CCminer(SP-MOD) Modded GPU kernels.

[AzzAz]

PASCAL GPU RELEASED

http://www.tomshardware.com/news/nvidia-pascal-tesla-p100-gpu,31557.html#xtor=EPR-8886

I'm surprised they released tesla first, I would expect them to beta test with the consumer line.
Maybe they are confident enough in the quality.

same ...

though nvidia do thngs in the most unorthodox ways ...

cant wait to see how cuda 8 performs with it ...

#crysx

If I understood properly: they will hit the market NEXT YEAR??

[crysx]

PASCAL GPU RELEASED

http://www.tomshardware.com/news/nvidia-pascal-tesla-p100-gpu,31557.html#xtor=EPR-8886

I'm surprised they released tesla first, I would expect them to beta test with the consumer line.
Maybe they are confident enough in the quality.

same ...

though nvidia do thngs in the most unorthodox ways ...

cant wait to see how cuda 8 performs with it ...

#crysx

If I understood properly: they will hit the market NEXT YEAR??

now that tesla pascal is out - it will be in production for the corporates ...

so chance are - nvidia already has the commercial consumer line already produced and is waiting on the influx of cashflow from the corporate range before they release the consumer line ...

this MAY mean that they are only a few months down the track - if that long at all ...

the consumer line will most definitely be the most expensive cards on the market when that happens ... so for me - it will be a good 6months before i buy any pascal in bulk for thefarm - though a single card for testing maybe on the books ...

#crysx

[pokeytex]

Just tested the Alexei miner. He is not submitting all the found solutions to the pool. (only one nonce per warp) This meens lower hashrates on low difficulties. Managed to gain 10MHASH on the gtx960M. Release 9 (alexus sp-mod) will be sendt to the decred donators.

Sendt.

Report your hashrates. I have only tested it on the gtx960M (768 cuda cores)

@sp - please resend #9 to me - I did not receive it - thanks - pokeytex

[GarlukKY]

Report your hashrates. I have only tested it on the gtx960M (768 cuda cores)

980 TI -i 31 1641mhz I'm getting about 3285 which is +42ish vs Mod#7. -i 31.9 bumps it to 3292.   edit: these are numbers reported by the miner.

Further testing....
980TI
Yiimp reports roughly the same hash rate (must run at least diff 2 which causes reported hash rate to vary depending on luck solving the job).
Suprnova poolside (diff 1) reports a MAX of around 2.65GH with an avg of roughly 2.45 at -i 31 (and I did let it run for over an hour and kept an eye on it the entire time) using Mod9 whereas Mod7 poolside shows 2.7-3.5 averaging around 3.1 (data taken today and from DAYS of monitoring).  Mod9 -i 31.9 was averaging about 2.3GH (see offtopic discussion at bottom of post for more info).


750TI
-i 24.5 1339mhz (primary display) Mod7 614MH, Mod9 620     (Mod7 -i 24.2 reports about 602 with MUCH less screen lag)
I am getting better poolside reports with Mod9 580-720 vs 470-630 Mod7.  I am also significantly more likely to solve a job with Mod9.  This is just an eyeball estimate but solve rate seems to have gone from 60-75% to around 85% on my 750TI.  It is uncommon on Mod9 to have a string of 3 or more jobs fail to find a solution whereas Mod7 often had runs of 5-7+.  The 980 TI also reports a higher solve rate on Mod9 but it's not as profound.



Conclusion: 
On my equipment, Mod9 appears to be a gain on compute 5.0 but something is broken on 5.2 cards, the poolside hash as reported on Suprnova is significantly lower.
Also, the --show-diff flag is broken in Mod9.  It always shows (diff 0.000) for every solution.  This flag was also broken in the 1.5% miner fee version from a few days ago.



Offtopic:  For the 980TI I actually get better throughput using -i 31 than -i 31.9 (as reported poolside).  The typical job time is about .8 seconds vs 1.1-1.2 for the higher intensity.  These numbers I can see when I have a run of jobs for which it did not find a solution (don't use -q).  I am a bit more likely to find a solution meeting the difficulty requirement of the job at the higher intensity but the slower processing time lowers throughput.  This made me scratch my head more than a little bit...why the higher solve rate at the higher intensity?  But I've watched the numbers stream past for hours.  Tried -i 31.5 and it yields same results as 31.9 for throughput.

[sp_]

Offtopic:  For the 980TI I actually get better throughput using -i 31 than -i 31.9 (as reported poolside).  The typical job time is about .8 seconds vs 1.1-1.2 for the higher intensity.  These numbers I can see when I have a run of jobs for which it did not find a solution (don't use -q).  I am a bit more likely to find a solution meeting the difficulty requirement of the job at the higher intensity but the slower processing time lowers throughput.  This made me scratch my head more than a little bit...why the higher solve rate at the higher intensity?  But I've watched the numbers stream past for hours.  Tried -i 31.5 and it yields same results as 31.9 for throughput.

The intensity control how many hashes you compute in the gpu before work is restarted..  
In version #9 only one solution is kept and sendt to the pool. In version #7 and #8 2 solutions are sendt if found.¨

I=25

2^25= 33554432

Up to 262 144 Hashes are computed on the gpu before work is restarted.

i=31

2^26= 2147483648

Up to 16 777 216 Hashes are computed on the gpu before the work is restarted.

On low difficulty the probability of finding 2 solutions is high (since we are checking up to 16 MHASH in one go), but ccminer #9 is only sending one solution to the pool.

On low difficulties and  high intensity version #9 will only send 50% of the solutions found. On high difficulty and high intensity version #9 will send up to 100% of the solutions found, but If the gpu is too fast, the miner might be faster than the pool, and work will be lost.

I use default intensity of 29 for compute 5.0 and 30 for compute 5.2 and up

[sp_]

Hey, does anyone have a Windows compiled version of CCminer optimized for 960 and 950 cards for Lyra2re2 algo.
The current version, I have is producing exactly the same has output as for those cards as my 750ti's?
10.1mhs 2x 750ti's
950 and 960 together are 10.1mhs

Wich ccminer are you using. My private (0.1BTC will give a boost of 15-20% on the gtx 950 and gtx 960 compared to the original DJM34 release)

[Ayers]

PASCAL GPU RELEASED

http://www.tomshardware.com/news/nvidia-pascal-tesla-p100-gpu,31557.html#xtor=EPR-8886

I'm surprised they released tesla first, I would expect them to beta test with the consumer line.
Maybe they are confident enough in the quality.

same ...

though nvidia do thngs in the most unorthodox ways ...

cant wait to see how cuda 8 performs with it ...

#crysx

If I understood properly: they will hit the market NEXT YEAR??

last news were talking about this summer in july, for bulk access for the customers, but don't dream too much there will be at best a 2x boost, and i don't know at what price

[crysx]

Offtopic:  For the 980TI I actually get better throughput using -i 31 than -i 31.9 (as reported poolside).  The typical job time is about .8 seconds vs 1.1-1.2 for the higher intensity.  These numbers I can see when I have a run of jobs for which it did not find a solution (don't use -q).  I am a bit more likely to find a solution meeting the difficulty requirement of the job at the higher intensity but the slower processing time lowers throughput.  This made me scratch my head more than a little bit...why the higher solve rate at the higher intensity?  But I've watched the numbers stream past for hours.  Tried -i 31.5 and it yields same results as 31.9 for throughput.

The intensity control how many hashes you compute in the gpu before work is restarted..  
In version #9 only one solution is kept and sendt to the pool. In version #7 and #8 2 solutions are sendt if found.¨

I=25

2^25= 33554432

Up to 262 144 Hashes are computed on the gpu before work is restarted.

i=31

2^26= 2147483648

Up to 16 777 216 Hashes are computed on the gpu before the work is restarted.

On low difficulty the probability of finding 2 solutions is high (since we are checking up to 16 MHASH in one go), but ccminer #9 is only sending one solution to the pool.

On low difficulties and  high intensity version #9 will only send 50% of the solutions found. On high difficulty and high intensity version #9 will send up to 100% of the solutions found, but If the gpu is too fast, the miner might be faster than the pool, and work will be lost.

I use default intensity of 29 for compute 5.0 and 30 for compute 5.2 and up

#9??? ...

sp - why are you not sending ANYTHING my way now? ...

the emails have been fixed - and you can send to the email you have on your list ...

just because i have a newbie account doesnt mean crap - especially when bct have done NOTHING in helping me with getting the chrysophylax account back ...

if you dont trust newbie accounts - fine ... send the updates to the email address you have always sent it to - and i will get it ... im NOT a fake bullshit account that is here to play ...

what gives mate? ...

#crysx

[Velgelm]

Now tested the latest miner from SP Decred#9 and Alexis78.
All GPU overclocked to 1600Mhz
Results Decred#9



And Alexis78 Miner results :

In ideal conditions so you have :
970 = +10Kh
960 = +13Kh
The difference is very small. Currently buying the paid miner a dubious idea.
Waiting for further improvements of the code of the miner from SP, to achieve a significant gap.

[sp_]

#9??? ...
sp - why are you not sending ANYTHING my way now? ...

I am sending to the crysx email adress adress.

[crysx]

#9??? ...
sp - why are you not sending ANYTHING my way now? ...

I am sending to the crysx email adress adress.

tanx sp ...

ill let you know when i get it ...

ill be off to bed soon - as ive had a HUGE day today ... but if you are sending it now - i should get it soon after you send ...

im getting very upset with bct and their lack of care for the people in my situation and profile lockouts and forum bans ... i got hit with both - and STILL nothing from them ...

pathetic really ...

#crysx

[sp_]

tanx sp ...
ill let you know when i get it ...

I have already sendt it yesterday. You are on the mailing list.

[sp_]

The difference is very small. Currently buying the paid miner a dubious idea.
Waiting for further improvements of the code of the miner from SP, to achieve a significant gap.

But my buyers last month have got a nice profit. The only difference is that Alexei used a couple of months to match my speed of the privateminer, and opensourced his work (11-april-2016). What Can I do?
I added a few MHASH'es to the donators..(12-april-2016)

[crysx]

tanx sp ...
ill let you know when i get it ...

I have already sendt it yesterday. You are on the mailing list.

it was last night that the email was fixed - i was on the fone for an hour to get it running again with the isp ... they had 'fixed' it prior - but i still had issues ... should be all good now ...

so i probably didnt get it ... could you send again please? ...

is this both vanilla and decred - or just decred? ...

the last i got was #5 on the 25032016 at 2027 ...

#crysx

[thevictimofuktyranny]

Hey, does anyone have a Windows compiled version of CCminer optimized for 960 and 950 cards for Lyra2re2 algo.
The current version, I have is producing exactly the same has output as for those cards as my 750ti's?
10.1mhs 2x 750ti's
950 and 960 together are 10.1mhs

Wich ccminer are you using. My private (0.1BTC will give a boost of 15-20% on the gtx 950 and gtx 960 compared to the original DJM34 release)

The #5 only does 11mhs on both cards for lyra2re2, which is only 10% better than the 750TI's and Vertcoin's own free miner does exactly the same hash output.

On other algos, the 960 and 950 are doing 35%-46% better, so a Lyra2re2 does not appear to be particular well optimized for this algo.

[Velgelm]

You need to make better, faster and more efficient!
You did a good job, many bought your miner, and myself included, now have a free alternative to your paid miner, so your task is to improve your miner to bought it and attract new buyers. New buyers and have already bought miner interested in the CURRENT position and not what it was a week or a month ago, an important current speed and results for paid and free miner. Or am I wrong ?

[sp_]

You need to make better, faster and more efficient!
You did a good job, many bought your miner, and myself included, now have a free alternative to your paid miner, so your task is to improve your miner to bought it and attract new buyers. New buyers and have already bought miner interested in the CURRENT position and not what it was a week or a month ago, an important current velocities and results for paid and free miner. Or am I wrong ?

And I did in #9, my Decred is the fastest isn't it.. He released his miner 11th of april, and I released an optimized version on the 12th of april.

[sp_]

The #5 only does 11mhs on both cards for lyra2re2, which is only 10% better than the 750TI's and Vertcoin's own free miner does exactly the same hash output.
On other algos, the 960 and 950 are doing 35%-46% better, so a Lyra2re2 does not appear to be particular well optimized for this algo.

It's because the code from djm34 is already pretty good.
And the cards you are using the 950 and the 960 are not the target cards. I do have a few, but I mainly do mod's for the 750ti.

[Velgelm]

Yes, the result is, but at the moment the gap is very small, but still progress is being made

[sp_]

The problem is that the next percents are very difficult. Perhaps not possible. Much bether to do a 6% improvement in the x11 algo, or other profitable algos.
Decred @ $2.4 was very profitable. Decred @ 1.4 and a higher diff is not. FPGA's are taking over now.. 1 pass simple algos can't be mined on gpu's for long.