Topic: XMR-stak-JK 2.10.7 Compiled with no devfee (Updated: 2019-8-18) - page 4. (Read 28606 times)

2.10.2 give  me an error about failing to open config file .openclcache/.

the previous version crash my rig in a few hours, thanks

your settings may need to be trimmed a little;   there were changes on the AMD side of 2.10.2.

Just a heads up that I found 2.10.1 to be stable with my AMD Radeon RX580 cards but 2.10.2 wasn't.

Dan

Compiled and uploaded version 2.10.2.

see build notes on github for compiling using --flags to disable options....   best bet i can give you.


Did you follow the instructions for being able to compile for openCL?  Install blockchain drivers/SDK, etc?

*note*  2.10.2 is compiling, I will have it uploaded when its finished.

I am able to compile just fine with a Nvidia GPU, but I get errors doing it with an AMD GPU. It seems to still want CUDA according to the error message.

What exactly do you have to do differently when compiling with an AMD GPU or even just using the CPU?

https://github.com/fireice-uk/xmr-stak/releases/tag/2.10.2  

This release fixes a memory leak in the OpenCL (AMD) backend if cryptonight_r is mined which filled the full main memory and crashed the miner after a few hours.

Sorry for being slightly OT, first time Monero miner.
Looking at XMR-Stak benchmarks page for Nvidias, it seems to me that the same OC method that works best for ethash is also best for cryptnonight: maximum memory OC and minimum core.
So something like 80PL, +700 memory and -100 core would be best?
Thanks.

I deleted your post;  as it contained links and advertisements that are unrelated to my thread 100%.  This makes me think you are spamming on this website and potentially posting useless content since your post was 100% out of context of this thread; set aside the only correlation that my thread is about a specific build of xmr-stak.  You arent even sharing my build; but in fact one with a reduced devfee that is most likely redirected to your personal account.

To put it simply: I do not know why you posted here, or how it even pertained to this thread.   Please do not do this on bitcointalk.  Its bad manners.

Sorry;  forgot to include them.  updated the link on the main page with a new file.

compiled and uploaded 2.10.1.  Many bugfixes in this release;  I suggest everyone update.

No prob, glad to help.


Thats about what I get on my reference style 1080.

you already checked my main recommendation: https://www.xmrstak.com/

You can look up other users settings for optimizations;  I did that for mine and only had to trim a tiny bit on my 1070;  I just left my 1080 on auto since it rarely mines now and is in my daily use PC.

Look at any 1080 ones you see; as well as 1070;  they will probably be similar; just take into count when transposing the 1080 having ~8-10% more computing power.

So I've tried this on my "regular" GTX 1080's and I'm getting hashrates just below 700 per GPU if I use the automatically generated parameters and the ETHlargement pill on the new Cryptonight/r algo.

Has anybody got better results? Any recommendations for tweaked parameters? I've checked that XMR-Stak bench site but they've got very little results for 1080's...

I installed new cuda and affined the threads to one cpu-core, working now. Thx for the tip and your work  

This version works well for me, thank you   

Compiled and uploaded.

2.10.0 is out - https://github.com/fireice-uk/xmr-stak/releases/tag/2.10.0

Dan

Ill have the recompiled test version added to this reply once it finishes compiling and being uploaded.  Stay tuned.



Here is the updated file: xmr-stak-JK-2.9.0-1.zip
Be sure you have generated new .txt files for V 2.9.0.

https://github.com/fireice-uk/xmr-stak/issues/2271
Looks like 2.9.1 will have it fixed.... seems like a known issue being worked on.  some have success, others don't.  I am attempting a recompile with the changes noted in that issue; but i probably wont jump any further down the rabbit hole and just wait for V2.9.1+ if this recompile with the minor changes doesn't help.



could be an issue with a system overclock;   those test functions do work and haven't changed since day one....  They pass on my machines.

could be you need to get your amd.txt from v 2.7.x/2.8.x, and copy the optimizations to the new amd.txt generated by 2.9.0;  so be sure you didnt copy your old .txt files;  you need to edit the new ones.

I had to let mine redo the nvidia and amd txt more than once for some reason.


My test rig:
B250D8P I3-6100 8Gb
GTX 1070, GTX 650Ti, RX560

I did update my video drivers to the current version for Nvidia because of the security bug that was found;  and I left my AMD blockchain drivers alone as its been since I installed the RX560.

Ill zip up and upload the whole folder including my configurations directly from that miner;  and post a link in a minute.  You can see if its any different for you.


I do suggest updating the nvidia drivers though;  the security bug is a real threat; as well as I think needing the updated driver is a reality (I did compile under CUDA 10)