information.

Any other logs I should look at?

[DiabloD3]

You're supposed to start DiabloMiner in a terminal. It is not a GUI app.

I did start it from a terminal.

Please post the entire error, then.

I can't - the terminal displays two messages, which look normal, then disappears along with my login session. :-(

I just checked the syslog, and there are no errors logged, but I did notice on login I get some warnings:

Code:

Apr 24 18:43:08 griphook AptDaemon: INFO: Initializing daemon
Apr 24 18:43:31 griphook acpid: client 9131[0:0] has disconnected
Apr 24 18:43:31 griphook acpid: client connected from 10969[0:0]
Apr 24 18:43:31 griphook acpid: 1 client rule loaded
Apr 24 18:43:32 griphook gnome-session[10994]: atk-bridge-WARNING: AT_SPI_REGISTRY was not started at session startup.
Apr 24 18:43:32 griphook gnome-session[10994]: atk-bridge-WARNING: IOR not set.
Apr 24 18:43:32 griphook gnome-session[10994]: atk-bridge-WARNING: Could not locate registry
Apr 24 18:43:32 griphook gdm-simple-greeter[11010]: Gtk-WARNING: /scratch/build-area/gtk+2.0-2.20.1/gtk/gtkwidget.c:5628: widget not within a GtkWindow

/var/log/Xorg.0.log has some interesting, and possibly useful info:

Code:

(EE) fglrx(1): Multiview is only supported with acceleration; this screen will now shutdown.
(EE) fglrx(1): PreInit failed
(II) fglrx(1): === [xdl_x750_atiddxPreInit] === end
(EE) fglrx(0): Failed shutdown interrupts. Error -9
(II) UnloadModule: "fglrx"
(II) UnloadModule: "fb"
(II) UnloadModule: "fglrxdrm"
(II) UnloadModule: "fglrxdrm"
(--) Depth 24 pixmap format is 32 bpp
(EE) fglrx(0): atiddxDriScreenInit failed, GPS not been initialized. 
(WW) fglrx(0): ***********************************************************
(WW) fglrx(0): * DRI initialization failed                               *
(WW) fglrx(0): * kernel module (fglrx.ko) may be missing or incompatible *
(WW) fglrx(0): * 2D and 3D acceleration disabled                         *
(WW) fglrx(0): ***********************************************************

Code:

(II) Module amdxmm: vendor="X.Org Foundation"
        compiled for 1.4.99.906, module version = 2.0.0
(EE) fglrx(0): XMM failed to open CMMQS connection.(EE) fglrx(0): 
(EE) fglrx(0): XMM failed to initialize

Code:

(WW) fglrx(0): board is an unknown third party board, chipset is supported
(EE) fglrx(0): ACPI: DRM connection failed
(WW) fglrx(0): Hasn't establisted DRM connection
(II) fglrx(0): [FB] MC range(MCFBBase = 0xf00000000, MCFBSize = 0x40000000)
(WW) fglrx(0): No DRM connection for driver fglrx.

and this one:

Code:

(II) fglrx(0): Modeline "1440x900"x0.0  136.75  1440 1536 1688 1936  900 903 909 942 -hsync +vsync (70.6 kHz)

Backtrace:
0: /usr/bin/Xorg (xorg_backtrace+0x28) [0x4acd58]
1: /usr/bin/Xorg (0x400000+0x61a09) [0x461a09]
2: /lib/libpthread.so.0 (0x7f7c7c125000+0xef60) [0x7f7c7c133f60]
3: /usr/lib/xorg/modules/drivers/fglrx_drv.so (xdl_x750_swlDriOpenConnection+0x25) [0x7f7c784a28d5]
4: /usr/lib/xorg/modules/drivers/fglrx_drv.so (0x7f7c780bb000+0x3d7150) [0x7f7c78492150]
5: /usr/lib/xorg/modules/drivers/fglrx_drv.so (0x7f7c780bb000+0x3d3d82) [0x7f7c7848ed82]
6: /usr/bin/Xorg (0x400000+0x48c84) [0x448c84]
7: /usr/bin/Xorg (0x400000+0x25c35) [0x425c35]
8: /lib/libc.so.6 (__libc_start_main+0xfd) [0x7f7c7ac29c4d]
9: /usr/bin/Xorg (0x400000+0x257e9) [0x4257e9]
Segmentation fault at address 0x8

Fatal server error:
Caught signal 11 (Segmentation fault). Server aborting


Please consult the The X.Org Foundation support 
         at http://wiki.x.org
 for help. 
Please also check the log file at "/var/log/Xorg.0.log" for additional information.

Any other logs I should look at?

Something is very wrong with your fglrx installation. Which video card do you own?

[Jim Hyslop]

Something is very wrong with your fglrx installation. Which video card do you own?

Two Radeon 6870s.

[DiabloD3]

Something is very wrong with your fglrx installation. Which video card do you own?

Two Radeon 6870s.

Make sure you have at least 11.1, and it was installed right, and you built the kernel module and its being loaded before X starts. Your 2D and 3D acceleration is dead, and OpenCL isn't going to work either. And since you'd be on 11.1 or later, make sure you're using SDK 2.4

[Jim Hyslop]

Make sure you have at least 11.1, and it was installed right, and you built the kernel module and its being loaded before X starts. Your 2D and 3D acceleration is dead, and OpenCL isn't going to work either. And since you'd be on 11.1 or later, make sure you're using SDK 2.4

Thanks for the tips. I have been a bit busy lately, so it might be a day or two before I get to this. Meanwhile, it's limping along CPU-mining :-/