Mine on a pool without extranounce support. F.ex wafflepool. What rates are you getting on benchmark mode?
benchmark (970) is:
X11 - 7.3
X13 - 5.4
X15 - 4.7
Topic: CCminer(SP-MOD) Modded NVIDIA Maxwell / Pascal kernels. - page 1223. (Read 2347601 times)
benchmark (970) is:
X11 - 7.3
X13 - 5.4
X15 - 4.7
i made a fix for yaamp today... but todays disconnections are not related to the code. yaamp doesnt use extranonce, only nicehash and westhash
Mine on a pool without extranounce support. F.ex wafflepool. What rates are you getting on benchmark mode?
@sp_ I'm mining at yaamp.com and release 9 doesn't seem to connect very well to the mining pool. No problems with release 8.
Same here. GTX 970, tried both exe files different pools (X11,13,15) it's a hit and miss.
At least I couldn't find any dependency on when it connects and when it doesn't.
+1 same thing , lots of connection interupts...
@sp_ I'm mining at yaamp.com and release 9 doesn't seem to connect very well to the mining pool. No problems with release 8.
Same here. GTX 970, tried both exe files different pools (X11,13,15) it's a hit and miss.
At least I couldn't find any dependency on when it connects and when it doesn't.
@sp_ I'm mining at yaamp.com and release 9 doesn't seem to connect very well to the mining pool. No problems with release 8.
Edit: CCminer(release9) sends 1 share then doesn't do anything else after. However gpu is at 99% load. Same for every algo. It does work for nicehash/westhash and p2pool.
Edit: CCminer(release9) sends 1 share then doesn't do anything else after. However gpu is at 99% load. Same for every algo. It does work for nicehash/westhash and p2pool.
yes, i need to double check the commit, i lost 1ms, edit: ok fixed 
The simd fix for 3.0/3.5 is nice. I only fixed x11 in my fork, so yours is bether. When comparing, you should also check the Wattage. Some changes does not improve hashrate, but they do reduce power.
nice work, that give me a lot of things to check for 3.0 simd, check my latest commit...
for 3.0 simd, check my latest commit... I've made a pull request on github.
The Hashrate didn't increase after your commit, but the power consumption is down. Memory controller load is now 14-15% and power 65-100%. I have also removed /merged some kernals, so this can also be the reason. Memory load in the previous release on 980 was above 20%.
release8 had 28% [email protected] MHASH
Thanks. Checked in a fix for 650 now.
I also managed to make x11 and x13 25KHASH faster by moving the final test into the (echo and fugue) kernal. @github.
Will prepare an exe for test. Build 9.
I also managed to make x11 and x13 25KHASH faster by moving the final test into the (echo and fugue) kernal. @github.
Will prepare an exe for test. Build 9.
The 650 is doing pretty bad. 727KHASH@x11. The merged simd code (compress1/compress2) into one kernal, reduced the performance by 10KHASH. (1.3%) But Cudaarch is not possible to check in host code.
In fact it is
check fugue :int intensity = (device_sm[device_map[thr_id]] > 500) ? 22 : 19;
I've made a pull request on github.
Looks good, I have merged you changes.
The 650 is doing pretty bad. 727KHASH@x11. The merged simd code (compress1/compress2) into one kernal, reduced the performance by 10KHASH. (1.3%) But Cudaarch is not possible to check in host code.
works fine with the echo Revert... and the change was not faster
What is the perf on x11 for the 650 ? i had one last year, just curious, it was a good card for games.
What is the perf on x11 for the 650 ? i had one last year, just curious, it was a good card for games.
I've made a pull request on github.
do you have a 970 test for every algo, or x algo at least?
I am testing on 650,750,750TI,970 and 980. but mostly on 750ti.
Yeah, had to rollback the echo change. It broke the chain.
wow... you missed to add x11/cuda_x11_luffa512_Cubehash.cu in Makefile.am
Sorry, i made some other changes for pools and jansson but should be easier to cherry-pick
but hmm this commit seems to break x11 algo (tested on your repo on linux 750Ti) and miss the 80 max reg tweak
Sorry, i made some other changes for pools and jansson
but should be easier to cherry-pickCode:
x11/cuda_x11_shavite512.cu x11/cuda_x11_simd512.cu x11/cuda_x11_echo.cu \
+ x11/cuda_x11_luffa512_Cubehash.cu \
x13/x13.cu x13/cuda_x13_hamsi512.cu x13/cuda_x13_fugue512.cu \
+ x11/cuda_x11_luffa512_Cubehash.cu \
x13/x13.cu x13/cuda_x13_hamsi512.cu x13/cuda_x13_fugue512.cu \
but hmm this commit seems to break x11 algo (tested on your repo on linux 750Ti) and miss the 80 max reg tweak
Jump to: