ATI OpenCL MultiBeam app (rev177)

[Raistmer]

So what experiment with card swap gave?
Especially interesting what with host where NV GPU disappeared from output...

[Claggy]

So what experiment with card swap gave?
Especially interesting what with host where NV GPU disappeared from output...

It was where you couldn't run any ATI Brook+ or ATI OpenCL apps standalone (with a Nvidia GPU fitted in the 1st slot),
all offline app runs Crashed, to do any of the Initial AP OpenCL testing i had to remove the Nvidia GPU,
and with Boinc 6.10.56 and earlier, the Boinc client crashed, Boinc 6.10.58 just detects the SIGGEV violation and ignores the ATI GPU,
i've had to use a workaround (renaming two ATI drivers) to get ATI & NV GPU's to work in Boinc,
Ghost and others have had their NV & ATI GPU's fitted the other way round, and it worked in that configuration apparently,

Claggy

[Raistmer]

And now both GPU positions work OK ?
With Cat 11.2 ?

[Claggy]

And now both GPU positions work OK ?
With Cat 11.2 ?

Yep, it's fixed in my system with NV GPU in 1st slot, ATI in 2nd, not tried the other way round, or tried Cat 11.1,
I think Ghost already had it working the other way round,

Claggy

[corsair]

good day to you all,

my first try with MB ATI:

http://setiathome.berkeley.edu/result.php?resultid=1887858163

if could helps, good.

it tooks 26m 41s in a AMD 4770 - 512 RAM the machine infor is in the report.

[corsair]

Hi all,

just one small question, why is not a native version for systems running win x64??

(well and if they were optimized form AMD64 that would be great)

[skildude]

Seti isnt helped by running in 64 bit mode.  Check their forums its been asked and answered there numerous times in the last year alone

[corsair]

Seti isnt helped by running in 64 bit mode.  Check their forums its been asked and answered there numerous times in the last year alone

But, why not to have the software unified in the same architecture, could be a cosmetic thing, but a unifying way for the near future of x64 o.s. and software

it's still early (some more years to come with old build)  but we should have been thinking about R.I.P. of x32 buildings

[Raistmer]

Seti isnt helped by running in 64 bit mode.  Check their forums its been asked and answered there numerous times in the last year alone

But, why not to have the software unified in the same architecture, could be a cosmetic thing, but a unifying way for the near future of x64 o.s. and software

it's still early (some more years to come with old build)  but we should have been thinking about R.I.P. of x32 buildings

The main reason - severely limited resourses. Most important - development time.
So only high priority improvements will go at resonable speed.

[Fredericx51]

I do have a host, i7-2600+2x HD5870, here are some results, last validated, are ther more, f.i. cmd line
params, last validated MB.

Also, ran out of AP work, done with rev.516 U=8;Blck=8196; Fetch 4096. Any Unroll above 10,
 depending also of card-model, gives inmediat screen lag and is not necessary to keep GPU load
high, ffa_block and ffa_block_fetch, are equally important, but also card dependeble.

Whigh, I say cards, cause, in a lot of cases a 5850 is a 'stripped/or some cores not connected' version,
of the card.

Unfortunatly, GPU load stays low, 0%-31%- long tine zero, then 31%-61%-99%-11%-0.

Which also causes the 5870, to clock down, 850-1200MHz., to 500- 800 and goes back to 157-400,
until anost 'burst'comes along. Times between, those bursts, are also unpredictable(?!)

Also a screen-lag is there, too heavy, compaired to the GPU load, IMHO.

Is there yet another version, that needs, tuning,(testing?) . (Well also slowly recovering fron surgery)

I also still, use your AP ATI rev.516, but ran out of AP's.  (Got hundreds or more, well I like them!)