funakoshi2718 / funakoshi-miner

Funakoshi Equihash Cuda Miner
62 stars 47 forks source link

Seem to be slower than DSTM #3

Closed StrobFX closed 6 years ago

StrobFX commented 6 years ago

Hi,

nice to see a new miner but right now it seems to be slower than DSTM (I still have to try it longer to get a good mean). I hope you can increase the speed a bit

But since it seems to be slower than DSTM. I would suggest to remove the dev fee or at least offer the option to remove the dev fee without slowing down the miner for those who set the dev fee to zero. At least until you can find a way to make it faster. That way you could maybe get a user base and then remove tha ability to set dev fee to zero once you get it faster. Or else I fear no one will use your miner.

funakoshi2718 commented 6 years ago

Thanks Jocelyn I have already improved performance by 1% (release v1.1). In the following days I will work hard and try to improve performance even more.

StrobFX commented 6 years ago

ok sounds great I will try it.

funakoshi2718 commented 6 years ago

Hey @StrobFX I am happy to inform you that after months of hard work, Funakoshi miner is now the fastest equihash cuda solver.

StrobFX commented 6 years ago

Wow! how much faster?

funakoshi2718 commented 6 years ago

It depends on the specific nvidia card. On gtx 1080 between 5 to 20 sol/s

On Thu, Apr 5, 2018 at 4:05 PM, Jocelyn Strob Simard < notifications@github.com> wrote:

Wow! how much faster?

— You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub https://github.com/funakoshi2718/funakoshi-miner/issues/3#issuecomment-378928704, or mute the thread https://github.com/notifications/unsubscribe-auth/Ah-xu9YbpnJpEby2wt3lk6UclPINns3oks5tlhZ9gaJpZM4SeCDx .

StrobFX commented 6 years ago

On my Titan Xp collector's edition, v2.1 seems to add 50 sol/s from v1.2 (which is a whopping 7%!) GOOD job!!! This is from your display of speed. I will see if it reflects on the pool side.

funakoshi2718 commented 6 years ago

My miner reports true meassured hashes/s and sol/s. dstm also reports true values. By between 5 to 20 sol/s for 1080, I meant compared to dstm, not compared to v1.2. Thanks very much for sharing your findings with me.

funakoshi2718 commented 6 years ago

I would like to add that you are very kind, thanks.

StrobFX commented 6 years ago

Because I'm also a Funakoshi Sensei fan! do you practice Karate?

tech3ric commented 6 years ago

slower compare to dtsm with gtx 1060..any enhancement for gtx 1060?

funakoshi2718 commented 6 years ago

You are right, thanks. I haven't previously run my miner on 1060 and it is slower. I have started working on improving performance on 1060. When finished I will publish a new release.

tech3ric commented 6 years ago

i'm looking forward..thanks

funakoshi2718 commented 6 years ago

I have improved the performance of Funakoshi-Miner on GTX 1060.

kire3 commented 6 years ago

nice..ill try and let you know..

kire3 commented 6 years ago

hi, i'm getting an error on my GTX 1060..already upgraded with latest driver. 390.59

CUDA error 'CUDA driver version is insufficient for CUDA runtime version' at line 61

funakoshi2718 commented 6 years ago

Please upgrade Cuda runtime from Nvidia website

tech3ric commented 6 years ago

hi, i'm still getting the same error after installing the Cuda Toolkit after reboot. Do i need to install the Nvidia Accelerated Driver that comes with it? Won't it replace my existing driver?

funakoshi2718 commented 6 years ago

The error msg 'CUDA driver version is insufficient for CUDA runtime version' is returned by Cuda and not by my code. This error means that Nvidia driver is too old or that Cuda runtime is too old. One of them need to be upgraded. On my ubuntu I only upgraded the Cuda runtime not the driver and it worked.

Release v2.6 of Funakoshi-Miner is compiled with Cuda 9.2 If your Cuda runtime is too old you need to upgrade it from Nvidia website

If your driver is too old you need to upgrade it

tech3ric commented 6 years ago

Hi, Im still getting the same error after upgraded to the latest Cuda and Nividia drivers. Btw, Im using GTX1060 3Gb. Could this be the reason of this error?

funakoshi2718 commented 6 years ago

I don't think that the reason is the memory size otherwise the error would have been: out of memory.

I have successfuly run the current version of the miner on 2 different machines with GTX 1060 6GB, one of them my physical PC and the other a cloud machine. In both cases all I had to do was to upgrade the Cuda runtime from Nvidia website.

I have no clue why you get 'CUDA driver version is insufficient for CUDA runtime version'.

tech3ric commented 6 years ago

Hi, finally its working now with nvidia driver 396.24 and cuda 9.2. But it's still slow for GTX1060 3Gb.

funakoshi2718 commented 6 years ago

Thanks for notifying me. Can you tell me how much I/secs (iterations per second, not solves per second) Funakoshi-Miner reports on your system compared to the competition. The speed is determined by I/secs because solves per seconds changes all the time and not stable and thus only avarage over a few hours should be checked. But Interations per seconds is stable and can be used for checking performance. Please tell me the avarage reported I/s over 1 minutes.

tech3ric commented 6 years ago

Hi, Here is my average.

[2018-May-27 12:46:39] [MINER] Received new job #e008daca0b14f3d29609 [2018-May-27 12:46:55] [GPU-0 (15 sec)] 154 I/s, 285 Sols/s [2018-May-27 12:46:55] [GPU-1 (15 sec)] 157 I/s, 291 Sols/s [2018-May-27 12:46:55] [GPU-2 (15 sec)] 144 I/s, 260 Sols/s [2018-May-27 12:46:55] [GPU-3 (15 sec)] 144 I/s, 271 Sols/s [2018-May-27 12:46:55] [GPU-4 (15 sec)] 142 I/s, 271 Sols/s [2018-May-27 12:46:55] [GPU-5 (15 sec)] 146 I/s, 281 Sols/s [2018-May-27 12:46:55] [TOTAL (15 sec)] 887 I/s, 1659 Sols/s, 0 Accepted, 0 Rejected

Thanks

funakoshi2718 commented 6 years ago

You are right. I need to improve the speed of my miner on GTX 1060. There was a big jump but more is needed.

tech3ric commented 6 years ago

I'll wait for your next release. If possible please include the Temp and Power on the output. Thanks