3% faster mining with phoenix+phatk for everyone

[bitless]

I just tried this and got >3% improvement in mining speed. On my 6870, I was getting 299 MHash/sec, and now I'm getting 308 or so. The change is simple enough for anyone to do it - you don't need to be a programmer to use it.


EDIT - Thank you! This has been added to git (not by me), so now everyone can use this speed-up without editing any files. https://github.com/m0mchil/poclbm/commit/3d43c4e4c2c476cf367822868202f02d168d6f3b Also, please read the comments here http://www.reddit.com/r/Bitcoin/comments/i9yba/2_3_increase_in_mining_per_gpu/; the person who added it to git knows more than I do, you should listen to him about SDK versions etc. Thanks for all the donations too!

EDIT2 - Thanks to the awesomeness of reddit and these forums, I now know that this change has been applied to Diablo miner as well. https://github.com/Diablo-D3/DiabloMiner/commit/ae16718dfa538aeef31f34e19542f3c18501efe5 I guess we can stop this thread now, and everyone can enjoy the speedup without editing any files. Thank you for trying this out.


You can go to phatk's kernel.cl file (don't worry, it just sits there in the open, no need to recompile anything), find this line
  #define Ma(x, y, z) amd_bytealign((y), (x | z), (z & x))
and change it to this line
  #define Ma(x, y, z) amd_bytealign( (z^x), (y), (x) )
Once you've done it, restart the miner.

Technically, this is 1 less instruction for the Maj function in the hash, which is called ~128 times for each nonce value, so we get +3% to mining speed. This will ONLY WORK if you're running with BFI_INT. I'm using phoenix with phatk kernel on Ubuntu, so YMMV, but I see no reason for this to not work with other setups. As always, do play around with aggression and other settings after you've applied the change. Deepbit seems to be accepting my shares generated this way, but it comes AS IS, without any warranty whatsoever - if it doesn't work for you, or has been posted already, please don't blame me

If this helps you mine faster, please share your MHash/sec results, before and after. You can also donate to 15igh5HkCXwvvan4aiPYSYZwJZbHxGBYwB . I hear people are getting 50 BTC for things like this, and it would be nice to get some.

If you want to verify the correctness of the change, here's the truth table for the new Ma() function

x y z   Ma
0 0 0   0
1 0 0   0
0 1 0   0
1 1 0   1
0 0 1   0
1 0 1   1
0 1 1   1
1 1 1   1

and here's C code that was used to generate it

#include <stdio.h>
int main()
{
#define CH(src0,src1,src2)   (  ((src1)&(src0)) |  ((src2)&~(src0))  )
   unsigned int x,y,z;
   for(z=0;z<=1;z++) for(y=0;y<=1;y++) for(x=0;x<=1;x++)
   {
      printf("%d %d %d   %d\n",x,y,z, CH(z^x,y,x) );
   }
  return 0;
}

Finally, I am newbie, so it would be nice if someone reposted this to the proper Bitcoin mining thread/board and/or checked-in this change into the source control.

Thanks and have fun!

[1714577207]

1714577207

[1714577207]

1714577207

[burger]

Very nice!

Thanx for sharing.

[bitless]

Very nice!

Thanx for sharing.

You're welcome. Did it help you? I wanna know because nobody but me has tried it so far... so I wanna know if it works for other people

[willphase]

Worked for me - boosted around 2%.  Very impressive.  Where can I send donation?

Will

[bitless]

Worked for me - boosted around 2%.  Very impressive.  Where can I send donation?

Will

Yay! Donate to 15igh5HkCXwvvan4aiPYSYZwJZbHxGBYwB if you feel like it. Your BTCs will (or may) help develop an even faster miner.

[willphase]

Worked for me - boosted around 2%.  Very impressive.  Where can I send donation?

Will

Yay! Donate to 15igh5HkCXwvvan4aiPYSYZwJZbHxGBYwB if you feel like it. Your BTCs will (or may) help develop an even faster miner.

Exact results, on my 2x5870 (each one overclocked slightly differently due to temperature profiles)

1st - 422.13 -> 430.25 (1.9% increase)
2nd - 411.80 -> 422.13 (2.5% increase)

so around 2% increase.

Will

[bitless]

Worked for me - boosted around 2%.  Very impressive.  Where can I send donation?

Will

Yay! Donate to 15igh5HkCXwvvan4aiPYSYZwJZbHxGBYwB if you feel like it. Your BTCs will (or may) help develop an even faster miner.

Exact results, on my 2x5870 (each one overclocked slightly differently due to temperature profiles)

1st - 422.13 -> 430.25 (1.9% increase)
2nd - 411.80 -> 422.13 (2.5% increase)

so around 2% increase.

Will

Excellent, thanks for the info... now it is back to the drawing board to me, I'll poke around and see if there's any other easy changes that can speed things up - so far everything I see will require a massive rewrite of the miner but ok, we'll see

[bcforum]

FYI, You can make the identical change to phoenix+poclbm with the same 2% increase in speed.

[bitless]

FYI, You can make the identical change to phoenix+poclbm with the same 2% increase in speed.

Cool, thanks for trying it out! I figured the same logic of using BFI_INT to build the MAJ function would apply to other miners and kernels as well, I just haven't tried it, so thanks for trying it out

[BiggieJohn]

Thank you

donation will be coming you way

[hart]

4.53% increase here. Thanks!

[USSJoin]

188.2 to 192.1 MHash/sec, so almost 2.1% increase-- thanks!

[mike678]

Went from ~241 to ~248 thanks.

[xjosx]

FYI, You can make the identical change to phoenix+poclbm with the same 2% increase in speed.

Not getting those speed increases w/poclbm-gui running an opencl miner. Still around 686 Mhash/s for me.

[coinmonger]

Thanks, Went from 288.8Mhash/sec to 295.96, 2.48% increase on all 3 5830's.

[huanix]

Thanks bitless - i went from 296 to 303 on my 5830's. Sent you a nickel and encouraged others on reddit to do the same!

[sabe]

303 -> 309 for each 5830, 2% as advertised!

[fitty]

FYI, You can make the identical change to phoenix+poclbm with the same 2% increase in speed.

Not getting those speed increases w/poclbm-gui running an opencl miner. Still around 686 Mhash/s for me.

You need to edit BitcoinMiner.cl in the guiminer folder. It works as advertised. Donation coming. =)

[murfshake]

Thanks, from 292 to 295.

[bitless]

Everybody, thanks for trying this out. I guess it really does work  for other people. Keep donating, I like it

Now that we know it works, it would be great to check this change into the source code repository somewhere so that everyone gets this change by default, without having to manually edit any files. How is this done? Also, how do I post into the non-newbie thread? I feel that this post it doesn't belong here.

[Opsamk]

The more efficient you make miners, the higher the difficulty rises.

[PcChip]

Wow, really awesome find. You're gonna be a minor celebrity around here soon no doubt.

I salute your ingenuity, and am firing up all the VNC sessions to edit my miners kernel files

edit: You added ten megahashes per second to one of my 5850's.  I'm now hitting 400 MH/s on only 1.162 volts.  I love you.


That was a very selfless act giving it away on the forums like that.  Unfortunately well over 60% of the community would have wanted to charge a fee for their optimization.

[eusor]

Nice!

On a 6950 : 349 MH/s -> 366 MH/s (3.75%)

Thanks!

[fitty]

Everybody, thanks for trying this out. I guess it really does work  for other people. Keep donating, I like it

Now that we know it works, it would be great to check this change into the source code repository somewhere so that everyone gets this change by default, without having to manually edit any files. How is this done? Also, how do I post into the non-newbie thread? I feel that this post it doesn't belong here.

I believe after 5 posts you can post in the miner section. Don't worry this will get around quick and it will I'm sure get added within a week or two at most. Great job! I hope your donation made it. =)

[PcChip]

I hope it doesn't get added and we keep it a secret.  Delete this post

 

Edit: To the OP - if you want to see the results of your kernel edit to my 5770:

210 MH/s -> 217 MH/s , check the graph (dip is when I swapped kernels)
One degree cooler and more stable usage

[JDog]

Excellent advice, thanks!!!

[godofal]

got a massive increase on openCL miner, from 159 to 166 max
well, okay, not massive, but in terms of % its above what the others got

and it'l come in handy for my mining rig

thanks for the tip and effort, gonna send you a thank you

[bitclown]

2.6% increase on Phoenix' poclbm kernel. Nice work!

[PcChip]

2.6% increase on Phoenix' poclbm kernel. Nice work!

I wasn't aware that poclbm supported BFI_INT ?

Why not use PhatK with the 2.4 Stream SDK ?

[icolsuineg]

Thanks, donation send, I got a total of about 100 MH/s from all my rigs, works as advertised.

[naukop]

Went from 193.7Mh/s to 195.1Mh/s on my 5770 (900/600) after I edited BitcoinMiner.cl and restarted guiminer.

 running guiminer phoenix.exe VECTORS BFI_INT FASTLOOP=false AGGRESSION=12 -k phatk WORKSIZE=128

Hmm, but now has rate has actually dropped into 186-188 range, not sure what is going on.

[Opsamk]

So bitless, how far are you in Computer Science?

[bitless]

So bitless, how far are you in Computer Science?

Naukop - Could the drop in the hashrate be due to overheating? I've noticed that my card heats up more as a result of the change (and then either it hangs or I have to overclock it less ), so could the same be happening to you?

Opsamk - m.sc., then asm programming for 10 years or so; i'm not looking to be-rich myself on bitcoins, just playing around with these hashes and ATI's hardware

[coblee]

Worked for me. 380 -> 392 (3.16%)
Thanks. This is worth donating for!

[jprockbelly]

Thanks so much for sharing!

I only got about 2% increase, but hey, thats pretty good for 30 seconds work

[msb8r]

From 339.7MH -> 351.7MH (+3.5%) on my 6950@860 using Guiminer (poclbm).

Very nice find, bitless. Donation on it's way 

[Crs]

Works for me too.
I have two 5850s:

bitcoin0
[27/06/2011 12:38:36] Result: 09c2f420 accepted
[330.86 Mhash/sec] [38843 Accepted] [992 Rejected] [RPC]nmc@nmcpc:~$
new bitcoin0
[27/06/2011 12:43:09] Result: 805929c4 accepted
[340.86 Mhash/sec] [4 Accepted] [0 Rejected] [RPC]nmc@nmcpc:~/phoenix$

bitcoin1
[27/06/2011 12:38:45] Result: d593d475 accepted
[27/06/2011 12:38:49] Result: f23c0c10 accepted
[332.14 Mhash/sec] [39162 Accepted] [970 Rejected] [RPC]nmc@nmcpc:~$
new bitcoin1
[27/06/2011 12:43:45] Result: 7b29a04b accepted
[342.27 Mhash/sec] [11 Accepted] [0 Rejected] [RPC]nmc@nmcpc:~/phoenix$

Thanks for sharing this!

[-ck]

Thanks for your code. It's nice to see people get rewarded for their efforts (though it doesn't always happen). Well, I've tried it on my own miner (minerd) and have some very strange findings.

https://forum.bitcoin.org/index.php?topic=21275.0

First of all, minerd supports up to 4 vectors, and when I add this change to my kernel, it actually _slows down_ the 4 vector version. But when I override it to set 2 vectors, it speeds it up. However, once it's sped up, I then get runs of rejected shares. I tried it multiple times with and without and it does appear to be just this change that causes it, so I'm not sure what's going on.

[being]

Thank you! Works great!

[t3h]

Wasn't the claimed improvement with hdminer 5% anyway? I wonder if Phoenix is now faster than hdminer... or was it before anyway?

5770 from ~200 to ~210

[bunnybare]

5770 from ~200 to ~210

Same with me!

Edited BitcoinMiner.cl on poclbm.

[manifold]

works perfectly. Many thanks!!!

[klaaster]

 Very nice hack, gives me 2% to 3% more on HD5850/30.

Great find!  

[forexmasterja]

This works great !!! got an extra 100 MH with all my cards combined. I'm very grateful for this but this is definitely gonna push the difficulty up when this gets more public.

Worth a donation definitely!

6870 295mh => 312mh
5850 350mh => 362mh
6870 411mh => 426mh

Again great find.

[syb3ria]

It's good to know that helps fellow miners, yet i get 16-20 mh\s more with that. http://forum.bitcoin.org/index.php?topic=19051.0. All that is going on Powercolor 5770 @ 990/171 with -f 1 and -f 20 on win 7 x64. Anyway keep up the good work. We all need optimisations

[heroBitcoin]

cool, will try these method

[bitless]

Thanks for your code. It's nice to see people get rewarded for their efforts (though it doesn't always happen). Well, I've tried it on my own miner (minerd) and have some very strange findings.

https://forum.bitcoin.org/index.php?topic=21275.0

First of all, minerd supports up to 4 vectors, and when I add this change to my kernel, it actually _slows down_ the 4 vector version. But when I override it to set 2 vectors, it speeds it up. However, once it's sped up, I then get runs of rejected shares. I tried it multiple times with and without and it does appear to be just this change that causes it, so I'm not sure what's going on.

I honestly do not know what's up with that, I saw ATI asm yesterday for the first time and can't tell you exactly what's wrong yet. All I know is that the truth tables match for the Ma() function with and without my modifications. Yet, here's a couple of ideas -
1 glancing through the doc, radeons are VLIW5 = 4+1, with 4 'normal' pipelines and one transcendental pipe, which can do a restricted set of instructions. I don't know where BFI_INT gets executed, but if it is only in the trans. pipe, then doing too many BFI's can hurt the performance by making that pipe a bottleneck. Check the docs and let us know, if you don't mind.
2 if (z^x) isn't already used in other places in your code, then it may be pushing up the register usage and you're running less threads in parallel. Again, I don't know much about ATI, but it would be the first thing I'd check if we were on nVidia/CUDA.
3 something else altogether...

Not sure. Sorry. If I think of anything else, I'll post it In the meantime, it would also suck if people started getting more rejected shares... hmmm. I don't, it does work for me, but I encourage everyone to check their results (the actual amount of accepted shares that they get).

[bitless]

Thanks for your code. It's nice to see people get rewarded for their efforts (though it doesn't always happen). Well, I've tried it on my own miner (minerd) and have some very strange findings.

https://forum.bitcoin.org/index.php?topic=21275.0

First of all, minerd supports up to 4 vectors, and when I add this change to my kernel, it actually _slows down_ the 4 vector version. But when I override it to set 2 vectors, it speeds it up. However, once it's sped up, I then get runs of rejected shares. I tried it multiple times with and without and it does appear to be just this change that causes it, so I'm not sure what's going on.

Also, a very silly thing here -> I've posted   #define Ma(x, y, z) amd_bytealign( (z^x), (y), (x) ), but I really should have put all expressions in parentheses. So, ((z)^(x)) and not (z^x) etc.... you know

Regardless. You got me worried about rejected shares. I HIGHLY encourage one of us to go through the truth table in my original post and the C code that was used to generate it to make sure there's no errors. That's why I posted it to begin with. It would really suck if it didn't work.

[czz]

Did not notice any increase in the reject ratio. Using 5870, came from 439 to 449 MH/s.
Many thanks.

[bitless]

Heeey! This change is now in git repository! I'd still parenthesize (z^x) as ((z)^(x)), but regardless, thanks m0mchill and everybody!

https://github.com/m0mchil/poclbm/commit/3d43c4e4c2c476cf367822868202f02d168d6f3b

[themike5000]

Boosted me from 319MH/sec on my 6870 to 326MH/sec.  Not bad.

I"m using GuiMiner v2011-06-09 because v2001-06-14 didn't have a BitcoinMiner.cl file.  Anybody know where to find this file in the newest GUIminer release?

[Kauwi]

I HIGHLY encourage one of us to go through the truth table in my original post and the C code that was used to generate it to make sure there's no errors. That's why I posted it to begin with. It would really suck if it didn't work.

You posted the truth-table (and modified the code) for the MA-function, but your C-code is for the CH-function.
But Your thruth-table for MA-function is correct.

[brunoshady]

not too much increase here in 5850... about 4/5 mhash =(

[ananas5]

Thanks! HD6850 270MHash/s -> 279Mhash/s

[bitless]

I HIGHLY encourage one of us to go through the truth table in my original post and the C code that was used to generate it to make sure there's no errors. That's why I posted it to begin with. It would really suck if it didn't work.

You posted the truth-table (and modified the code) for the MA-function, but your C-code is for the CH-function.
But Your thruth-table for MA-function is correct.

Sorry about the confusion. CH = BFI_INT, since BFI does the same thing as the CH function in the hash spec. MA is built using CH/BFI, so there's a define for CH (copied from and docs) and the MA built on top of CH. Thank you for verifying this stuff. 

[Kauwi]

I HIGHLY encourage one of us to go through the truth table in my original post and the C code that was used to generate it to make sure there's no errors. That's why I posted it to begin with. It would really suck if it didn't work.

You posted the truth-table (and modified the code) for the MA-function, but your C-code is for the CH-function.
But Your thruth-table for MA-function is correct.

Sorry about the confusion. CH = BFI_INT, since BFI does the same thing as the CH function in the hash spec. MA is built using CH/BFI, so there's a define for CH (copied from and docs) and the MA built on top of CH. Thank you for verifying this stuff. 

Ah, i see. Didnt look close enough to the printf();
Sorry for that.

[Keninishna]

woot works for me with guiminer. I edited the kernel.cl file in the phatk directory.

[burp]

10MH improvement, thanks! 

[theunforgiven]

Gigabyte 5850 @ 1000/300 - from 390MHash/s to 400MHash/s with the mod applied.

Got a question, which boolean operator is used to sum all the outputs?

((y),(x|z),(z&x))

"Y" , "X or Z", "Z and X"

xyz| (x|z) | (z&x) | y ? (x|z) ? (z&x)  | Ma
000|   0    |    0    |              ?             |  0
001|   1    |    0    |              ?             |  0
010|   0    |    0    |              ?             |  0
011|   1    |    0    |              ?             |  1
------------------------------------------------
100|   1    |    0    |              ¿             |  0
101|   1    |    1    |              ¿             |  1
110|   1    |    0    |              ¿             |  1
111|   1    |    1    |              ¿             |  1

[BombaUcigasa]

I have rewarded you with half of my increased bitcoin revenue margin, well done sir!

[bitless]

Gigabyte 5850 @ 1000/300 - from 390MHash/s to 400MHash/s with the mod applied.
Got a question, which boolean operator is used to sum all the outputs?
((y),(x|z),(z&x))

It is the CH function; given 3 arguments, a,b,c, CH returns B when A==0, and C when A==1. This is the same as (a&b)|(~a&c), which is the way BFI_INT is defined in ATI's docs.

[Hawkix]

Great, went up promised 3%. So I sent you 3% of my daily income as a tip! Thanks!

Its great someone tries to improve on the OpenCL code. Maybe, if we could check the result of OpenCL compilation (VLIW assembler?), we could fine tune a bit more.

[PcChip]

So m0mchill (or somebody?) already updated the code repository, and hasn't sent you a "thank you" PM yet ?

[bitless]

So m0mchill (or somebody?) already updated the code repository, and hasn't sent you a "thank you" PM yet ?

No, but it is me who should be thanking them for writing these awesome miners to begin with!

[heroBitcoin]

yeah, awesome information, thx guy

[theunforgiven]

http://k.min.us/jdQi1Q.jpg

can't do it

[Soak]

WoW

310 to 380 Mhash (+18.42%) ! AWESOME! Thank you!

Code:

phoenix -u http://x:x@x:1234/;askrate=10 -k phatk VECTORS BFI_INT AGGRESSION=7 DEVICE=0
phoenix -u http://x:x@x:1234/;askrate=10 -k phatk VECTORS BFI_INT AGGRESSION=7 DEVICE=1

6970 @ Crossfire

Proof:

Before



After

[Fiyasko]

FYI, You can make the identical change to phoenix+poclbm with the same 2% increase in speed.

Do tell, If theres a simple file to replace that you can upload i'll definitly Tip you
Winx86 GUIminer-poclbm

[Rob P.]

Just tossing my confirmations out there:

Sapphire 5830s:

Pair 1:  264 Mhash/sec -> 271 Mhash/sec (each) at 875 MHz GPU, 900 MHz RAM (2.58% increase)
Pair 2:  273 Mhash/sec -> 281 Mhash/sec (each) at 900 MHz GPU, 600 MHz RAM (2.84% increase)

I'm also tracking Rejected shares.

Before Patch:
Pair 1:  ~2.6% rejects (after 800+ shares)
Pair 2:  ~3.2% rejects (after 800+ shares)

After Patch:
Pair 1:  ~3.0% rejects (after 650+ shares)
Pair 2:  ~4.0% (after 750+ shares)

So, the increase appears to also cause additional rejects that should be discounted from total increase gain, cutting it down a full 1% in my case.

[bitless]

FYI, You can make the identical change to phoenix+poclbm with the same 2% increase in speed.

Do tell, If theres a simple file to replace that you can upload i'll definitly Tip you
Winx86 GUIminer-poclbm

at the office now, no access to the miners, I'll take a look once I get home. In the meantime, search for .cl files and see if you have any Ma (or, even better, amd_bytealign) strings in them and take it from there.

[bitless]

Just tossing my confirmations out there:
...
So, the increase appears to also cause additional rejects that should be discounted from total increase gain, cutting it down a full 1% in my case.

This is really not good. I honestly do not know why this would happen; are you sure it is related to the patch and isn't a result in the general randomness when searching for solutions? Perhaps someone else, more familiar with bitcoin mining than me, can chime in?

[bitless]

can't do it

Yes, you can what you've done is the CH function, not MA. Now you can build MA on top of it. I don't think this belongs to this thread though, so you can pm me and we'll figure it out. Alternatively, you can ask someone around you for help - two heads are better than one.

[bitless]

WoW
310 to 380 Mhash (+18.42%) ! AWESOME! Thank you!

Neat-o-rama! Somehow I still don't believe you (and my eyes)

[PcChip]

WoW
310 to 380 Mhash (+18.42%) ! AWESOME! Thank you!

Neat-o-rama! Somehow I still don't believe you (and my eyes)

I noticed he's using Crossfire on those cards, perhaps the kernel patch somehow makes up for % loss due to using crossfire?

[Soak]

Neat-o-rama! Somehow I still don't believe you (and my eyes)

I do 370 Mhashes with poclbm and the new formula (365 without).
I do 380 Mhashes with phatk and the new formula (310 without).

So, really, +2.63% between poclbm and phatk with the new formula.

[TurdHurdur]

Just tossing my confirmations out there:
...
So, the increase appears to also cause additional rejects that should be discounted from total increase gain, cutting it down a full 1% in my case.

This is really not good. I honestly do not know why this would happen; are you sure it is related to the patch and isn't a result in the general randomness when searching for solutions? Perhaps someone else, more familiar with bitcoin mining than me, can chime in?

I'm not seeing this issue with my miners. They range from 0.12% to 2.07% stale. A variance that I'd say rules out your patch being the cause of any supposed increased stales.

[theunforgiven]

I want to minimize the expression applying Karnaugh. So I need the main expression that is used on the kernel. But i can't figure out with it.

[Hawkix]

Seriously, the Ma() function is soo deeply burried, so that if it was wrong, by the principle of good hashing function like SHA256 is, all the hashes would fail.

So, if it can find some good results, all are, with 99.999999999999% probability correct, too.

[bitless]

Seriously, the Ma() function is soo deeply burried, so that if it was wrong, by the principle of good hashing function like SHA256 is, all the hashes would fail.

So, if it can find some good results, all are, with 99.999999999999% probability correct, too.

More like 2^-128 if not less

[Clipse]

Nice, avg about 2% mhash boost at same clocks. I notice no additional stales, still hovering at 1%

[mrb]

I was wondering when the community would discover this optimization... nice one bitless :-)

For the record, hdminer has implemented this maj() optimization since day 1:

Code:

        # ibit_extract patched to BFI_INT at runtime
        $code .=
        "    ixor $tmp0, $a, $b\n".
        "    ibit_extract $tmp0, $a, $c, $tmp0\n";

Phoenix is probably very close to hdminer's performance now, on HD 69xx.

[Wildvest]

from 402 -> 415 thanks

[Bert]

Sapphire 5870 @ 1000/347 before 442.7 Mhash/sec after 452.1 Mhash/sec overall an increase of 2.1%

[themike5000]

I'm still looking at the same rejection statistics after 1100 shares.  Thanks!

[sanchaz]

5850 890/300: 345 -» 355
5850 825/300: 318 -» 325
5850 900/300: 344 -» 355

donation on its way thx.

[KiLeRrosS]

Thank you

5830 - 314 -> 324 (1025/300)
5830 - 308 -> 316 (1000/300)
5850 - 370 -> 379 (930/300)

I've 57MHps more - almost 3%

I sent small donations.

[AtlasONo]

FYI, You can make the identical change to phoenix+poclbm with the same 2% increase in speed.

Do tell, If theres a simple file to replace that you can upload i'll definitly Tip you
Winx86 GUIminer-poclbm

http://www.mediafire.com/?jth8eicn9iifnlf

Extract contents to the location of your guiminer folder and allow them to replace the existing files.

[Rob P.]

Just tossing my confirmations out there:
...
So, the increase appears to also cause additional rejects that should be discounted from total increase gain, cutting it down a full 1% in my case.

This is really not good. I honestly do not know why this would happen; are you sure it is related to the patch and isn't a result in the general randomness when searching for solutions? Perhaps someone else, more familiar with bitcoin mining than me, can chime in?

Not saying it's the patch, just reporting the results.  That's why I posted the population of shares each measurement was done over.

If my reject rates are higher over a prolonged period, then I may change one of my cards running on the same rig, back to the old kernel values and compare them head to head over a few days.

[Freakin]

412 to 423 on my 5870
290 to 302 on my 5830s

[bitless]

I was wondering when the community would discover this optimization... nice one bitless :-)

For the record, hdminer has implemented this maj() optimization since day 1:

Code:

        # ibit_extract patched to BFI_INT at runtime
        $code .=
        "    ixor $tmp0, $a, $b\n".
        "    ibit_extract $tmp0, $a, $c, $tmp0\n";

Phoenix is probably very close to hdminer's performance now, on HD 69xx.

Dude, awesome  so this accounts for 3% out of your 6.4% improvement, where are the other 3.4% coming from? I don't have 250 btc, in fact I only have ~2 that have been donated to me so far. (Please note - if you decide to share this, post it,don't pm me )

[c_k]

A small donation coming your way for the excellent work

Many thanks!

[PcChip]

I was wondering when the community would discover this optimization... nice one bitless :-)

Makes me wonder how many other private modifications are in use right now

[bitless]

I was wondering when the community would discover this optimization... nice one bitless :-)

Makes me wonder how many other private modifications are in use right now

Also makes me wonder why people don't post them for everyone to see. So, 3% or even 5% speed increase doesn't really matter all that much if you're just a regular guy with a couple of ATI cards, right? Or am I wrong?

PcChip - give me your best advice, should I keep my next optimization to myself or post it?

[PcChip]

Also makes me wonder why people don't post them for everyone to see. So, 3% or even 5% speed increase doesn't really matter all that much if you're just a regular guy with a couple of ATI cards, right? Or am I wrong?

PcChip - give me your best advice, should I keep my next optimization to myself or post it?

I can't answer that my friend, only you can.

If you're the type of person who loves tinkering with asm/opencl/hardeware code just for fun, and improving things for the community (which it seems like you are, and a redditor as well - highfive! [Did you go to a reddit meetup saturday?]) then it makes sense to continue posting them like you are.

If you're in bitcoin for profits, and want to use your intelligence to tweak the code to give yourself a (totally fair) advantage over the competition, then obviously it's not in your best interest to post it.  Note that I'm not saying there's anything wrong with this scenario, because I can't even say which route I'd pick myself.

I can say however, that I very much appreciated that speedup you posted, as did many others, and I hope you get many more BTC than the five you currently had donated to you at http://blockexplorer.com/address/15igh5HkCXwvvan4aiPYSYZwJZbHxGBYwB

[BombaUcigasa]

I can say however, that I very much appreciated that speedup you posted, as did many others, and I hope you get many more BTC than the five you currently had donated to you at http://blockexplorer.com/address/15igh5HkCXwvvan4aiPYSYZwJZbHxGBYwB

He actually had 1.4 BTC from other sources, but surely for such a small and important contribution he is rewarded decently.

This is one of those moments when you think... is it right to make over 60 bucks for 1 LOC? Is it right to make just a measly 60 bucks for giving EVERYONE 3% more money?

[PcChip]

This is one of those moments when you think... is it right to make over 60 bucks for 1 LOC? Is it right to make just a measly 60 bucks for giving EVERYONE 3% more money?

Do the math on how much extra income hundreds of people are getting, and then answer that.

Also, I didn't want to say anything at first since it could have been an anomaly, but I've now realized this on four separate cards - my GPU usage (and thus hashrate) has always fluctuated like a small sine wave from 99% - 90% , after this patch, they all stay a straight line 99% , so that's way more than a 3% increase in hashrate for me.

[coblee]

I agree Bitless deserves more donation. But the truth about mining is that if EVERYONE gets 3% improvement, the difficulty level will just go up by 3% and everyone ends up making the same. So that's why some people may be hiding their private optimizations so that only they alone benefit. Anyways, but I truly appreciate what Bitless has done. Hope you get more donations.

[BombaUcigasa]

my GPU usage (and thus hashrate) has always fluctuated like a small sine wave from 99% - 90% , after this patch, they all stay a straight line 99% , so that's way more than a 3% increase in hashrate for me.

What's weird, mine went from 90%-96% to 96% straight but the added performance adds 1.5C to the temperature, I might have to clock down a bit to get tot he previous 90C stable line, which means I am forfeiting the benefits of the above optimization. At least I use less energy, the Earth will thank us all...

[simplecoin]

Donation sent. not much, but hopefully we can get a bunch of little donations headed your way!

[bitless]

my GPU usage (and thus hashrate) has always fluctuated like a small sine wave from 99% - 90% , after this patch, they all stay a straight line 99% , so that's way more than a 3% increase in hashrate for me.

What's weird, mine went from 90%-96% to 96% straight but the added performance adds 1.5C to the temperature, I might have to clock down a bit to get tot he previous 90C stable line, which means I am forfeiting the benefits of the above optimization. At least I use less energy, the Earth will thank us all...

Same here, actually. I was ok with 6870 overclocked to 1000 MHz, but after the change it became unstable so I had to go down to 975, which is kinda sad. Yet, I blame it on poor ventilation in my case, so may be I'll get it running at 1000 again, eventually

[Rob P.]

Just tossing my confirmations out there:

Sapphire 5830s:

Pair 1:  264 Mhash/sec -> 271 Mhash/sec (each) at 875 MHz GPU, 900 MHz RAM (2.58% increase)
Pair 2:  273 Mhash/sec -> 281 Mhash/sec (each) at 900 MHz GPU, 600 MHz RAM (2.84% increase)

I'm also tracking Rejected shares.

Before Patch:
Pair 1:  ~2.6% rejects (after 800+ shares)
Pair 2:  ~3.2% rejects (after 800+ shares)

After Patch:
Pair 1:  ~3.0% rejects (after 650+ shares)
Pair 2:  ~4.0% (after 750+ shares)

So, the increase appears to also cause additional rejects that should be discounted from total increase gain, cutting it down a full 1% in my case.

Bitless, sent my donation to you as well.  Thank you for sharing your findings, giving to the community is what this should be about.  There will be plenty of time for profits later, here's to hoping you'll be adequately rewarded for your finding and work.

[bitless]

Bitless, sent my donation to you as well.  Thank you for sharing your findings, giving to the community is what this should be about.  There will be plenty of time for profits later, here's to hoping you'll be adequately rewarded for your finding and work.

Thanks a lot! I hope you'll figure out the issues that you've having with your mining after the patch. If I think of anything that can be of help, I'll post it too.

[MiningBuddy]

I got ~10mhs increase on my 5870's and surprisingly my stale rate actually is a LOT LOT lower then before the mod.

I will certainly be donating as soon as I have some cleared funds bitless, thank you so much for your hard work and I look forward to your next modification you speak of previously. (Do you have any more details on that atm?)

[bitless]

I got ~10mhs increase on my 5870's and surprisingly my stale rate actually is a LOT LOT lower then before the mod.

I will certainly be donating as soon as I have some cleared funds bitless, thank you so much for your hard work and I look forward to your next modification you speak of previously. (Do you have any more details on that atm?)

Everything that I've heard so far suggests that the stale rate is completely unrelated to the mod - it is somewhat random (esp. given that hdminer uses something that's equivalent to the mod as well). However, I'm glad that it is low for you at the moment.

No, I don't have any ETA on other modifications. They are complex and may never materialize into code... in which case I'll just post my thoughts on the subject.

Thanks for the donation!

[Gameover211]

Yup worked for me exactly as you said. My 6870 went from 300 to 308. Thanks!

[goxed]

Yup worked for me exactly as you said. My 6870 went from 300 to 308. Thanks!

It definitely worked for my 5830! Awesome thanks a bunch bitless

[phorensic]

This community is awesome.  Nice little speed boost.  I'm at the ragged edge on core voltage and clock speeds, so this boost was very cool!

[abyrnes81]

Thanks for sharing, it worked well for me.

AB

[X68N]

my GPU usage (and thus hashrate) has always fluctuated like a small sine wave from 99% - 90% , after this patch, they all stay a straight line 99% , so that's way more than a 3% increase in hashrate for me.

What's weird, mine went from 90%-96% to 96% straight but the added performance adds 1.5C to the temperature, I might have to clock down a bit to get tot he previous 90C stable line, which means I am forfeiting the benefits of the above optimization. At least I use less energy, the Earth will thank us all...

Same here, actually. I was ok with 6870 overclocked to 1000 MHz, but after the change it became unstable so I had to go down to 975, which is kinda sad. Yet, I blame it on poor ventilation in my case, so may be I'll get it running at 1000 again, eventually

Just run the card with open Case.
Also a good improvement is add 2x 120mm 10db-Silent Fans, that cool your card more Silent.

I run my 6870 @ 1005Mhz , 60 degrees Celsius , 60% Fanssped (card fan) @ 302Mhash/s.
(now with speedfix @ 311Mhash)

[zeezy]

Thanks for sharing, it worked great for me. About 2.5% it looks like!

[ahitman]

How does one get an accurate reading of their rates? My rates seem to jump randomly between 363 to 367 on one of my 5850 and 361 to 364 on my other 5850. Should they both be the same? Should they be jumping this much?

[sc8nt4u]

How does one get an accurate reading of their rates? My rates seem to jump randomly between 363 to 367 on one of my 5850 and 361 to 364 on my other 5850. Should they both be the same? Should they be jumping this much?

I run headless w/ aggression=19, so my hash rate is  pretty stable.

[Junglebrother]

Wow, have to try, thanks for sharing. 

[Bert]

How does one get an accurate reading of their rates? My rates seem to jump randomly between 363 to 367 on one of my 5850 and 361 to 364 on my other 5850. Should they both be the same? Should they be jumping this much?

It depends on the mining software you use.
I use phoenix and it by default averages every 10, so I increased this (-a 50). It takes longer to settle down, but once there it will not jump about as much.

[BombaUcigasa]

How does one get an accurate reading of their rates? My rates seem to jump randomly between 363 to 367 on one of my 5850 and 361 to 364 on my other 5850. Should they both be the same? Should they be jumping this much?

-r 3

[PcChip]

my GPU usage (and thus hashrate) has always fluctuated like a small sine wave from 99% - 90% , after this patch, they all stay a straight line 99% , so that's way more than a 3% increase in hashrate for me.

What's weird, mine went from 90%-96% to 96% straight but the added performance adds 1.5C to the temperature, I might have to clock down a bit to get tot he previous 90C stable line, which means I am forfeiting the benefits of the above optimization. At least I use less energy, the Earth will thank us all...

Same here, actually. I was ok with 6870 overclocked to 1000 MHz, but after the change it became unstable so I had to go down to 975, which is kinda sad. Yet, I blame it on poor ventilation in my case, so may be I'll get it running at 1000 again, eventually

That means your 1000 MHz core wasn't actually stable to begin with, but you were getting by because it wasn't being used fully.  After the patch it's getting fully loaded which puts more stress on it; Because of that, at 975 MHz with the new patch you might still be getting a higher overall hashrate than before the patch at 1000 MHz

[RedLine888]

5870

970\365

430,5 -> 439,4 ~2.07%

W7x64 SP1, Phoenix 1.5 + Phatk, APP 2.4, Catalyst 11.6

 

THANX A LOT

[jaebird]

Worked for me as well. Thanks.

[grid]

OC'd 6970 went from 400 to 417MHash/s. Sent some mBTC your way!

[jones]

awesome thank you!

[Existence]

Good job. Thanks a lot!

[glassine]

Went from 380 to 391 on my 5870 @900MHz.

Nice work!

[jdebunt]

is it possible this doesnt work on nVidia cards? both with OpenCL miner & CUDA miner i'm still getting the exact same hashrate, even after restarting Guiminer.

or am i missing something? (using windows 7 btw)

[bitless]

is it possible this doesnt work on nVidia cards? both with OpenCL miner & CUDA miner i'm still getting the exact same hashrate, even after restarting Guiminer.

or am i missing something? (using windows 7 btw)

Sorry, this only works for ATI cards that support BFI_INT. In general, nVidia cards are much worse for mining, so I'm not sure if there's effort at all going into making nVidia's stuff mine faster...

[Tartarus]

Got the expected bump on my 6950, thanks!

[dovewing2000]

will give this a try after i learn to setup command line miner... thanks!

[KissMate]

Did the job there too with a 6850, solid 2.5% increase. Thanks!

[theunforgiven]

Yay! got the expression:

Y&(Z^X)|X&(~Z^X)

[mrb]

Dude, awesome   so this accounts for 3% out of your 6.4% improvement, where are the other 3.4% coming from? I don't have 250 btc, in fact I only have ~2 that have been donated to me so far. (Please note - if you decide to share this, post it,don't pm me )

I think I deserve the right to keep this remaining ~3% a secret of mine. I have contributed a lot to this community. AFAIK I was the...

* first to show how to down-plug x16 cards in x1 slots by cutting PCIe slots and/or shorting pins A1-B17 http://blog.zorinaq.com/?e=42
* first to use BFI_INT, and document how to patch a kernel with it: http://blog.zorinaq.com/?e=43
* first to document the PCIe 12V rail redirection hack http://blog.zorinaq.com/?e=44
* first to present a cheap solution to cool 4 double width cards http://blog.zorinaq.com/?e=47

Now all the miners are using BFI_INT; shops are selling pre-modified PCIe extenders; etc. Enjoy :-)

[jdebunt]

is it possible this doesnt work on nVidia cards? both with OpenCL miner & CUDA miner i'm still getting the exact same hashrate, even after restarting Guiminer.

or am i missing something? (using windows 7 btw)

Sorry, this only works for ATI cards that support BFI_INT. In general, nVidia cards are much worse for mining, so I'm not sure if there's effort at all going into making nVidia's stuff mine faster...

Yeah i know Nvidia cards aren't too good for mining, but i was hoping for a slight increase... oh well

Thanks for pointing it out tho, much appreciated

[bitless]

Dude, awesome   so this accounts for 3% out of your 6.4% improvement, where are the other 3.4% coming from? I don't have 250 btc, in fact I only have ~2 that have been donated to me so far. (Please note - if you decide to share this, post it,don't pm me )

I think I deserve the right to keep this remaining ~3% a secret of mine. I have contributed a lot to this community. AFAIK I was the...

* first to show how to down-plug x16 cards in x1 slots by cutting PCIe slots and/or shorting pins A1-B17 http://blog.zorinaq.com/?e=42
* first to use BFI_INT, and document how to patch a kernel with it: http://blog.zorinaq.com/?e=43
* first to document the PCIe 12V rail redirection hack http://blog.zorinaq.com/?e=44
* first to present a cheap solution to cool 4 double width cards http://blog.zorinaq.com/?e=47

Now all the miners are using BFI_INT; shops are selling pre-modified PCIe extenders; etc. Enjoy :-)

Sure, of course you do, dude! It was just a joke sorry! By the way, I absolutely love your blog - I've read it a lot, just didn't know it was yours.

[robin007bond]

I tried using Phoenix, but my GPU isn't recognized. Neither with phatk.

I'm running Arch Linux with an Ati Radeon HD4650. DiabloMiner detects my GPU, and I installed AMDStream, so that isn't the problem.

[stealth17]

5870 (930/350), 410->421, 2.6%, thx 

[fizzisist]

I did this with phatk in phoenix-1.50 and I got a 2.4% increase on both my 5870 and my 5770. Thanks!

[hennessy]

Getting some good results, but one out of 7 workers always locks up after some minutes. GPU usage goes down to zero, shown MHash rate stays locked at last value, cursor is still flashing. Occurs only when the cards are overclocked, but the same and even higher OC values work flawless without this fix.

[HolodeckJizzmopper]

6970 @ 975/400 from ~420 Mhash/S to ~432 Mhash/S

[nemesis51]

400 -> 412 on my 5870. great!

[Patheos]

Anyone tried this using guiminer?  I can see the file but not sure if I just edit the portion:

// Ma can also be implemented in terms of BFI_INT...
   #define Ma(x, y, z) amd_bytealign((y), (x | z), (z & x))

[Krypta]

Hy all, I'm new here and pretty new with this bitcoin thing.

I tried to modify phatk kernel as explained and I was able to speed up my mining as all of you.

Really don't know when and how this amd_bytealign function is called, anyway I think it should be better to write it as follows

#define Ma(x, y, z) amd_bytealign(((z) ^ (x)), (y), (x))

That said, I took a look at the rest of the phatk kernel and found what seems to be a sintax error. In my file it's on line 74:

#define r1(x) ((rot(x],15)^rot(x,13)^((x)>>10U)))

Mmmm... that poor and lonely square bracket... rewrite it as follows

#define r1(x) ((rot(x,15)^rot(x,13)^((x)>>10U)))

These modifications did not speed up my mining power but it seems that my rejecteds almost halved.

If this helps you as well, donations are obviously welcome

1CbRmFAYrGYhzQibmE5A1CWMp7srFT4upy

Krypta

[hughes]

Before: 353 MHps
After:  364 MHps
Difference: 11 MHps (3.0%)

Thanks!

[redcodenl]

Krypta:

I've seen that before as well, I didn't change it, but it sure looks like it shouldn't even compile!
I haven't tried the alterations yet, but will do so when I come home!

UPDATE: It works great, increase on both 6870's from 287 to 296 (> 3%!)

[Jen4538]

thank you very much , this has increased my rate


Jen

[msb8r]

Anyone tried this using guiminer?  I can see the file but not sure if I just edit the portion:

// Ma can also be implemented in terms of BFI_INT...
   #define Ma(x, y, z) amd_bytealign((y), (x | z), (z & x))

This patch works fine with guiminer as well. And yes, just edit the portion you posted.

[Fiyasko]

Seeing all these success stories makes me wonder if the file i downloaded was actually assembled correctly.
When apply this little 'tweak'
my mh/sec rate increses by seeming 2mhash/s per 6870 core But, the value at what im hashing at jumps all the fuck over the place. like 20-150-270-279-410-4-89.
It's rediculos

[Fiyasko]

Seeing all these success stories makes me wonder if the file i downloaded was actually assembled correctly.
When apply this little 'tweak'
my mh/sec rate increses by seeming 2mhash/s per 6870 core But, the value at what im hashing at jumps all the fuck over the place. like 20-150-270-279-410-4-89.
It's rediculos

Yup, T'was compiled wrong.

Everyone update your GUIminer client
This beautiful tweak fix has been intergrated into the newest update, No weird fiddling required

And i can confirm that i've gained 10mh/s/core now 6870BE

[ercolinux]

It works really well:: +3MHs on my 5770 at stock frequencies.

[bitless]

Krypta - it looks like r0() and r1() aren't used; try deleting them if you like.

Note that they aren't used by the phatk kernel that I've got, and even though I can't think of circumstances in which r0() will compile correctly, still, be careful if your kernel is different!

[Krypta]

Thank you bitless, it seems you are right, they are never used, so I really don't know why these modifications halved my stales.

Here's another strange thing:

Code:

256	Vals[0]=H[0];
257	Vals[1]=H[1];
258	Vals[2]=H[2];
259	Vals[3]=H[3];
260	Vals[4]=H[4];
261	Vals[5]=H[5];
262	Vals[6]=H[6];
263	Vals[7]=H[7];
264
265	Vals[7] = 0xb0edbdd0 + K[0] + W[64] + 0x08909ae5U;
266	Vals[3] = 0xa54ff53a + 0xb0edbdd0 + K[0] + W[64];

I think we can comment out

Code:

259	//Vals[3]=H[3];
263	//Vals[7]=H[7];

because they are changed on lines 265 and 266 before they are used, am I right?

Thanks.
Krypta

[bitless]

You may be right. May I suggest you run the kernel analyzer and see the disasm of both versions and see if the generated code is any different?

(This is because things like x=1; x=2; x=3; are easily converted by the compiler to just x=3;  )

[joulesbeef]

put my 5830 up to 323.6.. very nice

[Krypta]

You may be right. May I suggest you run the kernel analyzer and see the disasm of both versions and see if the generated code is any different?

(This is because things like x=1; x=2; x=3; are easily converted by the compiler to just x=3;  )

You are right, it's the same 

No improvement then...

K.

[Palmdetroit]

Thanks boosted mah 6770s ~3%

[karoshi]

wow, thanks.

each of my cards is making 357 now instead of 347. I thought these kernels were optimized to the max.

[dafuq]

hehe will try it thx

[Gui_0]

my GPU usage (and thus hashrate) has always fluctuated like a small sine wave from 99% - 90% , after this patch, they all stay a straight line 99% , so that's way more than a 3% increase in hashrate for me.

What's weird, mine went from 90%-96% to 96% straight but the added performance adds 1.5C to the temperature, I might have to clock down a bit to get tot he previous 90C stable line, which means I am forfeiting the benefits of the above optimization. At least I use less energy, the Earth will thank us all...

You're still being more efficient and saving money by using less energy. His optimization made your load a constant 96% instead of 90%-96%, so your GPU has less time to cool. You should still get as many hashes as you did before, if not more, plus less energy

[Diapolo]

Did you guys try my phatk kernel-mod from here: http://forum.bitcoin.org/index.php?topic=25135.0
I really would like to get some feedback .

Thanks,
Dia

[Alan Lupton]

Did anybody experience more rejected packages? I used to be a 99.xx% guy, but now I'm down to 96.5 - 97%. I'm talking about 6 workers on 2 different systems. Coincidence?

[xurious]

This worked for me on all my cards. Roughly 2-4% on 3x 6850's, 5870, and 5770. Thanks!

[Rob P.]

put my 5830 up to 323.6.. very nice

Hmmm, that's a helluva rate for a 5830.  Can you share your voltage, GPU, and RAM clocks please?

[klaaster]

I'm getting 327 with this on a Sapphire Xtreme 5830.

Ubuntu11.04_64, 1030/300/1.2V/66°C

[carbonc]

I'm getting a stable/cool stock voltage setting 323 MH/s
Win 7 64-bit
975/300/stock volt/69°C

Check this post on how:
http://forum.bitcoin.org/index.php?topic=28825.msg362947#msg362947

[XIU]

Thanks for the patch, got a few extra MH/s

[giv]

Hi guys.
On my 6950 this patch DECREASES performance about ~10MH/s. So if you have 69xx be careful and check your performance.

[norulezapply]

Perfect, thanks!

[film2240]

This is gr8.

I got a boost as well whilst saving energy on my Pc by 11Watts as well from GPU side of things

My card Radeon Hd6950 is clocked 895MHz and 947MHz (due to stability issues the clocks are here for mem)
This has been boosted from ~394 to ~398MHash/s on the card.

This is what happens when we improve on programming code (you save system resources,use less energy and increase performance as well.Every1 wins :-))

[film2240]

Hi guys.
On my 6950 this patch DECREASES performance about ~10MH/s. So if you have 69xx be careful and check your performance.

I used to have this issue,it turns out you need to apply this change to the kernel.cl file in phatk,poclbm folders (both which u find in kernels folder) and apply this chaneg to this file BitcoinMiner.cl in the GUIminer root folder as well,then you'll be boosted as well.

Hope this helps :-)

[DiamondPlus]

Just tossing my confirmations out there:

Sapphire 5830s:

Pair 1:  264 Mhash/sec -> 271 Mhash/sec (each) at 875 MHz GPU, 900 MHz RAM (2.58% increase)
Pair 2:  273 Mhash/sec -> 281 Mhash/sec (each) at 900 MHz GPU, 600 MHz RAM (2.84% increase)

I'm also tracking Rejected shares.

Before Patch:
Pair 1:  ~2.6% rejects (after 800+ shares)
Pair 2:  ~3.2% rejects (after 800+ shares)

After Patch:
Pair 1:  ~3.0% rejects (after 650+ shares)
Pair 2:  ~4.0% (after 750+ shares)

So, the increase appears to also cause additional rejects that should be discounted from total increase gain, cutting it down a full 1% in my case.

-DiamondPlus