Research on graphics

[riodoro1]

Sadly there are no advanced settings like RAM timings in our setup

 

[leostone]

Too bad, but maybe investigation will lead to a fix that can be applied on the OS side ...

 

[leostone]

it happened already ...
if anyone likes to play: TRY THIS

 

[BigDonkey]

it happened already ...
if anyone likes to play: TRY THIS

I usually have lines on fresh installs. Haven't had one in three weeks. I modified timing for our patched kexts from 1ms to 5ms as per your description. I hope I got "the offset" right for each version.
Posting below for everybody else to try.

 

[mac4mat]

EDIT: Disregard, no difference

 

[leostone]

@BigDonkey: You said you've searched the 10.6.8 binary for 0x07C70424. What i'd like to do
is to disassemble it and compare the logic in that function as long as there is the same function. Just to know if I hit a real change from 10.6.8 or if the disappearing of the lines might be just a lucky side-effect of changing this timeout. Do you think you could povide me with the 10.6.8 binary?

 

[trusteh]

it happened already ...
if anyone likes to play: TRY THIS

I usually have lines on fresh installs. Haven't had one in three weeks. I modified timing for our patched kexts from 1ms to 5ms as per your description. I hope I got "the offset" right for each version.
Posting below for everybody else to try.

Thanks. I'm going to test it this afternoon, let's see what happens.

 

[riodoro1]

It seems better with this kext, i will stress it at home, as i noticed that several games and software increases glitches.

 

[BigDonkey]

Hey guys!

After so many failed attempts to keep this beast under control let's hope leostone is into something this time.

@leostone
here's the binary from 10.6.8 kext
AppleIntelSNBGraphicsFB.zip

 

[leostone]

thx! Well, as i was afraid, this timing is exactly the same in 10.6.8 ....
BUT a bit deeper down in the code they introduced some check on the display or somthing, which consists of a even smaller timeout, and this looks odd to me, cause if there is a fail? they just abort ... -->the abort==lines??? So the bigger timeout further up, changed by us causes the check to succeed more often, or always????

Anywayz, what i will do is ... jump over this check and continue just as in 10.6.8 ...
and set the timeout back to where it was .....

I'll keep you posted