Differences in modelines legacy vs ADL

[haynor666]

Some time ago I've spot difference in target vs received modeline when I compare legacy timings vs new ones. For example with Radeon HD4350 and target 256x240@60 I receive (did I ?) 256x240@60.03 but with new timings I get 256x240@59,977.

My question is are switchres timigs posted before I run game are target timings that groovymame wants to get or timings calculated by system ? Why there are differences between those two solutions ?

[Calamity]

Not sure what you mean.

If you mean the refresh that is shown in the information screen (GroovyMAME), that's the target refresh calculated by the modeline generator, which is totally independent from the specific implementation (AMD ADL, ATI legacy, Powerstrip or xrandr).

The actual achieved refresh cannot be known directly, you need to measure it (as Arcade OSD does when you press "5").

[haynor666]

Yes, I wa thinking about switchres information at the bottom of information screen. So if this is target then why those two differ?

[Calamity]

So if this is target then why those two differ?

They can't differ. It could be you're using a different monitor preset in each case.

[haynor666]

Well, I use exactly the same driver, the same tools with the same configuration and somehow differ.

[Calamity]

The relevant preset is in mame.ini.

[haynor666]

I didn't touch mame.ini as well and I in both cases used groovymame 170 

[Calamity]

Then a log taken from both situations should show us something.

[haynor666]

Here they are.

[Foxhole]

I've just installed a 7850 sapphire card, and i'm experiencing exactly the same thing.
Any game that i ran using an hd 4350 gave me almost perfect timings, if said game runs in 320x224@59.185606, switchres would show 2560x224@59.186.
With the 7850, switchres shows 59.178 Hz. This happens with every game.
Also tried changing presets, didn't help.
Using GM ASIO 0.169.
I've also noticed this in your guide, Calamity.
http://aburamushi.net/calamity/img/5450/20160201_200304.jpg

[haynor666]

Thanks for confirming FoxHole. I'm glad it's not another my strange problem

[Calamity]

Thanks, I'm seeing the problem now

I was right regarding both implementations are using the exact same modeline. However the ADL implementation is special because it reads the actual modeline from the driver right after setting it, so a rounded dotclock value is read back and ends out showing as a modified refresh in the Switchres information info. In the legacy implementation I was using the original dotclock all the time so people wouldn't complain, I need to fix the ADL one so the theoretical refresh is shown too.

[haynor666]

So in ADL those values are actually what we got from driver? If yes than I think it's better to get actual timings rather desired.

[Calamity]

So in ADL those values are actually what we got from driver? If yes than I think it's better to get actual timings rather desired.

I see what you mean, this has already been suggested a number of times. The problem is, that dotclock value isn't the real one either, it's just the desired one aligned to 10 kHz. The real one can only be guessed by measuring the actual refresh. So showing that value only serves to further confuse people.