DiabloMiner GPU Miner

[arcalium]

It gives me the same result each time I run it.

It appears in my process list but I never see any CPU consumption at all and GPU Caps doesn't show anything changing as I launch it.  I'm on Windows Vista Ultimate 64 SP2, if there are any known issues.

If I run it with -D 0, it doesn't write this line to the terminal:

[6/9/11 10:12:49 AM] Added GeForce 9600 GSO (#1) (12 CU, local work size of 512)

and instead writes "waiting..." and then replaces that with "0/0 khash/sec"

Is there any diagnostic info that I can provide that would tell you what's going wrong?

I'm getting the same kind of issue with Window 7 64 bit.

If I run Diablo-Windows.exe with no params, it will start hashing using the GPU, but can't connect to bitcoin. 
If I use the -u -p params (Diablo-Windows.exe -u uid -p pwd) it just sits there after adding the GPU
Added GeForce GTX 580 (#1) (16 CU, local work size of 1024)

[1714218165]

1714218165

[1714218165]

1714218165

[1714218165]

1714218165

[DiabloD3]

It gives me the same result each time I run it.

It appears in my process list but I never see any CPU consumption at all and GPU Caps doesn't show anything changing as I launch it.  I'm on Windows Vista Ultimate 64 SP2, if there are any known issues.

If I run it with -D 0, it doesn't write this line to the terminal:

[6/9/11 10:12:49 AM] Added GeForce 9600 GSO (#1) (12 CU, local work size of 512)

and instead writes "waiting..." and then replaces that with "0/0 khash/sec"

Is there any diagnostic info that I can provide that would tell you what's going wrong?

-D 0 isn't valid, thats why. It starts at 1, and you shouldn't be using -D anyhow unless you have a very specific reason to.

[krom]

I had the same problem when setting up the miner. There seems to be a problem with the most recent version of bitcoin (0.3.22-beta). Try version 0.3.21-beta and it should work - given that the rest of your setup is okay.

[DiabloD3]

I had the same problem when setting up the miner. There seems to be a problem with the most recent version of bitcoin (0.3.22-beta). Try version 0.3.21-beta and it should work - given that the rest of your setup is okay.

What, seriously? Thats bad.

[frisco]

I got yesterday a MSI 6950 rev.1 I was difficult to find a unlockable 6950 but still there are some models out there.
So as it was going to be my main miner for some time (still have a 5770 but I can have both on my computer due to the PSU) I made a real table of doom with every vector option and most of the worksizes for the best -vs most of the measures are no more than 5 minutes enough to see the possible outcome with more fine tunning.
I remeasured the best during the whole process, stock -> unlock shaders -> be aggresive -> OC
Catallist 11.5 SDK 2.4 Windows 7 Ultimate 5Gb DDR2-800

V	W	Config 1	Config 2	Config 3	Config 4
1	64	290
1	128	290
1	192	291
1	256	287
2	64	308	338		370
2	128	305	337		364
2	192	308	307		340
2	256	265	282
3	64	263
3	128	256	307
3	192	305	339	347	373
3	256	264
4	64	285
4	128	242
4	192	187
4	256	241
5	64	236
5	128	240
5	192	178
5	256	240
6	64	242
6	128	238
6	192	176
6	256	242
17	64	291
17	128	291
17	192	291
17	256	270
18	64	303	330		355
18	128	274
18	192	274
18	256	232
19	64	278
19	128				259
19	192				195
20	64	115
21	64	177
22	64	doesn't work
23	64	doesn't work
33	64	287
34	64	303	330		357
34	128	360
34	192	363
34	256	249
35	64	278
36	64	262
37	64	290 22% fail
38	64	180
39	64	183
40	64	182
41	64	106
42	64	116
43	64	119

Config 1 - Stock -f 60 -dd 800/1250@1100
Config 2 - Unlocked shaders -f 60 -dd 800/1250@1100
Config 3 - Unlocked shaders -f 5 -dd 800/1250@1100
Config 4 - Unlocked shaders -f 15 -g 4 -dd 900/1200@1129

I have been able to reach 408Mhs being more aggresive on the OC and -f but after 2 hours miner crashed

Also with this card I can not see flash videos (I hade no problems with 5770 and SDK 2.1) becouse the miner also crashes.

From the data I am unable to reach to any conclussions as with some config OC or unlicking doesn't improve a bit, -v 34 is a good number, -v 37 just gives 22% of HW errors... Anyway my best setup is -v 3 -w 192 similar to -v2 -w 64 but a hell more stable

There is a significative drop in MHs when the nonces are exhausted, it seems a thread just sits waiting while we must have some unused nonces from other threads, and also the spureus errors but ina lesser extent.

Not sure if this info will be valuable for someone else but here it is for reference.

[DiabloD3]

I have been able to reach 408Mhs being more aggresive on the OC and -f but after 2 hours miner crashed

Also with this card I can not see flash videos (I hade no problems with 5770 and SDK 2.1) becouse the miner also crashes.

From the data I am unable to reach to any conclussions as with some config OC or unlicking doesn't improve a bit, -v 34 is a good number, -v 37 just gives 22% of HW errors... Anyway my best setup is -v 3 -w 192 similar to -v2 -w 64 but a hell more stable

There is a significative drop in MHs when the nonces are exhausted, it seems a thread just sits waiting while we must have some unused nonces from other threads, and also the spureus errors but ina lesser extent.

Not sure if this info will be valuable for someone else but here it is for reference.

If it crashes after 2 hours of OC, you might be OCing too far.

Im surprised any of the 30s do anything reasonable at all.

Most people find -v 2, 3, and 19 to be the best with some -w setting tuning it.

It doesnt nonce steal because it costs nothing to get more... if you're pool mining and your network has a lot of lag this can screw stuff up, but it looks like you were solo mining, which leads me to believe low -f values on Windows are screwing over bitcoin due to Window's craptastic scheduler.

Also, benchmarking on low -f values are difficult. Due to Windows' broken clock implementation, the first number (the 15 second avg) reports rather random values. The only way to stop that is to run for at least 5 minutes and use the second number (the forever average) only.

Other than that, interesting table.

[DiabloD3]

Diablo,

In a previous post you did slash the average hash/s function on deepbit. I am a bit puzzled though: I manage close to 800Mh/s according to dm's output. However, whenever I look at deepbit, I see roughly dm/2. This is fully reflected in the share I receive. Using poclbm, I only managed about 90% of dm's rate, however, the values on deepbit and my share nearly doubled...

Make sure you're using the absolutely most up to date version of the miner. mtrlt's patch screwed up dropping about 1/3rd of shares for some people. I've fixed it since then.

[frisco]

I was mining with deepbit for the entire time, maybe laggy sometimes but I have 66ms ping right now.

And about the OC, I tried diferent config that caused visual corruption or even instant freeze this time the app crashed but everything remained working and I think I was able to start the miner again, anyway I will try improve that.

And maybe we don't need nonce stealing but ask ahead when they are running low, that will have no impact on slow miners and should help to avoid short stops, the bigger problem comes when every thread stops at the same time, usually during the first minutes after a connection or block change, not a big deal but it can be an improvement.

[claw]

I had the same problem when setting up the miner. There seems to be a problem with the most recent version of bitcoin (0.3.22-beta). Try version 0.3.21-beta and it should work - given that the rest of your setup is okay.

What, seriously? Thats bad.

Well, it might not be good, but it is progress. 

Now -- using the previous version of the Bitcoin client, after reporting that it's added the GPU, it has started throwing "Can't connect to Bitcoin: Connection refused: connect" errors.  It throws a bunch (~12) of them with "waiting..." between the first several and then khash/sec figures between them for a while (in the 11-13000 range) and then reports the fatal error and drops out of running.

I'm launching the app with the -u and -p flags each followed by the values in the bitcoin.conf file.  Am I supposed to be doing something different because I'm using 0.3.21 now?

[DiabloD3]

I was mining with deepbit for the entire time, maybe laggy sometimes but I have 66ms ping right now.

And about the OC, I tried diferent config that caused visual corruption or even instant freeze this time the app crashed but everything remained working and I think I was able to start the miner again, anyway I will try improve that.

And maybe we don't need nonce stealing but ask ahead when they are running low, that will have no impact on slow miners and should help to avoid short stops, the bigger problem comes when every thread stops at the same time, usually during the first minutes after a connection or block change, not a big deal but it can be an improvement.

What you described sounded like  you OC'ed too far.

Asking ahead is hard to do right and it also causes increased stale work. It fetches new work more than once a minute for most users. Also, this uses 3 different sets of work in parallel, its unlikely all 3 will block at the same time.

It really isn't the issue that people claim it is, other miners suffer from it much more.

[qed]

Radeon users: Catalyst 10.9 through 10.11 + SDK 2.1 seems to be the golden combination for 4xxx and 5xxx miners. Catalyst 11.5a + SDK 2.4 seems to be golden combination for 6xxx. 2.4 does not suffer from the CPU use bug 2.2/2.3 have. Do not use 2.2/2.3.

100% CPU usage still present with Catalyst 11.5a + SDK 2.4.

[DiabloD3]

Radeon users: Catalyst 10.9 through 10.11 + SDK 2.1 seems to be the golden combination for 4xxx and 5xxx miners. Catalyst 11.5a + SDK 2.4 seems to be golden combination for 6xxx. 2.4 does not suffer from the CPU use bug 2.2/2.3 have. Do not use 2.2/2.3.

100% CPU usage still present with Catalyst 11.5a + SDK 2.4.

Its a different bug this time. They moved the CPU usage bug into the driver itself. It only effects SOME Windows users, but not all, and no Linux users.

[Bartbarian]

I'm a bit late to the party but i like magic, and it would be awesome if the landlord would actually pay me to live here (fixed rent baby

I have a Vista 64 work/game rig with a Intel Core 2 e8500 and a HD4870 X2, which lists in GPU Caps OpenCL as two RV770's and a Core2 (thats 3 devices, mining will be fun; it's know for being a hot bitch). I updated my ancient drivers today with newest Catalyst 11.5 and OpenCL 2.4. After all the basic mistakes i now seemed to have DiabloMiner running:

C:\BitCoin\DiabloMiner>DiabloMiner-Windows.exe -u login -p secret -w 64
[11-6-11 11:53:59] Started
[11-6-11 11:53:59] Connecting to: http://127.0.0.1:8332/
[11-6-11 11:53:59] Using AMD Accelerated Parallel Processing OpenCL 1.1 AMD-APP-SDK-v2.4 (595.10)
[11-6-11 11:54:00] Added ATI RV770 (#1) (10 CU, local work size of 64)

but then.. nothing.. the wiki says: A few lines of output will scroll by, followed by "waiting..." or error messages, but this just sits there.

I look into GPU Caps and it says GPU load 2%, Catalyst panel tells me temp stays cool.

I tried fiddling with options but i dont' really understand what's going on. Anyone got suggestions or some angle of attack? (Besides installing linux that is.)

[krom]

I'm a bit late to the party but i like magic, and it would be awesome if the landlord would actually pay me to live here (fixed rent baby

I have a Vista 64 work/game rig with a Intel Core 2 e8500 and a HD4870 X2, which lists in GPU Caps OpenCL as two RV770's and a Core2 (thats 3 devices, mining will be fun; it's know for being a hot bitch). I updated my ancient drivers today with newest Catalyst 11.5 and OpenCL 2.4. After all the basic mistakes i now seemed to have DiabloMiner running:

C:\BitCoin\DiabloMiner>DiabloMiner-Windows.exe -u login -p secret -w 64
[11-6-11 11:53:59] Started
[11-6-11 11:53:59] Connecting to: http://127.0.0.1:8332/
[11-6-11 11:53:59] Using AMD Accelerated Parallel Processing OpenCL 1.1 AMD-APP-SDK-v2.4 (595.10)
[11-6-11 11:54:00] Added ATI RV770 (#1) (10 CU, local work size of 64)

but then.. nothing.. the wiki says: A few lines of output will scroll by, followed by "waiting..." or error messages, but this just sits there.

I look into GPU Caps and it says GPU load 2%, Catalyst panel tells me temp stays cool.

I tried fiddling with options but i dont' really understand what's going on. Anyone got suggestions or some angle of attack? (Besides installing linux that is.)

There's a problem with the most recent version of Bitcoin (0.3.22-beta), which results in the problem you describe. Use bitcoin 0.3.21-beta (link) and it should work.

@DiabloD3:
Maybe you should put that info on the first post as long as the problem persists. Although it's not a problem with your miner, ppl might think that.


edit: maybe i should add that i do use Win7 64bit and a HD 4870, accompanied by a AMD X2 @3Ghz. No clue if this problem exists in other OS versions / constellations.

[claw]

After installing 0.3.21 I wasn't sure if I could get it running with the -server flag...it runs, but I'm still not connecting with DiabloMiner so I tried running bitcoind but I had to copy the libeay32.dll into the \daemon directory.  Now that's chugging away in the background and when I run DiabloMiner I get several copies of

[6/11/11 5:50:33 AM] ERROR: Can't connect to Bitcoin: Bitcoin returned error mes
sage: <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN"
"http://www.w3.org/TR/1999/REC-html401-19991224/loose.dtd">
<HTML>
<HEAD>
<TITLE>Error</TITLE>
<META HTTP-EQUIV='Content-Type' CONTENT='text/html; charset=ISO-8859-1'>
</HEAD>
<BODY><H1>401 Unauthorized.</H1></BODY>
</HTML>

during which time it spiked up to 16Khash/sec and then it stopped returning those errors, threw

Exception in thread "DiabloMiner Executor (GeForce 9600 GSO (#1)/0)" Exception in thread "DiabloMiner Executor (GeForce 9600 GSO (#1)/1)" java.lang.NullPointerException
        at com.diablominer.DiabloMiner.DiabloMiner$DeviceState$ExecutionState.run(DiabloMiner.java:832)
        at java.lang.Thread.run(Unknown Source)
Exception in thread "DiabloMiner Executor (GeForce 9600 GSO (#1)/2)" java.lang.NullPointerException
        at com.diablominer.DiabloMiner.DiabloMiner$DeviceState$ExecutionState.run(DiabloMiner.java:832)
        at java.lang.Thread.run(Unknown Source)
java.lang.NullPointerException
        at com.diablominer.DiabloMiner.DiabloMiner$DeviceState$ExecutionState.run(DiabloMiner.java:832)
        at java.lang.Thread.run(Unknown Source)

but didn't quit.  It settled into displaying "0/yyyy khash/seccsecc" at the bottom of the terminal where yyyy started up in the 12000 range and quickly dropped into the low hundreds before I stopped it.

[krazzer]

I have an early 2009 imac with a Radeon 4850, latest updates running DiabloMiner with -v 2 -w64, i'm getting : 
Block 166 found on Radeon HD 4850 (#1)
33241/39200 khash/sec

Is this the best i can get?

[horrorshow]

Ok, so I installed catalyst 10.11 and I thought I had installed SDK 2.1 properly, however...

[6/11/11 3:31:08 PM] Using AMD Accelerated Parallel Processing OpenCL 1.1 AMD-AP
P-SDK-v2.4 (595.10)
[6/11/11 3:31:09 PM] BFI_INT patching enabled, disabling hardware checking
[6/11/11 3:31:09 PM] Added Cypress (#2) (14 CU, local work size of 128)
[6/11/11 3:31:12 PM] Using ATI Stream OpenCL 1.0 ATI-Stream-v2.1 (145)

Does anyone know why it says stream v2.1 and SDK 2.4? How can I fix this on win7 x64?

[mike503]

I have a couple headless Solaris 11 machines (or OpenSolaris maybe on one) I'd love to run miners on them if possible. CPU mode right now, since I don't have graphics cards for it right now. It has 2 PCIe slots and 2 open PCI-X slots (yes, PCI-X) if anyone had any ideas of decent-bang-for-buck mining cards t to throw in that are supported under Solaris...

They are headless, so I don't have X11 working.

Any ideas? Tried setting DISPLAY to 0:0 ... still no go. it actually didn't immediately fail, but also didn't work.

Code:

# DISPLAY="0:0" bash ./DiabloMiner-Solaris.sh 
[6/11/11 4:23:17 PM] Started

Exception in thread "main" java.lang.InternalError: Can't connect to X11 window server using '0:0' as the value of the DISPLAY variable.
        at sun.awt.X11GraphicsEnvironment.initDisplay(Native Method)
        at sun.awt.X11GraphicsEnvironment.access$100(X11GraphicsEnvironment.java:52)
        at sun.awt.X11GraphicsEnvironment$1.run(X11GraphicsEnvironment.java:155)
        at java.security.AccessController.doPrivileged(Native Method)
        at sun.awt.X11GraphicsEnvironment.<clinit>(X11GraphicsEnvironment.java:131)
        at java.lang.Class.forName0(Native Method)
        at java.lang.Class.forName(Class.java:169)
        at java.awt.GraphicsEnvironment.getLocalGraphicsEnvironment(GraphicsEnvironment.java:68)
        at sun.awt.X11.XToolkit.<clinit>(XToolkit.java:89)
        at java.lang.Class.forName0(Native Method)
        at java.lang.Class.forName(Class.java:169)
        at java.awt.Toolkit$2.run(Toolkit.java:834)
        at java.security.AccessController.doPrivileged(Native Method)
        at java.awt.Toolkit.getDefaultToolkit(Toolkit.java:826)
        at org.lwjgl.LinuxSysImplementation.<clinit>(LinuxSysImplementation.java:45)
        at org.lwjgl.Sys.createImplementation(Sys.java:113)
        at org.lwjgl.Sys.<clinit>(Sys.java:98)
        at org.lwjgl.opencl.CL.<clinit>(CL.java:51)
        at com.diablominer.DiabloMiner.DiabloMiner.execute(DiabloMiner.java:391)
        at com.diablominer.DiabloMiner.DiabloMiner.main(DiabloMiner.java:127)

[DiabloD3]

Ok, so I installed catalyst 10.11 and I thought I had installed SDK 2.1 properly, however...

[6/11/11 3:31:08 PM] Using AMD Accelerated Parallel Processing OpenCL 1.1 AMD-AP
P-SDK-v2.4 (595.10)
[6/11/11 3:31:09 PM] BFI_INT patching enabled, disabling hardware checking
[6/11/11 3:31:09 PM] Added Cypress (#2) (14 CU, local work size of 128)
[6/11/11 3:31:12 PM] Using ATI Stream OpenCL 1.0 ATI-Stream-v2.1 (145)

Does anyone know why it says stream v2.1 and SDK 2.4? How can I fix this on win7 x64?

You have both installed. Don't do that.

[DiabloD3]

I have a couple headless Solaris 11 machines (or OpenSolaris maybe on one) I'd love to run miners on them if possible. CPU mode right now, since I don't have graphics cards for it right now. It has 2 PCIe slots and 2 open PCI-X slots (yes, PCI-X) if anyone had any ideas of decent-bang-for-buck mining cards t to throw in that are supported under Solaris...

They are headless, so I don't have X11 working.

Any ideas? Tried setting DISPLAY to 0:0 ... still no go. it actually didn't immediately fail, but also didn't work.

Code:

# DISPLAY="0:0" bash ./DiabloMiner-Solaris.sh 
[6/11/11 4:23:17 PM] Started

Exception in thread "main" java.lang.InternalError: Can't connect to X11 window server using '0:0' as the value of the DISPLAY variable.
        at sun.awt.X11GraphicsEnvironment.initDisplay(Native Method)
        at sun.awt.X11GraphicsEnvironment.access$100(X11GraphicsEnvironment.java:52)
        at sun.awt.X11GraphicsEnvironment$1.run(X11GraphicsEnvironment.java:155)
        at java.security.AccessController.doPrivileged(Native Method)
        at sun.awt.X11GraphicsEnvironment.<clinit>(X11GraphicsEnvironment.java:131)
        at java.lang.Class.forName0(Native Method)
        at java.lang.Class.forName(Class.java:169)
        at java.awt.GraphicsEnvironment.getLocalGraphicsEnvironment(GraphicsEnvironment.java:68)
        at sun.awt.X11.XToolkit.<clinit>(XToolkit.java:89)
        at java.lang.Class.forName0(Native Method)
        at java.lang.Class.forName(Class.java:169)
        at java.awt.Toolkit$2.run(Toolkit.java:834)
        at java.security.AccessController.doPrivileged(Native Method)
        at java.awt.Toolkit.getDefaultToolkit(Toolkit.java:826)
        at org.lwjgl.LinuxSysImplementation.<clinit>(LinuxSysImplementation.java:45)
        at org.lwjgl.Sys.createImplementation(Sys.java:113)
        at org.lwjgl.Sys.<clinit>(Sys.java:98)
        at org.lwjgl.opencl.CL.<clinit>(CL.java:51)
        at com.diablominer.DiabloMiner.DiabloMiner.execute(DiabloMiner.java:391)
        at com.diablominer.DiabloMiner.DiabloMiner.main(DiabloMiner.java:127)

For one, DiabloMiner cannot CPU mine. The other thing is DiabloMiner requires X running, which isn't that huge of an issue since mining on Radeons requires X running.

You can still do it headless, but start X (with, say, startx &) and then export DISPLAY=:0. All can be done from ssh.

By the way, no one has tested my miner on Solaris, you would be the first.