Trial 1... [OPENCL]:[OPENCL]:Using platform: Using platform: AMD Accelerated Parallel Processing AMD Accelerated Parallel Processing [OPENCL]:Using device: Hawaii(OpenCL 2.0 AMD-APP (1800.11)) [OPENCL]:Using device: Hawaii(OpenCL 2.0 AMD-APP (1800.11)) [OPENCL]:Printing program log [OPENCL]: [OPENCL]:Creating one big buffer for the DAG [OPENCL]:Loading single big chunk kernels [OPENCL]:Mapping one big chunk. 0 Trial 2... [OPENCL]:Printing program log [OPENCL]: [OPENCL]:Creating one big buffer for the DAG [OPENCL]:Loading single big chunk kernels [OPENCL]:Mapping one big chunk. [OPENCL]:Creating buffer for header. [OPENCL]:Creating mining buffer 0 [OPENCL]:Creating mining buffer 1 [OPENCL]:Creating buffer for header. [OPENCL]:Creating mining buffer 0 [OPENCL]:Creating mining buffer 1 21321045 Trial 3... 42991616 Trial 4... 42729472 Trial 5... 42467328 min/mean/max: 0/29901892/42991616 H/s inner mean: 35505948 H/s Phoning home to find world ranking...
@sebstar I think you should be able to get more out of the 660s using my cudaminer-uint64 branch. I dropped Compute 3.0 support from the (main) cudaminer-frontier branch, but that branch has some nice CPU saving code.
@sebstar I think you should be able to get more out of the 660s using my cudaminer-uint64 branch. I dropped Compute 3.0 support from the (main) cudaminer-frontier branch, but that branch has some nice CPU saving code.
I would like to try your custom ethminer build, not sure where to start with it though , do you have updated windows binaries out there ? or do we have to build from githhub which seems to be a pain in Windows.
Also to use your miner do we need the run that dagapolalpypse thing or just run the miner ? I have the build you posted on the forums on another theread but it requires some Cuda DLLs which i did download but not sure if I am running the miner properly.
so I have higher rates on the benchmark because my DAG file is smaller? I noticed the benchmark DAG is about 1.1 GB while the mining DAG is about 1.3GB.
Will the DAG file keep growing? If so does that mean our mining speed would decrease as time increase?
Whats the max size the DAG can grow to?
Also why does the benchmark in windows shows slower rate? Benchmark in windows shows 58mhash while in ubuntu shows 65mhash on the same cards? 3x 7950s.
Lastly how can I get my real hash rate from my miner on ubuntu? It does not show the mining speed like it does on windows.
Hmm, seems lot's of posts here with 280's that perform much more hashes. Isn't 380 supposed to be rebranded 280's? It should be the same or faster (from OC) than 280's? Have you tried it some more?
[EDIT:] - Nevermind! I realize now that the memory bus width on the 360 is 256bit vs. the 384bit on the 280 - thereby even making 4GB a complete waste on the 380 :-/ [/EDIT]
@sebstar I think you should be able to get more out of the 660s using my cudaminer-uint64 branch. I dropped Compute 3.0 support from the (main) cudaminer-frontier branch, but that branch has some nice CPU saving code.
I would like to try your custom ethminer build, not sure where to start with it though , do you have updated windows binaries out there ? or do we have to build from githhub which seems to be a pain in Windows.
Also to use your miner do we need the run that dagapolalpypse thing or just run the miner ? I have the build you posted on the forums on another theread but it requires some Cuda DLLs which i did download but not sure if I am running the miner properly.
I have a mix of 7950, 280xs and 390s
I have new binaries posted today in the release folder on github. But really there's no point in using it instead of the stock miner if you have AMD, it uses the same kernel in OpenCL mode. Well it does display hashrate in the console, that's something many people want.
Anyone know why I get 0 for the min? Is this something I can fix?
This is what prints when I run the bench:
[OPENCL]:Creating one big buffer for the DAG [OPENCL]:Loading single big chunk kernels [OPENCL]:Mapping one big chunk. Trial 1... 0 Trial 2... 0 Trial 3... [OPENCL]:Creating buffer for header. [OPENCL]:Creating mining buffer 0 [OPENCL]:Creating mining buffer 1 1660245 Trial 4... 25427968 Trial 5... 25515349 min/mean/max: 0/10520712/25515349 H/s inner mean: 17534520 H/s
Any ideas to what is going wrong with the min? Thanks for the help!
@Film54 the benchmark does not wait until your GPU's RAM is completely loaded and its CU's saturated. Try again using --warmup-time 10 or longer.
Do you know of a way to get the benchmark to use current or custom DAG sizes instead of the ~1GB test DAG?
I've been meaning to do some fine tuning again, but the benchmark is no longer indicative of current mining stresses, hashrates seem to be far more tied to memory timings/performance than they were 6 months back.
Some cards seem to be suffering far more than others, 7950/280x with Elpida memory seem the most impacted thus far
has anyone experienced dropped hashrate? before my 2 r9-280x runs @45-50 MHs. but yesterdayi noticed it dropped to 30Mhz. windows 10 64bit , 15.12 driver
Windows 8.1 PRO / XFX Radeon R9-290 2GB (stock settings) : 23.5MH (O/C best 27MH) Windows 8.1 PRO / XFX Radeon R7-370 2GB (stock settings) : 16.2MH (O/C best 17MH)
Comments
AMD Accelerated Parallel Processing
[OPENCL]:Using device: Hawaii(OpenCL 2.0 AMD-APP (1800.11))
[OPENCL]:Using device: Hawaii(OpenCL 2.0 AMD-APP (1800.11))
[OPENCL]:Printing program log
[OPENCL]:
[OPENCL]:Creating one big buffer for the DAG
[OPENCL]:Loading single big chunk kernels
[OPENCL]:Mapping one big chunk.
0
Trial 2... [OPENCL]:Printing program log
[OPENCL]:
[OPENCL]:Creating one big buffer for the DAG
[OPENCL]:Loading single big chunk kernels
[OPENCL]:Mapping one big chunk.
[OPENCL]:Creating buffer for header.
[OPENCL]:Creating mining buffer 0
[OPENCL]:Creating mining buffer 1
[OPENCL]:Creating buffer for header.
[OPENCL]:Creating mining buffer 0
[OPENCL]:Creating mining buffer 1
21321045
Trial 3... 42991616
Trial 4... 42729472
Trial 5... 42467328
min/mean/max: 0/29901892/42991616 H/s
inner mean: 35505948 H/s
Phoning home to find world ranking...
update: now it seems stable at 20xxxxxx h/s most of the time.
Running latest git version of webthree ethminer
--> 16 MH/s <--
Unfortunately, I also have 2x 100% CPU usage (both in constant TurboMode 3.3GHz)
@Conibeast, which software are you running to achieve 18.3 MH/s
Also to use your miner do we need the run that dagapolalpypse thing or just run the miner ? I have the build you posted on the forums on another theread but it requires some Cuda DLLs which i did download but not sure if I am running the miner properly.
I have a mix of 7950, 280xs and 390s
Will the DAG file keep growing? If so does that mean our mining speed would decrease as time increase?
Whats the max size the DAG can grow to?
Also why does the benchmark in windows shows slower rate? Benchmark in windows shows 58mhash while in ubuntu shows 65mhash on the same cards? 3x 7950s.
Lastly how can I get my real hash rate from my miner on ubuntu? It does not show the mining speed like it does on windows.
Thanks.
Have you tried it some more?
[EDIT:] - Nevermind! I realize now that the memory bus width on the 360 is 256bit vs. the 384bit on the 280 - thereby even making 4GB a complete waste on the 380 :-/ [/EDIT]
min / mean / max: 0 / 10.5 / 25.5
Inner mean: 17.5 MH
Anyone know why I get 0 for the min? Is this something I can fix?
This is what prints when I run the bench:
[OPENCL]:Creating one big buffer for the DAG
[OPENCL]:Loading single big chunk kernels
[OPENCL]:Mapping one big chunk.
Trial 1... 0
Trial 2... 0
Trial 3... [OPENCL]:Creating buffer for header.
[OPENCL]:Creating mining buffer 0
[OPENCL]:Creating mining buffer 1
1660245
Trial 4... 25427968
Trial 5... 25515349
min/mean/max: 0/10520712/25515349 H/s
inner mean: 17534520 H/s
Any ideas to what is going wrong with the min? Thanks for the help!
--Neil
min/mean/max: 0/5050641/25253205 H/s
inner mean: 8417735 H/s
[OPENCL]:Creating mining buffer 0
[OPENCL]:Creating mining buffer 1
786432
Trial 3... 1223338
Trial 4... 1223338
Trial 5... 1223338
min/mean/max: 0/891289/1223338 H/s
inner mean: 669923 H/s
I've been meaning to do some fine tuning again, but the benchmark is no longer indicative of current mining stresses, hashrates seem to be far more tied to memory timings/performance than they were 6 months back.
Some cards seem to be suffering far more than others, 7950/280x with Elpida memory seem the most impacted thus far
http://forum.ethereum.org/discussion/comment/17961/#Comment_17961
I have 3 x R9 270x but my average under 30MH/s
windows 10 64bit , 15.12 driver
min / mean / max: 25.6 / 26.2 / 28.0
Inner mean: 27.2553 MH
Windows 8.1 PRO / XFX Radeon R7-370 2GB (stock settings) : 16.2MH (O/C best 17MH)