further improved phatk_dia kernel for Phoenix + SDK 2.6 - 2012-01-13

[actudoran]

I got a pair of 5830 and they run diferently as one is closer to the CPU cooler that's pumping out about 125 w of residue heat from cpu ...
the "hot one" 74 - 80 C ( varying upon day/night ambiental temperature ) runs at 990/300 getting about 308 - 314 mhas  stok volts
the cool one 70-77 C running flat out 1040/360 332mhas avg 318 - 338 @ 1.163 V

boy they do give out some heat ...  at night the temp didn't drop under 30 C where I keep em ... about 18 Sqm of a well ventilated room ( big window )

Any of you keeping rigs in the basement / garrage ? what's that like tempwise ?

[ditchmagnet]

Really loving this kernel!  Looking forward to the next release!

So does the 11.7 preview give much of a performance increase with this?  Currently getting about 329 Mhash/s @ 1030/350 on 5830 Xtremes.

I'm assuming you jumped on the newegg bandwagon too? I just started (wasnt aware of a dummy plug so I only have 2 running tonight, will have 5 after dummy plugs), and I can only get 970/300 stable @ 100% fan and 1.195 core. Temperatures are 78c., 309 mhash

How can you push the core so far? Did I just get a bad card? ty in advance

I run my fans at 75%.  As for taking off the shroud, it will hurt cooling if the card has access to good air, but when they are sandwiched, I found this setup to work decent:

http://imageshack.us/photo/my-images/219/img20110628120156.jpg/

Also, try putting on some fresh TIM.

I wouldn't say you have a bad card since you are still OCed.  And ya, got most of mine from newegg when they had the $110 shipped cards - and 1 I have is a used one I traded for.

[ssateneth]

I finally got some dummy plugs and got 3 5830s working in 1 PC, though its not "in the open" like yours is, Furthermore, they are sandwiched very tightly (only 1 "blank" slot between the 16x slots instead of 2), so much so that when initially testing it before dummy plugs, the fan from 1 video card loudly grinded against the back of the card in front of it. It was very sudden (at 100% fan speed no less) and I pulled the power plug in a panic. I've since made all 3 "naked" with a drop of AS5 I've had lying around, but I'm not sure if its good anymore (its really hard to force out of the tube, but its still very tacky, like a slow running glue). Regardless, I also tied the PCI-E Power cables to parts of the case so it bends the top and bottom card away from the middle one a little bit to rpevent the fans from grinding again, and to give access to better air. Finally, when I get a drill that actually works, I'm going to screw this directly on the case side so I can close it. http://www.youtube.com/watch?v=RyMUddH0qzY (not my video, but exact same fan)

[simonk83]

Very nice, thanks for this

2 x 5870's, both running at 970/320
11.6 drivers, 2.4 SDK

Before:  429Mh each
After:  443Mh each

[Diapolo]

Very nice, thanks for this

2 x 5870's, both running at 970/320
11.6 drivers, 2.4 SDK

Before:  429Mh each
After:  443Mh each

Another great result ... I had other things to do, but during the next week I will release a new version.

Dia

[P1n3apqlExpr3ss]

Works nicely, went from 298~ to 301~ on a 6870 @ 975/440

[ssateneth]

What happened to the estimated huge gains with your unreleased patch? I could've sworn the first page had a difference of about 7 ALU less between current version and unreleased version, now it's only 2. O_o

[1984]

355 Mh/s to 370 Mh/s on a stock hd6950, I've seen no difference apart from the increased hash rate, keep up the good work!

[checkmate]

Nice... I went from 295 to 307! Thanks!

[Diapolo]

What happened to the estimated huge gains with your unreleased patch? I could've sworn the first page had a difference of about 7 ALU less between current version and unreleased version, now it's only 2. O_o

I can explain this, I made a change in the kernel, checked it via KernelAnalyzer and it looked good. Then it was time for some sleep, checked the next day with Phoenix and observed, that all shares were invalid, so I had to revert the changes and the numbers on page 1. Sorry, my mistake!

Dia

[Diapolo]

Important: since version 2011-07-17 a modified version of __init__.py (for the Phoenix 1.5 miner) is included in this package and has to be used!
The kernel won't work for other Miners without modifications to them, see kernel.cl for further infos.


The new version 2011-07-17 is ready for download . Should be faster on 58XX and 69XX cards again.
This version will only work, if you use it with Phoenix and the supplied __init__.py file because of modifications to kernel variables!

A very big thank you goes to user Vince for input and ideas!

Download here:
http://www.mediafire.com/?317u0y93u7mnbys

Have fun,
Dia

[bcheck555]

roughly .5Mhash/s increase with latest. Thanks.
337.20>337.86
5830 1050/350 -k phatkmod DEVICE=1 VECTORS BFI_INT FASTLOOP=false WORKSIZE=256 AGGRESSION=13

[urstroyer]

very nice work, thanks a lot!

[BeeTeaSea]

Hi,

I am getting this error with a ATI 5870. The modified kernel from July 6th works fine.

./phoenix.py -u http://XXX DEVICE=1 -k phatk VECTORS BFI_INT WORKSIZE=256 AGGRESSION=14 FASTLOOP=true -q 8
[17/07/2011 16:58:36] Phoenix 1.50 starting...
[17/07/2011 16:58:36] Connected to server
[0 Khash/sec] [0 Accepted] [0 Rejected] [RPC]Traceback (most recent call last):
  File "/usr/lib/python2.6/threading.py", line 504, in __bootstrap
    self.__bootstrap_inner()
  File "/usr/lib/python2.6/threading.py", line 532, in __bootstrap_inner
    self.run()
  File "/usr/lib/python2.6/threading.py", line 484, in run
    self.__target(*self.__args, **self.__kwargs)
--- <exception caught here> ---
  File "/usr/lib/python2.6/dist-packages/twisted/python/threadpool.py", line 207, in _worker
    result = context.call(ctx, function, *args, **kwargs)
  File "/usr/lib/python2.6/dist-packages/twisted/python/context.py", line 59, in callWithContext
    return self.currentContext().callWithContext(ctx, func, *args, **kw)
  File "/usr/lib/python2.6/dist-packages/twisted/python/context.py", line 37, in callWithContext
    return func(*args,**kw)
  File "kernels/phatk/__init__.py", line 391, in mineThread
    self.output_buf)
  File "/usr/lib/pymodules/python2.6/pyopencl/__init__.py", line 202, in kernel_call
    self.set_args(*args)
  File "/usr/lib/pymodules/python2.6/pyopencl/__init__.py", line 224, in kernel_set_args
    self.set_arg(i, arg)
pyopencl.LogicError: clSetKernelArg failed: invalid arg size

[Tartarus]

Hi,

I am getting this error with a ATI 5870. The modified kernel from July 6th works fine.

./phoenix.py -u http://XXX DEVICE=1 -k phatk VECTORS BFI_INT WORKSIZE=256 AGGRESSION=14 FASTLOOP=true -q 8
[17/07/2011 16:58:36] Phoenix 1.50 starting...
[17/07/2011 16:58:36] Connected to server
[0 Khash/sec] [0 Accepted] [0 Rejected] [RPC]Traceback (most recent call last):
  File "/usr/lib/python2.6/threading.py", line 504, in __bootstrap
    self.__bootstrap_inner()
  File "/usr/lib/python2.6/threading.py", line 532, in __bootstrap_inner
    self.run()
  File "/usr/lib/python2.6/threading.py", line 484, in run
    self.__target(*self.__args, **self.__kwargs)
--- <exception caught here> ---
  File "/usr/lib/python2.6/dist-packages/twisted/python/threadpool.py", line 207, in _worker
    result = context.call(ctx, function, *args, **kwargs)
  File "/usr/lib/python2.6/dist-packages/twisted/python/context.py", line 59, in callWithContext
    return self.currentContext().callWithContext(ctx, func, *args, **kw)
  File "/usr/lib/python2.6/dist-packages/twisted/python/context.py", line 37, in callWithContext
    return func(*args,**kw)
  File "kernels/phatk/__init__.py", line 391, in mineThread
    self.output_buf)
  File "/usr/lib/pymodules/python2.6/pyopencl/__init__.py", line 202, in kernel_call
    self.set_args(*args)
  File "/usr/lib/pymodules/python2.6/pyopencl/__init__.py", line 224, in kernel_set_args
    self.set_arg(i, arg)
pyopencl.LogicError: clSetKernelArg failed: invalid arg size

Both the new kernel.cl and __init__.py need to go into kernels/phatk/

[Diapolo]

Both the new kernel.cl and __init__.py need to go into kernels/phatk/

That's it, I thought I made it clear enough .

Dia

[Bert]

Thanks for the hard work, sent you another donation. 458.88Mh/sec up to 459.41 with the last update on overclocked 5870's running at 1000/347.

[Diapolo]

Thanks for the hard work, sent you another donation. 458.88Mh/sec up to 459.41 with the last update on overclocked 5870's running at 1000/347.

Thanks Bert, you are one of the frequent donators and I really appreciate it ! Have fun with this latest version!

Dia

[Soak]

388 to 391 with last update (Radeon HD 6970). Now equal with original poclbm miner. Thank you!

[streetuff]

just wanted to report....

HD5770 from 220 -> 222MH/s
HD5850 from 355 -> 359MH/s

with 2011-07-17