[ANN] Yet another GPU miner release! [YAC]

[Kosmatik]

I keep getting the same block found 0 while solo mining and wallet is not increasing, -D drops me back to command prompt. 6970, tc 8000, worksize 256 or 64, lg 2. Hashes about 90 kh/s so I know something's not right when it's only doing double of a 6750. Also  pool shares not increasing on yac.ltcoin.net, but I am getting "block solve" accepted shares.

[1715022408]

1715022408

[1715022408]

1715022408

[bitdwarf]

Same. Right now I find more reliable to GPU mine other coins and trade for YAC, specially with the cheap price.

[hanzac]

I keep getting the same block found 0 while solo mining and wallet is not increasing, -D drops me back to command prompt. 6970, tc 8000, worksize 256 or 64, lg 2. Hashes about 90 kh/s so I know something's not right when it's only doing double of a 6750. Also  pool shares not increasing on yac.ltcoin.net, but I am getting "block solve" accepted shares.

See. Later, after about two days, I'll try to migrate to cgminer 3.2.1, to see if it can fix this issue.

[Kosmatik]

I keep getting the same block found 0 while solo mining and wallet is not increasing, -D drops me back to command prompt. 6970, tc 8000, worksize 256 or 64, lg 2. Hashes about 90 kh/s so I know something's not right when it's only doing double of a 6750. Also  pool shares not increasing on yac.ltcoin.net, but I am getting "block solve" accepted shares.

See. Later, after about two days, I'll try to migrate to cgminer 3.2.1, to see if it can fix this issue.

Latest stable is 3.1.1, 3.2.x was made for usb asics for bitcoin, why not stick with 3.1.1?

[centergate]

32bit version release?

[hanzac]

I keep getting the same block found 0 while solo mining and wallet is not increasing, -D drops me back to command prompt. 6970, tc 8000, worksize 256 or 64, lg 2. Hashes about 90 kh/s so I know something's not right when it's only doing double of a 6750. Also  pool shares not increasing on yac.ltcoin.net, but I am getting "block solve" accepted shares.

See. Later, after about two days, I'll try to migrate to cgminer 3.2.1, to see if it can fix this issue.

Latest stable is 3.1.1, 3.2.x was made for usb asics for bitcoin, why not stick with 3.1.1?

... because I always think that newer version will have good features ... anyway, it can still work to be normal after awhile. I'll fix that after I have time.

[hanzac]

32bit version release?

Currently I didn't have much energy for building a 32-bit version. Thirtybird [https://bitcointalk.org/index.php?topic=232241.msg2448039#msg2448039] has built one, you may contact him.

I'll consider building one for x86 after two days when I find some free time.

[megablue]

you did it!! congrats! 

[sairon]

I'm getting segfaults

Code:

 [2013-06-12 11:37:45] Started cgminer 3.2.0
 [2013-06-12 11:37:45] Started cgminer 3.2.0
 [2013-06-12 11:37:45] CL Platform 0 vendor: Advanced Micro Devices, Inc.
 [2013-06-12 11:37:45] CL Platform 0 name: AMD Accelerated Parallel Processing
 [2013-06-12 11:37:45] CL Platform 0 version: OpenCL 1.2 AMD-APP (1214.3)
 [2013-06-12 11:37:45] Platform 0 devices: 1
 [2013-06-12 11:37:45]  0       Caicos
 [2013-06-12 11:37:45] Probing for an alive pool
 [2013-06-12 11:37:45] Popping work to stage thread
 [2013-06-12 11:37:45] Testing pool [...]
 [2013-06-12 11:37:45] Probing for GBT support
 [2013-06-12 11:37:45] JSON-RPC call failed: {
   "data": null,
   "code": -32601,
   "message": "Method not found"
}
 [2013-06-12 11:37:45] Failed to connect in json_rpc_call
 [2013-06-12 11:37:45] No GBT coinbase + append support found, using getwork protocol
 [2013-06-12 11:37:45] X-Roll-Ntime expiry set to 100
 [2013-06-12 11:37:45] Successfully retrieved and deciphered work from pool 0 [...]
 [2013-06-12 11:37:45] Pushing pooltest work to base pool
 [2013-06-12 11:37:45] Network diff set to 49.1K
 [2013-06-12 11:37:45] New block: d9c3b6dc1e97ac28... diff 49.1K
 [2013-06-12 11:37:45] Pushing work to getwork queue
 [2013-06-12 11:37:45] Popping work to stage thread
 [2013-06-12 11:37:45] Pool 0 [...] alive
 [2013-06-12 11:37:45] Long-polling activated for [...]Segmentation fault (core dumped)

The reason for segfault seems to be a dangling pointer issue, or some sort of buffer overflow.

Code:

--- SIGSEGV {si_signo=SIGSEGV, si_code=SEGV_MAPERR, si_addr=0x2904908} ---

[hanzac]

you did it!! congrats! 

Ha, thanks.

[hanzac]

I'm getting segfaults

Code:

 [2013-06-12 11:37:45] Started cgminer 3.2.0
 [2013-06-12 11:37:45] Started cgminer 3.2.0
 [2013-06-12 11:37:45] CL Platform 0 vendor: Advanced Micro Devices, Inc.
 [2013-06-12 11:37:45] CL Platform 0 name: AMD Accelerated Parallel Processing
 [2013-06-12 11:37:45] CL Platform 0 version: OpenCL 1.2 AMD-APP (1214.3)
 [2013-06-12 11:37:45] Platform 0 devices: 1
 [2013-06-12 11:37:45]  0       Caicos
 [2013-06-12 11:37:45] Probing for an alive pool
 [2013-06-12 11:37:45] Popping work to stage thread
 [2013-06-12 11:37:45] Testing pool [...]
 [2013-06-12 11:37:45] Probing for GBT support
 [2013-06-12 11:37:45] JSON-RPC call failed: {
   "data": null,
   "code": -32601,
   "message": "Method not found"
}
 [2013-06-12 11:37:45] Failed to connect in json_rpc_call
 [2013-06-12 11:37:45] No GBT coinbase + append support found, using getwork protocol
 [2013-06-12 11:37:45] X-Roll-Ntime expiry set to 100
 [2013-06-12 11:37:45] Successfully retrieved and deciphered work from pool 0 [...]
 [2013-06-12 11:37:45] Pushing pooltest work to base pool
 [2013-06-12 11:37:45] Network diff set to 49.1K
 [2013-06-12 11:37:45] New block: d9c3b6dc1e97ac28... diff 49.1K
 [2013-06-12 11:37:45] Pushing work to getwork queue
 [2013-06-12 11:37:45] Popping work to stage thread
 [2013-06-12 11:37:45] Pool 0 [...] alive
 [2013-06-12 11:37:45] Long-polling activated for [...]Segmentation fault (core dumped)

The reason for segfault seems to be a dangling pointer issue, or some sort of buffer overflow.

Code:

--- SIGSEGV {si_signo=SIGSEGV, si_code=SEGV_MAPERR, si_addr=0x2904908} ---

Caicos? what's this GPU brand?

Can you also post the command line? the thread-concurrency? Maybe you set the parameter not match with your hardware.

[paulthetafy]

hanzac, what's the relationship between intensity and thread concurrency, and should there be a way to run this at > I=12?  I tried 13 and it seemed to drop the hash rate and 14 died.  My TC is 32768

[feeleep]

Can anyone suggest a good pool?  I was using yac.coinmine.pl but I think it's a scam as my "estimated per payout" was really low (0.1) after a couple of hours at 500 kh/s

I think it's legit. YAC needs 520 confirmations to payout. So maybe you need to wait for tomorrow to start getting payouts.
Just check Unpaid Shares - Your Valid: -----, if it's normal.

Or you go to yac.ltcoin.net, it seems having the highest hash rate so far.

Again I have to definitely disagree of calling our pool scam - as mentioned earlier - to confirm first block you started mining we need ~8-10 hours - and this is approx the time you should have full payouts. Please note that even if you stop mining you will be receiving payouts until you are out of last Nshares. This has been proved by me several times here...

feeleep

[paulthetafy]

Can anyone suggest a good pool?  I was using yac.coinmine.pl but I think it's a scam as my "estimated per payout" was really low (0.1) after a couple of hours at 500 kh/s

I think it's legit. YAC needs 520 confirmations to payout. So maybe you need to wait for tomorrow to start getting payouts.
Just check Unpaid Shares - Your Valid: -----, if it's normal.

Or you go to yac.ltcoin.net, it seems having the highest hash rate so far.

Again I have to definitely disagree of calling our pool scam - as mentioned earlier - to confirm first block you started mining we need ~8-10 hours - and this is approx the time you should have full payouts. Please note that even if you stop mining you will be receiving payouts until you are out of last Nshares. This has been proved by me several times here...

feeleep

My sincere apologies, I've only ever mined with p2pool so was not expecting the long wait.  All is working now and I'm in your top 10 for contributors / hashrate.

[hanzac]

hanzac, what's the relationship between intensity and thread concurrency, and should there be a way to run this at > I=12?  I tried 13 and it seemed to drop the hash rate and 14 died.  My TC is 32768

Currently the intensity is used to control the global worksize (2^I, if I = 12, global worksize is 4096) of the opencl kernel execution. This means the concurrent executions pushed to the GPU kernel. The total memory buffer for chacha mixing will be 2^I * 128 kB (nfactor = 9, check the tech info here: http://yacexplorer.tk/graphs.htm), if I = 12, it will need 512MB of GPU's global memory. If you use I = 13, it will consume 1GB GPU memory, but considering some GPU's local memory need, it will need more than 1GB.

Even if you have enough memory, you still have limitation on the GPU stream cores. For 7970, it has 2048 GPU stream cores, so set the intensity to make concurrent executions to 2x GPU cores will be approaching optimal, that's why I = 12 works better than I = 11. I think concurrent executions works well in range from 2x GPU cores - 4x GPU cores.

The reason it died, it because memory used up, pointer to some memory address not exist. So if you insist on achieve higher intensity, the cl kernel needs to support look-up gap or some other memory shrink/share technique, but current cl kernel is quite plain & simple and doesn't support these.

The thread-concurrency(=32768) here is for allocating the memory pad buffer, 32768 * 64K, is allocated 2GB memory. To allocate more, you can try some other lookup-gap setting to allocate the maximized memory. For example, your card total memory is 3GB you set the lookup-gap = 2, and set thread-concurrency to 49152? I think I have tried once, but not succeed for my card.

[Maniac]

What hashrate has 6930 ?
What optimal memory clocks?
Is it profitable to mininig yac or ltc?

[paulthetafy]

Regarding rejects, just FYI I'm getting about 33% rejects across both 7950's I'm running after running after a few hours.  Would this be down to the pool?

[hanzac]

Regarding rejects, just FYI I'm getting about 33% rejects across both 7950's I'm running after running after a few hours.  Would this be down to the pool?

Me too, the rejects are quite high (1/5 rejects), back to when I used it mining on p2pool, the rejects are not that high. So I assume the pool is blocking fast, make a lots of stales. But considering if you use p2pool, the fee caused by small payments. It's all really trade-offs. I think pushpools should be better than p2pools.

[Boing7898]

Regarding rejects, just FYI I'm getting about 33% rejects across both 7950's I'm running after running after a few hours.  Would this be down to the pool?

I'm getting 6% of rejects with a HD6770, using yac.ltcoin.net.

[centergate]

waiting for your new version for x86.