ethminer-0.9.41-genoil-1.1

[Genoil]

My first custom Kernel for AMD HD7850 Pitcarn :

Code:

  m  23:30:25|ethminer-1.1.8  Mining on PoWhash #b558a53c : 16.74MH/s [A0+0:R0+0:F0]
  m  23:30:26|ethminer-1.1.8  Mining on PoWhash #b558a53c : 16.74MH/s [A0+0:R0+0:F0]
  m  23:30:26|ethminer-1.1.8  Mining on PoWhash #b558a53c : 16.74MH/s [A0+0:R0+0:F0]

YEAY

Nice! How much increase from my kernel?

I wanna know if it gets shares, first

Easy check add -Z after cmd line for quick verification against low diff

[1715020007]

1715020007

[1715020007]

1715020007

[1715020007]

1715020007

[1715020007]

1715020007

[joaogti36]

no 1.1.8 for windows?|!

[Genoil]

no 1.1.8 for windows?|!

It's a fork by LeChuckDe. There will be no 1.1.8 from me. 1.2 maybe late August with fast loading binaries and +1% on GCN1.0 cards.

[restless]

Is there repo for his clone?
Any reason why you dont wanna build 118 for Win? It should be just recompile, no?

[scryptr]

My first custom Kernel for AMD HD7850 Pitcarn :

Code:

  m  23:30:25|ethminer-1.1.8  Mining on PoWhash #b558a53c : 16.74MH/s [A0+0:R0+0:F0]
  m  23:30:26|ethminer-1.1.8  Mining on PoWhash #b558a53c : 16.74MH/s [A0+0:R0+0:F0]
  m  23:30:26|ethminer-1.1.8  Mining on PoWhash #b558a53c : 16.74MH/s [A0+0:R0+0:F0]

YEAY

Nice! How much increase from my kernel?

R7 265 (PITCAIRN)--

I purchased 4 R7 265 cards, 2GB memory, shortly after they were released.  At the time, they were billed as "750ti Beaters",  and I put them in a rig for mining scrypt.  They were a re-issue of the HD 7850, with tweaks.  They only mined scrypt at 1/2 the rate of a 750ti. I boxed them up and put them away, rebuilding the rig with 750ti cards.

Well, I am re-assembling a new rig with them.  With your miner, the R7 265 cards mine ETH at 10-11MH/s each (2X 750ti rate).  However, I had been expecting a hash rate around 14-15MH/s. 

I'd like to know if LeChuckDe is mining ETH, or EXP, or another Dagger-Hashimoto clone.  I'd also like to try his version of cpp-ethereum.  Is his kernel a public release?       --scryptr

[Genoil]

Is there repo for his clone?
Any reason why you dont wanna build 118 for Win? It should be just recompile, no?

Because I don't have it. I have 1.1.7 publicly available and 1.2.0 in the works. If LeChuckDE decides to release his changes, it would be ethminer-0.9.41-lechuckde-1.1.8

[giagge]

Once again low hashrate with Nvidia and windows 10 ? .

[Naimon]

My first custom Kernel for AMD HD7850 Pitcarn :

Code:

  m  23:30:25|ethminer-1.1.8  Mining on PoWhash #b558a53c : 16.74MH/s [A0+0:R0+0:F0]
  m  23:30:26|ethminer-1.1.8  Mining on PoWhash #b558a53c : 16.74MH/s [A0+0:R0+0:F0]
  m  23:30:26|ethminer-1.1.8  Mining on PoWhash #b558a53c : 16.74MH/s [A0+0:R0+0:F0]

YEAY

Nice! How much increase from my kernel?

My HD7850 (OC'd to 1050/1500) running your v.1.1.7 gets 12MH/s in Win10 with 15.12 drivers.  16.7MH/s is HUGE.  Seriously, me want!

[LeChuckDE]

Code:

  
  m  09:29:39|ethminer-1.1.8  Mining on PoWhash #6b5a9b6c : 15.22MH/s [A303+0:R0+0:F0]
  m  09:29:39|ethminer-1.1.8  Mining on PoWhash #6b5a9b6c : 15.22MH/s [A303+0:R0+0:F0]
  m  09:29:40|ethminer-1.1.8  Mining on PoWhash #6b5a9b6c : 15.06MH/s [A303+0:R0+0:F0]
  m  09:29:40|ethminer-1.1.8  Mining on PoWhash #6b5a9b6c : 15.22MH/s [A303+0:R0+0:F0]
  m  09:29:41|ethminer-1.1.8  Mining on PoWhash #6b5a9b6c : 15.22MH/s [A303+0:R0+0:F0]
  m  09:29:41|ethminer-1.1.8  Mining on PoWhash #6b5a9b6c : 15.22MH/s [A303+0:R0+0:F0]
  m  09:29:42|ethminer-1.1.8  Mining on PoWhash #6b5a9b6c : 15.22MH/s [A303+0:R0+0:F0]
  m  09:29:42|ethminer-1.1.8  Mining on PoWhash #6b5a9b6c : 15.22MH/s [A303+0:R0+0:F0]
  m  09:29:43|ethminer-1.1.8  Mining on PoWhash #6b5a9b6c : 15.22MH/s [A303+0:R0+0:F0]
  m  09:29:44|ethminer-1.1.8  Mining on PoWhash #6b5a9b6c : 15.22MH/s [A303+0:R0+0:F0]
  m  09:29:44|ethminer-1.1.8  Mining on PoWhash #6b5a9b6c : 15.22MH/s [A303+0:R0+0:F0]

yes it is hashing ! but drops down to stable 15.2 MH/s without OC, 16,7 MH/s was with a OC of 10 % but with this tempratures Outside it is to risky for my to use the OC Settings.
Hint: "Was modifications with the worksize and the addaped optimization routines for pitcarn from Wolfs optimized public Kernels for the Algoprocessing!"
This is mining ETH @ nanopool

I am not CUDA or OpenCL Programmer, so that the code isn't clean
The Version is not stable enough to publish it. Genoils Version get me with same setup without OC 12 MH/s and with OC 10% 12,9 MH/s

will only support Linux in case of publishing, Windows WDDM sucks, resets everytime the displaydriver after DAG Load is finished at Linux runs like a charm

This was only a training to get a little bit looking inside OpenCL programming, only a training-project to verify that the learned things are working

[Genoil]

What is the benchmark speed with the 1GB epoch 0 DAG? If that is also 27% above my kernel, you would be hashing at a higher speed than the bandwidth of the card permits

[Q_R_V]

I'm willing to bet money 1/4th of the results are actually HW errors (miner not showing them, though.)

How to verify if they are HW errors? Counting shares on the pool? Comparing profits from this miner with non modified one?

[LeChuckDE]

I'm willing to bet money 1/4th of the results are actually HW errors (miner not showing them, though.)

How to verify if they are HW errors? Counting shares on the pool? Comparing profits from this miner with non modified one?

Hi Q_R_V, no that are not HW errors ...
Mining at open-ethereum-pool with following stats :

ID	Hashrate (rough, short average)	Hashrate (accurate, long average)	Last Share
HD7850	14.44 MH	15.27 MH	10 seconds ago

no rejected shares

[Q_R_V]

Hi Q_R_V, no that are not HW errors ...
Mining at open-ethereum-pool with following stats :

ID	Hashrate (rough, short average)	Hashrate (accurate, long average)	Last Share
HD7850	14.44 MH	15.27 MH	10 seconds ago

no rejected shares

Hi LeChuckDE, when you say "that are no HW errors" do you have a some method of searching for HW errors? When you push card to the limits and above, are HW errors somehow noticeable? Because that hashrate is looking awesome

[restless]

Eth pools reject NO shares, unless stale. Dupes they consider too compute-heavy to return negative responses for.

You mean that from miner's POV the hashrate will be higher, but from pool's POV, this will be kind of cheating?

[LeChuckDE]

Eth pools reject NO shares, unless stale. Dupes they consider too compute-heavy to return negative responses for.

You mean that from miner's POV the hashrate will be higher, but from pool's POV, this will be kind of cheating?

From pool's POV, most will silently drop dupes. The ethminer tends to handle HW errors by not submitting at all - of course, this isn't important enough to tell you about it.

is there a way to measure HW errors ?

[Naimon]

Eth pools reject NO shares, unless stale. Dupes they consider too compute-heavy to return negative responses for.

You mean that from miner's POV the hashrate will be higher, but from pool's POV, this will be kind of cheating?

From pool's POV, most will silently drop dupes. The ethminer tends to handle HW errors by not submitting at all - of course, this isn't important enough to tell you about it.

is there a way to measure HW errors ?

I use SGMiner - I built in regenhash, a CPU implementation of Ethash used to verify before submission. You can also use abort() in the area where it happens in Genoil's ethminer. You could let me take a look at it, if you trust me.

Uh, yeah - dude, do that.

[adaseb]

I think instead of trying to making ETH mining faster its better to just make a miner that lets you dual mine many algorithms like what Claymore did.

Since SIA and ETH generally are making exacts profits with certain cards like Tahiti and Hawaii. You make 40% more profits by dual mining.

So if a dual miner could be made for Lyra, Quark, Qubit it would be nice however I have no idea how those algos differ from DCR and SIA.

[Genoil]

I think instead of trying to making ETH mining faster its better to just make a miner that lets you dual mine many algorithms like what Claymore did.

Since SIA and ETH generally are making exacts profits with certain cards like Tahiti and Hawaii. You make 40% more profits by dual mining.

So if a dual miner could be made for Lyra, Quark, Qubit it would be nice however I have no idea how those algos differ from DCR and SIA.

My 380 disagrees with you (NSFW): https://ottrbutt.com/miner/eth-380-23.4MHs.png

Ethminer has always had CPU validation, there is no reason to assume it suddenly isn't working correctly. Poolside hashrate also looks good, so it looks like we have a winner! If I were confident about this I'd sell it. And get yourself some Tahiti and Hawaii to see if you can repeat the magic.

[LeChuckDE]

By the way dualmining works also !
use SGMiner with Decred as second Coin ins seperated session. -I 16  let drop the Hashrate on ETH from constant 15,27 MH/s to 15,22 MH/s and every 8 - 9 report a drop down to 12,7 MH/s,
but i think it is a timing problem with --farm-recheck 400 ...
next hint ...
Local-Worksize 512 and global-worksize 16384 as std. and modify the checks in oclkernel.
Put out the not needed if statments, this brings round about 2-4 % then look at the needed algos,
for all there are optimized ones in the net as opensource....
bring it all together and you will get the hashrates ....

My code is to dirty to publish it.
Will get a try at weekend to implement the same HW-Failure based output from SGminer to the status console log but not shure if it possible for daggerhashimoto.

learned many things in this time, my reason was to get intouch to port from cuda to opencl and backwards, then i hit this as a surprise,
like the guy who has developed the PostIt's Glue,
he would like to build a super glue and get a nearly non sticky glue ....

It wasn't my goal to improve hashrates at DaggerHashimoto. it was an accident.

[Genoil]

By the way dualmining works also !
use SGMiner with Decred as second Coin ins seperated session. -I 16  let drop the Hashrate on ETH from constant 15,27 MH/s to 15,22 MH/s and every 8 - 9 report a drop down to 12,7 MH/s,
but i think it is a timing problem with --farm-recheck 400 ...
next hint ...
Local-Worksize 512 and global-worksize 16384 as std. and modify the checks in oclkernel.
Put out the not needed if statments, this brings round about 2-4 % then look at the needed algos,
for all there are optimized ones in the net as opensource....
bring it all together and you will get the hashrates ....

My code is to dirty to publish it.
Will get a try at weekend to implement the same HW-Failure based output from SGminer to the status console log but not shure if it possible for daggerhashimoto.

learned many things in this time, my reason was to get intouch to port from cuda to opencl and backwards, then i hit this as a surprise,
like the guy who has developed the PostIt's Glue,
he would like to build a super glue and get a nearly non sticky glue ....

It wasn't my goal to improve hashrates at DaggerHashimoto. it was an accident.

You obviously got a talent. But you need to work on your marketing skills .