Topic: [ANN][YAC] YACoin ongoing development - page 51. (Read 379983 times)

I noticed that some peers consistently download the same block range from my client and after some checking it looks like those guys stuck on an old version and can't get past block 431490 or somewhere around:



Now the problem is that they are burning my bandwidth and my cpu (a lot of both!).
Though I think of myself as a generous person, I think they had enough.
If someone knows na easy way to disable communications with old version peers, I will be glad to hear (I have no problems implementing simple changes in source code and rebuilding client).
Or is there any blacklist parameter I could use?

-R 2550 -g 1  -B 2560 --lookup-gap 16
try this

Many thanks, but buffer don't work 
What is the correct hashrate for 280x ?
[/quote]

Many thanks, but buffer don't work  
What is the correct hashrate for 280x ?
[/quote]

If you have R9 280x  3Gb memory right buff value is normally (afaik) 2560  (least if system memory is enough! for whole rig)  but..   if lookup gap is 8 and -g1, your -R is far too high. Start tuning with low -R. (also you can try -g 2 but then agen other parameters need be related to it)   I will recommend (because problems)  that first run only with 1 to 4 r9 280x  connected to board and look how you can adjust it for work.

If I remember right it is far below 500hash/s for one 280x.

I use R7 240 4Gb and  Nvidia 750Ti's for YAC. (time when I test r9 280x there was too many difficulties and also (perhaps partially due to yacminer when pool diff drops below 1), high power consumption related to hashing speed.

R7 240 and Nvidia 750Ti beats it clearly (with N16, if N is lower situation turn very different)

Example: N14 chacha, My 4x750Ti rigs give around 12.5khash/s (some loss due to risers what affect with Nvidia but not with AMD radeons)   and  4xR9 280x give around 17.5khash/s.  But Nvidia use lot of less power related to hashrate.  With N16 situation is even more bad for R9 280x.

If it is even over 500h/s for one r9 280x then I still use these for YAC and immediately turn my R9 280 rigs to YAC  )

And with N16 it is fun but R7 240 4G start show its advantages. 0.35khash/s and  perhaps something like ~25W power (if moderate overclocked). (not system total power but card itself)




Add: just fast tiny test without any fine adjusting for best result. 250-260 hash/s one r9 280x, clocks e 1100  and m 1500 (changing m to 1600 did not give anything)

g 1, l-g 8, buff 2560, R 1260 (crash if try 1280) exept gpu 0 (display) what crash even with this but it can run with R 1220.
I did not test with g 2 or other settings. Rig have 4x r9 280x. and mobo is Asrock H81 BTC and 8Gb mem.

YACoin N=16 today!

you overrun your GPU's memory with these settings! (R= far too high)
Perhaps these settings work earlier with lower N value but really not with N16.
Look N tables what tell how much memory N16 need for one calc then you know what is perhaps ok R with lookupgap 8 and your buffer size (your buffer size is ok for 3Gb GPU's).

Also this may give some help http://www.ultracoin.net/configgen_raw.html
(remember select N16)

Btw...how fast hashing you expect with R9 280x with N16?





[/quote]

Many thanks, but buffer don't work 
What is the correct hashrate for 280x ?

I personally like yac.m-s-t.org but yac.coinmine.pl is popular as well.

I have a question,which pool is the best ?

5*280x
1*7870
8Gb Memory

[/quote]

YACoin N=16 today!

you overrun your GPU's memory with these settings! (R= far too high)
Perhaps these settings work earlier with lower N value but really not with N16.
Look N tables, it tell how much memory N16 need for one calc then you know what is perhaps ok R with lookupgap 8 and your buffer size (your buffer size is ok for 3Gb GPU's).

This time N=16.  (N14 need 4Mb, N15 need 8Mb and N16 need 16Mb. Next step, N17 need 32Mb etc...

(if stay with -g 1)
Lookup gap 8 means that this 16Mb (or what ever it is with current N value) is divided to 8 "gaps". So, one is 2Mb.
Memory (buffer) 2560Mb can take 1280  2Mb "gaps" at once.  So  -R max is 1280.

then this "hassle" (--temp-cutoff 95 --gpu-fan 0-100 --auto-gpu --temp-target 85 --temp-overheat 95 ) any idea for these paramaters? Why run so hot, and why --auto-gpu?

If I set these for my R9 280x rigs (Sapphire Toxic R9 280x) I will set --temp-cutoff 90 --auto-fan --auto-gpu --temp-target 70 --temp-overheat 80 --temp-hysteresis 2

I do not understand what for you use --auto-gpu ?? it is for  autoadjust GPU clocks (for temperature),   really strange in this situation. (perhaps mistake)

And if really need some details. I'm not GPU specialist! So these are only some winks for starting finer adjusting.




Also this may give some help http://www.ultracoin.net/configgen_raw.html
(remember select N16)

Btw...how fast hashing you expect with R9 280x with N16?

Hello,
I do not understand why I am unable to use yacminer.
Even my old .Bat do not want to work.
I use this bat:




5*280x
1*7870
8Gb Memory

anyone have an idea?
Thanks

Edit : ok it's works without "--buffer-size" but 450h for a 280x.
How to choose my buffer size ?
Thanks

Do you mean that now I should turn on my amd rigs for YAC. And use nvidia's for other things...  

Yes, these nvidia rigs can use also nicely with N14 chacha. 4x750Ti give around 12-13kh/s and 4x R9 "toxic" 280x  house heating rigs around 16.5-17.5kh/s (gross)  But, if need pay electric and if heat goes to waste, these nvidia beats these amd radeons clearly. With N16 chacha, nvidia 750Ti even more better than R9 280x what start really get hick-up with N16 chacha.

Previously I did some exercise with my R7 240 4Gb test card, ands result was poor and then also I did not have time to do more deep testings. Also with R9 280x I give hands up with N16 chacha.
 
Perhaps after this change I can try agen if result is better.

btw, also I have noted this situation time ago but no one react to this what is really going on.




Edit|Add:


After pool diff set to 1  (previously 0.25)  I have tested agen my R7 240 4Gb with yacminer and YAC. Result: without any problems easy to 350h/s 340hash/s (gross) (Engine 1140 1100, Mem 1000). It can more but stability is not perfect and start generate HW'errs or just crash randomly.
Now yacminer also show accepteds ok.


Edit Add 2:

Are there some kind of bug in yacminer or is it only that I do not understand something.
If set yacminer for debug display, it display N17 for every works. (but top part of display of course show N16 as it is)

Qxw, you shouldn't cry so much! Because with your nvidia miner you had constantly get two times more yacs since 0.25 difficulty. So you owe all the amd gpu's miners (such as Grim) literally! 

Great?  My earnings drops dramatically. 50% drop or even more. Without any markable change in difficulty what have been around 0.0012. Something is now very odd if compare time before this change. Also now there is lot of more times when my workers drops offline. Previously my 2 rigs (GPU rigs) never show in pool "my workers" that they are off. Now many times. Now my other miners, (i7 cpu's) are mostly  "off".   Previously system (pool) works just ok. Now it is so terrible that perhaps need try solo. I do not accept over 50% drop without any markable change in Network and my hashrate. Rule no 1. is that do not "repair" and "adjust" well working system.

Please change pool hashrate display to display hashes/s instead of khashes/s.  khash/s display is too rough for N16.
With this peoples can check what miner display and pool thinks about hashrate. If there is too big difference then people need find problem in his own end.

I can not yet proof it but now it really looks like that pool do not count every accepted shares. (if I look my miners displays and pool dashboard "valid" count. They do not match and after this change situation looks like  more bad. (lest in my case)

Now it works great. My earnings from last blocks are up for x2 or x3 from previous one.
The problem started 1 month ago, but  solution was obviously fast and simple. Maybe next time   

Hi - sorry - I could not follow this thread. I was not aware of miner's limitation to diff 1 so I will change stratum settings right now.

feeleep

EDIT: diff changed to 1

I'll shed some light on this - For some reason feelep lowered the share difficulty to 0.25 (a non-integer).  From my memory, the mining software doesn't even validate any share unless its difficulty is 1 or higher - and it's all integer based, so it would be a major undertaking to make it work for his pool.  So, you are definitely losing out compared to people who can submit shares lower than 1 (maybe cudaminer or ccminer can?).  I spent an afternoon looking through the yacminer code for where it made that determination to only validated shares of diff 1 or higher, but never found it (it may have been in the CPU miner software).  That's when I decided to solo mine, and started netting almost double compared to the pool.

I set up my own pool after latest N change and my coin generation doubled after that.
However I have issues with my network that only affect my gpu miner.

I get at least 10% of these errors with yaccminer (gpu):
Rejected 810ea74e Diff 1/0 GPU 0  (job not found)

Every couple of minutes I get "Stratum connection to pool 0 interrupted".

No errors from cpu miner working on the same machine, connecting to same port with same pool difficulty.
Nevertheless yacminer's sucessfull submission share rate is almost double than my cpus (so is it's hash rate).

But my gpu miner consistently!!! (over a week now) finds only half (or even less) of blocks of what my cpu miner does inspite of it's doubled hash (and sucessfull share submission) rate. Nobody can convince me this is by chance - something is favoring cpu mining. Favoring big time.

My friend also mines on this pool with all cpu miners - he is also finding far more (2x-3x) blocks with his miners than I do with my gpu miner and he's comulative hashing rate isn't much higher than my gpu's. Besides more of my gpu mined blocks get orphaned (50% at least).

I am sure coinmine.pl is loosing at least half of what it could produce with that hash rate. I don't know why.

Also I think CPU miners on that pool are loosing even more: Pool difficulty was probably not ment to go that far below 1 than it was "pushed" with N factor increments. I tried setting my pool difficulty to 0.001, 0.000001, even 1e-15 and dual core athlon 64bit for example still needed hours to hit a share. That means pools don't know exactly "when" cpu miners were working because their share submissions are so rare. And gpu miners submit far more shares and rarely find a block - at least that is what I am observing.

So there are many issues combined here, but I think current block rewards this high are due problems that GPU miners are observing. Pools also have issues with setting/calculating fair share difficulty. Perhaps miner software also have issues with recognizing difficulty (difficulty number get's rounded and numeric precision isn't accounted properly?).

I don't have time to check all the code now, my own priority is to fix my network issues. And after that gpu miner's picture could improve a bit.
But I can also confirm that my friend who is now mining on my pool with cpu miners gets 2-3x what he got on the public pool before.

I mean together we have less than 1kh and we hit 5-10 blocks per day. Plus orphans.
In a way that is encouraging, because I think we all could be mining on "private" pools until reward get's below 40 or even 25.

I will be glad to hear if someone has any recommendations for my gpu miner's problems (yacminer --scrypt-chacha --lookup-gap 6 -R 1280).

My ping time to coinmine.pl is short (ok)
to m-s-t.org I have nearly 200ms ping. (too long)

Good news! Yac/Usd is cool, hopefully they don't kill crypto-trade this time.

yep just switched to http://yac.m-s-t.org and everything works out fine ...


https://www2.coinmine.pl/yac has a share count issue !!!!

Only the first 6 miners at https://www2.coinmine.pl/yac show a hashrate above 1 khash,
I do believe many miners there share the same issues as I do at that pool.


It seems to me that the pool doesn't count every share.