BitMinter.com * Optional Custom Miner, PPLNS, Merged mining, Newbie-Friendly! *

[ToriAmos1963]

I signed up and the miner appears to be faster than GUIMiner.  I was getting around 405-410 and now I'm getting 415-420.  I'm worried about security because this is closed source, but I only have a couple of bitcoins, nothing too bad to worry about.  I have to agree with others about the safety of my wallet.  Even if you encrypt the wallet, we don't know what your miner is doing in the background in terms of key logging.

Time will tell.  If the speed is accurate, I would be glad to make a donation once I get paid.

Is there any reason why you are against releasing your code?  You're not selling your software and the more people look at it, the better chances of finding and fixing bugs.

I have a question and a couple of suggestions.

First, what is the "work unit" number displayed in the bottom right side?

I think additional info about the number of work units done and invalids in the past hour is helpful in making sure the graphic cards are working properly.

I think an option to minimize to notification tray would be a nice feature to keep the task bar clean and tidy.

[1713483188]

1713483188

[Xephan]

Yes, I will at least add OpenCL mining on CPUs in the near future, which should increase speed by a lot. But I'm not sure if it can compete with the fastest CPU miners available - I haven't tried OpenCL on CPU yet. We'll find out soon.

I installed the Intel OpenCL SDK on my laptop and tried mining using openCL/SSE4 with pocblm and SSE2 with ufasoft. I got about 1.2~1.3Mh with openCL and 6.4Mhash or so with SSE2. Not sure if that means that an additional openCL wrapper introduces way too much overheads compared to directly using the CPU.

[LordTreasurer]

Hm, with your miner I have about 70MH/s.
With my default opencl-miner I reach 140MH/s...

I am using a single 5750.
I did not change any options so far but I dont think it would though...

But nice approach to make mining more easy

[DrHaribo]

I signed up and the miner appears to be faster than GUIMiner.  I was getting around 405-410 and now I'm getting 415-420.  I'm worried about security because this is closed source, but I only have a couple of bitcoins, nothing too bad to worry about.  I have to agree with others about the safety of my wallet.  Even if you encrypt the wallet, we don't know what your miner is doing in the background in terms of key logging.

Glad you are getting good performance.

I understand that you are cautious. It's a good thing to be careful about what you download and run off the internet. I hope to become a person that people in the bitcoin community can trust. But I'm sure that will take some time. There's nothing bad in the miner (except probably some bugs). I made it to promote my mining pool. With the 5% promotion I am also giving away a total of 150 bitcoins to promote the mining pool.


You can also run a different miner with my pool. You won't get the extra speed of my miner, but you still get the 5% bonus pay, and very few stales.

Time will tell.  If the speed is accurate, I would be glad to make a donation once I get paid.

That reminds me, I have to implement voluntary donations soon.

Is there any reason why you are against releasing your code?  You're not selling your software and the more people look at it, the better chances of finding and fixing bugs.

At the moment I'm just losing 150 BTC. But I hope in the future people will be willing to set a voluntary donation percentage in their account and donate some coins. The client is part of my strategy to make my mining pool better than the other pools. That way I will hopefully make some bitcoins from my software in the future.

First, what is the "work unit" number displayed in the bottom right side?

That's how much work the client is holding at the moment. 1 work unit = 1 "getwork" from the server. The client tries to hold a few ready at all times so the GPU can always be kept 100% busy. If there is a network issue, the number will drop to zero after a while, and shortly after the GPU will become idle.

I think additional info about the number of work units done and invalids in the past hour is helpful in making sure the graphic cards are working properly.

Yeah, a lot of people request additional statistics in the miner. Adding this to the TODO list.

I think an option to minimize to notification tray would be a nice feature to keep the task bar clean and tidy.

Good idea - added it to my list!

[DrHaribo]

I installed the Intel OpenCL SDK on my laptop and tried mining using openCL/SSE4 with pocblm and SSE2 with ufasoft. I got about 1.2~1.3Mh with openCL and 6.4Mhash or so with SSE2. Not sure if that means that an additional openCL wrapper introduces way too much overheads compared to directly using the CPU.

I suspect it will help a bit to make a separate kernel for CPUs, with different optimizations. But I doubt it will beat ufasoft. Still, it will be much faster than my CPU code today - it's java bytecode and I don't think the JIT-compiler makes use of SSE at all.

Hm, with your miner I have about 70MH/s.
With my default opencl-miner I reach 140MH/s...

I am using a single 5750.

Right now the miner usually doesn't give good performance on Radeon HD 4000 and 5000 series cards. I bought a 5970 a couple days ago and I hope I can speed up the miner for those GPUs this weekend. I would suggest using a different miner on those GPUs for now. I'll post some updated speed comparisons after I have a new version of the miner ready.

[Xephan]

I have similar performance experience as Lord Treasury.
On my 5870 I get regularly get about 370~380Mh/s and my system remains very responsive.
However on the bitminter Java miner, my system is sluggish and I only get about 355Mh/s after observing for 30 minutes (performance mode with details minimized). The work done, 143units, is in line with the reported speeds; I get about 150~155 blocks done in 30 minutes with poclbm. Seems like your algorithm might be excellent but undone by possible JVM overheads.

[DrHaribo]

I have similar performance experience as Lord Treasury.
On my 5870 I get regularly get about 370~380Mh/s and my system remains very responsive.
However on the bitminter Java miner, my system is sluggish and I only get about 355Mh/s after observing for 30 minutes (performance mode with details minimized). The work done, 143units, is in line with the reported speeds; I get about 150~155 blocks done in 30 minutes with poclbm. Seems like your algorithm might be excellent but undone by possible JVM overheads.

Nah, it's not JVM overhead. As far as I know it's the fastest miner on VLIW-4 (newest Radeon cards) and GeForce GPUs. It was tested and optimized on Radeon 6990 and GeForce GTX 580 during development. So it became very fast on those chips.

But I know there are many many VLIW-5 GPUs out there, and it is a priority issue for me to fix performance on those cards - so much so that I went out and bought one. There's a 5970 next to my 6990 now and I hope to make it go 100 mhps faster (50 mhps more for each of the two GPUs in the 5970).

The reason for the difference in performance is simply that VLIW-4 and VLIW-5 are different architectures and they need different optimizations.

[Xephan]

Nah, it's not JVM overhead. As far as I know it's the fastest miner on VLIW-4 (newest Radeon cards) and GeForce GPUs. It was tested and optimized on Radeon 6990 and GeForce GTX 580 during development. So it became very fast on those chips.

But I know there are many many VLIW-5 GPUs out there, and it is a priority issue for me to fix performance on those cards - so much so that I went out and bought one. There's a 5970 next to my 6990 now and I hope to make it go 100 mhps faster (50 mhps more for each of the two GPUs in the 5970).

The reason for the difference in performance is simply that VLIW-4 and VLIW-5 are different architectures and they need different optimizations.

Cool, looking forward to that since I've got two 5800 series card which means I might just get another 80~100Mhps out of them too after clocking them up a bit

[LordTreasurer]

Sounds convincing, but am I (and Xephan) the only one who suffers from a very sluggish desktop behavior?

Can this issue be fixed or is does everyone experiences this?
Or does this happen because I use an "old" radeon card?

[ToriAmos1963]

I have another suggestion.  I only use my GPU so I have removed the CPU and total work windows and resized the window to only show the graphic card.  The problem I have run into is that when I go back and forth between performance and "pretty" modes, the window resizes with a large blank area in the middle.


After resizing
http://i33.photobucket.com/albums/d58/ToriAmos1963/prettymode.jpg


Changing back from performance mode
http://i33.photobucket.com/albums/d58/ToriAmos1963/prettymode_after_performancemode.jpg

[BitSense Informatics]

Source code? Not to be that guy, but running random programs is asking for wallet theft, even in Java.

It's closed source, sorry. The only file access is through the webstart muffin interface for storing settings (webstart muffins are similar to browser cookies). Unfortunately it still cannot run sandboxed because it needs access to native libraries for GPU mining.

Is it based on or containing any code from any of the open-source miners? If you have even a tiny amount of their code in there you have to release the source. Unless you open-source this I'm never going to use it. You could easily lie about the hash rates or expropriate more than your stated comission. Heck, it could do anything.

This is not necessarily true.  It depends on the open source licenses that the other libraries are using.  Something like the GPL you can include the open source library in your application and only need to release code for that library if you modified the library itself, but you are *not* required to release the code for your own application.  Look at Oracle, they make heavy use of Apache web server in the ERP applications, but they are most definitely not open source.

[Soak]

Please include on the website the total hashes rate of the pool (no problem with pool hoppers I think with only this stat). Well, I will can calculate my contribution to the pool in % (in waiting more stats on the website)

[Xephan]

Sounds convincing, but am I (and Xephan) the only one who suffers from a very sluggish desktop behavior?

Can this issue be fixed or is does everyone experiences this?
Or does this happen because I use an "old" radeon card?

Good question there since I was wondering if it was worth the aggravation for a potential extra 50Mh. 30 minutes of sluggish mouse movements was a mild annoyance but hours of it might just drive me crazy. Currently poclbm is quite well behaved, it drops usage immediately if I start up a graphics intensive application. I don't know if the Java miner might continue to grab the GPU and make it impossible to play a game without manually stopping the miner.

[DrHaribo]

Sounds convincing, but am I (and Xephan) the only one who suffers from a very sluggish desktop behavior?

Try going in the "tune & tweak" settings for the devices you are running and set a shorter "break interval". This setting determines how often the OS will get a chance to use the GPU to update the desktop. Maybe 50 ms is too high a default.

I have another suggestion.  I only use my GPU so I have removed the CPU and total work windows and resized the window to only show the graphic card.  The problem I have run into is that when I go back and forth between performance and "pretty" modes, the window resizes with a large blank area in the middle.

Hmm, looks like it's making room for the devices that have been hidden. Added to bug list - thanks!

Please include on the website the total hashes rate of the pool (no problem with pool hoppers I think with only this stat). Well, I will can calculate my contribution to the pool in % (in waiting more stats on the website)

Coming right up!

[Test Bank Guy]

looks fancy

[Exo]

Considering using this, how many blocks have you already found?

[Xephan]

Try going in the "tune & tweak" settings for the devices you are running and set a shorter "break interval". This setting determines how often the OS will get a chance to use the GPU to update the desktop. Maybe 50 ms is too high a default.

Not sure why but I run poclbm with -f20, which if I read the help about 1/f seconds correctly would put that at the same 50ms refresh. If I push it down to f10, or 100ms, then I get a bit of sluggishness that is just slightly better than the default on the minter miner. I'll admit my bias and would like to blame Java for this

[DrHaribo]

Considering using this, how many blocks have you already found?

We only found 2 blocks so far. So plenty of +5% bonus left (promotion runs till we find 60 blocks).

Not sure why but I run poclbm with -f20, which if I read the help about 1/f seconds correctly would put that at the same 50ms refresh. If I push it down to f10, or 100ms, then I get a bit of sluggishness that is just slightly better than the default on the minter miner. I'll admit my bias and would like to blame Java for this

Hehe In theory 50 ms in bitminter should be the same as 20 fps in poclbm. You could try 20 fps with DiabloMiner - it's also Java.

Maybe I have to take another look at the code that determines the OpenCL global work size, which again determines these intervals. I have a feeling it sometimes creates longer intervals with multiple GPUs running.

[ToriAmos1963]

I have another suggestion.  I only use my GPU so I have removed the CPU and total work windows and resized the window to only show the graphic card.  The problem I have run into is that when I go back and forth between performance and "pretty" modes, the window resizes with a large blank area in the middle.

Hmm, looks like it's making room for the devices that have been hidden. Added to bug list - thanks!


[/quote]

There is one more thing.  When I restarted my computer and the client, the CPU module and total module were back.  I think it would be better if the client remembered the state of modules for subsequent runs.

[DrHaribo]

There is one more thing.  When I restarted my computer and the client, the CPU module and total module were back.  I think it would be better if the client remembered the state of modules for subsequent runs.

Yes, at the moment it only remembers the login credentials (if you check the "remember me" at the login screen). I know it's annoying setting everything up each time - saving all the setup stuff is on my list!