CCminer(SP-MOD) Modded GPU kernels.

[bensam1231]

Did you have the minerhash vs poolhash bug before? Did it fix it?

The bug is only present on some computers with slow cpu'.'s. On one of my rigs. 6x 750ti celeron windows 8 release #5 used 50% cpu. Release #6 is using 25% cpu. If I reduce the the intensity, the cpu usage is down.

It's not the ccminer process that use CPU but the system. (driver?)

All my windows 7 rigs are fine even with the same CPU.

It's not a CPU utilization bug. As I mentioned I have quad cores and they're running at about 1-5% CPU utilization. I just decommissioned my last A4-4000. I know what a CPU bottleneck looks like, I bought quads over duel specifically to prevent something like this from happening.

The only other possibility is I'm running on W8.1 x64. I haven't had problems up till this point on using W8 over W7, however, sometimes I get better hashrate because of it. I still don't know why this would be the problem as the miner is showing a higher hashrate... which should include a CPU slowdown if my CPU was limiting the GPUs.

[1714763026]

1714763026

[1714763026]

1714763026

[sp_]

The only other possibility is I'm running on W8.1 x64. I haven't had problems up till this point on using W8 over W7, however, sometimes I get better hashrate because of it. I still don't know why this would be the problem as the miner is showing a higher hashrate... which should include a CPU slowdown if my CPU was limiting the GPUs.

You reported that the original 1.7.4 had the same problem. It could be the stratum code that is not working correct. You need to mine on a pool with difficulty>1

I will run a kernal test to make sure my fork isn't missing any hash. I will run 1.7.4 kernal, and then my modded kernal and check if they find the same solutions..

[scryptr]

The only other possibility is I'm running on W8.1 x64. I haven't had problems up till this point on using W8 over W7, however, sometimes I get better hashrate because of it. I still don't know why this would be the problem as the miner is showing a higher hashrate... which should include a CPU slowdown if my CPU was limiting the GPUs.

You reported that the original 1.7.4 had the same problem. It could be the stratum code that is not working correct. You need to mine on a pool with difficulty>1

I will run a kernal test to make sure my fork isn't missing any hash. I will run 1.7.4 kernal, and then my modded kernal and check if they find the same solutions..

POOL STATS--

I had a problem with the original tpruvot v1.7.4 stratum connection also.  Currently, the best connection I get at SuprNova is on port 9115 9112,  an "http" connection with a difficullty of 64 8.  The port uses Long-Polling, and the poolside hashrate stats fluctuate both over and under the actual hashrate of ~540MH/s per 750ti card  I tried the "stratum+tcp" port of 2252 (see cryptomining-blog), but it was not as good.  My rig posts 3240MH/s for 6 cards at the local console, and reads from 2K to 4K MH/s poolside while using the high-difficulty Long-Polling port of 9115.  I mine about 5 DCR per day at this rate, but the poolside stats fluctuate.       --scryptr

EDIT:  There is a DECRED proxy available at https://github.com/bitbandi/decred-proxy , it may solve things.  It is a true proxy, unlike the proxy stratum  implementation in CCminer v1.7.4 .  I use a proxy for all my ETHminer rigs.      --scryptr

EDIT:  I fell back on difficulty some and started using the Long-Polling port of 9112, with a difficulty of 8.  My small rig was not able to submit shares often enough at a difficulty of 64 (port 9115).       --scryptr

[tbearhere]

This is your competition Suit lamers...

https://www.youtube.com/watch?v=MMXcjhQ5Diw

What is that?

[Bombadil]

Hey I'm running ccMiner-(spmod) 1.5.78 which says it is optimised for Maxwell based cards.

Any reason why my NVIDIA Quadro M6000 12GB is only getting about 11.7MH/s @ x11 ?

This is a pretty high end card and I hear ATI high ends go waaaay better so is it just the fact that it's an NVIDIA card or can I push this thing faster? No overclocking and running default settings on ccMiner. I tried tweaking different things like work sizes etc but no real gains.

Any one else mined with such a card? I didn't buy it for mining, just utilising it while it's sitting idle in my box so not really crucial just thought I'd see if I can squeeze more out of it

Tried playing around with the -i setting too? Default is -i 20 IIRC, but you can probably get a lot higher.
Anyways, Quadros are differently designed than consumer-based cards and ccminer might probably need some code tweaks for such cards. Maybe if you donate to a dev, they might want to help you ^^

Also, X11 isn't worth it anymore There are more profitable algos out there to mine.

[sp_]

Hey I'm running ccMiner-(spmod) 1.5.78 which says it is optimised for Maxwell based cards.
Any reason why my NVIDIA Quadro M6000 12GB is only getting about 11.7MH/s @ x11 ?

Use version 1.5.74 and compile with cuda 6.5. Or buy my private kernal. (0.1BTC) The 980ti (6 GB) is doing 13MHASH x11, and 28MHASH quark ( @1280Mhz core) X11 has not been protifable on Nvidia cards since wolf0's binaries was stolen and spread. I only optimize the most profitable algos.

[sp_]

This is your competition Suit lamers...
https://www.youtube.com/watch?v=MMXcjhQ5Diw

What is that?

This was made by the best hackers in the world. Back in 1988. on the commodore 64.

[bensam1231]

The only other possibility is I'm running on W8.1 x64. I haven't had problems up till this point on using W8 over W7, however, sometimes I get better hashrate because of it. I still don't know why this would be the problem as the miner is showing a higher hashrate... which should include a CPU slowdown if my CPU was limiting the GPUs.

You reported that the original 1.7.4 had the same problem. It could be the stratum code that is not working correct. You need to mine on a pool with difficulty>1

I will run a kernal test to make sure my fork isn't missing any hash. I will run 1.7.4 kernal, and then my modded kernal and check if they find the same solutions..

If it were just a me issue other people wouldn't be getting it... Still other people wouldn't be getting correct speeds at pool.

[Ayers]

which intensy for 970 and 960 ?

what algo? usually you don't need intensity setting for etheruem or decred

[Velgelm]

decred - which intensy without freezing?

[AzzAz]

decred - which intensy without freezing?

Try 30, 29, etc...

[sp_]

I did a regression test comparing my mod #4 and the original kernal. Some of the nounces are not found resulting in a lower hashrate on the pool. Bensam was right again..New fixed version will be sendt out today.

[Velgelm]

ok i ll be waiting

[scryptr]

I did a regression test comparing my mod #4 and the original kernal. Some of the nounces are not found resulting in a lower hashrate on the pool. Bensam was right again..New fixed version will be sendt out today.

SUPRNOVA--

I have been mining with LongPoll on SuprNova with beta #6.  I cannot get the stratum implementation to work well on port 2252, but the original tpruvot code was the same.  It under-reports hash poolside.

The LongPoll is simply less efficient.  I get readings poolside that are both under- and over- my local console reported hashrate for DCR.  Over 24 hours, it appears more under-reported hash poolside than otherwise.

Proxies are supposed to improve efficiency in hash results, maybe by 10-20%.  The DECRED proxy I listed above does not compile on Linux easily, it has dependencies that I don't understand.

I am earning about 5 DCR a day with beta #6, a local hashrate of 540MH/s per 750ti FTW card, or 3240MH/s for my 6 card rig.  I have only been miniing on SuprNova.       --scryptr

[duboisi]

SUPRNOVA--

I have been mining with LongPoll on SuprNova with beta #6.  I cannot get the stratum implementation to work well on port 2252, but the original tpruvot code was the same.  It under-reports hash poolside.

The LongPoll is simply less efficient.  I get readings poolside that are both under- and over- my local console reported hashrate for DCR.  Over 24 hours, it appears more under-reported hash poolside than otherwise.

Proxies are supposed to improve efficiency in hash results, maybe by 10-20%.  The DECRED proxy I listed above does not compile on Linux easily, it has dependencies that I don't understand.

I am earning about 5 DCR a day with beta #6, a local hashrate of 540MH/s per 750ti FTW card, or 3240MH/s for my 6 card rig.  I have only been miniing on SuprNova.       --scryptr
[/quote]

Use the Decred Mining Calculator to check your earning. Previously (Tpruvot 1.7.3) I mined DCR at Suprnova Stratum port, although the pool stat is way out, the earning is more than what the calculator indicated. So i think is the pool stat error.

[sp_]

There was a bug in my testcode. #6 seems to give correct results. At least on yiimp.ccminer.org

[sp_]

do the right thing SP and look after your donators and fix and update the decred miner.  

Yes. I will send out a fixed final version.

Decred sp-mod #7 sendt to the 0.1BTC donators. Check your inboxes.
decred+vanilla sendt to the 0.2BTC donators.

[bathrobehero]

do the right thing SP and look after your donators and fix and update the decred miner.  

Yes. I will send out a fixed final version.

Decred sp-mod #7 sendt to the 0.1BTC donators. Check your inboxes.
decred+vanilla sendt to the 0.2BTC donators.

Thank you.

[bensam1231]

I did a regression test comparing my mod #4 and the original kernal. Some of the nounces are not found resulting in a lower hashrate on the pool. Bensam was right again..New fixed version will be sendt out today.

...and it looks like it is fixed. Coinmine seems like it's getting ddos'd, doesn't work properly. MNpool was working properly, but then I get 'unknown work' and lose shares. Suprnova seems to be giving me full speed right now, not sure how long that's going to last. Using diff 8, 4, 8 respectively.

Not sure what the 'unknown work' submit error is, I've gotten it a few times though over the months trying to use decred. It'll usually work fine and then crap out a bunch of those all in a row.

This is a confirmed bug in cgminer on windows. It affects all pools. This flags seem to help:

--no-submit-stale --scan-time 2 --queue 0

Unknown-work means you are sending work back to the pool that was never requested - thats why it gets rejected, on windows cgminer seems to wrong calculation of the hash sometimes which results in this behavior.

It will probably get fixed in one of the future releases, or simply try the flags above.

https://forum.decred.org/threads/unknown-work-rejected-shares-pool-mining.511/

Please fix this.

Please at least try your releases on some of the other decred pools instead of just yiimp. The yaamp clones seem weird for some reasons, try longpoll over stratum instead as well.

[coinut]

do the right thing SP and look after your donators and fix and update the decred miner.  

Yes. I will send out a fixed final version.

Decred sp-mod #7 sendt to the 0.1BTC donators. Check your inboxes.
decred+vanilla sendt to the 0.2BTC donators.

will test thanks SP