Topic: SILENTARMY v5: Zcash miner, 115 sol/s on R9 Nano, 70 sol/s on GTX 1070 - page 6. (Read 209334 times)

What is the url to your fork?

By the way, I found out the root cause of a low number of shares.
It turned out that there is a bug in AMD's drivers, so you need to so specify the "-O1" build option to disable most of optimizations when you call clBuildProgram if the miner is running on AMD cards. The optimizer must have incorrectly eliminated s_waitcnt. What a wonderful world.

Here are some stats for my fork. (The red rectangle marks Rounds 0 through 8.) It seems that I could at least squeeze out another 10-15% speedup. The current implementation of SA doesn't use shared memory effectively, so it is not all that surprising.

 I've noted the same thing on my "extensive testing" HD 7870 rig.

 I was already aware that the 15.12 drivers did not work with RX 4xx series cards (cards came out AFTER the driver), so I went straight to 16.9.2 for the newly rebuilt RX 470 based rig.

 It might be time to finally move to XUbuntu 16.04 or 16.10 from 14.04 as my base LINUX build system.

I wonder where Marc is at with v6.
My fork is running at 130 sol/s on 480, only after a ridiculous amount of work.
I'm now using NR_ROWS_LOG=16 as the new sorting algorithm I implemented seems to prefer a larger NR_SLOTS.
I'm trying to set up NR_ROWS_LOG=14 and 15 to see how that would affect performance.

Just wait...there's nothing wrong to use closed source miner. After all higher hashrate is priority for ZEC at the moment. It may take more than we expected for competitive open source miner to step on the scene, but eventually it will happen. Fast OSS ethereum OCL/CUDA miner was released almost 1 year after ethereum's launch

really would like to see v6 ... i know its a lot of effort and hopefully it will see the light soon ... just its enough of claymore & optiminer ... im actually using v5 so far !!!

Just FYI, but RX 4xx cards aren't supported on the 15.12 (fglrx) drivers - it's amdgpu-pro (16.xx drivers) only.

 I'll keep that in mind for early next week when my first pair of RX 470s are due to appear in my mailbox.

I compared the performance difference between the 15.12 and 16.9.2 for my R9 390 and 7970. There is  no difference.

Hey Marc,
Any ETA for v6?

I'm not impressed with that 16.12.1 driver even when it does work - did some extensive testing and I find it loses hashrate compared to the 15.12 drivers on my older hardware.

 I'm not exactly shocked it doesn't work well with new gear too.

Congratulation mate! I'm really happy when I know that you won that challenge.
I'm waiting for new version of Silent Army, can't wait to coming back to linux 

One more reason competitive Linux miner should appear soon.

if anyone cares i deleted my posts i don't want any hard feeling direct at me right or wrong i should never had made the comments i did , im not sorry for making the post and not saying I'm sorry for making them i just shouldn't have made the comments and I'm not asking to forgive me i meant what I said i just should have kept it to my self. .


i got the drivers installed after i put my test 480 back to default bios.

SO i guess you can say these driver won't work at all if you have a modded bios sigh . so no way to test them with a modded bios unless you try to install them with out using driver signing or with driver sighing off , which I did not try . or you driver sign your modded bios which i read a while back is possible with a fee program, i found, i can't find now .


interesting the drivers do save power but not like they say at default settings so far but do heat up a lot and my hash did increase at defaults with these drivers i went from 210 cmV9 to 221 cmV9 with these drivers. i have deep freeze or whatever it's called on . using about the same power i was with the modded bios if GPU-z is right 88 watts tops .

Everything is at defaults with these drivers .



it hashes better or more  uses the same power as my modded bios did but gets hotter. so that make me wonder how it will do with modded bios owell till some one figures out a good way around the driver signing without a lot reboots il stay at defaults and hope for the best for the new SA windows version that is coming .
 

The new AMD Crapson ReBleep driver are not able to support even two cards. All cards get disabled when you plug in 2nd card..and cont. with all cards.
AMd is just crap in terms of drivers..

if anyone cares i deleted my posts i don't want any hard feeling direct at me right or wrong i should never had made the comments i did , im not sorry for making the post and not saying I'm sorry for making them i just shouldn't have made the comments and I'm not asking to forgive me i meant what I said i just should have kept it to my self. .


i got the drivers installed after i put my test 480 back to default bios.

SO i guess you can say these driver won't work at all if you have a modded bios sigh . so no way to test them with a modded bios unless you try to install them with out using driver signing or with driver sighing off , which I did not try . or you driver sign your modded bios which i read a while back is possible with a fee program, i found, i can't find now .

cya
thanks top

Yeah, those 16.12.1 drivers really messed up my rig! Went back to 16.9.2

Thanks for heads up.  I remember being puzzled for a while as to why my bios mod didn't work on one rig, even though it worked flawless on another.  Had the 16.8 drivers, which also check bios signature.

@toptek
deleted his post about his insider info and AMD hatred and 20% less power usage in mining 
new driver 16.12.1 is total failure for AMD miners, don`t try them they also now require signed bios for anyone who has modded RX 4xx si beware