Topic: [ANN] sgminer v5 - optimized X11/X13/NeoScrypt/Lyra2RE/etc. kernel-switch miner - page 185. (Read 877844 times)

The big fluctuations are caused because the pool of your renter is letting your rig hash at a very high difficulty. The average should work out OK but the graphs look shitty. Check your SGminer and you will see they are running at diff 0.5 or something in that area. Nothing you can do about it, other than asking your renter to lower the diff at the pool.

No I'm with you bullus.  Hear me out though.

I have every build of aznboy84's including the very first one where he cloned the 6-25 prerelease off of nicehash and created bitblock.cl

I'm trying to isolate which version of his miners can reproduce the darkcoin-mod .bin UPON CREATION, reliably, and see if we can't track down why, with the matching source.  The problem with his github is he didn't know how to use git until about the 30th, so all of his files are LOCAL UPLOADS with no versioning.

If I'm on 14.7 and do all those mods and still have now dropped to 5.6 - 5.7, then this has to be a combination of build factors AND the drivers that were available at that time.  Since 14.6RC2 came out late during that weekend, I wasn't using it, I was using 14.6RC1; which means there may be a reason why that uploaded .bin of yours ONLY WORKS STABLY on ONE RIG.  If you put that .bin across to other cards like a Tri-X OC 290X, it wrecks them SICK in a matter of 15 minutes, or even worse -> HUNG RIG.

I want to know which version built that badass .bin, and then recreate and build bins for my 2 other rigs that contain Tri-X OC, and one that has mixed Tri-X OC and standard Sapphire 290X.

This latest build from 7-25 is rock fucking stable, but now we're barely coming up with 5.6 across 3 cards, which is a vast improvement from getting 5.4|5.4|5.0 which is why I had opened https://github.com/sgminer-dev/sgminer/issues/347

This can't be some mystery miner; one of us has to have this shit; and it's hard for you to share only your bin when it might not apply to others; I'm lucky in that I have one rig clocking in at 6.1|6.0|5.9 but 2 other rigs hodling only at 5.6|5.6|5.5 (which, is still an improvement)

Yes I've done luffa 1 and keccak unroll 6 and hamsi 7 (irrelevant to x11)

About betarig. I change the no-submit-stale to false and even lowerd the intensity and use my 290x for the total of 10mh/s ,but stil fluctuate ..I did what I can so it's got to be annother issue not at my side.

I rather suggest u lower ur memclocks..I was running at 1500, but after several hours it wasn't stable anymore, even at idle (booting into windows)..

both BF4 bought at the same time

edit: broke out the sapphire trixx and boosted my vddc by +25, now running 1050/1500 xI 30 lets see if it will stay stable

Could be, is your 1 GPU a Sapphire 290X, red box or BF4 edition?

Asus dcII for the win

xI=30 =6050kh/s
xI=64 =6160kh/s

Could be, is your 1 GPU a Sapphire 290X, red box or BF4 edition?

The big difference is that I only use 1 gpu, u both uses 3 or more cards. Is that causing the instability of your rigs?
Maybe try 1 gpu for test with xI=64?

The big difference is that I only use 1 gpu, u both uses 3 or more cards. Is that causing the instability of your rigs?
Maybe try 1 gpu for test with xI=64?

it is strange that undoing an "optimization" gives better hash rate....

6|6|6, with last card lagging on curses readout, but increasing up to its potential.

Fucking weird.

I thought ystarnaud/slik had knocked the credibility of replacing those 2 loops.  I didn't change darkcoin-mod.cl at all; just replaced in groestl.cl.

Gonna get a morning nap and hope to not wake up to hung rigs.

Much appreciated; still very interested in bullus' specific miner as well.

EDIT:  So much cleaner building them with my specific cards, than applying another pre-built bin which is notorious for crashing my drivers on 2 rigs.

i have found the majority of the increase comes from the change i made in groestl.cl on mine, found this in aznboy84's darkcoin-mod.cl
also try xI 30

I think you're on Linux right?  Do you have the ability to get the Windows binaries too?

No I'm with you bullus.  Hear me out though.

I have every build of aznboy84's including the very first one where he cloned the 6-25 prerelease off of nicehash and created bitblock.cl

I'm trying to isolate which version of his miners can reproduce the darkcoin-mod .bin UPON CREATION, reliably, and see if we can't track down why, with the matching source.  The problem with his github is he didn't know how to use git until about the 30th, so all of his files are LOCAL UPLOADS with no versioning.

If I'm on 14.7 and do all those mods and still have now dropped to 5.6 - 5.7, then this has to be a combination of build factors AND the drivers that were available at that time.  Since 14.6RC2 came out late during that weekend, I wasn't using it, I was using 14.6RC1; which means there may be a reason why that uploaded .bin of yours ONLY WORKS STABLY on ONE RIG.  If you put that .bin across to other cards like a Tri-X OC 290X, it wrecks them SICK in a matter of 15 minutes, or even worse -> HUNG RIG.

I want to know which version built that badass .bin, and then recreate and build bins for my 2 other rigs that contain Tri-X OC, and one that has mixed Tri-X OC and standard Sapphire 290X.

This latest build from 7-25 is rock fucking stable, but now we're barely coming up with 5.6 across 3 cards, which is a vast improvement from getting 5.4|5.4|5.0 which is why I had opened https://github.com/sgminer-dev/sgminer/issues/347

This can't be some mystery miner; one of us has to have this shit; and it's hard for you to share only your bin when it might not apply to others; I'm lucky in that I have one rig clocking in at 6.1|6.0|5.9 but 2 other rigs hodling only at 5.6|5.6|5.5 (which, is still an improvement)

Yes I've done luffa 1 and keccak unroll 6 and hamsi 7 (irrelevant to x11)

Try using 14.4 with the 14.7 opencl files copied to the sgminer folder.

The 14.7's have issues with overvolting/throttling random gpus..

Will upload the whole miner I downloaded it, when I get home.